EEVblog
Vložit
- čas přidán 6. 09. 2024
- Dave looks at some traps in chips and their modes of operation. In this case the TI LM2776, a look a the classic 7660 charge pump voltage inverter, and output ripple and ways to reduce it.
Parametric searching, and a look at some interesting rail inverters.
The uSleeve: www.eevblog.co...
Forum: www.eevblog.com...
EEVblog Main Web Site: www.eevblog.com
The 2nd EEVblog Channel: / eevblog2
Support the EEVblog through Patreon!
/ eevblog
Stuff I recommend:
kit.com/EEVblog/
Donate With Bitcoin & Other Crypto Currencies!
www.eevblog.co...
T-Shirts: teespring.com/s...
Likecoin - Coins for Likes: likecoin.pro/@...
💗 Likecoin - Coins for Likes: likecoin.pro/@...
Seriously Dave. I don't think people like man can stress the importance and value of videos like this. I'm a software engineer by trade, but consider myself an advanced hobbyist. Because I don't work in the industry or along side other electrical engineers, these "gotchas" aren't things that people like me get to see. Honestly, I'd almost be prepared to pay for your time to do a video each week just like this one. Just a brain dump of little tips and tricks just like this. Not only do we get to learn something new, but we also get introduced to parts that maybe we didn't know were even out there or how to use them properly.
Thanks, glad you liked it. Unfortunately these types of videos aren't that popular, it seems that people would rather see me review a $20 multimeter
EEVblog - So when is the next cheap DMM review? ;-P
EEVblog still important videos. each video is a resource that didn't exist before. and when it's needed it's REALLY needed
ElmerFuddGun - I'll get right on it. How about a $5 one?
EEVblog - $5? Can't you find one under $4? And be sure to compare it, safety and all, to the 121GW and some nice Flukes... ;-)
I used a 555 as an inverter once. It wasn't critical but did the job.
Me too. Kinda bad but it did the trick.
If all else fails, simply add a 74LS04 gate inverter chip too. he-he
you think designing circuits on PCB's is tough, you'll find silicon even harder to design in. In my 15+ years in Semiconductors, I'll openly tell you, EVERY part has gotcha's. EVERY part has compromises that were made for time to market, cost, power or size. Good product definers at Semi manufacturers do their best to minimize those gotcha's impacts in their target markets.
The test of a good design engineer is to find what compromises were made, and decide if it impacts your product negatively.
Indeed
Let alone the tolerance variations of components which aren't always exactly what the data sheets shows. So it's a good practice to measure and verify all components individually as much as possible (especially the passive ones).
Agreed.
I have learned to not just download the Datasheet, but to always download the latest Errata Sheet for the chip too.
The Errata Sheets list all the "known" gotchas, and typically list the recommended work-around for each one too.
New segment maybe? Part surfing Saturdays?
"The problem is that we're literally only getting the 7660's here..."
Search Entry: 7660
This shows why Dave mentioned spending days/weeks/months in part searches for projects in some of his other videos. Dave is right about searching by function for the 7660. I did notice that if you search by the Charge Pump entry under Topology you do find the parts. Part selection is so much easier when you just want to make something for yourself that isn't going to have a major production run.
Great video. I didn't even know about those bias generators. There are so many ICs out there; always something new and interesting to find!
Yeah and I just finished a project where I could have used one. Darn. Maybe V2.0
I also liked that dual output +/- chip. That could have saved me a +5v regulator in another project where I used the 7660 and an opamp.
Thanks Dave
bitshiffed Just too bad at least 2 of the big wholesalers omit or hide the category so designers can't find or discover them. But that seems a general problem with their public catalog systems: Too much messed up by office workers who can barely tell plus from minus (just try parametric search for p-mosfets, office workers not stripping the inherent minus from half the data sheets ruins the database).
Dave, you have goofed the searching filter. You have "7660" in the filter, so it's not surprise that no other chips than 7660 were found.
I'm using LM27762 Dave mentioned at 24:31 to generate clean bipolar supply in photodiode amplifier. Smallest BOM area and count definetly, and works like a charm.
At 2:40 you're only finding 7660 on digikey because you had that typed into the search within results box and it applied a "7660" filter at the same time as the charge pump filter. Without that extra filter, your LM2776 is the 7th result.
Come on Dave! When you searched digikey you still had 7660 as a search term. Also you could have selected negative under the output configuration selection to get the inverters... Anyway just 5 minutes in and that was bugging me. Seems like a good video otherwise. (Although I could certainly be missing something, just a basic hobbyist)
Honestly, what scares me of that chip LM2776 is that the ripple changes its frequency over several decades. Great video!
And that is the problem, it's difficult to tune the filter over several decades without introducing side effects (apart that you can't be sure what some of those frequencies can excite).
Well, think of the case the load increases suddenly for a while, your low pass filter will behave like a high impedance to the source, and the voltage will drop immediately. You are at the mercy of what capacitors connected to the line can provide.
thanks for this dave. we recently found the high speed dc/dc converter and we're so stoked waiting for it to arrive. the low output ripple is everything for us.
you're a hero to us mortals. giz kiss
So the moral of the story is learn to read data sheets before you use a part.
PhilfreezeCH Way back when I started in electronics the data sheets were more like a guide line on how the part might work. The application circuits were not even tested or simulated. Today most of the data sheets are directly from the design itself so they should work and most companies have demo boards to prove it. Some times you can get models and simulate yourself.
The data sheets sometimes lose something in the translation but not like the crap they use to be. There is a 2 letter companies in Texas that I find always works different than the data sheet. They are very creative spec writers and support sucks.
Or rather: beware of PFM charge pumps (and in fact many charge pumps which change ratio dynamically). Other than that, nothing new under the sun.
I now couple a traps you can not figure out from datasheet. And one of them even can kill almost every IC that is behind the trappy part.
Read the application notes as well, often so much more in those!
Also, it's a good practice to measure and verify all components individually as much as possible (especially the passive ones).
Regarding the LM2776 chip, just make sure to use a minimum load resistor of the correct value on the output so the chip doesn't go into that weird low current PFM mode. Easy. A little extra mA current draw doesst matter too much anyway. Most switching power supplies don't work properly without a minimum load and this chip is no different.
More datasheet analysis/explanation videos Dave! We learn a lot from these.
Two excellent engineering related videos in a row! Thanks a lot. I know most people want the mailbag (I do love it as well), but the good technical videos are why I subbed years ago. It's also cool to see you get your nerd on. Great job!
Dave, can you please make a video on capacitance being affected by dc bias and the type of capacitor it is, as mentioned in this clip@ 31min, looks like it is a very interesting subject, if you haven't done one already.Thanks again for these awesome types of videos.
Great advice. There are so many traps in selection of parts. It's great to hear your views and recommendations.
I had to deal with a bunch of devices where the power supplies failed, and I remember seeing something like the PFM mode in the datasheet for the switching controller when under low/no load conditions. This video makes me wonder if high ripple could have been the reason diodes on the buck converters kept burning with no load.
I've always said: we as electronics design engineers spend a whole lot of time reading. 50%+ of the day is reading datasheets. I don't trust the marketing overview - I only trust it once I see the figures and believe them. No wonder we like using the same parts over and over!
"... that that PFM regulator... your output ripple goes to buggery"... make me crack up. xD
And I'm just designing a noise sensitive project with this part. Nice Timing. Thanks Dave !
Thank you. Wonderful video.
I love the videos you make that potentially save us all so much time. No point in all of us falling in the same traps...
Please, make more real life lessons videos, like tis one.
Cheers!
ooh a blind spot here, thanx Dave for teaching!!! :) :) I didn´t know that these exists. Usually i go for a recom ROE line device when i need a separate/negative rail.
You will find them at digikey, trust me. You just need to use different search criteria. I will check later how. Although I am not a huge fan of these charge pumps. There are other ways that work much better.
But yeah, good point and good video. Sometimes it can be a real pain to find these little quirks.
Minimum load is also a similar example.
The conclusion i got from this video is when in doubt about your part always add a load resistor to meet minimum current draw for less ripple or more stability..
@DLS
i also thought that as soon as he mentioned it :-)
DLS Try that in a battery based design or in a standby circuit subject to legislation against power waste.
Hey Doe,
I will implement that in a battery based design if the stability and noise performance of my system is effected by the switching device ripple.
As for the legislation part, i believe i don't have to worry about that for my hobby projects. That is something to check at the end when i intend to sell my design for a consumer or medical or military entity.
Hi Dave, thanks for this informative and important video. This is why I watch your videos, because they get me a step further. Thanks for letting me participate on your huge knowledge. Online they sell this cheap current to Voltage 4 - 20mA, to 0 to 5/10 Volt converters. They claime they are galvanically isolated. I bought a couple of them, and indeed they are. They have two chips on board, a 7660 and a quad op amp. I wonder what trickery they used and what extend it is galvaic isolated. If you could make a little movie about this magic part I would really love it. Thanks for your work, catch you later, Lisa
Thanks for the Video. DAVE! I like the videos where you can show your knowledge as a electrical engineer/designer, because you shine!
Really nice, and informative video. I would happily welcome more of these. I've learned the hard way that sometimes I have to read not only the data sheet and appnote, but the errata too. It cost me a few days and nerve cells - I'm looking at you Microchip.
I really liked this deeply technical video. I am almost watched all of your videos multiple times. Really they are good.
This is why companies with a reputation to maintain employ graduate engineers.
Regarding the LM7705, I'm wondering if just using a diode or two for a ground offset with a Jelly Bean/Joe Blogs part would allow for sensing down to ground. If you can't find a simple solution with common parts, you failed it.
Digikey's search has an output voltage selection of '-vin' which made finding parts pretty easy.
7660's are used a bit in guitar effects to double up the voltage from a 9v battery
Datasheet curves also depend a lot on the measurement instruments used. If the manufacturer is ISO certified, the company should have a record of the particular instruments used in measurement. For example, ripple measurements may be lower if a power supply is used instead of an SMU.
Justin Spencer I have seen data sheets specify very specific instrument models in a way that's just annoying. As a designer, I want specs that apply consistently over a parameter range (such as the specified operating conditions on page 1 or 2), not numbers that only apply to the manufacturers demo set up, with huge "appnotes" that explain away but don't answer the questions.
As my grand dad always says, "all that glitters are not gold", so never believe that first page in any data sheet because it is a sheet allocated for the marketing department. Just wanted to screw it bit more, I guess you guys can make out the rest. Bingo!
Please make more of this digikey surfing videos. Such an amazing video
Didnt you have „7660“ entered as a search term on digikey (look at the filters area bottom left)? 3:19
Oh man, where was this a year ago?? I had some problems with a project that used the LM2776 with an LT1675. Breadboarded, the circuit worked fine, but on the PCB I was getting all kinds of bad effects. I couldn't figure out the problem, my confidence was shot (it was my first design and I'm very much a beginner hobbyist) and I kinda let it slide. Now that I think about it, noisy -5V could explain the issues I was having.
The LT1675 maximum current draw is 42mA, the LM2776 probably was spending a lot of time in that damn PFM mode. I wonder if a bodged 50 Ω resistor, drawing a permanent minimum 100mA would do the trick?
Daves the best!!!
I wish i could watch this before starting design of my PhD thesis. Now i'm looking a pin-compatible alternative to 7660 because it requires very close capacitor due its high frequency. Probably will end up using low ESR expensive capacitor.
ad9854 - no bypass cap possible on vref diode while it is possible in ad9851. Thanks ad
yeah, I looked for a -V supply circuit/IC for a simple Low noise amp some time ago, and ran into a few problems with most stuff available. Nothing was able to operate from the supply of 10vdc to 15vdc input (13.8vdc nominal), The few that could was way more complex than the problem using a negative rail would have elevated, and nothing really gave the needed +/-12V output that was needed to drive the load. At the end of the day, the needed to have input isolation via a large capacitor dictated an electrolytic anyway, so the negative supply thing just never happened, lol. It was just easier to do a lot of filtering on the 13.8Vdc power-in, use an LM78L06 for the virtual ground for the amp internals, and slap large caps on the inputs and outputs, lol.
As for them Chip Traps, I've seen it just looking for a good output buffer IC. Some praise a few chips for one spec, however, they require a ton of extra junk to work with a single power rail, and/or have horrid response curves once you exceed 100mA out of them, lol. Sometimes a simple BUF634 bolted to a heatsink is still the better choice, lol.
Great vid Dave and crew. B)
Pure-F'ing-Magic operation. Sounds like my kind of chip!
Use one of these, NMA 1W Isolated DC-DC Converter.
Not cheap, but it fixes the problem.
the dip only makes it better :v
I think I'll use the Lm7705 in a project...
Recently I heard about effect of DC bias on capacitor's capacitance, but had no futher information how big this lose is - it would be great, if you could make a video on it. People say that using e.g. 16 V caps for 12 V supply rail is waste of space and money, because these caps are fraction of their nominal capacity.
I did not channel my inner Admiral Akbar when I skimmed through datasheet and missed that the op-amp selected couldn't handle more than 2V differential voltage on inv/non-inv input. Wow, I was perplexed for a few hours with the results being good.. good... good.. f****k up! as soon as slightly elevated signal strenght!
Selecting parts for uSupply? Nice tips on finding alternates (Digikey is my only source, so I'd be sol without trying mouser).
Is there any good information out there about how voltage ripple influences other components like an ADC?
And how the PSRR helps to reduce the influences?
For example, if I had to decide between regulator A and B, one with 10mV ripple, the other with 50mV ripple, beside the "lower is better" argument, how can I determine what influence the different regulators would have on my ADC and ultimately the conversion results of that ADC?
What are you building Dave? I have noticed that you talk more about inverter ICs these days
Did mouser clean their categories lately?
In a bipolar analog application, would you recommend an inverter IC or a center tap transformer followed by regulators for positive and negative rails?
This has to be the most frustrating thing as a beginner in electronics. Choosing a part can be so difficult without hours of studying data sheets!
Wait what, where is the TS100 vs TS80 video? I just watched 1min, when I decided to got to bed and watch it the next morning as breakfast TV and now its gone :(
This happend with a few videos before, and it is very very sad, so sad..
Did you really make that many errors, that you had to take it down? :O
What I am supposed to watch now?
He took it down because too many people were upset and wouldn't stop making a fuss, after he showed that the TS80 is overall the better iron.
Sadly, these people seem to have completely missed the point, that their TS100s will still work just as well as they ever did, and as long as they're happy with the performance, why worry?
You can find a re-upload of the video of you just search Google...
I have one query other than this:
I am using TC7662, its working on breadboard giving me -10V from +10V but when i use it on zero PCB its not. Tried everything still not working. Is it also some kind of startup problem. I am posting this problem to every electronic expert on youtube. Its freaking me out :(
Man do I love part surfing
Maxim was the brand that made me suffer with datasheet tricks.
max964eee comparator that apparently has internal clamping didoes...
so during events where the voltage difference on the comparator is higher than 2.1volts current consumption from the input jumps drastically... causing circuits to fail...
maor malka Yeah, they earned a reputation for popular but expensive unique parts like their RS232 signal level converters or the 1-wire stuff they got from Dallas and locked down to block other manufacturers from introducing additional functions not in their catalog.
WAFFLED ON !! YOU KNOW IT LOL
Dave, if you notice on digikey when you selected Charge Pump at 2:40 it came up with a "Search Entry: 7660" so the word charge pump directly related to 7660 on their site, so that explains why all your results were limited to 7660, poor filtering choice on the part of digikey.
Hey dave what are your thoughts on the Quick 957dw+ Rework station?
Jesus... and I thought I was going to hear how an inverter works.
OMG
I live this shit every day. The problem is that there is not a parametric search engine 'to rule them all'... It doesn't exist... if someone could just find a way to solve this problem we would all start bitching about something else entirely!
good video
Do i need to care about 2MHZ ripple in an analog audio circuit?
As always in engineering, the issue is usually a PEBKAC error, better known as READ THE FUCKING MANUAL.
Just don't try Arrow's part selector. OMD.
*Fun fact:* saying _"double 6"_ or _"double 7"_ takes *double the time* than just repeating the number. Double LOL. ;-P
ElmerFuddGun actually 50% not 100% longer. 3 syllables vs 2
May take longer, but it is more clear. Also standard way to say repeated numbers or letters in English for brits or ausies. in any case it makes for an easier time when entering long chains of numbers read by another person.
John - I gotta disagree. When I was listening it was more difficult to "see" which IC he was talking about. For some things like "1.0005V" I can see it maybe being better to say "triple zero" or "triple O" but here I found it harder to follow along as they all seemed to have double numbers. Also consider why "911" is not spoken "nine double one." Thanks for your thoughts though, John.
Sure, but it avoids confusion when you wonder "did that guy stutter, or is it actually double zero"
Waw, thats a video!
Trap = Bloke convincingly dressed like a Sheila.
Can you help me to find a trap chip
Of course none of that is mentioned in the "Features" 😁
19:33 does it say 2mm x 2mm? these components are definitly tiny.
Are you saying you handsoldered it with a soldering iron? o.o
I mean, 0402's are smaller, but those only have two pins. The 2mm x 2 mm chip there in the video got 8!
Those smd resistors should be quite a bit easier than those leadless packages.
6 pins on 1mm x 1mm is quite insane too, of course.
manufacturers should be forced to manufacture one version of any ic that is smaller than dust lol.
Well, you can solder lead-less, or whatever thats called parts by hand with an iron. but it already is a pain in the butt when they are not microscopically sized.
Hot air is a better idea though, lol.
You should try 01005 resistors one day: 0.4 mm x 0.2 mm
oh i know those exist. Dust sized pretty much, but at least those only have 2 connections.
Devil is in the detail!
Why manufacturers provide misleading/difficult to read information? What's the benefit? To hide it is actually not that great? To sell?
to sell crappy parts for upmarket prices, why else
gezuz christ dave. you know fucking eeeverything! this is so valuable for me. thank you so much. i give you screw all money every month and you give me brilliant education like this. if i wasn't such a poor piece of shit i'd pay more. thanks once again. take care amongst all that fire & smoke.
tl;dr: Read the full datasheet?
"I'm sorry Dave", but I don't see what part of this isn't covered in EE101. Basic power supply stuff that hasn't significantly changed since the days we were using 6X4 rectifier choobs.
I have made a board and used the ADS1274 in a project of mine... you got nothing on me brah!
dont wanna come a gutza
"You" should be on the right side of the video. I know you are _looking_ towards the right (our) and the data sheets but more often than not you are over top of stuff we want to see. ie. at 18:17 Not a biggy, just a suggestion.
Yeah, I tried to avoid that but goofed at that spot. It's always a trade-off.
Oh well. Just a reminder for next time.
Thumbs up if your CZcams is broken.
Yeah, I don't seem to get notifications anymore to replies etc.
So RTFM...
Sometimes the gotchas are in the documentation, or insufficiency thereof. Firmware has a huge number of these as IDEs will do all kinds of things to you rather than for you. Duplicity of state can happen in IDE specific project config files, which are undocumented as they are ordinarily not supposed to be user facing. Then something breaks cryptically and you have no idea what but it involves Debugging your debug settings!!! When innovating in ease of development, many vendors did things that make it really easy to break ISO9001, revision control, and/or your engineering time budget, all at the same time. That was my day today. FML!!!!
I strongly dislike most of the closed source tools for this reason. If I can't see how it works, it darn well better work in a well documented fashion. I would very much prefer a Unix style of separate but integrable (do one thing well) tools and documentation with a well documented and scripted workflow to make everything user readable so that I can fix ALL the code that wrongs me. Sadly that is not a standard my workplace allows.
As for hardware, read the datasheet (and the errata) and interpret it literally, then apply a whole shaker of salt. You still will only know 90% of what the part will do as it takes a knowledgeable and creative guru electronics guy to really guess good enough on a datasheet 99% of the time. Then design, build a prototype, and test thoroughly. You may yet be writing your own errata on a part you bought!!! There is a manufacturer which is blacklisted as a vendor at work for insufficient documentation and support. They make good parts, but only sometimes documented them sufficiently, nor provided correct support. This once cost a month of engineering time for multiple engineers. Boss put his foot down on the matter.
not amouse Did you by any chance battle the SiLabs "Simplicity" toolset. Their horrible web-first spyware happy approach killed the brand for me after already doing a board layout. How hard would it be to just publish PDFs, samples and the minimum add-ons for using generic 8051 toolchains? But no, you have to install a huge IDE with a spyware license just to download the firmware samples and get the tools for the debug/programming interface.
Keil, actually. Its actually one of the better IDEs when it works and isn't a new version. MPLabX is decent too. I may soon have to deal with "simplicity" in all of its complexity. ;-) I also dealt with the previous version of that, which was good enough. My actual worst experiences with engineering software were Altium's community edition: "so other people can make your footprint mistakes for you... ouch". and various forms of academic fool oriented Spice softwares that aren't LTSpice: "So there are things that nobody actually needs and we put them in the front page for all to see while you need to go through 3 layers of menus to access anything with a semblance of usefulness". For the love of god, never "give me a menu of common passive component values when I can remember and type them". Certainly don't crash 10x as often.
90 percent of all closed source engineering software is utter garbage. The reason is because it is marketed to simpletons, the lesser of the technicians I have known, internal and external marketing wankers, academics, and those who lack the capacity to comprehend usability of a software as they only check "has a feature" boxes. It is also rarely tested for its ability to cause lowered grades and increased stress. Keil during my microprocessors course failed one way consistently during a debug session causing me to rage-quit on my last homework of the semester. Thus I got a B.
What I want from engineering software vendors is simple:
1) Scripting support where it makes sense: either python or Javascript or any arbitrary language through a set of well documented commands.
2) A user interface which doesn't hide anything we engineers should use with any frequency. Unless there is a need for heirarchy, type of parts for example, don't put anything more than 1 menu deep. The old windows file, edit, etc drop down menu style was ideal. No second level in right click menus unless to shrink to fit the user's screen. Don't change UI design between versions or offer an old version mode if you really need to do so. Especially if you make it less efficient for me to use. e.g. LTSpiceIV UI was PERFECT, XVII was not wholly an improvement in that direction.
a) It shall preferably be adjustable.
b) It had better have configurable hotkeys.
3) make the "cloud" optional, and less nebulous in its benefits towards your consumers. Many of us in corporate environments can't use the cloud for anything R&D or production related as the security isn't there.
4) It had better all work with version control. Especially external version control programs to replace whatever junk got piled in with the software. Especially git for software.
5) (in gandalf's voice) "You Shall Not Lag". Nor hang. Not acceptable, not unless what you are doing should be computationally intensive. And even then: Never lag the UI thread more than 50ms on an engineering student grade laptop that meets decent spec, EVER.
6) Educational engineering software should NOT have any feature that a good engineer would not use on the job. The oscilloscope thing in the front page of Multisim that was useless compared to a net node label comes to mind. This so called "oscilloscope" didn't need to be a first class feature!!! Schools should teach people electronics with a breadboard, $20-50 in parts, and a well input protected oscilloscope and multimeter. Or do schools not think their students future is worth that!!! Simulate with something which approaches the simplest abstraction that works. Node names and autogenerated netlists and recording and displaying waveforms using Spice.
7) Don't screw up the drawing of wires by failing to follow the principle of least surprise. LTSpice does it better than any other simulator.
/rant
some one didnt like the video he get banned eventually :D
chips arent gay
Yes, they gay.
interesting info...thanks Dave, you shill! :P
does this make me gay?
bs aside this is useful. thank you
Great video but ditch the groovy stuff...
What is "the groovy stuff"?
Digikey is terrible. I ordered 1 mosfet for $0.98 and had to pay $28.00 with postage in the end.
Most suppliers have minimum order values for free or discounted shipping, that's nothing new. As with Dave's video - always read the shipping terms and conditions!
How did you end up doing that? Whenever I price things up it always gives me the total cost including shipping on the confirm order page.
The Kaiser Digikey is wholesale, not hobbyist. Their business is selling huge piles of components for your 1000+ device mass production, but allows lower counts when needed.
Some places in Europe have free postage on everything including sub $5 orders.