Wow, easy to do, well explained and sounds great! Headroom correction needed in Roon: 0 dB! However, I have 3 questions: - I recently bought the Umik 2 microphone. In the IR windows, entering the ms corresponding to the distance of my speakers (3.2m = 9.41 ms, REF time 0.007), to see the behavior on the high frequencies, as you did, I can't see frequencies below 106 Hz. How is this possible? - I did not correct the microphone calibration with Minimum phase. Where can I find the procedure? - Multiplying 1/A with Harman Curve results in a mess of delay between the 2 channels in addition to the foreseeable need to use about 8db of Headroom. I can't understand where I'm wrong.
1. You shouldn't be able to see the response below 106Hz because a short right window size of 9.41ms means the minimum frequency you can detect is 1/(9.41/1000)=106.27Hz - frequency is just 1/time in seconds. I can see it with a trick but it's just an extrapolation by REW for convenience, is not correct by any means (Preferences/Analysis/Use right window for min frequency unticked!) 2. Somebody located it and linked it down here in the comments. 3. I don't recommend applying a target curve, I suggest turning up the volume if the bass is too weak not to your liking :) Again, I explained and suggested a way (LS filter) to increase bass which seemed to work with the channel member down below in these comments.
@@ocaudiophile Just for information: I applied the MP correction to the microphone and redid the procedure. Now to avoid clipping I have to give -4dB in Roon, before 0. What I changed is only the correct phase of the microphone.
How does the MP version of the mic cal file look like? If it's deviating from 0 degrees too much (more 20-25 degrees at the extremes) than you must have done something wrong during MP generation. Try adding tails and/or removing IR delays from the mic cal MP version.
Woah, This guy is at it again. I did the previous excess phase video and this one sounds like it will be a million times easier. Cool explanations on what Minimum phase is. Thanks Bro!
There was a little bit of Ringing with the 225Hz setting after I tried it out with increasing the range. Have to reduce it to below 200Hz I think to lessen Ringing on my end. It was not noticeable until I tried higher frequencies and then compared it On & Off a few times to note it was ringing even with 225Hz cut-off. You mentioned you might have to change it on our end and I can confirm this to be true. Increasing the range just causes LOTS of ringing for me. And ~200Hz seems to not have it.
You are probably right but HF is not a known cause of ringing. Precision becomes quite important with inversion. Also filters are longer than usual because of the zero padding required for the FFT (even after trimming). I would suggest to vector average multiple measurements with 1M length before starting to create the filters for best results.@@nighthog7003
Hello! Love your stuff, great logic - but a little fracmented I think, difficult to get the golden overview and understand what comes first. Will you maybe make a step-by-step tutorial for a complete system setup, with all the various steps from the differnt videos, in the correct order, to get the best results? Huge thanks in advance 🙂
Another great video, personally learning a lot and my musical enjoyment has been taken to another level. This video is very similar to the very first video I watched and made some major improvements to my system titled "REW (Room EQ Wizard) Top Tricks: Convolution with Inversion (no EQ filters, all FIR!)". This video cuts out the use of a House Curve so looking forward to the upcoming video re-introducing a House Curve\LS EQ Filter. A few questions/general comments for the masses: 1 - For my 1/A Inversion (From L/R-MP Copy), the IR Window sizes where huge, Left Window greater than 10,000ms. Before exporting I trimmed the 1/A window to R=100ms/L=500ms. Curious as to why so large or what I did wrong? 2 - For my 1/A Inversion (From L/R-MP Copy) I was getting pre-ringing with Roon. Creating a 1/A MP copy resolved that issue. 3 - A few comments below noted lack of bass, to resolve I imported the Harmon House Curve and multiplied with the 1/A MP Copy (L/R - 1/A MP * Harmon House Curve) which boosted my bass to better match my frequency level aboves 225Hz. I'll note to avoid pre-ringing weirdness must do against the MP copy and I had to adjust the IR Windows sizing from being super large.
Large window sizes are normal with inversion (it requires double the zero padding of the FFT in the background). In fact the cause for pre-echo is still the insufficient window sizes as a perfect inversion wouldn't ring.
@@ocaudiophile Thanks, not doing something wrong then. My default window size in preferences for left is 1000ms, right 2000ms. Do I have those values default correct?
It's amazing... The final filtering is great at improving bass. However, in some trial listening clips, the bass will produce a delayed ringing effect. If the music clip ends with bass, the party music will end, but the bass will continue in a background music like rumbling sound for about 1 second For example: ANNE-SOPHIE MUTTER "CARMEN-FANTASIE" Zigeunerweisen op.20 IV. Allegro molto vivace
@@aland8195 Ok, sorry I thought the comment was under Evo video :) AFter you create the final 1/A filters, generate their "Minimum phase versions" and use these as filters instead. That will completely remove the ringing.
@@ocaudiophile Thank you very much for sharing and answering I just checked the data of REW again, and the original curve (A) * 1/A filtering (B) shows a significant increase in RT60 data in the 150Hz frequency band. This may be the reason for the strange listening experience, which is somewhat strange.
This is really fantastic and I am currently experimenting with the inverted response as FIR filter. I am also starting with rephase and I have following question: When I invert without Minimum Phase the computed 1/A filter has the same phase characteristic as my measurement. Could I load this inverted response into rephase an do (only) a phase calibration there (no gains)? Then I would generate a kind of phase correction filter in rephase, export it, import into REW and then multiply it via Trace Arithmetic with the inverted response. This would then be my new phase-corrected FIR filter. I assume that your approach with the Minimum-Phase version in REW is still more efficient and the best for measurements from the listening position. My phase-correction idea is on my mind for speaker measurements and would be applied fullrange (I would even use this for digital crossover design with measurements of single speaker drivers) - basically as the optimized, starting filter for an acitve speaker system. When the system is in place, I would add a room correction filter with your method for the bass response. Sorry, I am really new to this and not firm (yet) with Rephase and I am not a happy user yet. I might be getting it totally wrong ... but am really inspired from you. Thanks a lot!
Always work with excess phase of a speaker measurement in rePhase, not the phase itself. Usually crossover phase corrections will suffice but you can also do port/box corrections if you have enough FIR taps to deal with bass frequencies. Check this video: czcams.com/video/ChPu0u3nZxc/video.html
I have no idea what you're talking about but you sound really intelligent so I'll keep watching. I'm confident that my IQ will increase as I watch your videos 😂😂
Hello sir, I'm most impressed and grateful for the content you're sharing. I've some questions (disclaimer: I'm a rookie so I may be very wrong). My priority is to flatten the low bass, because I have insane drop in 34-57 Hz range. 1. If I was to leverage all your latest methods for a 2.0 system with Roon (no audyssey), I suppose that would be combination of VBA + xover/box/port correction + the inversion below Schroeder. Is that right? 2. In what order should these filters be created and applied? My guess: 1) create VBA filter based on original measurement, 2) create xover/port filter based on VBA-applied trace, 3) create "inverted below Schroeder" filter based on xover/port-applied trace, 4) convolve inverted filter * xover/port filter, 5) convolve result of "4)" * vba filter 3. What REW measurement and analysis settings should be used for creation of each of these 3 filters? You specified one here czcams.com/video/ChPu0u3nZxc/video.html (with mic pointing at speakers and minimum phase cal file) and another one here czcams.com/video/zmIqJ_jlZPw/video.html . I'm guessing the same measurement needs to be used for creation of all 3 filters, right? So which one? 4. What's the optimal measurement length for these corrections? Will 1M be good enough? 5. Does it make sense to aim for nearly flat (at zero degrees) result of maximum phase when correcting fox xover/box? Let's assume such correction won't ring or anything. Because I can make it much flatter compared to what you've shown, and even with quite small Q values. Or are the corrections you've shown sufficient? Should we only care about 100-1000 Hz range when "flattening" the max phase response (as it's the range that's audible phase-wise)? 6. I have quite pronounced phase differences between left and right speaker in my listening position. Should I rather them separately (rather than based on LR averaged)? Because my LR average would have quite high std. Deviation, so I would be correcting against a max phase trace which isn't very applicable to neither left nor right speaker.
Thanks for the gesture. 1M length is good enough, the louder and the longer the better but rew will slow down and inversions etc will be extremely large beyond that. VBA, phase correction, inversion order seems right, don't forget to generate minimum phase of the final vba and inversion filters. 100-1000hz is good for phase but remember to correct only excess phase. If certain correction causes pre echo, try to just fix what's different in excess phase of left and right speaker.
Hmmm, just tried this following along on real time. I went back and loaded my microphone calibration and created a minimum 26:24 version. This new process is fairly straight forward. However, I am getting a ton of pre-ringing from the convo filter. Where did I go wrong?
Always respect you OCA. 😊 Minphase Inversion is simple, fast, and produces good results. Also, messurement graph is improved to make it look good to the eye. However There is a limit to one listening position on one speaker (or subwoofer) and eventually we will use up to the limit of drc when active treatment, inversion, and time domain calibration are performed simultaneously with multiple speakers (or dual, quad subwoofer). Thank you for making the video.
I watched this video maybe 30 times now :) and the others too . I made everything , sound very good , but lost a lot of base... First time i did it with the sub connected and just followed a long but the sub was working i had LFE + Main set . Second time i only made left and right (subwoofer unpluged) . Remains to test the next days . I would really appreciate if you would explain a little more about adding the harmon curve , i dont understand : ,,Drop the .txt file on to REW - you mean in the EQ ? 3. AxB (trace arithmetic) multiply that with each of the "1/A" filters" really lost me here Thank you and God bless you for sharing the knowledge
I understand you are using FIR filters in a PC and using an AV receiver as an amp (LFE+Main reminds Denon/Marantz gear) because otherwise you cannot use these filters with Audyssey. Anyway, the inversion is done on a flat line so there's no bass tilt. I've developed some solution for that already and its video is in the pipeline. But in the mean time, you can A x B multiply each of the 1/A filters with a low shelf filter and increase the bass as much as you want. You can generate a low shelf filter with REW in the EQ window: open EQ filters, select "LS" from the drop down list, frequency i.e.100Hz, gain +2dB and then "Generate measurement from filters" under "Filter tasks". This is your "B" in "A x B" trace arithmetic. You will need to adjust clipping headroom. PS if you hear any pre-ringing effect, generate minimum phase versions of each of the 1/A filters and multiply these MP versions with the LS filter to generate your final convolution filters.
Thank you very much . I have this : Pc - Smsl Su1 Dac - analog to Denon Avr 3500H and 2.1 speakers . Front sp cut 80 Hz , i tried 110 hz but no big difference . Bass i have Lfe + Main 120 Hz . Direct mode in Avr , i understand this is the best for music . I wanted to make the music listening good . I will try , and i will wait for your next video . Respect and wish you all the best
Always a pleasure to see a new video coming from you. A question: The Virtual bass array filters from previous videos, are complementary options or are somehow also substituted by this methodology? Regards ;)
You must have missed. I applied 7.11ms and is the time it takes sound to arrive directly at the mic from the source. My first floor reflection travels 0.75m longer to arrive at the mic +2.19ms after that. Ceiling reflection takes even longer (+5.24ms) There are early reflections arriving within the first 7 ms but these are right from the speaker surfaces hence the weakness of the technique.
Thank you very much. "REW (Room EQ Wizard) Top Tricks" and this video were the two greatest videoes that I have ever seen about music. I don't think I can distinguish between the sound of the two filsters. But I think this filter makes music softer. I'm a biginner. So I could not understand the part of the video after making the filters. But I want to study about REW. I have one question. I took about 7 measurements for each SP before Cross Correlation align and vector average. The length of the sweep was 1M, Is the length suitable?
If the sub is properly time aligned and crossed over with the speakers, you can use the method as is with the combined sub+speaker response. If you have the possibility to separately apply convolution filters to the sub and speakers (most people don't), you can also apply the method to each first and combine them afterwards according to the resulting curves.
Having viewed most of your previous videos, I immediately had the same question as @wmac296 after watching this new video. From your response to his question, I infer that, for my HT setup, it may be preferable to limit IIR correction of my LCR’s to below ~200hz, but to potentially perform full frequency IIR correction to my less-than-ideally-placed surround/height speakers. Is that your recommendation or would there still be a benefit to a full frequency correction of the LCR’s too? Also, you’ve inspired me to manually calibrate my new Monoprice HTP-1 instead of using Dirac. The HTP-1 seems perfectly suited to your calibration methodologies, since it has 16 PEQ slots for each of its 16 outputs, the ability to set distance values with millimeter precision, dynamic “loudness” controls, as well as the ability to output REW sweeps to its atmos channels for setting proper time alignment. However, one thing it can’t due is import a FIR filter. I currently use a MiniDSP Flex to align and correct my two subs using an IIR filter, but it is also capable of importing FIR filters. In your experience, would a FIR filter generated using the method in this video yield better results than an IIR filter using the method in your subwoofer alignment video from last year? Thanks in advance! I truly enjoy learning your techniques!
Definetely stick with 20-200Hz PEQ correction with LCR and use gated windowing techniques explained in detail in Audyssey ART video for the rest of the channels. I guess, with 16 band parametric equalizers for each channel, you can do a better manual job than Dirac. Check Dolby Atmos manual tuning videos for custom configuration of REW auto-EQ filters for your particular unit.They also have Atmos sweeps that play from Atmos channels although your AVR doesn't seem to need them from what I understand. MinidSP FIR capacity (1024 or 2048 taps depending on the sampling rate) is not enough to make any meaningful phase correction below 100Hz so no use for subs. You could use MiniDSP FIR filters to apply crossover phase correction to your speakers since their crossovers will be above 200Hz but that requires different wiring on your end I guess. It was in my notes but I forgot to mention in the video, the only drawback with these inverse filters is that inversion requires a lot of zero padding of the FFT and final filters will need many taps (ie 65k+) so defo needs a PC as the source.
Dear OCA, I'm indebted to you because of the knowledge that I have garnered from your videos within a span of just 2 hours. Do you have a step by step video like this for a Dolby Atmos setup too in case I missed? If not then when will you be coming up with one?
@@ocaudiophile I did find this in your video Phase Match Your Speakers with AllPass Filters czcams.com/video/ChPu0u3nZxc/video.htmlsi=oM9uAEoz8bII3jut at 5:15. 00:05:04 Calibration microphone phase correction. This may have been the video you were referring on the U Mike 1. Thanks! By the way, the RME-2 Dac fs is german engineering at its finest. Thanks for the tip.
Great Video OCA! Thank you very much. Is it preferable to take measurements with the microphone file calibrated to minimum phase, or is it acceptable to use the original calibration file and then apply minimum phase correction to the microphone afterward?
Hi OCA First of thanks a lot for the great content. I am really eager to try this out. Following your instructions, I have created the Filters. They are looking great. But there I havent found a way to import them to minidsp SHD. Am I missing something? or is there a way to manually type in? Just got a roon subscription. This solved the problem! Man, your filter worked wonders!!!
Serkan, I have had great results with this simply correction method. Only issue I am having is on DSD files, I now hear a quick noise prior to each DSD file playing. It is the result of my filter. Any ideas on what I did or how I can eliminate it? TIA!!
Thanks for sharing your new findings! 1. This seems to be the method shown in the short video only without the min phase version with var smoothing? 2. In your opinion, is this version better than the more complicated version of your VBA, xo and lowshelf filter in the last video? 3. Now it would be interesting for me how exactly do I proceed if the playback is from the PC to the AVR and the sub behind it is connected to a MiniDSP? Should everything be measured in Puredirect and the sub aligned to the front beforehand and then proceed as shown here? Or do I leave the sub with EQ and the xo set and only apply the front filters? But maybe this will result in a new xo that is better?
1. Yes, but no min phase conversion of the final filter completely changes the results 2. I tried to implement VBA into this but nothing other than crossover phase correction (and only to a small extent) could improve its results. It really already does the excess phase and doable bass dip corrections as much as possible. 3. Do you have an Atmos set up? If it's only up to 7.1 then you can apply FIR filters to every channel in the PC, you can achieve out of this world results. Otherwise, with neither Audyssey and nor MiniDSP, you will not have the resolution to apply the phase inversion properly.
Hi! Very inspiring video! Have 2 questions if I may. I would like to integrate the convolution DSP into my system. My current setup: Apple TV, PS (HDMI) -> TV (S/PDIF) -> miniDSP 2x4HD. miniDSP is used for XO, time aligning, and PEQ (20-200Hz). To achieve convolutional DSP I guess I need a Linux computer with HDMI input and S/PDIF output (could use USB as input for miniDSP, but then it would be hard to operate miniDSP). So the future setup could look like this: Apple TV, PS (HDMI) -> TV (HDMI ARC) -> ConvolutionDSP (S/PDIF) -> miniDSP. miniDSP is used for XO. What do you think about such a setup? Will XO using miniDSP introduce phase changes achieved by the convolution filtering? if so, I guess I need to ditch the miniDSP all together? What would be the hardware requirements for such a computer (max number of taps)? All your recent research suggests to me that the convolutional DSP is the future, but how much better it would be compared to my previous PEQ-based setup in your opinion? Thanks!
You haven't mentioned your speaker count but I believe it's 2.1? Even a Rasperry PI can easily do over 16k filter taps without significant delays. Yes, you will not need a MiniDSP if you use a PC for convolution and stay within 7.1 channels.
Thanks for the videos! I previously followed your Digital Room Correction Workshop videos but with an uncalibrated microphone just to try and feel how room correction in REW feels like. I recently got a UMIK-1 and decided to follow this video instead. I managed to get the filter but I have to apply -8dB of headroom management in Roon to prevent it from clipping. Is this right? I generated a minimum phase version of my calibration file as per your previous video but my phase measurements still do not look as clean as your measurements. Could my speakers using DSP internally be causing this? Look forward to your new video!
That's totally normal and expected with phase filters and Roon headroom. Min phase versions of the inversion filters will not match the clarity of the real filters but they will guarantee no ringing. Play around for now, I'm working on an Audyssey automation video now but I've developed something even better for stereo. I guess it's the next one after in the pipeline.
Can you please point to the video where you create a minimum phase version of the UMIK calibration file? Should we use the standard calibration or the vertical calibration?
Thanks for the great video, it really helped a lot and easy to do compared to the FIR before. I have also tried Harman curve you attached. However I have the same issue as others, I have to reduce -3db or 4db to prevent the clipping. Is that normal?
Thank you for your great work! This is very easy to do and my Subs now sounding much better, but.... Today I heard music very loud and noticed preringing at lower frequencies with "your" MP-filter. With some Tracks (also "Beat it" - Michael J.) it was very intensive. Now that I know it is there, I can hear it also when music is at lower volume. Could you please give me some advice how to avoid the preringing? Ah, I forgot: There is also a constant noise when music has stopped. It needs 4-5 seconds till this noise stops. It sounds like a tesla. (I use camilladsp) With biquads(peak) I have absolutely no preringing or postnoise.
How many FIR taps are you using in Camilla? These inversion filters are very sensitive and require a lot of taps. When some sharp phase filters are rounded by lower number of taps pre-echo can be introduced. You can generate minimum phase versions of your final filters and use them as the new convolution filters. These will have zero pre-ringing and adequate phase correction. I'm working on the issue as you're not the only one with that result and I've made progress. That video is in the pipeline.
Correcting seperate speaker phase seems quite tricky to me. Have you also tested how the end result is when both (or all your ) corrected speakers are used simultaneously? Phase so far has been my biggest enemy with most DRC methods. After having wasted weeks with DIRAC BC only to conclude that system is completely unsuitable for HT purposes. (they actually create more phase issues than solving them.. since they linearalize speaker groups, yet don't align the groups resulting in huge nulls. Worse yet, they seperately adjust SW phase per speaker group, causing even nulls in the actual SW-out signal.. unless they fix this huge bug i would recommend to stick to XT32 (with perhaps REW help to straighen curve) for HT content for those who have no external DSP for AVR)
I agree the whole method is flawed but they're also quite limited with the number of FIR taps available for bass correction. Audyssey is not much different but at least it's fully customizable by the apps. XT32 will also stay my choice until someone comes up with an AVR (wishful thinking) with user configurable FIR filters. The correct way of creating these filters are quite well known and agreed upon among experts for sometime now but somehow the market monopoly don't bother updating theirs. Audiophiles are stuck between ancient automation and GEQ filters with sound united gear.
@@ocaudiophile Weird my last comment was removed, did you do that or youtube? Its okay if you dont want to discuss about some things, just let me know :) Appreciate all the work you put into your channel. Just bit strange i get no notification of the removal though, then again never used youtube to comment so idk whats normal.
FYI, I didn't delete i a single comment from this channel to this day. Also, most probably (to the best of my knowledge) didn't leave a single question unanswered.
Hi OCA! Can you share a house curse that is between the "Harman" (to much bass) and the "Dr. Olive Toole" with a sligh increase in bass? Thanks a lot again for your great videos! :)
Thanks for the video. With REW's light background it is very difficult to follow the instructions. In the next videos can you put the dark background? I use Equalizer APO. Can the filter be exported in .txt? Written with translator Greetings
I am way too familiar with REW light colors to make a change but I will consider that ;) EQ APO accepts wav files for convolution so you can use the filters in the video with it. You need to save them in a certain directory and direct the config files to them.
Hey there, im freshly venturing into the world of dsp. Thank you for the video! I have a desk setup with a sub and 2 single driver speakers. You mentioned creating min phase crossovers at the end of this vid. Before trying this method would you first suggest I ensure my sub to mains crossover is min phase, and if so which of your videos would be best to learn?
If your speakers are really single driver units, there is no crossover in them. You will only need to time align them with the sub. I have a video for aligning two subs between them but not much else on this subject. It's quite easy though, you need to sync group delays of the sub with the speakers. After that, this inversion method should work fine with the combined system.
Hi, sorry if this is a noob question, I assume we should be using the shroder frequency cut off for all previous videos now. Is there any difference in creating a full range filter and cutting it to 200hz in the Audyssey app or creating the filter in rew only to 200hz and using it full range in the app. Thanks.
Not really, this is for HIFI systems. With surround speakers, you might need to do HF correction due to absurd positions of surround/ceiling speakers. Also you cannot do phase correction with Audyssey. Still, inversion filters described here gave very good results in my Audyssey set up (even without their phase information and up to Shroeder's limits)
Final correction should be done on the combination but subwoofer and speaker interaction should be taken into consideration. What crossover filters are being used at what order all factor in.
Dear OCA, thanks for this tutorial. The last tutorial I followed was the Supreme Audyssey calibration. The new videos show many different things and different approaches. Hence, a few questions in context of home theater: 1. Does this tutorial apply to receiver level calibration? I don't run anything on PC, nor do I have any DSP like miniDSP etc. 2. Is Audyssey ART video the successor to Supreme Audyssey Calibration? Meaning Audyssey ART should deliver better results? Thank you for your amazing work.
1. Unfortunately, you cannot apply these filters with phase corrections to a multichannel receiver although exporting just the frequency response filters even improved the sound of Audyssey in my own tests. 2. Yes, slightly better HF performance compared to Supreme Audyssey method.
I thought i followed everything correctly. When i load the wav file in the convolution engine in Roon i am not getting any sound....its very very faint. Has this happend to anyone, is this is common issue ?
Great video and a much simplified process. I am experiencing an issue when using the Roon Convolution Engine. Using “No Son of Mine” by Genesis as a reference track the heartbeat in the opening seconds is not as ‘focused’ as without the convolution. I have a recording of the difference, but I’m not sure where I can upload it. Thoughts on while the lack of focus?
@@ocaudiophile That definitely brought cohesion to the thump of the heartbeat. Feels like over all result lacks bass. I'll reprocess and add 3dB at the inversion. (REW updated the "1/A Screen" to use dB instead of cycles in the latest version.)
@ocaudiophile, My Measured LR and Vectored LR are not the same. My vectored LR is about 6 db less than measured with the nulls dropping/ending at the same point. Its like my Vectored isnt summing properly. I took 5 measurents for each speaker separately and LR together for 15 total. I cross cor. each speaker to itself and vector averaged the measurements to give left, right, and LR. I vectored my left right together to get LR Vectored. Seems pretty basic so I dont know where im off at.
6dB difference is totally normal .Sorry, I forgot to mention that in the video. My measured LR response was decreased by -6dB. When sound sources are combined you get additional 3dB/added speaker. But outside that, I'd expect measured and calculated averages to be quite similar if left and right speakers were time aligned correctly. If removing IR delays don't bring their impulse peaks on top each other, then there might be some delay REW can't remove for some reason with one of the speakers.
@ocaudiophile , Right on. Yeah, the 6db was what I expected and makes sense combined with the nulls not summing. So many places for one to make an error I wanted to check with the Guy that knows. I watch a lot of these type of videos and most questions go unanswered. Thanks for being so responsive. Appreciate you.
A couple of stupid questions: - Would the low frequency dips be gone and response be perfectly flat if I had some magic corner bass traps behind the sub(s) that would unrealistically remove all reflections at these frequencies combined with this method? - Would you recommend using JRiver rather than Equalizer APO for a non-server windows computer too? I don't see the benefits if you are streaming music and video.
Most dips are not as audible as one would think looking at the charts. I don't have much experience with bass traps but you would need extremely large absorbers to deal with anything below 150Hz which is unpractical and it might absorb higher frequencies as well which might dim mid frequencies below acceptable levels. But as I said, I am no expert when it comes to bass traps. Equalizer APO is free, that's a big plus and and it does the job but it digs itself deep into the Windows sound output and you need to restart the PC even to temporarily turn it off. JRiver is more flexible. Also I have seen very small latency variances with EQ APO at times. JRiver DSP is much easier to use and has one of the best digital volume controls in the market plus it's a great tool for videos & photos.
The method does not require a target curve itself but it can be manipulated to match a curve if needed although not recommended. There's a curve example in the video description.
The overwhelming amount of my source material is ripped from CD, so it's 44.1k native sampling rate, with very few 48k, 96k, and almost no 88.2k files. I use Foobar2000 on a PC with the convolver plug-in. To date, I have generated headphone correction filters in REW using suggested PEQ settings or downloaded others for all the required sample rates that I use for nearly all of my listening and only switch when necessary. Do you resample all your source material to 48k in JRiver and simply use that single filter? Thanks!
My library is similar. REW can export any .wav file in any sampling rate (up to 1536kHz) but I believe most convolution engines do a better job of resampling them on the go, so I keep REW exports at the native 48kHz sampling rate and leave resampling them to 44.1kHz (or any other required by the tracks) to the DSP engine i.e. JRiver.
What was done to get to your L0, R0 and LR0 measurements? Would you also do this same technique for your subwoofer response? Lastly, there is a chance I may be moving from an AVR with Audyssey to one that uses YPAO - if I wanted to continue to use your calibration methods, how would one get this information transferred to that kind of receiver (i.e. a Yamaha R-N800A)?
Vector average of all measurements of each speaker after cross corr align'ing them. Yamaha has the best PEQ among AVRs, I'd go for manual calibration rather than YPAO. You can customize REW auto-EQ (configurable Generic EQ) with the limits of the Yamaha PEQ and it will optimize the filters for your AVR. I have older videos on manual Atmos calibration, have a look at them.
Amazing video as always. I am, however, experiencing around a second(!) of audio latency on windows inputs when I use the filters created following the video in EQ APO - i.e. a mouse click only stops playback after almost a second + all audio arrives later than video. The filters generated according to your old convilution+inversion tutorial didn't cause this problem. Any idea what might be causing this? My computer isn't strained running eq APO and the player, which is musicbee, plus another 10 chrome tabs
Multiplying the filter with a target curve results in a filter that also boosts in the bass, which is something you are avoiding the creation of the original filter. Is there a trace-arithmetic that allows us to do the multiplication, and then limit the filter to 0db (i.e, no boost)?
I don't suggest to multiply with a target curve and I don't do that myself, I shared a curve and a method to embed it for the bass freaks out there ;) If inversion is not resulting in a flat line at the bass frequencies, it's because that bass will be delayed bass composed of wall reflections. There's no TA operation to lift the bass response. You can try applying a low shelving filter (i.e LS 100Hz, +3dB) to the 1/A filter in EQ window and generate predicted from that!
Thank you very much, your videos are awesome. I made everything exactly as you said but i m able to hear a strange preringing (i think) - sounds like a sinus frequency thats playing just before something starts like an open mic that feedbacks lightly on a frequency, that frequency is around 160 Hz... any ideas?
It helps to get all speaker volumes level with each other and see abnormalities if any with certain speakers. For example it can determine how many dBs a subwoofer is too loud. Also ignoring the phase response, only sending the frequency correction to Audyssey target curves improved the sound somewhat in my personal system but still, you have a valid point.@@alexruspantini3484
I just bought a x3800h i have a 5.1.2 with the atmos being upfiring on my front mains. So much with rew has changed since the first video, where do i start with the first video then dolby then art?
Hello OCA! Congratulation! Great job as usual. Can I ask you for assistance? I made "1/A" (mono, 24bit PCM, 96kHz) filter as you presented in this vid but Brutefir (volumio convolver) does not accept it. The message appears every time: "Error with filter size. Filter is not usable! Filter size must be power of 2!". Brutefir accepts filters made by rePhase but in rePhase I can change the taps parameters or FFT lenght. How can I change these parameters in REW to adjust the filter length? Thank you in advance. I wish you all the best.
Under "Export Impulse Response as WAV" window, there is an option "Export this sample count", when you tick it, REW will give you number FIR taps options for power of 2. But I'd careful not to decrease the tap count too much, these inverse filters require a lot of taps to work correctly.
Sorry maybe for stupid question. I did all steps to create final 1\A filter then I've read in comments below that someone had the same problem with ringing (after I applied filter in Equlaizer APO) and then you advised to generate minimum phase version of 1\A again - I did that but still have this strange ringing. also when i turn off audio it fades, doesnt stop immediately. One more problem I have is that with this filter - volume is very low, i have to apply preamplification +24 dB to make it sound normal. But I use very cheap lavalier microphone from Aliexpress for measurement, ofcourse I dont have any calibration files to create minimum phase cal files for this microphone, I just wanted to see if it will improve FR a little (I have very bad speakers). Can my mic cause ringing problem or i did something wrong?
@@ocaudiophile I dont know what exactly helped but I installed new version of REW, also I changed settings on 48 kHz instead of 44.1 and did all measurements again. I also set upper limit in trace arithmetic to 17 Khz (not just 220 Hz) and maximum gain 6 dB so entire FR got corrected. 1/A filter still has ringing but much less but 1/A MP version almost no ringing at all (it still hearable when you turn off music). But overall sound got so much better that I cant believe it. It almost sound like in IEM's after I equalize it with many filters (I listen tone generator and equalize by my ear). Anyway, I will try maybe improve it after watching your other videos, your method is gold, thanks man!
Dear OCA, I'm a happy audyssey evo user since 2 months😀 I'm planning to have a party in an old barn in September with a 2kW PA system. I'd like to control it with a Windows PC, REW and APO equalizer... Which method for room correction would you please recommend? Because I read in the comments on Audyssey Evo that you now seem to prefer the REW PEQ filters instead of this Mastering Digital Room Correction Video method? Many thanks in advance and best wishes from the Austrian Alps, Simon 😀 @@ocaudiophile
@vortexor1 so long as you use MP version of the response, results will be good but yes PEQ has the edge in terms of group delay (at least for Audyssey measurement resolutions)
Thanks again for a great video. Three questions: 1. Is this method only doable with two speakers, or why is this method so different from the Audyssey ART method? 2. Is there a way to calculate the Schroeder frequency for your room using measurements? 3. How do you click and mark in the REW graphs? I've tried right click, left click, +ctrl, +alt etc., and I don't get the same info when clicking and dragging in a graph. + A comment: I wish your Dolby Atmos sweeps were integrated in REW. I don't get why REW does not have that when Windows supports it.
1. It can be applied to multichannel. In fact, I have done such Audyssey calibrations for some members of the channel and achieved excellent results. The problem is with surround/height channels, the speaker placement is so odd compared to a hifi system that, they do need HF correction. Also Audyssey cannot apply phase corrections and the frequency correction with this method is very similar to auto-eq method. PS I do run my surround system speakers all calibrated with this new system. 2. You can calculate it by dividing your room’s reverberation time in seconds by the volume of your room in cubic meters, then multiplying the square root of that sum by 2000! If you cannot, just take it 200Hz and you will not be too wrong. 3. Keyboard CTRL+mouse right click and then drag (and you should be in the correct REW window/tab) There's nothing REW can do about it, Windows doesn't truly support Atmos. The Atmos sweeps bypass Windows and go to the receiver un-decoded through HDMI.
Hey OCA, great video, as always. Very interesting approach. Want to add a question, what about the HFR1 of Audyssey, how can this be handeled using this method? Creating the filters, multiply them with your HFR0 correction file, then export/save as txt.?
I think there's no harm in multiplying anti-HF1 with the 1/A filter created with this method as it will just be a reversal of what was done to the HF response. However, keep in mind that Audyssey cannot apply phase corrections and the application of this method in Audyssey will be limited to frequency content of the filters only although in my experiences, it still sounded better than any other filter before.
@@ocaudiophile Thank you. 1. OK. Interesting. Looking forward to future videos in case those are in the making :) 2. Thank you. So, the average RT60 decay time (e.g. calculated in the range 100-11500 Hz, 1/6 octave decay window width as per Audioholic's recommendation)? My room has some openings (staircase, opening to a hall etc.), that will increase the estimated Schroeder frequency somewhat right? 3. Hmm, in which tabs does this work? I tried many, can't get it to work. Weird. Other apps can send Atmos signals. I hope they find a way. Lastly, would you consider helping out with a calibration (payed of course)?
A follow up on Schroeder's. What decay time do I use? The maximum, at a certain frequency, the average etc? I tried it different ways and figured out that a larger room volume results in a lower Schroeder's frequency for a given reverberation time. My RT60 Decay is about 450 ms on average with a room volume of 96 cubic meters (ignoring openings to other rooms), resulting in a Schroeder's frequency of 136 Hz. I was expecting a higher number, given what I have heard people argue about the upper limit where you can/should apply EQ.
@@ocaudiophile thanks, so you mean that one can replace it in every measurement later, otherwise there will be no harm when taking the measurements with the MP version of the mic cal file directly, correct?
REW doesn't export in .bin format and although it's possible to convert a wav file to bin format (I recently wrote a short script which can do that), MiniDSP FIR filter capacity is 1024 taps per channel which is not enough for these filters. With MiniDSP your best way is to use REW autoEQ filters 20-200Hz and also apply a crossover phase correction for your speakers designed in rePhase for the 1024 taps. rePhase can export in bin format, too. Check this link: audiosciencereview.com/forum/index.php?threads/limitations-of-minidsp.51254/post-1867895
@@ocaudiophile Yes, the HD version of the 2x4 has 2048 taps per channel but I realize even that is not good enough for a filter like this. Btw, the crossover phase correction in rephase should be applied to the auto eq filter exported from rew?
@@paulsoumya you can export auto EQ results as biquads directly to minidsp. Then you make xo phase linearization filter for your speakers in rephase and save it as a 2042 tap.bin file and upload to minidsp fir. They work in paralel in minidsp.
@@ocaudiophile Was it referring to this video? czcams.com/video/i7AaBpqv_p8/video.html I watched it, and must have missed the part that describes how to make a minimum phase version of an existing cal file. In that video, it seemed like you were creating more accurate cal files using the EarthWorks as a reference. Is it as simple as loading the cal file in rew as a filter and generating a minimum phase version as described here?
How would you recommend going about SW integration with this method? I have only 2 front fullrange speakers and 1 SW. I am using CamillaDSP with a RME UCX-II interface, so can set each filter individually per channel. I created the mp filters for L, R, and SW. How should I go about aligning LR with SW?
You should be seeing something close to your speakers' spec bass roll off curve after the inversion and the crossover frequency must be kind of obvious now. I'd time/volume align the inverted SW/speaker bass responses to the LP and that should do it. I don't think you will need additional EQ after the crossover with the correct XO point. You'll need to sync impulse peaks at the LP for time alignement. It should also be obvious from phase responses and impulse peak directions if SW needs a polarity inversion. Let me know how it goes.
@@ocaudiophile Thanks for the direction. I created the 3 separate filters for L/R/SW based on the inversions of the minimum phase measurement (as described here). I "tested" them in REW with A*B on the original measurement and they do indeed flatten the response until 200Hz. I also applied in CamillaDSP and it sounds good. That said, I have not time aligned them yet, not sure of the best approach here. The first peak of each IR seems to be very close to t=0, I assumed that this is due to acoustic ref but could be wrong. Also,I was using a UMIK-1 until now but just got a M23R so will retake all the measurements anyway with the new mic. Lastly, I used the alignment tool to try to align SW to L/R. To do this, I made a vector average of L/R after applying 1/MP filter, and aligned with SW. Interestingly, the best response seems to be when I use the measured SW, and not SW after applying the 1/MP filter, which has some dips. I hope my explanation was clear.
@itsik_ Congrats on the M23R, I also own one and I never looked back since. You've to just time align a single sub to MLP especially for music. Alignment tool is for multiple sub relative alignment. I'm about to launch a video on subwoofer alignment btw.
@@ocaudiophile Oh nice, looking forward to it. Will that cover subwoofer alignment to main speakers? In the previous comment, you wrote "You'll need to sync impulse peaks at the LP for time alignment." Is these between Left and right, or between them to the sub?
Any chance, that you'll explain how to implement a simple tone-adjustment to the speakers with audyssey? I really like Dyn. Eq and Dyn. vol, but I don't like the sound of an eq'ed speaker, because i listen to the system at so many different locations. I'd like to simply make a +1 bass and -1 treble adjustment to the whole thing, as I find the speakers a tad too bright.
If you replace speaker responses with "perfect speaker" settings in the ady file (a script in Audyssey ART video does that automatically) and set your Audyssey to "Reference" in the AVR, DYnamic EQ and Dynamic volume will work but there will be no EQ applied.
@@ocaudiophile Thank you - but i can already do this by setting the eq-range in audysseyapp to 0(or 20 Hz). What i want is audyssey to "correct" the perfect speaker with +1 db in the bass range, and -1db in the treble range. Again, Thank you so much for replying.
Like telling audyssey, that i have a "perfect speaker" that plays bass exactly 1 db too low, and treble 1 db too loud, so audyssey corrects it to +1 bass and -1 treble. Like a tone control
Before watching every video your channel provides, all of these methods require an AVR with at least the support for the Audyssey App right? Because I have an Denon AVR X1200W, so no Audyssey App :(
@@ocaudiophile That's what I thought :/ Sadly my PC is upstairs and 100% used as a PC for PC stuff, currently using a Firestick 4K Max and a few consoles on my AVR, maybe a little home theater PC will work too? EDIT: Just came into my mind that only the music coming from the PC can be modified, the Game Consoles and whatnot can't, at least with my current AVR!
@@ocaudiophile Which software? I use Equalizer APO to EQ my headphones, is there a more sophisticated one for these kinds of applications? Oh and, when I want to DSP my PS5 or Switch for example, what is needed for these kind of filters? In my past I only ever had a Omnitronic Smard 24, t.racks DSP Mini and internal DSP's from various pro amps, but I sold them before I gained knowledge of advanced DSP techniques :/
Discovered your channel through exploring the inverse filter technique for my full range speakers, and the results have been transformative-full control and unparalleled sound quality. However, I'm facing challenges integrating DSP with my JBL 590s, which sound nice but not superior with DSP. Every attempt to enhance them through DSP paradoxically detracts from their performance, falling short of the no-DSP setup. Any insights or suggestions on this would be greatly appreciated!
Frequency domain filters should work below the room transient but that's not enough to avoid sound degradation. Hard to say what exactly causes it in your case without seeing the actual measurements. I can have a look if you share.
@@ocaudiophile so grateful🙏for this opportunity and for all you share. Please find attached two .mdat files for your review. The first file contains measurements and the work leading up to the creation of inverse MP filters for my setup. The second file comprises fresh measurements taken today, without moving the microphone, from a single listening position. These were done after slightly adjusting the speaker placement, as I'm still in the process of experimenting with their optimal positioning, which results in them being moved around. Thank you so much for taking the time to look into this drive.google.com/drive/folders/1yQxKyoWKHtQu2-b-kgZ7wUBxjO35HeSv
@ocaudiophile I've shared the original and minimum phase microphone calibration files in the same folder for your convenience. Additionally, I've included a file for the 90-degree measurements that I had on hand to compare. Thank you once again for looking in drive.google.com/drive/folders/1yQxKyoWKHtQu2-b-kgZ7wUBxjO35HeSv
You've done everything right (one of the best implementations of my methods I have seen to date) but your speakers are not the easiest to calibrate. Left speaker seems to be closer to adjacent walls than the right. Most measurements had some distortion here and there but there were enough of them so I could pick good ones. 4M measurement length was tough to process even for my PC which is a beast, had to trim them to open in rePhase. Anyways, it's all done now and I'm sure they will sound great. Removing IR delays from L & R was a bit tricky, I guess that's where your filters failed. I've also embedded phase linearization (XO590.wav) into the filters. StereoConvolutionFilters.wav is the filter you need. In the highly unlikely case that, you will hear pre-echo with this one, use ConvolutionFIlters_2.wav. That's 110% echo safe but will not sound as good as the other. REW and rePhase files with all details are included in the zip below. Enjoy! drive.google.com/file/d/1xjEu6LmNIySQnuBVw85xTVJMs4Wf5VtJ/view?usp=sharing
Hard to say without seeing the actual measurements but I am interested to find out. Can you share your mdat file? You can post google drive links here (most others will be deleted by YT).
@@ocaudiophile Sure. I have some strange dips that don't appear in other measurements. I have a CSL Umik-2 but I want to take new measurements with the REW specific cal file that I made minimum phase. I used these measurements along with changing the cal file in REW to make the XO phase filter and that works perfectly though. I sent the link via PM in the roon community forums. Thanks
Looks like it's your XO filter that caused the pre-echo. Here's a fixed correction for your system: drive.google.com/file/d/1guMQoUQ3Ow3I7kYX4B7-5rn8KTyTVKtD/view?usp=sharing
@@ocaudiophile Hi, I apologize for the interruption. I encountered the same issue. I followed each step carefully, but pre-ringing still occurred. Unfortunately, I am currently at a loss regarding a solution.
Hi OCA, could you possibly help me with your procedure? I'm not sure my measurements look right. Things are not equal on both sides of my listening room. PSB Synchrony One speakers seem to be canceling each other at 35 to 60hz when measured together but not separately. I did the mic min. phase, 1LP for all measurements - measured each speaker 3 times using length 2M, vector averaged L and R separately. My impulse response start looks like a saw-tooth: i cant seem to find where to align left with right because of this. Thank you
You've changed the acoustic timing reference channel from right to left during measurements, it should always be the same channel for both speaker measurements. But this is not the reason for the dips in combined response, it's to do with asymmetric wall proximity between speakers. I fixed most of it by applying an extra allpass phase filter to the right speaker (Rxo.rephase which produces Rxo.wav filter). The optimal Q value is 5 but this could cause ringing so I left it at a safe 2. You can experiment with higher Q values and improve the dip even better. FL-trimmed_FR_trimmed-48k.wav is the final stereo convolution filter which should sound good: drive.google.com/file/d/1mgnWTh2GQdGbF-7cckRoImDgGL3V1Xk-/view?usp=sharing
Your videos are awesome, some stuff goes way above my head though. Just to be clear, your previous video: czcams.com/video/5YcH7j2-L1Y/video.htmlsi=s_U4Wh3Qpb3ZsRaJ named: "REW (Room EQ Wizard) Top Tricks: Convolution with Inversion (no EQ filters, all FIR!)" is not relevant anymore? I followed that one many times, but now I'm confused if I should leave that and try the video above? Or are they connected?
They both invert the speaker response at the LP. The original one, uses a minimum phase version of the inversion. The newer technique inverse the min phase version of the speaker and uses it as a filter as is. The latter has better phase correction but requires more taps. The original one has the advantage of being able to be inverted over a target curve. The other iverts over a flat target. You wouldn't lose much by trying and comparing them both.
@@ocaudiophile - Good explanation and looking forward to the upcoming video re-introducing the House Curve/LS Filter. You've put out a number of techniques, this on correcting Schroder (up to 225Hz) against a flat response, the mentioned FIR video to correct against the entire range against a House Curve, and your Digital Room and Speaker correction workshops of which I only used the EQ FIR technique to correct from 20-200Hz, 200-500Hz and 500-20,000 along with IR window size adjustment to lower my TAPS. That being said, I hope the upcoming video addresses benefits or preferred techniques such as Schroder 0db/Flat technique, Schroder with db boost/house curve technique, full range against house curve technique (updates to FIR Inversion video vs Workshop EQ FIR video). Thanks for being the Man and educating us!
I finally got round to trying this. This one gives me way better results. Other one tamed my high mids/highs a bit more. But also had some weird effect on the sound stage for some reason (maybe I did it wrong, who knows). For my system, in my room, with my speakers adding the Harman curve sounds just a bit richer. But I got to find one with a 2 maybe 3 db rise in the bass. Now it's sometimes overwhelming. Could I add a small low pass shelve in ROON to tame it a bit? Or would that ruin the filter? Great work once again!🎉
Wow, easy to do, well explained and sounds great!
Headroom correction needed in Roon: 0 dB!
However, I have 3 questions:
- I recently bought the Umik 2 microphone. In the IR windows, entering the ms corresponding to the distance of my speakers (3.2m = 9.41 ms, REF time 0.007), to see the behavior on the high frequencies, as you did, I can't see frequencies below 106 Hz. How is this possible?
- I did not correct the microphone calibration with Minimum phase.
Where can I find the procedure?
- Multiplying 1/A with Harman Curve results in a mess of delay between the 2 channels in addition to the foreseeable need to use about 8db of Headroom. I can't understand where I'm wrong.
1. You shouldn't be able to see the response below 106Hz because a short right window size of 9.41ms means the minimum frequency you can detect is 1/(9.41/1000)=106.27Hz - frequency is just 1/time in seconds. I can see it with a trick but it's just an extrapolation by REW for convenience, is not correct by any means (Preferences/Analysis/Use right window for min frequency unticked!)
2. Somebody located it and linked it down here in the comments.
3. I don't recommend applying a target curve, I suggest turning up the volume if the bass is too weak not to your liking :) Again, I explained and suggested a way (LS filter) to increase bass which seemed to work with the channel member down below in these comments.
@@ocaudiophile Thank you so much!
@@ocaudiophile Just for information: I applied the MP correction to the microphone and redid the procedure. Now to avoid clipping I have to give -4dB in Roon, before 0.
What I changed is only the correct phase of the microphone.
How does the MP version of the mic cal file look like? If it's deviating from 0 degrees too much (more 20-25 degrees at the extremes) than you must have done something wrong during MP generation. Try adding tails and/or removing IR delays from the mic cal MP version.
@@ocaudiophile about 19 degrees at 10Hz and -3 degrees at 20kHz, looks nice.
Woah, This guy is at it again. I did the previous excess phase video and this one sounds like it will be a million times easier. Cool explanations on what Minimum phase is. Thanks Bro!
You're spoiling me. Thanks!
Awesome work. Great to have a much simpler process that also delivers improvements over the EQ/Target Curve process.
Good to see you back, great video. You got my mind wondering again, so thank you.
Cheers m8!
Great to see you working on this. Makes it easier and quicker to get great results.
Absolutely!
There was a little bit of Ringing with the 225Hz setting after I tried it out with increasing the range. Have to reduce it to below 200Hz I think to lessen Ringing on my end. It was not noticeable until I tried higher frequencies and then compared it On & Off a few times to note it was ringing even with 225Hz cut-off.
You mentioned you might have to change it on our end and I can confirm this to be true.
Increasing the range just causes LOTS of ringing for me. And ~200Hz seems to not have it.
You are probably right but HF is not a known cause of ringing. Precision becomes quite important with inversion. Also filters are longer than usual because of the zero padding required for the FFT (even after trimming). I would suggest to vector average multiple measurements with 1M length before starting to create the filters for best results.@@nighthog7003
Hello! Love your stuff, great logic - but a little fracmented I think, difficult to get the golden overview and understand what comes first. Will you maybe make a step-by-step tutorial for a complete system setup, with all the various steps from the differnt videos, in the correct order, to get the best results? Huge thanks in advance 🙂
Noted!
Always perfecting the art of sound.!! Thanks OCA
My pleasure!
This is really wonderful advice. Thank you. Very happy to subscribe and continue to learn more!
Welcome aboard!
Another great video, personally learning a lot and my musical enjoyment has been taken to another level. This video is very similar to the very first video I watched and made some major improvements to my system titled "REW (Room EQ Wizard) Top Tricks: Convolution with Inversion (no EQ filters, all FIR!)". This video cuts out the use of a House Curve so looking forward to the upcoming video re-introducing a House Curve\LS EQ Filter.
A few questions/general comments for the masses:
1 - For my 1/A Inversion (From L/R-MP Copy), the IR Window sizes where huge, Left Window greater than 10,000ms. Before exporting I trimmed the 1/A window to R=100ms/L=500ms. Curious as to why so large or what I did wrong?
2 - For my 1/A Inversion (From L/R-MP Copy) I was getting pre-ringing with Roon. Creating a 1/A MP copy resolved that issue.
3 - A few comments below noted lack of bass, to resolve I imported the Harmon House Curve and multiplied with the 1/A MP Copy (L/R - 1/A MP * Harmon House Curve) which boosted my bass to better match my frequency level aboves 225Hz. I'll note to avoid pre-ringing weirdness must do against the MP copy and I had to adjust the IR Windows sizing from being super large.
Large window sizes are normal with inversion (it requires double the zero padding of the FFT in the background). In fact the cause for pre-echo is still the insufficient window sizes as a perfect inversion wouldn't ring.
@@ocaudiophile Thanks, not doing something wrong then. My default window size in preferences for left is 1000ms, right 2000ms. Do I have those values default correct?
It's amazing... The final filtering is great at improving bass.
However, in some trial listening clips, the bass will produce a delayed ringing effect. If the music clip ends with bass, the party music will end, but the bass will continue in a background music like rumbling sound for about 1 second
For example: ANNE-SOPHIE MUTTER "CARMEN-FANTASIE" Zigeunerweisen op.20 IV. Allegro molto vivace
Thanks. How was your front speakers set up by Evo?
@@ocaudiophile Sorry, I loaded the filtering wav file based on the ROON software and did not use the EVO ONE software you recently shared
@@aland8195 Ok, sorry I thought the comment was under Evo video :) AFter you create the final 1/A filters, generate their "Minimum phase versions" and use these as filters instead. That will completely remove the ringing.
@@ocaudiophile Thank you very much for sharing and answering
I just checked the data of REW again, and the original curve (A) * 1/A filtering (B) shows a significant increase in RT60 data in the 150Hz frequency band. This may be the reason for the strange listening experience, which is somewhat strange.
@@aland8195 Inverting a response properly requires a lot of filter taps for accuracy.
This is really fantastic and I am currently experimenting with the inverted response as FIR filter. I am also starting with rephase and I have following question: When I invert without Minimum Phase the computed 1/A filter has the same phase characteristic as my measurement. Could I load this inverted response into rephase an do (only) a phase calibration there (no gains)? Then I would generate a kind of phase correction filter in rephase, export it, import into REW and then multiply it via Trace Arithmetic with the inverted response. This would then be my new phase-corrected FIR filter. I assume that your approach with the Minimum-Phase version in REW is still more efficient and the best for measurements from the listening position. My phase-correction idea is on my mind for speaker measurements and would be applied fullrange (I would even use this for digital crossover design with measurements of single speaker drivers) - basically as the optimized, starting filter for an acitve speaker system. When the system is in place, I would add a room correction filter with your method for the bass response. Sorry, I am really new to this and not firm (yet) with Rephase and I am not a happy user yet. I might be getting it totally wrong ... but am really inspired from you. Thanks a lot!
Always work with excess phase of a speaker measurement in rePhase, not the phase itself. Usually crossover phase corrections will suffice but you can also do port/box corrections if you have enough FIR taps to deal with bass frequencies. Check this video:
czcams.com/video/ChPu0u3nZxc/video.html
Thanks for the video ,pk harmonic + Your filter + Ubacch filter is best for me.
Thanks for sharing
I have no idea what you're talking about but you sound really intelligent so I'll keep watching. I'm confident that my IQ will increase as I watch your videos 😂😂
😂
Hello sir, I'm most impressed and grateful for the content you're sharing. I've some questions (disclaimer: I'm a rookie so I may be very wrong). My priority is to flatten the low bass, because I have insane drop in 34-57 Hz range.
1. If I was to leverage all your latest methods for a 2.0 system with Roon (no audyssey), I suppose that would be combination of VBA + xover/box/port correction + the inversion below Schroeder. Is that right?
2. In what order should these filters be created and applied? My guess: 1) create VBA filter based on original measurement, 2) create xover/port filter based on VBA-applied trace, 3) create "inverted below Schroeder" filter based on xover/port-applied trace, 4) convolve inverted filter * xover/port filter, 5) convolve result of "4)" * vba filter
3. What REW measurement and analysis settings should be used for creation of each of these 3 filters? You specified one here czcams.com/video/ChPu0u3nZxc/video.html (with mic pointing at speakers and minimum phase cal file) and another one here czcams.com/video/zmIqJ_jlZPw/video.html . I'm guessing the same measurement needs to be used for creation of all 3 filters, right? So which one?
4. What's the optimal measurement length for these corrections? Will 1M be good enough?
5. Does it make sense to aim for nearly flat (at zero degrees) result of maximum phase when correcting fox xover/box? Let's assume such correction won't ring or anything. Because I can make it much flatter compared to what you've shown, and even with quite small Q values. Or are the corrections you've shown sufficient? Should we only care about 100-1000 Hz range when "flattening" the max phase response (as it's the range that's audible phase-wise)?
6. I have quite pronounced phase differences between left and right speaker in my listening position. Should I rather them separately (rather than based on LR averaged)? Because my LR average would have quite high std. Deviation, so I would be correcting against a max phase trace which isn't very applicable to neither left nor right speaker.
Thanks for the gesture. 1M length is good enough, the louder and the longer the better but rew will slow down and inversions etc will be extremely large beyond that. VBA, phase correction, inversion order seems right, don't forget to generate minimum phase of the final vba and inversion filters. 100-1000hz is good for phase but remember to correct only excess phase. If certain correction causes pre echo, try to just fix what's different in excess phase of left and right speaker.
this is explained very well, thanks
Glad it was helpful!
Hmmm, just tried this following along on real time. I went back and loaded my microphone calibration and created a minimum 26:24 version. This new process is fairly straight forward. However, I am getting a ton of pre-ringing from the convo filter. Where did I go wrong?
Generate minimum phase versions of the final 1/A filters and export these as filters instead.
Always respect you OCA. 😊
Minphase Inversion is simple, fast, and produces good results.
Also, messurement graph is improved to make it look good to the eye.
However There is a limit to one listening position on one speaker (or subwoofer) and eventually we will use up to the limit of drc when active treatment, inversion, and time domain calibration are performed simultaneously with multiple speakers (or dual, quad subwoofer).
Thank you for making the video.
Thanks.
I watched this video maybe 30 times now :) and the others too . I made everything , sound very good , but lost a lot of base...
First time i did it with the sub connected and just followed a long but the sub was working i had LFE + Main set .
Second time i only made left and right (subwoofer unpluged) .
Remains to test the next days .
I would really appreciate if you would explain a little more about adding the harmon curve , i dont understand :
,,Drop the .txt file on to REW - you mean in the EQ ?
3. AxB (trace arithmetic) multiply that with each of the "1/A" filters" really lost me here
Thank you and God bless you for sharing the knowledge
Maybe should use spl allign ? Like in the first video ...i have a difference in the convolution filter 0.2 db
I understand you are using FIR filters in a PC and using an AV receiver as an amp (LFE+Main reminds Denon/Marantz gear) because otherwise you cannot use these filters with Audyssey. Anyway, the inversion is done on a flat line so there's no bass tilt. I've developed some solution for that already and its video is in the pipeline. But in the mean time, you can A x B multiply each of the 1/A filters with a low shelf filter and increase the bass as much as you want. You can generate a low shelf filter with REW in the EQ window: open EQ filters, select "LS" from the drop down list, frequency i.e.100Hz, gain +2dB and then "Generate measurement from filters" under "Filter tasks". This is your "B" in "A x B" trace arithmetic. You will need to adjust clipping headroom.
PS if you hear any pre-ringing effect, generate minimum phase versions of each of the 1/A filters and multiply these MP versions with the LS filter to generate your final convolution filters.
Thank you very much .
I have this :
Pc - Smsl Su1 Dac - analog to Denon Avr 3500H and 2.1 speakers .
Front sp cut 80 Hz , i tried 110 hz but no big difference .
Bass i have Lfe + Main 120 Hz .
Direct mode in Avr , i understand this is the best for music .
I wanted to make the music listening good .
I will try , and i will wait for your next video .
Respect and wish you all the best
Which previous video covers how to create the proper mic cal file? Thanks!
czcams.com/video/ChPu0u3nZxc/video.html
Thanks for a great video. Can you explain in more detail how to apply the Harman curve?
I just added an explanation of the steps and a minimum phase version of the Harman Target Curve to the video description.
Always a pleasure to see a new video coming from you. A question:
The Virtual bass array filters from previous videos, are complementary options or are somehow also substituted by this methodology?
Regards ;)
I am curious about the same thing. This correction is so much simpler than the 4-5 steps necessary to do the former corrections.
I tried hard to add a VBA filter to this solution but none was increasing clarity but in some very problematic rooms, it might help.
@@ocaudiophile thks for the feedback ;)
Yes and it does the excess phase correction automatically as much as it can be done, it does volume leveling and it requires no target curve.
The window you apply for 11ms gating isn't anechoic. It contains reflection from floor and ceiling.
You must have missed. I applied 7.11ms and is the time it takes sound to arrive directly at the mic from the source. My first floor reflection travels 0.75m longer to arrive at the mic +2.19ms after that. Ceiling reflection takes even longer (+5.24ms)
There are early reflections arriving within the first 7 ms but these are right from the speaker surfaces hence the weakness of the technique.
Thank you very much. "REW (Room EQ Wizard) Top Tricks" and this video were the two greatest videoes that I have ever seen about music. I don't think I can distinguish between the sound of the two filsters. But I think this filter makes music softer.
I'm a biginner. So I could not understand the part of the video after making the filters. But I want to study about REW.
I have one question. I took about 7 measurements for each SP before Cross Correlation align and vector average. The length of the sweep was 1M, Is the length suitable?
Yes, I also regularly use 1M measurement length. It's a good compromise between practicality and SNR ratio.
Thank you very much, I am going to study
about your previous videoes.
Great video...❤
Thank you 🤗
Another question from an amateur: What to do with the subwoofer during measurements?
If the sub is properly time aligned and crossed over with the speakers, you can use the method as is with the combined sub+speaker response. If you have the possibility to separately apply convolution filters to the sub and speakers (most people don't), you can also apply the method to each first and combine them afterwards according to the resulting curves.
Having viewed most of your previous videos, I immediately had the same question as @wmac296 after watching this new video. From your response to his question, I infer that, for my HT setup, it may be preferable to limit IIR correction of my LCR’s to below ~200hz, but to potentially perform full frequency IIR correction to my less-than-ideally-placed surround/height speakers. Is that your recommendation or would there still be a benefit to a full frequency correction of the LCR’s too?
Also, you’ve inspired me to manually calibrate my new Monoprice HTP-1 instead of using Dirac. The HTP-1 seems perfectly suited to your calibration methodologies, since it has 16 PEQ slots for each of its 16 outputs, the ability to set distance values with millimeter precision, dynamic “loudness” controls, as well as the ability to output REW sweeps to its atmos channels for setting proper time alignment. However, one thing it can’t due is import a FIR filter. I currently use a MiniDSP Flex to align and correct my two subs using an IIR filter, but it is also capable of importing FIR filters. In your experience, would a FIR filter generated using the method in this video yield better results than an IIR filter using the method in your subwoofer alignment video from last year?
Thanks in advance! I truly enjoy learning your techniques!
Definetely stick with 20-200Hz PEQ correction with LCR and use gated windowing techniques explained in detail in Audyssey ART video for the rest of the channels.
I guess, with 16 band parametric equalizers for each channel, you can do a better manual job than Dirac. Check Dolby Atmos manual tuning videos for custom configuration of REW auto-EQ filters for your particular unit.They also have Atmos sweeps that play from Atmos channels although your AVR doesn't seem to need them from what I understand.
MinidSP FIR capacity (1024 or 2048 taps depending on the sampling rate) is not enough to make any meaningful phase correction below 100Hz so no use for subs. You could use MiniDSP FIR filters to apply crossover phase correction to your speakers since their crossovers will be above 200Hz but that requires different wiring on your end I guess.
It was in my notes but I forgot to mention in the video, the only drawback with these inverse filters is that inversion requires a lot of zero padding of the FFT and final filters will need many taps (ie 65k+) so defo needs a PC as the source.
Dear OCA, I'm indebted to you because of the knowledge that I have garnered from your videos within a span of just 2 hours. Do you have a step by step video like this for a Dolby Atmos setup too in case I missed? If not then when will you be coming up with one?
Check my Audyssey playlist. You will find everything from full manual Atmos calibration to fully automated one and inbetween ;)
@@ocaudiophile Thank you dear. You are so generous. I am sure your hardwork will pay off in the most wonderful way soon.
TOP!!!
Which video contains the min phase mic cal? Great video thank you for sharing your DSP knowledge.
Thanks.
I think it's this one:
czcams.com/video/i7AaBpqv_p8/video.html
I have the U mike 1, is this a different procedure from the U mike 2 example?@@ocaudiophile
No, the exact same procedure
@@ocaudiophile I did find this in your video
Phase Match Your Speakers with AllPass Filters
czcams.com/video/ChPu0u3nZxc/video.htmlsi=oM9uAEoz8bII3jut
at 5:15. 00:05:04 Calibration microphone phase correction.
This may have been the video you were referring on the U Mike 1.
Thanks!
By the way, the RME-2 Dac fs is german engineering at its finest. Thanks for the tip.
Great Video OCA! Thank you very much. Is it preferable to take measurements with the microphone file calibrated to minimum phase, or is it acceptable to use the original calibration file and then apply minimum phase correction to the microphone afterward?
It would be the same thing as long as you apply the mic phase correction to the responses right.
Hi OCA
First of thanks a lot for the great content. I am really eager to try this out. Following your instructions, I have created the Filters. They are looking great. But there I havent found a way to import them to minidsp SHD. Am I missing something? or is there a way to manually type in?
Just got a roon subscription. This solved the problem! Man, your filter worked wonders!!!
Gr8!
Serkan, I have had great results with this simply correction method. Only issue I am having is on DSD files, I now hear a quick noise prior to each DSD file playing. It is the result of my filter.
Any ideas on what I did or how I can eliminate it? TIA!!
DSD files require a lot of CPU power, decreasing filter sizes might help or upgrading your PC ;)
Thanks for sharing your new findings!
1. This seems to be the method shown in the short video only without the min phase version with var smoothing?
2. In your opinion, is this version better than the more complicated version of your VBA, xo and lowshelf filter in the last video?
3. Now it would be interesting for me how exactly do I proceed if the playback is from the PC to the AVR and the sub behind it is connected to a MiniDSP?
Should everything be measured in Puredirect and the sub aligned to the front beforehand and then proceed as shown here?
Or do I leave the sub with EQ and the xo set and only apply the front filters? But maybe this will result in a new xo that is better?
1. Yes, but no min phase conversion of the final filter completely changes the results
2. I tried to implement VBA into this but nothing other than crossover phase correction (and only to a small extent) could improve its results. It really already does the excess phase and doable bass dip corrections as much as possible.
3. Do you have an Atmos set up? If it's only up to 7.1 then you can apply FIR filters to every channel in the PC, you can achieve out of this world results. Otherwise, with neither Audyssey and nor MiniDSP, you will not have the resolution to apply the phase inversion properly.
Hi! Very inspiring video! Have 2 questions if I may.
I would like to integrate the convolution DSP into my system.
My current setup: Apple TV, PS (HDMI) -> TV (S/PDIF) -> miniDSP 2x4HD. miniDSP is used for XO, time aligning, and PEQ (20-200Hz).
To achieve convolutional DSP I guess I need a Linux computer with HDMI input and S/PDIF output (could use USB as input for miniDSP, but then it would be hard to operate miniDSP).
So the future setup could look like this:
Apple TV, PS (HDMI) -> TV (HDMI ARC) -> ConvolutionDSP (S/PDIF) -> miniDSP. miniDSP is used for XO.
What do you think about such a setup?
Will XO using miniDSP introduce phase changes achieved by the convolution filtering? if so, I guess I need to ditch the miniDSP all together?
What would be the hardware requirements for such a computer (max number of taps)?
All your recent research suggests to me that the convolutional DSP is the future, but how much better it would be compared to my previous PEQ-based setup in your opinion?
Thanks!
You haven't mentioned your speaker count but I believe it's 2.1? Even a Rasperry PI can easily do over 16k filter taps without significant delays. Yes, you will not need a MiniDSP if you use a PC for convolution and stay within 7.1 channels.
Danke!
🙏
Thanks for the videos! I previously followed your Digital Room Correction Workshop videos but with an uncalibrated microphone just to try and feel how room correction in REW feels like. I recently got a UMIK-1 and decided to follow this video instead. I managed to get the filter but I have to apply -8dB of headroom management in Roon to prevent it from clipping. Is this right? I generated a minimum phase version of my calibration file as per your previous video but my phase measurements still do not look as clean as your measurements. Could my speakers using DSP internally be causing this? Look forward to your new video!
That's totally normal and expected with phase filters and Roon headroom. Min phase versions of the inversion filters will not match the clarity of the real filters but they will guarantee no ringing. Play around for now, I'm working on an Audyssey automation video now but I've developed something even better for stereo. I guess it's the next one after in the pipeline.
Can you please point to the video where you create a minimum phase version of the UMIK calibration file? Should we use the standard calibration or the vertical calibration?
Standard horizontal.
Thanks for the great video, it really helped a lot and easy to do compared to the FIR before. I have also tried Harman curve you attached. However I have the same issue as others, I have to reduce -3db or 4db to prevent the clipping. Is that normal?
Yes, Roon alone adds 3dB for example.
Thank you for your great work!
This is very easy to do and my Subs now sounding much better, but.... Today I heard music very loud and noticed preringing at lower frequencies with "your" MP-filter. With some Tracks (also "Beat it" - Michael J.) it was very intensive. Now that I know it is there, I can hear it also when music is at lower volume. Could you please give me some advice how to avoid the preringing? Ah, I forgot: There is also a constant noise when music has stopped. It needs 4-5 seconds till this noise stops. It sounds like a tesla. (I use camilladsp) With biquads(peak) I have absolutely no preringing or postnoise.
How many FIR taps are you using in Camilla? These inversion filters are very sensitive and require a lot of taps. When some sharp phase filters are rounded by lower number of taps pre-echo can be introduced. You can generate minimum phase versions of your final filters and use them as the new convolution filters. These will have zero pre-ringing and adequate phase correction. I'm working on the issue as you're not the only one with that result and I've made progress. That video is in the pipeline.
Fabulous video, a question: why don't you use the filter for full range instead of 225Hz upper limit?
Won't be the whole response flatter?
Thank you!
👍 The answer to that is about directionality of high frequency sound waves and explained in the video
Correcting seperate speaker phase seems quite tricky to me. Have you also tested how the end result is when both (or all your ) corrected speakers are used simultaneously?
Phase so far has been my biggest enemy with most DRC methods. After having wasted weeks with DIRAC BC only to conclude that system is completely unsuitable for HT purposes.
(they actually create more phase issues than solving them.. since they linearalize speaker groups, yet don't align the groups resulting in huge nulls. Worse yet, they seperately adjust SW phase per speaker group, causing even nulls in the actual SW-out signal.. unless they fix this huge bug i would recommend to stick to XT32 (with perhaps REW help to straighen curve) for HT content for those who have no external DSP for AVR)
I agree the whole method is flawed but they're also quite limited with the number of FIR taps available for bass correction. Audyssey is not much different but at least it's fully customizable by the apps. XT32 will also stay my choice until someone comes up with an AVR (wishful thinking) with user configurable FIR filters. The correct way of creating these filters are quite well known and agreed upon among experts for sometime now but somehow the market monopoly don't bother updating theirs. Audiophiles are stuck between ancient automation and GEQ filters with sound united gear.
@@ocaudiophile Weird my last comment was removed, did you do that or youtube? Its okay if you dont want to discuss about some things, just let me know :) Appreciate all the work you put into your channel. Just bit strange i get no notification of the removal though, then again never used youtube to comment so idk whats normal.
It was definitely not me. Let me check "held for review" section....@@user-pt1sn9nc4p
No, there's nothing being held by YT. I received your comment by email btw but I can't see it here.
FYI, I didn't delete i a single comment from this channel to this day. Also, most probably (to the best of my knowledge) didn't leave a single question unanswered.
Hi OCA!
Can you share a house curse that is between the "Harman" (to much bass) and the "Dr. Olive Toole" with a sligh increase in bass?
Thanks a lot again for your great videos! :)
Why don't you add a low shelving filter (LS in REW EQ) to the 1/A filter according to your taste? Remember it'll require some clipping headroom.
Which video do you talk about creating minimum phase versions of microphone calibration files?
should be this one:
czcams.com/video/i7AaBpqv_p8/video.html
Thanks for the video.
With REW's light background it is very difficult to follow the instructions.
In the next videos can you put the dark background?
I use Equalizer APO.
Can the filter be exported in .txt?
Written with translator
Greetings
I am way too familiar with REW light colors to make a change but I will consider that ;)
EQ APO accepts wav files for convolution so you can use the filters in the video with it. You need to save them in a certain directory and direct the config files to them.
Which video was used to create the smallest phase version of the microphone calibration file?
czcams.com/video/ChPu0u3nZxc/video.html
Hey there, im freshly venturing into the world of dsp. Thank you for the video! I have a desk setup with a sub and 2 single driver speakers. You mentioned creating min phase crossovers at the end of this vid. Before trying this method would you first suggest I ensure my sub to mains crossover is min phase, and if so which of your videos would be best to learn?
If your speakers are really single driver units, there is no crossover in them. You will only need to time align them with the sub. I have a video for aligning two subs between them but not much else on this subject. It's quite easy though, you need to sync group delays of the sub with the speakers. After that, this inversion method should work fine with the combined system.
Hi, sorry if this is a noob question, I assume we should be using the shroder frequency cut off for all previous videos now. Is there any difference in creating a full range filter and cutting it to 200hz in the Audyssey app or creating the filter in rew only to 200hz and using it full range in the app. Thanks.
Not really, this is for HIFI systems. With surround speakers, you might need to do HF correction due to absurd positions of surround/ceiling speakers. Also you cannot do phase correction with Audyssey. Still, inversion filters described here gave very good results in my Audyssey set up (even without their phase information and up to Shroeder's limits)
Do you recommend or will it be worse if this is implemented as convolution in Roon after general parametric eq of subwoofers in the AV processor ?
You could do that if you have to but you need to measure the speakers with AVR EQ on first. Roon normally would do a lot better on its own though.
Question: should the initial measurements be made with or without subwoofers active?
Final correction should be done on the combination but subwoofer and speaker interaction should be taken into consideration. What crossover filters are being used at what order all factor in.
Dear OCA,
thanks for this tutorial. The last tutorial I followed was the Supreme Audyssey calibration. The new videos show many different things and different approaches. Hence, a few questions in context of home theater:
1. Does this tutorial apply to receiver level calibration? I don't run anything on PC, nor do I have any DSP like miniDSP etc.
2. Is Audyssey ART video the successor to Supreme Audyssey Calibration? Meaning Audyssey ART should deliver better results?
Thank you for your amazing work.
1. Unfortunately, you cannot apply these filters with phase corrections to a multichannel receiver although exporting just the frequency response filters even improved the sound of Audyssey in my own tests.
2. Yes, slightly better HF performance compared to Supreme Audyssey method.
@@ocaudiophile thank you. Does it make sense/is it possible to create minimum phase calibration file of UMIK1, in context of Audyssey ART method?
No, because Audyssey filters cannot correct phase.@@ElectroPineapple
I thought i followed everything correctly. When i load the wav file in the convolution engine in Roon i am not getting any sound....its very very faint. Has this happend to anyone, is this is common issue ?
No, not at all. If you share your mdat file, I'll have a look.
Great video and a much simplified process. I am experiencing an issue when using the Roon Convolution Engine. Using “No Son of Mine” by Genesis as a reference track the heartbeat in the opening seconds is not as ‘focused’ as without the convolution. I have a recording of the difference, but I’m not sure where I can upload it. Thoughts on while the lack of focus?
Try using minimum phase versions of final 1/A divisions as filters.
@@ocaudiophile That definitely brought cohesion to the thump of the heartbeat. Feels like over all result lacks bass. I'll reprocess and add 3dB at the inversion. (REW updated the "1/A Screen" to use dB instead of cycles in the latest version.)
Maybe I'll try applying the room curve first.
@@mweidlick You can simply add a lowshelf filter to the 1/A MP
@ocaudiophile, My Measured LR and Vectored LR are not the same. My vectored LR is about 6 db less than measured with the nulls dropping/ending at the same point. Its like my Vectored isnt summing properly.
I took 5 measurents for each speaker separately and LR together for 15 total. I cross cor. each speaker to itself and vector averaged the measurements to give left, right, and LR. I vectored my left right together to get LR Vectored. Seems pretty basic so I dont know where im off at.
6dB difference is totally normal .Sorry, I forgot to mention that in the video. My measured LR response was decreased by -6dB. When sound sources are combined you get additional 3dB/added speaker. But outside that, I'd expect measured and calculated averages to be quite similar if left and right speakers were time aligned correctly. If removing IR delays don't bring their impulse peaks on top each other, then there might be some delay REW can't remove for some reason with one of the speakers.
@ocaudiophile , Right on. Yeah, the 6db was what I expected and makes sense combined with the nulls not summing. So many places for one to make an error I wanted to check with the Guy that knows.
I watch a lot of these type of videos and most questions go unanswered. Thanks for being so responsive. Appreciate you.
A couple of stupid questions:
- Would the low frequency dips be gone and response be perfectly flat if I had some magic corner bass traps behind the sub(s) that would unrealistically remove all reflections at these frequencies combined with this method?
- Would you recommend using JRiver rather than Equalizer APO for a non-server windows computer too? I don't see the benefits if you are streaming music and video.
Most dips are not as audible as one would think looking at the charts. I don't have much experience with bass traps but you would need extremely large absorbers to deal with anything below 150Hz which is unpractical and it might absorb higher frequencies as well which might dim mid frequencies below acceptable levels. But as I said, I am no expert when it comes to bass traps.
Equalizer APO is free, that's a big plus and and it does the job but it digs itself deep into the Windows sound output and you need to restart the PC even to temporarily turn it off. JRiver is more flexible. Also I have seen very small latency variances with EQ APO at times. JRiver DSP is much easier to use and has one of the best digital volume controls in the market plus it's a great tool for videos & photos.
Is this in addition to a Harmon Curve or is this replacing a HC to avoid EQ above Schroeder?
The method does not require a target curve itself but it can be manipulated to match a curve if needed although not recommended. There's a curve example in the video description.
The overwhelming amount of my source material is ripped from CD, so it's 44.1k native sampling rate, with very few 48k, 96k, and almost no 88.2k files. I use Foobar2000 on a PC with the convolver plug-in. To date, I have generated headphone correction filters in REW using suggested PEQ settings or downloaded others for all the required sample rates that I use for nearly all of my listening and only switch when necessary. Do you resample all your source material to 48k in JRiver and simply use that single filter? Thanks!
My library is similar. REW can export any .wav file in any sampling rate (up to 1536kHz) but I believe most convolution engines do a better job of resampling them on the go, so I keep REW exports at the native 48kHz sampling rate and leave resampling them to 44.1kHz (or any other required by the tracks) to the DSP engine i.e. JRiver.
What was done to get to your L0, R0 and LR0 measurements? Would you also do this same technique for your subwoofer response? Lastly, there is a chance I may be moving from an AVR with Audyssey to one that uses YPAO - if I wanted to continue to use your calibration methods, how would one get this information transferred to that kind of receiver (i.e. a Yamaha R-N800A)?
Vector average of all measurements of each speaker after cross corr align'ing them.
Yamaha has the best PEQ among AVRs, I'd go for manual calibration rather than YPAO. You can customize REW auto-EQ (configurable Generic EQ) with the limits of the Yamaha PEQ and it will optimize the filters for your AVR. I have older videos on manual Atmos calibration, have a look at them.
@@ocaudiophile Is there a specific video I should watch to be able to apply a manual calibration once I do get that Yamaha receiver?
Amazing video as always. I am, however, experiencing around a second(!) of audio latency on windows inputs when I use the filters created following the video in EQ APO - i.e. a mouse click only stops playback after almost a second + all audio arrives later than video. The filters generated according to your old convilution+inversion tutorial didn't cause this problem. Any idea what might be causing this? My computer isn't strained running eq APO and the player, which is musicbee, plus another 10 chrome tabs
Download and try with Jriver (it's free for a month I guess) and see if the latency problem persists.
@@ocaudiophile thanks, will do
just weird that the full range convolution didn't cause this problem
Multiplying the filter with a target curve results in a filter that also boosts in the bass, which is something you are avoiding the creation of the original filter.
Is there a trace-arithmetic that allows us to do the multiplication, and then limit the filter to 0db (i.e, no boost)?
I don't suggest to multiply with a target curve and I don't do that myself, I shared a curve and a method to embed it for the bass freaks out there ;)
If inversion is not resulting in a flat line at the bass frequencies, it's because that bass will be delayed bass composed of wall reflections. There's no TA operation to lift the bass response. You can try applying a low shelving filter (i.e LS 100Hz, +3dB) to the 1/A filter in EQ window and generate predicted from that!
@@ocaudiophile that worked very well. Thanks!
Thank you very much, your videos are awesome. I made everything exactly as you said but i m able to hear a strange preringing (i think) - sounds like a sinus frequency thats playing just before something starts like an open mic that feedbacks lightly on a frequency, that frequency is around 160 Hz... any ideas?
i think using a minimum phase version of 1/A solved the problem, but why is that?
Generate minimum phase versions of the final 1/A filters and use them instead. That will stop any ringing.
The filter resolution is insufficient to create the exact inverse of the phase response which is causing pre-echo.
@@ocaudiophile does using a minimum phase version has any disadvantages? Am i able to correct the phase afterwards with rephase?
Thanks for the video. Is it possible to export the target curve with created filters as text files and insert them into audyssey?
Pls check the answer to
@isak6626 below.
So with audyssey this method is useless because it doesn't manage the phase but only the frequency?@@ocaudiophile
It helps to get all speaker volumes level with each other and see abnormalities if any with certain speakers. For example it can determine how many dBs a subwoofer is too loud. Also ignoring the phase response, only sending the frequency correction to Audyssey target curves improved the sound somewhat in my personal system but still, you have a valid point.@@alexruspantini3484
I just bought a x3800h i have a 5.1.2 with the atmos being upfiring on my front mains. So much with rew has changed since the first video, where do i start with the first video then dolby then art?
Go to ART directly, it has the most automated scripts. You might need to refer to immersive audio video as well for certain basics.
Thank you@@ocaudiophile
Hello OCA! Congratulation! Great job as usual. Can I ask you for assistance? I made "1/A" (mono, 24bit PCM, 96kHz) filter as you presented in this vid but Brutefir (volumio convolver) does not accept it. The message appears every time: "Error with filter size. Filter is not usable! Filter size must be power of 2!". Brutefir accepts filters made by rePhase but in rePhase I can change the taps parameters or FFT lenght. How can I change these parameters in REW to adjust the filter length? Thank you in advance. I wish you all the best.
Under "Export Impulse Response as WAV" window, there is an option "Export this sample count", when you tick it, REW will give you number FIR taps options for power of 2. But I'd careful not to decrease the tap count too much, these inverse filters require a lot of taps to work correctly.
@@ocaudiophile You are great! Just as you suggested. "Export this sample count 64k" helped. Works great, Thanks :)
Sorry maybe for stupid question. I did all steps to create final 1\A filter then I've read in comments below that someone had the same problem with ringing (after I applied filter in Equlaizer APO) and then you advised to generate minimum phase version of 1\A again - I did that but still have this strange ringing. also when i turn off audio it fades, doesnt stop immediately. One more problem I have is that with this filter - volume is very low, i have to apply preamplification +24 dB to make it sound normal. But I use very cheap lavalier microphone from Aliexpress for measurement, ofcourse I dont have any calibration files to create minimum phase cal files for this microphone, I just wanted to see if it will improve FR a little (I have very bad speakers). Can my mic cause ringing problem or i did something wrong?
You must have done something wrong or your measurements are faulty, these are not expected outcomes.
@@ocaudiophile ok, thanks man, i guess i will try just simple EQ then
@@ocaudiophile I dont know what exactly helped but I installed new version of REW, also I changed settings on 48 kHz instead of 44.1 and did all measurements again. I also set upper limit in trace arithmetic to 17 Khz (not just 220 Hz) and maximum gain 6 dB so entire FR got corrected. 1/A filter still has ringing but much less but 1/A MP version almost no ringing at all (it still hearable when you turn off music). But overall sound got so much better that I cant believe it. It almost sound like in IEM's after I equalize it with many filters (I listen tone generator and equalize by my ear). Anyway, I will try maybe improve it after watching your other videos, your method is gold, thanks man!
Dear OCA, I'm a happy audyssey evo user since 2 months😀 I'm planning to have a party in an old barn in September with a 2kW PA system. I'd like to control it with a Windows PC, REW and APO equalizer... Which method for room correction would you please recommend? Because I read in the comments on Audyssey Evo that you now seem to prefer the REW PEQ filters instead of this Mastering Digital Room Correction Video method? Many thanks in advance and best wishes from the Austrian Alps, Simon 😀 @@ocaudiophile
@vortexor1 so long as you use MP version of the response, results will be good but yes PEQ has the edge in terms of group delay (at least for Audyssey measurement resolutions)
Thanks again for a great video. Three questions:
1. Is this method only doable with two speakers, or why is this method so different from the Audyssey ART method?
2. Is there a way to calculate the Schroeder frequency for your room using measurements?
3. How do you click and mark in the REW graphs? I've tried right click, left click, +ctrl, +alt etc., and I don't get the same info when clicking and dragging in a graph.
+ A comment: I wish your Dolby Atmos sweeps were integrated in REW. I don't get why REW does not have that when Windows supports it.
1. It can be applied to multichannel. In fact, I have done such Audyssey calibrations for some members of the channel and achieved excellent results. The problem is with surround/height channels, the speaker placement is so odd compared to a hifi system that, they do need HF correction. Also Audyssey cannot apply phase corrections and the frequency correction with this method is very similar to auto-eq method. PS I do run my surround system speakers all calibrated with this new system.
2. You can calculate it by dividing your room’s reverberation time in seconds by the volume of your room in cubic meters, then multiplying the square root of that sum by 2000! If you cannot, just take it 200Hz and you will not be too wrong.
3. Keyboard CTRL+mouse right click and then drag (and you should be in the correct REW window/tab)
There's nothing REW can do about it, Windows doesn't truly support Atmos. The Atmos sweeps bypass Windows and go to the receiver un-decoded through HDMI.
Hey OCA, great video, as always. Very interesting approach. Want to add a question, what about the HFR1 of Audyssey, how can this be handeled using this method? Creating the filters, multiply them with your HFR0 correction file, then export/save as txt.?
I think there's no harm in multiplying anti-HF1 with the 1/A filter created with this method as it will just be a reversal of what was done to the HF response. However, keep in mind that Audyssey cannot apply phase corrections and the application of this method in Audyssey will be limited to frequency content of the filters only although in my experiences, it still sounded better than any other filter before.
@@ocaudiophile Thank you.
1. OK. Interesting. Looking forward to future videos in case those are in the making :)
2. Thank you. So, the average RT60 decay time (e.g. calculated in the range 100-11500 Hz, 1/6 octave decay window width as per Audioholic's recommendation)? My room has some openings (staircase, opening to a hall etc.), that will increase the estimated Schroeder frequency somewhat right?
3. Hmm, in which tabs does this work? I tried many, can't get it to work.
Weird. Other apps can send Atmos signals. I hope they find a way.
Lastly, would you consider helping out with a calibration (payed of course)?
A follow up on Schroeder's. What decay time do I use? The maximum, at a certain frequency, the average etc? I tried it different ways and figured out that a larger room volume results in a lower Schroeder's frequency for a given reverberation time. My RT60 Decay is about 450 ms on average with a room volume of 96 cubic meters (ignoring openings to other rooms), resulting in a Schroeder's frequency of 136 Hz. I was expecting a higher number, given what I have heard people argue about the upper limit where you can/should apply EQ.
Hey OCA, does this mean that the measurements need to be taken with the MP Version of the Mic calibration file only?
You can also add it later, REW has tools for that
@@ocaudiophile thanks, so you mean that one can replace it in every measurement later, otherwise there will be no harm when taking the measurements with the MP version of the mic cal file directly, correct?
Ačiū!
Much appreciated, thanks!
Thanks for the great tutorial. Can i export these filters to use them with a MiniDSP 2x4HD?
REW doesn't export in .bin format and although it's possible to convert a wav file to bin format (I recently wrote a short script which can do that), MiniDSP FIR filter capacity is 1024 taps per channel which is not enough for these filters. With MiniDSP your best way is to use REW autoEQ filters 20-200Hz and also apply a crossover phase correction for your speakers designed in rePhase for the 1024 taps. rePhase can export in bin format, too.
Check this link:
audiosciencereview.com/forum/index.php?threads/limitations-of-minidsp.51254/post-1867895
@@ocaudiophile Yes, the HD version of the 2x4 has 2048 taps per channel but I realize even that is not good enough for a filter like this. Btw, the crossover phase correction in rephase should be applied to the auto eq filter exported from rew?
@@paulsoumya you can export auto EQ results as biquads directly to minidsp. Then you make xo phase linearization filter for your speakers in rephase and save it as a 2042 tap.bin file and upload to minidsp fir. They work in paralel in minidsp.
@@ocaudiophile thanks. will try it 👍
@ocaudiophile I use a Minidsp for my fronts. I was going to rephase and then autoeq but I'll switch the order up.
Thanks for the new video! Do you have a link to the specific video that explains how to make the min. phase mic cal file?
There's a link down in the comments for that video.
Thanks! Should you also take the measurements with this cal file or is it only needed for filter creation?
I'd suggest to re-measure for better precision of real phase response.@@itsik_
@@ocaudiophile Was it referring to this video? czcams.com/video/i7AaBpqv_p8/video.html
I watched it, and must have missed the part that describes how to make a minimum phase version of an existing cal file. In that video, it seemed like you were creating more accurate cal files using the EarthWorks as a reference.
Is it as simple as loading the cal file in rew as a filter and generating a minimum phase version as described here?
czcams.com/video/ChPu0u3nZxc/video.html
How would you recommend going about SW integration with this method? I have only 2 front fullrange speakers and 1 SW. I am using CamillaDSP with a RME UCX-II interface, so can set each filter individually per channel. I created the mp filters for L, R, and SW. How should I go about aligning LR with SW?
You should be seeing something close to your speakers' spec bass roll off curve after the inversion and the crossover frequency must be kind of obvious now. I'd time/volume align the inverted SW/speaker bass responses to the LP and that should do it. I don't think you will need additional EQ after the crossover with the correct XO point. You'll need to sync impulse peaks at the LP for time alignement. It should also be obvious from phase responses and impulse peak directions if SW needs a polarity inversion. Let me know how it goes.
@@ocaudiophile Thanks for the direction. I created the 3 separate filters for L/R/SW based on the inversions of the minimum phase measurement (as described here).
I "tested" them in REW with A*B on the original measurement and they do indeed flatten the response until 200Hz. I also applied in CamillaDSP and it sounds good.
That said, I have not time aligned them yet, not sure of the best approach here. The first peak of each IR seems to be very close to t=0, I assumed that this is due to acoustic ref but could be wrong.
Also,I was using a UMIK-1 until now but just got a M23R so will retake all the measurements anyway with the new mic.
Lastly, I used the alignment tool to try to align SW to L/R. To do this, I made a vector average of L/R after applying 1/MP filter, and aligned with SW. Interestingly, the best response seems to be when I use the measured SW, and not SW after applying the 1/MP filter, which has some dips.
I hope my explanation was clear.
@itsik_ Congrats on the M23R, I also own one and I never looked back since. You've to just time align a single sub to MLP especially for music. Alignment tool is for multiple sub relative alignment. I'm about to launch a video on subwoofer alignment btw.
@@ocaudiophile Oh nice, looking forward to it. Will that cover subwoofer alignment to main speakers? In the previous comment, you wrote "You'll need to sync impulse peaks at the LP for time alignment." Is these between Left and right, or between them to the sub?
Any chance, that you'll explain how to implement a simple tone-adjustment to the speakers with audyssey? I really like Dyn. Eq and Dyn. vol, but I don't like the sound of an eq'ed speaker, because i listen to the system at so many different locations. I'd like to simply make a +1 bass and -1 treble adjustment to the whole thing, as I find the speakers a tad too bright.
If you replace speaker responses with "perfect speaker" settings in the ady file (a script in Audyssey ART video does that automatically) and set your Audyssey to "Reference" in the AVR, DYnamic EQ and Dynamic volume will work but there will be no EQ applied.
@@ocaudiophile Thank you - but i can already do this by setting the eq-range in audysseyapp to 0(or 20 Hz). What i want is audyssey to "correct" the perfect speaker with +1 db in the bass range, and -1db in the treble range.
Again, Thank you so much for replying.
Like telling audyssey, that i have a "perfect speaker" that plays bass exactly 1 db too low, and treble 1 db too loud, so audyssey corrects it to +1 bass and -1 treble. Like a tone control
Copy and paste the below under "CustomTargetCurvePoints" for each detected channel in the .ady file in the Json editor:
[
"{15\t1.01}",
"{15.5\t1.01}",
"{16\t1.01}",
"{16.5\t1.01}",
"{17\t1.01}",
"{17.5\t1.01}",
"{18\t1.01}",
"{18.5\t1.01}",
"{19\t1.01}",
"{19.5\t1.01}",
"{20\t1.01}",
"{20.6\t1.01}",
"{21.200001\t1.01}",
"{21.799999\t1.01}",
"{22.4\t1.01}",
"{23\t1.01}",
"{23.6\t1.01}",
"{24.299999\t1.01}",
"{25\t1.01}",
"{25.799999\t1.01}",
"{26.5\t1.01}",
"{27.200001\t1.01}",
"{28\t1.01}",
"{29\t1.01}",
"{30\t1.01}",
"{30.700001\t1.01}",
"{31.5\t1.01}",
"{32.5\t1.01}",
"{33.5\t1.01}",
"{34.5\t1.01}",
"{35.5\t1.01}",
"{36.5\t1.01}",
"{37.5\t1.01}",
"{38.700001\t1.01}",
"{40\t1.01}",
"{41.200001\t1.01}",
"{42.5\t1.009}",
"{43.700001\t1.009}",
"{45\t1.009}",
"{46.200001\t1.009}",
"{47.5\t1.009}",
"{48.700001\t1.009}",
"{50\t1.009}",
"{51.5\t1.008}",
"{53\t1.008}",
"{54.5\t1.008}",
"{56\t1.008}",
"{58\t1.007}",
"{60\t1.007}",
"{61.5\t1.007}",
"{63\t1.006}",
"{65\t1.006}",
"{67\t1.005}",
"{69\t1.005}",
"{71\t1.004}",
"{73\t1.003}",
"{75\t1.002}",
"{77.5\t1.001}",
"{80\t1}",
"{82.5\t0.999}",
"{85\t0.997}",
"{87.5\t0.995}",
"{90\t0.994}",
"{92.5\t0.992}",
"{95\t0.99}",
"{97.5\t0.988}",
"{100\t0.985}",
"{103\t0.982}",
"{106\t0.979}",
"{109\t0.975}",
"{112\t0.972}",
"{115\t0.967}",
"{118\t0.963}",
"{122\t0.957}",
"{125\t0.951}",
"{128\t0.946}",
"{132\t0.938}",
"{136\t0.93}",
"{140\t0.921}",
"{145\t0.909}",
"{150\t0.895}",
"{155\t0.881}",
"{160\t0.866}",
"{165\t0.851}",
"{170\t0.834}",
"{175\t0.817}",
"{180\t0.798}",
"{185\t0.779}",
"{190\t0.76}",
"{195\t0.74}",
"{200\t0.72}",
"{206\t0.695}",
"{212\t0.669}",
"{218\t0.644}",
"{224\t0.618}",
"{230\t0.593}",
"{236\t0.568}",
"{243\t0.539}",
"{250\t0.51}",
"{258\t0.479}",
"{265\t0.452}",
"{272\t0.427}",
"{280\t0.399}",
"{290\t0.366}",
"{300\t0.336}",
"{307\t0.316}",
"{315\t0.295}",
"{325\t0.27}",
"{335\t0.247}",
"{345\t0.227}",
"{355\t0.208}",
"{365\t0.191}",
"{375\t0.175}",
"{387\t0.159}",
"{400\t0.143}",
"{412\t0.13}",
"{425\t0.117}",
"{437\t0.107}",
"{450\t0.098}",
"{462\t0.09}",
"{475\t0.082}",
"{487\t0.075}",
"{500\t0.069}",
"{515\t0.063}",
"{530\t0.058}",
"{545\t0.053}",
"{560\t0.049}",
"{580\t0.044}",
"{600\t0.04}",
"{615\t0.037}",
"{630\t0.035}",
"{650\t0.032}",
"{670\t0.03}",
"{690\t0.028}",
"{710\t0.026}",
"{730\t0.025}",
"{750\t0.024}",
"{775\t0.022}",
"{800\t0.021}",
"{825\t0.02}",
"{850\t0.02}",
"{875\t0.019}",
"{900\t0.019}",
"{925\t0.018}",
"{950\t0.018}",
"{975\t0.018}",
"{1000\t0.018}",
"{1030\t0.018}",
"{1060\t0.018}",
"{1090\t0.018}",
"{1120\t0.018}",
"{1150\t-0.019}",
"{1180\t-0.019}",
"{1220\t-0.02}",
"{1250\t-0.021}",
"{1280\t-0.021}",
"{1320\t-0.022}",
"{1360\t-0.024}",
"{1400\t-0.025}",
"{1450\t-0.027}",
"{1500\t-0.029}",
"{1550\t-0.031}",
"{1600\t-0.034}",
"{1650\t-0.037}",
"{1700\t-0.04}",
"{1750\t-0.044}",
"{1800\t-0.047}",
"{1850\t-0.051}",
"{1900\t-0.056}",
"{1950\t-0.061}",
"{2000\t-0.066}",
"{2060\t-0.072}",
"{2120\t-0.079}",
"{2180\t-0.087}",
"{2240\t-0.095}",
"{2300\t-0.104}",
"{2360\t-0.113}",
"{2430\t-0.124}",
"{2500\t-0.137}",
"{2580\t-0.151}",
"{2650\t-0.165}",
"{2720\t-0.18}",
"{2800\t-0.197}",
"{2900\t-0.22}",
"{3000\t-0.244}",
"{3070\t-0.261}",
"{3150\t-0.281}",
"{3250\t-0.307}",
"{3350\t-0.334}",
"{3450\t-0.361}",
"{3550\t-0.389}",
"{3650\t-0.416}",
"{3750\t-0.443}",
"{3870\t-0.476}",
"{4000\t-0.51}",
"{4120\t-0.541}",
"{4250\t-0.574}",
"{4370\t-0.602}",
"{4500\t-0.632}",
"{4620\t-0.658}",
"{4750\t-0.684}",
"{4870\t-0.707}",
"{5000\t-0.73}",
"{5150\t-0.755}",
"{5300\t-0.778}",
"{5450\t-0.799}",
"{5600\t-0.818}",
"{5800\t-0.841}",
"{6000\t-0.861}",
"{6150\t-0.875}",
"{6300\t-0.887}",
"{6500\t-0.901}",
"{6700\t-0.914}",
"{6900\t-0.925}",
"{7100\t-0.935}",
"{7300\t-0.943}",
"{7500\t-0.951}",
"{7750\t-0.959}",
"{8000\t-0.966}",
"{8250\t-0.972}",
"{8500\t-0.977}",
"{8750\t-0.981}",
"{9000\t-0.985}",
"{9250\t-0.988}",
"{9500\t-0.991}",
"{9750\t-0.994}",
"{10000\t-0.996}",
"{10300\t-0.998}",
"{10600\t-1}",
"{10900\t-1.001}",
"{11200\t-1.003}",
"{11500\t-1.004}",
"{11800\t-1.005}",
"{12200\t-1.006}",
"{12500\t-1.006}",
"{12800\t-1.007}",
"{13200\t-1.008}",
"{13600\t-1.008}",
"{14000\t-1.009}",
"{14500\t-1.009}",
"{15000\t-1.009}",
"{15500\t-1.01}",
"{16000\t-1.01}",
"{16500\t-1.01}",
"{17000\t-1.01}",
"{17500\t-1.01}",
"{18000\t-1.01}",
"{18500\t-1.01}",
"{19000\t-1.01}",
"{19500\t-1.01}",
"{20000\t-1.01}",
"{20600\t-1.01}",
"{21200\t-1.01}",
"{21800\t-1.01}"
]
If you want Audyssey only to do that but no correction itself, you will still need to convert them to perfect speakers.
Before watching every video your channel provides, all of these methods require an AVR with at least the support for the Audyssey App right? Because I have an Denon AVR X1200W, so no Audyssey App :(
If you use a PC as your music source, you can do a lot more than you can do with Audyssey and completely free.
@@ocaudiophile That's what I thought :/ Sadly my PC is upstairs and 100% used as a PC for PC stuff, currently using a Firestick 4K Max and a few consoles on my AVR, maybe a little home theater PC will work too?
EDIT: Just came into my mind that only the music coming from the PC can be modified, the Game Consoles and whatnot can't, at least with my current AVR!
@SeGin_ sure, I use a NUC for all my dsp
@@ocaudiophile Which software? I use Equalizer APO to EQ my headphones, is there a more sophisticated one for these kinds of applications? Oh and, when I want to DSP my PS5 or Switch for example, what is needed for these kind of filters?
In my past I only ever had a Omnitronic Smard 24, t.racks DSP Mini and internal DSP's from various pro amps, but I sold them before I gained knowledge of advanced DSP techniques :/
EQ APO is pretty good for PC. For PS5 you are stuck with multichannel AVRs and their own automated DSPs I am afraid.
Discovered your channel through exploring the inverse filter technique for my full range speakers, and the results have been transformative-full control and unparalleled sound quality. However, I'm facing challenges integrating DSP with my JBL 590s, which sound nice but not superior with DSP. Every attempt to enhance them through DSP paradoxically detracts from their performance, falling short of the no-DSP setup. Any insights or suggestions on this would be greatly appreciated!
Frequency domain filters should work below the room transient but that's not enough to avoid sound degradation. Hard to say what exactly causes it in your case without seeing the actual measurements. I can have a look if you share.
@@ocaudiophile so grateful🙏for this opportunity and for all you share. Please find attached two .mdat files for your review. The first file contains measurements and the work leading up to the creation of inverse MP filters for my setup. The second file comprises fresh measurements taken today, without moving the microphone, from a single listening position. These were done after slightly adjusting the speaker placement, as I'm still in the process of experimenting with their optimal positioning, which results in them being moved around. Thank you so much for taking the time to look into this drive.google.com/drive/folders/1yQxKyoWKHtQu2-b-kgZ7wUBxjO35HeSv
Can you share your original and min phase mic calibration files?
@ocaudiophile I've shared the original and minimum phase microphone calibration files in the same folder for your convenience. Additionally, I've included a file for the 90-degree measurements that I had on hand to compare. Thank you once again for looking in drive.google.com/drive/folders/1yQxKyoWKHtQu2-b-kgZ7wUBxjO35HeSv
You've done everything right (one of the best implementations of my methods I have seen to date) but your speakers are not the easiest to calibrate. Left speaker seems to be closer to adjacent walls than the right. Most measurements had some distortion here and there but there were enough of them so I could pick good ones. 4M measurement length was tough to process even for my PC which is a beast, had to trim them to open in rePhase. Anyways, it's all done now and I'm sure they will sound great. Removing IR delays from L & R was a bit tricky, I guess that's where your filters failed. I've also embedded phase linearization (XO590.wav) into the filters. StereoConvolutionFilters.wav is the filter you need. In the highly unlikely case that, you will hear pre-echo with this one, use ConvolutionFIlters_2.wav. That's 110% echo safe but will not sound as good as the other. REW and rePhase files with all details are included in the zip below. Enjoy!
drive.google.com/file/d/1xjEu6LmNIySQnuBVw85xTVJMs4Wf5VtJ/view?usp=sharing
I tried this, but the resulting filter had a right window size of over 10,000. It also causes pre-ringing. What am I doing wrong?
Hard to say without seeing the actual measurements but I am interested to find out. Can you share your mdat file? You can post google drive links here (most others will be deleted by YT).
@@ocaudiophile Sure. I have some strange dips that don't appear in other measurements. I have a CSL Umik-2 but I want to take new measurements with the REW specific cal file that I made minimum phase. I used these measurements along with changing the cal file in REW to make the XO phase filter and that works perfectly though. I sent the link via PM in the roon community forums. Thanks
Looks like it's your XO filter that caused the pre-echo. Here's a fixed correction for your system:
drive.google.com/file/d/1guMQoUQ3Ow3I7kYX4B7-5rn8KTyTVKtD/view?usp=sharing
@@ocaudiophile Thanks! I'll take a look at your corrections and incorporate them into my new measurements.
@@ocaudiophile Hi, I apologize for the interruption. I encountered the same issue. I followed each step carefully, but pre-ringing still occurred. Unfortunately, I am currently at a loss regarding a solution.
Hi OCA, could you possibly help me with your procedure? I'm not sure my measurements look right. Things are not equal on both sides of my listening room. PSB Synchrony One speakers seem to be canceling each other at 35 to 60hz when measured together but not separately. I did the mic min. phase, 1LP for all measurements - measured each speaker 3 times using length 2M, vector averaged L and R separately. My impulse response start looks like a saw-tooth: i cant seem to find where to align left with right because of this. Thank you
Difficult to say without seeing the measurements. Can you share your mdat file here?
@@ocaudiophileMuch appreciated OCA
mdat - drive.google.com/file/d/13DHhElknwMw4zuIafRdVZ13Z9RFdQBKY/view?usp=drive_link
UMIK-1 7026592_mp.txt - drive.google.com/file/d/1KlvO8fExAYBhf2dfS6A7nc7JIYDPnRrA/view?usp=drive_link
UMIK-1 7026592_original.txt - drive.google.com/file/d/10FYX33LtwaqM1aVbXjjmoiGd0eCphMbo/view?usp=drive_link
I need access!@@user-vf7eb3oy8k
@@ocaudiophile Hi, i have given full access. Sorry about that. i will email the files if you are still struggling.
You've changed the acoustic timing reference channel from right to left during measurements, it should always be the same channel for both speaker measurements. But this is not the reason for the dips in combined response, it's to do with asymmetric wall proximity between speakers. I fixed most of it by applying an extra allpass phase filter to the right speaker (Rxo.rephase which produces Rxo.wav filter). The optimal Q value is 5 but this could cause ringing so I left it at a safe 2. You can experiment with higher Q values and improve the dip even better. FL-trimmed_FR_trimmed-48k.wav is the final stereo convolution filter which should sound good:
drive.google.com/file/d/1mgnWTh2GQdGbF-7cckRoImDgGL3V1Xk-/view?usp=sharing
Hello from Dubai ! If you are in Turkey I like to pay a visit for a training to feed by audiophile hunger 😂
Maybe one day ;)
Thanks! So a VBA filter is not needed anymore? Or still better to include it?
If you can find the right VBA filter, it's the only way to remove dips, these filters only cut the peaks.
@@ocaudiophileThanks. Your Channel is eye (ear) opening honestly. I have been playing with IK ARC for many years but this is on another level!
Your videos are awesome, some stuff goes way above my head though.
Just to be clear, your previous video: czcams.com/video/5YcH7j2-L1Y/video.htmlsi=s_U4Wh3Qpb3ZsRaJ named: "REW (Room EQ Wizard) Top Tricks: Convolution with Inversion (no EQ filters, all FIR!)" is not relevant anymore? I followed that one many times, but now I'm confused if I should leave that and try the video above? Or are they connected?
They both invert the speaker response at the LP. The original one, uses a minimum phase version of the inversion. The newer technique inverse the min phase version of the speaker and uses it as a filter as is. The latter has better phase correction but requires more taps. The original one has the advantage of being able to be inverted over a target curve. The other iverts over a flat target. You wouldn't lose much by trying and comparing them both.
@@ocaudiophile - Good explanation and looking forward to the upcoming video re-introducing the House Curve/LS Filter. You've put out a number of techniques, this on correcting Schroder (up to 225Hz) against a flat response, the mentioned FIR video to correct against the entire range against a House Curve, and your Digital Room and Speaker correction workshops of which I only used the EQ FIR technique to correct from 20-200Hz, 200-500Hz and 500-20,000 along with IR window size adjustment to lower my TAPS.
That being said, I hope the upcoming video addresses benefits or preferred techniques such as Schroder 0db/Flat technique, Schroder with db boost/house curve technique, full range against house curve technique (updates to FIR Inversion video vs Workshop EQ FIR video).
Thanks for being the Man and educating us!
I finally got round to trying this. This one gives me way better results. Other one tamed my high mids/highs a bit more. But also had some weird effect on the sound stage for some reason (maybe I did it wrong, who knows).
For my system, in my room, with my speakers adding the Harman curve sounds just a bit richer. But I got to find one with a 2 maybe 3 db rise in the bass. Now it's sometimes overwhelming.
Could I add a small low pass shelve in ROON to tame it a bit? Or would that ruin the filter?
Great work once again!🎉