Understanding FFT in Audio Measurements
Vložit
- čas přidán 26. 02. 2021
- Frequency analysis in audio is a common technique (called "FFT"). How it works though is key to understanding its benefits and limitations. Importantly, without knowing how it works, you will have an impossible time comparing one frequency spectrum analysis to another. In this video I explain what an FFT is, and show in real time the effect of various parameters.
See this article on all aspects of audio measurements for devices such as DACs: www.audiosciencereview.com/fo... - Věda a technologie
Thanks for the educational content Amir. Signal processing and basic electrical theory are knowledge which most audiophiles lack, even though they are so very essential for understanding our equipment. You could make a whole video series about basic math, electrical theory, physics and psychoacoustics for audio applications. I'd watch every last video and try to spread the knowledge, I'm sure many others would as well.
It is refreshing to see knowledgeable folks spend their time to advance the general knowledge of folks...it enables their choice to follow a productive path for audio performance vs the myth based nonsense ( audiophoolery ) that so often yells for attention from the sidelines.
Great stuff, keep it up!
Glad we have people like you to cut through all the marketing bullshit and actually educate people.
I study computer engineering and find myself learning something new almost every new "debunking" video.
Thank you Amir, keep up the amazing content!
This is the best kind of video imho. Snake Oil win where there's no knowledge. Helping people understand basics is the best currency.
This is clearly a labor of love. I knew about your website but this has focused your efforts and mission for me. I donated through your website but clearly should increase my donation to support this work and your wonderful community. Thank you again. This is wonderful.
Great introductory video on FFT analysis Amir. I feel like I'm back in college. :) This video shows viewers that you really master electronics engineering and the testing of electronic systems. Some people that have called you a hack in dark basement with some cheap testing equipment, because they didn't like the outcome of the gear you reviewed, must now be feeling real small in their shoes and revising their opinions of your qualifications and honest intentions. Keep up the great work Amir! To those who still don't trust this empirical measurement path, knowledge will set you free... not blind faith in some glorified reviewer's seductive words.
The best thing to come out of your work is how manufacturers are taking measurements seriously and building better products and we can also see a trend in many reviewers leaning towards measurements too :). Had it not been for your website, I would have saved up a quite a bit of money and spent it buying things which got very positive reviews but performed horribly in measurements, will not name them here. Thanks for preventing us from wasting our money :)
Yes the power of measurements certainly cant be understated, and the improvements it brings to actually making better performing gear. Good example is a lot of the "audiophile" amplifiers like single-ended triodes, and even some solid state units, that have _worse_ distortion figures than my speakers at moderate listening volumes. At around 80 dB, the distortion is less than 1%, but there are a lot of very expensive tube amps that post up to 5-6% THD. Maybe you like that sort of thing, but its certainly not the paragon of fidelity they make it out to be. And oh my how quick they are to downplay actual measurements of their gear.
Thanks! This was helpful. I find these types of videos far more constructive than those that appear as battles in the war between subjectivists and objectivists. Knowledge that is conveyed for the value it has in itself is far more valuable than the one that is used as a club to prove somebody wrong.
So, damn,glad..you started this channel. Great vid
This video gave me flashbacks to my EE digital signals processing class.
It brought back the night sweats from university algebra classes. Lol
Excellent video,Amir
This video helps me understanding why MSB can get -200db noise floor!! They use 1M FFT length!
Amir your channel is full of treasure.
Great video Amir (nostalgic memories of my career 20 years back as DSP engineer in audio codecs - not in HiFi but GSM)
Enjoying all of your videos, Thank you.
Great video again... and Yes !! Don't hesitate to put out more of these educational videos... Thank you so much
Thanks for the refresher course, never a bad thing to go back and re-learn something.
Another great video Amir, having you explain it using this state of the art analyzer is quite interesting and informative. Thanks again.
Great video! Thank you. I would like to see more of these.
Top notch content and presentation! I'm going to build a microcontroller spectrum analyzer just for the heck of it and so I'm doing a crash course in FFT.
Super video! To not understand what a theorem or "laws" are / what is debateable, is fundamental. Very, very good Amir!
Superb, clear and informative; it’s definitely refreshed my knowledge! Love this sort of content
I was just noticing this the other day in your graphs. Excellent 👌
Like the poster below, you just flashed me back to grad school and caused me to go into cold sweats. My PhD involved acoustics but I stayed away from FFT as much as I could. Awesome video!
Great job, Amir, especially in the discussion of FFT Gain and the noise floor. There are many years of signal processing compressed in this video; thank you,
This kind of videos are awesome!!!
Great explanation of FFT. I would like to see more educational content such as this.
Wow, THANK YOU for the video, Amir !!! I've been wondering what the FFT size and the parameters means. GREAT VIDEO !! Keep on making them !!
Another excellent video! Many thanks...
Amir, this is superb. Talking about FFT through the point of instrumentation & measurement is educational as well as practical. I am sure all would want to see more of these types of tutorial sessions on physics fundamentals.
@Amir; Great content, appreciate your efforts !!! Please keep up! I am sure many undergraduate and graduate studies students will also appreciate these concepts and techniques workings in practice and also how to use them in practical day-to-day life! FFT forever !!! (Nyquist-Shannon/DSP/FPGA also ;)
Wonderfully helpful primer for me. Thanks.
Thanks, I found this very interesting. Keep it up.
Much better than those goldensound crap. True scholar and a gentleman.
I feel like I’m 20 years old again... 😄
Very good refresher for all those saying “bits are bits” and ignoring the very basic of Transmission Lines and Digital Signal Processing theory.
Thank you for this piece of knowledge.
I am involved in audio engineering 15+ yrs. And still I have learned something today.
Um, that was fantastic. You have an excellent communicative style and are a wonderful teacher. Moooooar please
What a great teacher .....
Yes please Amir. More educational videos. The square wave part of the video takes me back to my additive synthesis days trying to "recreate any sound in nature" with a Kawai K5. 126 harmonics didn't quite cut it :)
Thank you Amir... more please!
yes please we need more,, thank you in advance 👍
Great video, thanks!
Brilliant video - thank you!
wow now i have a headace :D keep up those great videos Amir! love the mix of reviews debunking and tutorials :) in interessting one would be SND+Noise is that calculated with fixed parameters so all have the same result? cheers janosch
Another great video! I think it could be a good idea to make some playlists and stuff current and future videos into them to keep the channel navigable in the long term.
Great explanation.
I was listening and thought the same thing when I heard the complaint of using sine waves to test audio equipment, in that your hearing basically works on the principal of the Fourier transform in that the cochlea essentially breaks the incoming sound into its spectral components that can then be sensed by cilia and nerve bundles, and then recombined in the auditory center. Really a nice example of how fundamental these concepts really are in physics in that they are also in nature.
Thank you very much. Hope to find tutorials on amplifiers measurements, which I find very tricky.
I was thinking, toward the end of this, that many may be baffled just by the concepts of 'frequency domain' and 'time domain'. But this is one very noble effort to communicate the terms and some of the 'why is that' to audiophiles who are still struggling to remember last year's manufacturer model designations.
And engineers aren't generally the dolts that tinkerers like to think they are. ;)
For folks that want to jump into this on their own, all you need is a sound card and software such as ARTA. For measuring higher power devices, there are interfaces out there that can normalize the voltage so that it is friendly to your soundcard, e.g Pete Millett's Sound Card Interface or Jan Didden's Autoranger
I really enjoyed that. In one of my jobs we had to attend several courses per year. I chose random vibration and spectral analysis one time. Never really used it and sadly I've lost the book ! As an aside, I read once that Hewlett Packard's first 16 bit general purpose computer came from their need to have an FFT computation engine. It became the HP 9800 series - I loved that kit.
Thanks for sharing!
Great lesson! One observation: an analogue spectrum analyser also displays noise in the resolution bandwidth so the noise level can be reduced by increasing resolution and scan time.
thanks alot so much great explanation!
I work for a well known Analyzed company that was a HP offshoot, this was spot on.
Thanks, I may have to watch this a few times to fully comprehend 😀
My Goodness, I have watched over and over videos and read books about dFFT vs TD and I finally get it.
Why oh why are these concepts not taught the way you do in my signals and systems class back in college?
Please more of these 'behind the science/measurement' videos!
Can you do an in-depth guide to how you review?
Great video, but unfortunately you forgot to say when distortion is audible.
Thank you so much Professor! If you keep up this great work, you may turn me into an audio big shot, kindest regards
Amir, I have a video idea for you. Could you show how we at home could use something like an Oscilloscope to do some of our own measurements? It's a FAR cry from your audio precision analyzer, but do you think we could still derive some ideas about the performance of our DACs/Amps with an oscilloscope?
I really appreciate these "back to school" videos you are doing for us here Amir. I watch every one, and I learn something new each time. We really need people like you and what you are doing for the community because otherwise "Audiophilia", "high end audio" or whatever you want to call it is doomed. It's already dying and has been for a long while and that's because people just don't take it seriously. The field is littered with so much garbage and snake oil it scares people off and they equate "Audiophools" to the same level as "UFO tin foil hats" people. Seeing all kinds of weird stuff in the sky and hearing all kinds of weird stuff in their speakers. It's actually kind of sad. I'm relatively young and relatively new to all this "all about sound quality-philia" and man... let me tell you. I almost didn't want anything to do with it. But I love music and I enjoy artifact free sound so... I stuck around. Anyhow... wall of text... Glad we have you Amir!
I LOVE being able to see it real-time.
Listening with Motu M2 > BH Crack > HD600 :]
This is great! Could you also give us a tour of your Audio Presision gear? :)
Thank you for this video!! So how do we measure the noise floor and know that we are not filtering it to something lower? If my scope has a 12 bit res then am I limited to 6*12=72dB?
Unfortunately yes, your dynamic range is at best 72 dB. You are much better off using an audio interface and its ADC for audio measurements. Scopes are only useful if you need their high bandwidth which we don't need for audio.
You mentioned perhaps the most common argument against the value of measuring audio gear - that good performance with 1kHz sine wave doesn't guarantee good music reproduction.
I'm intrigued myself whether the circuits might behave differently with a complex music signal than a one-frequency-component signal. Whether that might even show some design flaws.
Could you demonstrate this with a segment of a real music file in a future video? Is it as simple as running evaluated DAC -> analyser's ADC and comparing the result digital file with the original? Or analyser's DAC -> evaluated AMP -> analyser's ADC.
This can be done with two-tone IMD tests, or using frequency combs. Basically you look for the spectral components that are produced. Generally if you know the THD vs. frequency, you can get an idea of how good the linearity will be, provided you stay within the operating parameters of the device in question. With the exception of speakers and headphones, the distortion components are always well below the threshold of hearing in well designed products.
Thank you for sharing your knowledge and experience.
What kind of software are you using in this video?
What is your opinion about
- Audio Right Mark
- REW
- ARTA
when used with MOTU M2, or Focusrite Scarlett 2i2 Gen 3 as ADC/DAC ?
good explanation ty
I feel like I’m back in advanced engineering math.
FFT RTA is that all you care about wow how shallow
I think hollywood re-recording mixers ampas oscar winners sound editors mixes are stupid worthless trash mixes that should be trashed and their oscar stripped from them for doing near field mixes on blu 4k and don't you dare say its the director the studios excuse your fingers are on that mixing console you should be ashamed and should look for a pizza delivery job
@@andysummersthxcinemaandmyc7748 What are you talking about? Did you forget to take your meds today?
Ignorance is bliss. Yer right ha ha.
Thank you! How could one determine with more exactitude the amplitude of the signal conform changing the window and FTT size?
In sound synthesis, there is additive synthesis type. It is essentially bunch of sine generators which when you add up give you the sound you want. In Harmor (VST plugin) there is 512 sine generators which is enough to sample any sound with those (In addition to regular saw, square etc waves that come with synthesizers). The benefit of Harmor is that you have extreme control over the sound as you can manipulate the sine generators individually.
This may be a naive question, but where is the wave form show being computed, in the analyzer or your computer? Is your computer simply a display, or is the data collected being sent to the computer and the computer is generating the wave we are looking at? Also, thanks for offering what I would call "essential building blocks" to understanding the scientific methods which you use. I may have to watch this particular video a few more times to grasp all of the concepts, but for me it's time well spent. Many thanks for your time and patience!
Interesting.
Hello, great video, just to test if i understood the video:
If my fft lenght is set to 48K and the audio i want to analyze is on 48KHz too, would it take 1 second to compute the fft for each result?
hi, could you make a video about gaming motherboard that advertise good audio?
thanks for this, I'm trying to understand the FFT display of the korg NTS-2...
Great vid! So what is then the real noise floor in your example @6:55 - 130dB?
It'll average out to about -121dBr as that's roughly the level at which johnson noise from resistors takes over and really there's basically nothing you can do about that.
Can you do different mp3 rates? And compare 320kps mp3 with 128kps aac? Then flac?
Brilliant but very meaty one this!! A bit tough for me as there is a lot of knowledge required to fully understand it ... which I don't have. Thanks for trying to explain it though - I've been looking for an explanation of the FFT plots for a while! Given the jitter in modern DACs is mostly (nearly always) inaudible and the frequency responses are flat across the audible spectrum how can you measure the differences which are clearly perceived by hearing.... are we missing an important measurement? Or perhaps I need to listen "blind" with level matched output :-)
Yes, concept of FFT is not easy to digest. On your question, the jitter measurements themselves don't say what is or is not audible. I do however perform that analysis using psychoacoustics in my reviews.
Fourier transform.. brings back memories from University.
Thanks for the informative video. I am new to this channel and your website. I would like to know how this is applied to evaluating audio equipment.
@Douglas Blake Thanks for the explanation.
So can we use this to measure a complete system with a mic and pink noise in our homes via REW and RTA to fine tune placement of speakers?
I assume we could see comb filtering in ~real time, as we make changes to toe and distance? Would this be more precise than tape measure and protractor?
yes, of course you could use it, though you wouldn't wanna use pink noise since it falls of 10dB/decade, you would use white noise because it's flat (at low level, don't wanna blow out your tweeters). Don't get a heart-attack when you see the results measuring the room and the speakers like that and use sensible smoothing like 1/6 to make sense of the resulting chaos.
Can you please show/explain how the SINAD is calculated from the FFT data and compensated for FFT gain and bandwith, .?
HK used square waves as their base for measuring the quality of amplifier outputs. They sounded clean and the bass was nicely rendered.
I'm sure that a 'golden ear' decided it wasn't 'warm or musical'.
Which software are you using ?
More please.
In this video I hear a faint buzzing, I think the E below middle-C, and I'm surprised it doesn't bother you!
Great video. But you never explained how to measure the noise floor.
Amir. Great channel. Subscribed ! Can you please show us how to take simple measurements of our systems with REW? But for very basic things such as how to properly integrate a subwoofer using REW. How to set the crossover and phase.
Thank you !
Thanks. It is tough to do a lot with REW for electronics measurements. It is great for room measurements and I can show that in a future video.
@@AudioScienceReview yeah I would love that. Something more for regular stereo users. Most of the stuff on REW room measurements is for home theater. I tried using RTA on REW for my sub integration and I think it worked. Would be cool if you could show some basic measurements and how to read them.
Btw. Love your channel and measurements of gear. Finally some real science behind the gear and ugh cables. Ha.
Keep up the great work !
Did you ever test ifi hipdac? I hear lots of clicks at time
Amir do you make music??
what is the software ?
Amir error on your comment @20:03 , tilt (slope down )on top of square is because you put AP555 High-pass filter on AC(
That's one of the key points: 23:30 -> 24:00
♥
Trust is good, control is better.
It's unbelievable how much money is invested in useless high-end equipment these days.
Well, it probably secures jobs and fills the pockets with money. Anyway I also belive in good cables and high quality Hifi components and speakers, but I do not follow every new hype.
Music is what we want to enjoy at the end of the day! To many people are more and more focusing only on equipment and forgot to enjoy the music.
Your videos are excellent.
Can’t you test the noise floor of a device simply with a sound measurement device?
Hello amir. I hope this comment makes it to you. I am currently looking at the emotiva line of gear. Xmce pre pro and 7 and 2 channel amps and 2 of their most powerful subs. I know this is alot to take in. Im really the most interested in the xmc2. Could you help me. I really value your input. Thanks Allan from st george utah
From the back of my mind (not involved in that kind of electrical or audio-engineering), for a NON-periodic signal (a whole tune/song), isn’t Fourier Transform better trashed in favor of other frequency domain tools such as Power-Spectral Density Analysis ? Sure Sine waves and frequency sweeps give a good indication of anything going obviously wrong, but that doesn’t address transients (which seems to be the Holly Grail for « resolving » systems), does it ? Unless you analyse Goldfish EDM , can you really consider a time « slice » (not to use the word « sample ») as a pseudo-periodic signal for Fourier to hold ?
Great time listening to you !
(discrete) Fourier transform is what you use for non-repetitive signals/transients. I believe you are thinking of Fourier series, which describes the harmonics of a single cycle or a repeating function.
👍
Uh, I have to read up on FFT gain. It seems that a lot can be changed in displaying. Very confusing.
No fun, indeed, but educational, Thanks.
Mr Majidimehr, are you teaching a protege?
Asking for the whole community.