Audio Grade Electrolytics and the Ever Elusive ESR...

[Bunkey]

Hello peoples of the internet, you smart folk you.

Is anybody with an interest in the wider field of electronic engineering for audio familiar enough with the various audio-grade electrolytic capacitors to know what sort of ESR figures are associated with such caps?

I require a 220uF with an ESR of around 200-300mOhm to use on the output of an ideal LM317 regulator (which could see temps of up to around 50deg in an enclosure).
See the following if you're interested in what I'm obsessing over: http://www.tnt-audio.com/clinica/regulators2_impedance1_e.html

Personally I like the Nichicon FG and KZ series caps - I have a bunch of them here already just waiting to be used - but Nichicon don't publish their ESR spec, nor does any other manufacturer who's caps fall into the 200-300mOhm range I'm interested in; audio or not.
I wonder if these various popular audio caps do actually have an ESR in this region but it isn't published as it wouldn't look good on paper.

I appreciate the application isn't strictly in the audio path to necessitate a high grade cap (heck it's a guitar amp) but like I say, I like these caps and that accounts for more than it's worth!

Maybe one of you has an LCR meter and has investigated this before?

The alternative would be to use something like a Panasonic FC with its relatively 'high' low-esr and put a 100mOhm resistor in series - I'm not sure if there are other issues that could arise as a result of faking it though?


Thank you

[iainpunk]

if you aren't pulling huge currents, having up to an ohm isn't to concerning, if you need the low ESR, put smaller values parallel, like ten times 22uf is still 220uf, but with 1/10th of the esr (if the traces are big enough, the leads short enough and the solder job is propper)

cheers, Iain

[Bunkey]

I don't mind prototyping like that but when it comes to actually producing a finished piece of work what I need is a single cap of the correct capacitance with an ideal ESR of 200-300mOhm. Preferably a quality item specifically suited to decoupling in an audio amplifier.

I'm inclined to put a 100mOhm resistor in series with a Panasonic FC of 50-100mOhm but it'd be nice if these Nichicon audio caps I have already actually fit the bill without having to get all Heath Robinson on that shit.

[R.G.]

Engineering is as much about economics as it is about getting something to work. As Iain alludes to, there are ways to get ESR down to as low a level as you like. The trick is - do you know what the ESR is that you need to reach? Is 200mOhm low enough? Do you need 50mOhm? How about 10mOhm?  The price in both parts cost and board space and volume rises with every step. At some point, the ESL gets to being a bigger effect than the ESR. Do you know where that is? And where to find ESL specs?

One of the more subtle and sophisticated parts of audio engineering lies in making the circuit more immune to whatever it is you're trying to get low ESR for.

[Bunkey]

My 200-300mOhm requirement is based on the fact this will reduce to a little over 100mOhm in the theoretical 50degC temps of an unventilated enclosure.

I require this value to reduce the resonant peak caused by inherent inductance of the LM317.

By my understanding 200-300mOhm isnt low-ESR and that's the problem - Else I'd just use a Pana FC or FM and be done with it. They're both <100mOhm - the FM by an order of magnitude - and as you can see and is explained in the documentation in the link that induces a resonant peak on the output of the LM317.

[Bunkey]

I take no credit for that work btw, I'm literally just reading it off somebody else's site, but of course I'm going to try and implement that in what I'm doing if I can.

At the end of the day I doubt it's really going to make much of a difference in this application but as PRR (I think it was) stated in a topic I read as part of my research; the performance of the LM386 and similar chips that I intend on implementing in the design are really succeptible to the quality of the power supply they run on; therefore I'm trying to create an ideal power supply as far as the LM317 goes.
I could build a discreet regulator that was quieter but I'm working to the constraints of the enclosure and I mean, f-ck man, how far down that rabbit hole do you really wanna go?

Finding a cap with a 200-300mOhm ESR and running an LM317 seems like a reasonable solution all things considered.

...So that is my take on engineering economics.


I just want straightforward assistance from someone who might know the ESR range of the kinda caps I have to hand here man that's all but I appreciate your wider scope.

[PRR]

...ESR of around 200-300mOhm to use on the output of an ideal LM317 regulator..............................

I thought the 'Ideal' grade of LM317 was out of stock everywhere?

Why do you even need a regulator? Most audio is not voltage critical as long as variation is slow.

And as said, there are ways to design audio to ignore rail impedance.

[R.G.]

I take no credit for that work btw, I'm literally just reading it off somebody else's site, but of course I'm going to try and implement that in what I'm doing if I can.

Could you post a link to the site you found that info on? I'd like to do some thinking about it.

I looked at the LM317 datasheets - actually, a couple of them, and it looks like the LM317 is compensated in a manner that keeps it at a low output impedance up to about 500Hz, at which time the amplifier inside the internal feedback loop runs out of gain, and so the feedback can no longer keep the output impedance low. It's output impedance rises, and it's over to the output caps for any impedance reduction as frequency goes up. The blue curve looks pretty much like the LM317's output impedance with that linear rise up to the about-400mOhm flat. The green and red curves seem to reflect the loss of damping the "resonance" of the rising output impedance of the Lm317 and the falling impedance of the 220uF cap.

One way around this is to use a different kind of cap. Ceramic and film caps typically have ESRs in the 20mOhm or less range. A 22uF film or low ESR ceramic with 20mOhm would have a capacitive reactance of 100mOhm at the 8K resonance point, and would have that 20mOhm ESR, so the parallel ESR of the caps at that frequency (and on up in frequency) would nail the ESR to 120mOhm or lower even if the 220uF cap continued to have issues.

Another approach is to insert a 2.2 ohm to 4.7ohm resistor after the LM317 and put the 220uF cap on the outward side of that. It makes the DC regulation drop a little, perhaps acceptably little depending on the application, but forces the output impedance to the load to be no higher than the resistance at DC, but declining to whatever the impedance of the cap(s) on the output bypass happens to be.

One truth in designing low-impedance power supply outputs is that you have to fight inductance with locality - that is bypass caps right at the chip. The inductance of even a couple of inches of PCB trace can cause high power supply impedance local to the chip, even from a perfect voltage source far away. RF guys know this all too well. In the Golden Age of tubes, RF guys had charts of lead lengths of caps and frequency where above this frequency, the cap no longer matters, only the length/inductance of any lead wire.

But I'm wandering again. Can you post us a link?

[Bunkey]

Sorry I was just hoping for a general discussion on the characteristics of audio-grade electrolytics in the hope of picking out something suitable.
Besides, it's an interesting topic and I get the impression what sounds good is a far cry from what would be considered well spec'd
...maybe I should adopt that logic here, order a few different types and roll them through to see if any make an appreciable difference.

...ESR of around 200-300mOhm to use on the output of an ideal LM317 regulator..............................

I thought the 'Ideal' grade of LM317 was out of stock everywhere?

Why do you even need a regulator? Most audio is not voltage critical as long as variation is slow.

And as said, there are ways to design audio to ignore rail impedance.

I'm not sure what you mean, I'm trying to pair the LM317 with the ideal supporting components to get as clean a supply as possible.

This was based on an assertion of your's when troubleshooting an issue somebody else had with LM386 distortion in which you said the performance of the chip depended mostly on the quality of the power supply it ran off - It's not a problem I've encountered yet but I'm designing with that in mind.

Still working on the miniature amp with the single LM386 I started during my last thread (not the one I was designing the rectified supply for but the smaller thing I ended up breadboarding in the process);
For the purpose of practicality and the space available in the enclosure, as well as what is commonly available, I'm using an unregulated 18v wall supply (max 25v unloaded).
The supply is of low quality with a single 1000uF cap, so I'm adding an additional 1800uF smoothing/reservoir cap at the amp side feeding this regulator which supplies a clean 14.5v to the discreet pre-amp (cascaded BC108's) and LM386 power amp (or bridged pair).

I have no doubt there are better ways to design audio to ignore rail impedance but I'm afraid at this stage I'm nonethewiser, so I'm just trying to get this thing performing to the best of my ability based on all the information I've absorbed during the process of designing the thing.

[Bunkey]

Could you post a link to the site you found that info on? I'd like to do some thinking about it.

http://www.tnt-audio.com/clinica/regulators2_impedance1_e.html

The author seems to agree that the LM317 isn't exactly stable and turns inductive above 400hz. I'm sure you can make more sense of this information than me but it didn't trigger my BS alarm so I'm inclined to follow the findings as presented

See also:

http://www.tnt-audio.com/clinica/regulators_noise1_e.html

[PRR]

A LM386 is all current source except the bias resistor which should be bypassed.

A crappy power source may bother a LM386, but it does not need or deserve any "ideal" source. Get a nice voltage and put 500uFd across it.

[Bunkey]

Cheers, I'll keep that in mind.

Tbh the amp has been working reasonably well with an even worse 9v 100mA supply I've been using straight off the wall (measuring 12.5 under load) but it does seem to hit the limits a bit.

I'm sending the finished piece to an old friend and I can't guarantee what he'll plug into it (I know he has at least one SMPS for his pedalboard) so the overengineered and regulated supply should hopefully prevent any issues as long as it's fed with something around 18v+ and a current capability to suit.

This way I can copy-paste the idea into other designs too, once I know what's what..

That A&R A60 stereo (which seems to get a mention in every thread I publish - not intentional!) has a textbook LM317 supply for the pre-amp so I may well implement some of this stuff there; and again the 18v bridged LM380's and 12v heater on the larger valve hybrid design.

Spinning a lot of plates here man but they're all revolving around the same point so it's not too bad

[R.G.]

Still reading. No, the guy's not wild and raving, and the things he's saying don't refute common sense.

Here's one for you to go look at: the datasheet on the LM386 says that if you put a 10uF cap as close to it as you can reasonably put it, the chip has a 50db power supply rejection ratio from few hundred hertz on up. Looks like the LM386 can ignore power line disturbances by a factor of one part in 100,000 above say, 500Hz. That begs the question of why not use the PSRR of the LM386 to complement the production of power by an LM317. Perhaps use a 1 Ohm series resistor after the LM317 and put a 100uF to 220uF after that, feed it to the LM386 and a local 10uF decoupling cap and see if the audio quality is sufficient.

[Rob Strand]

By my understanding 200-300mOhm isnt low-ESR and that's the problem - Else I'd just use a Pana FC or FM and be done with it. They're both <100mOhm - the FM by an order of magnitude - and as you can see and is explained in the documentation in the link that induces a resonant peak on the output of the LM317.

No, it's not the problem it's the solution.

The output of impedance of the LM317 is inductive.   When that feeds the output filter cap, the cap resonates with the inductive impedance of the LM317 and creates a resonant peak.

*Increasing * the series resistance of the cap increases the damping and reduces the peak.    The side-effect is the impedance of the supply at high frequencies increases - as that is determined by and is pretty much equal to the ESR of the cap.

The lower frequency impedance is set by the regulator output impedance.

So here's the dilemma:
- Reducing high frequency output impedance need a low ESR
- the low ESR exaggerates the peaking

To reduce the peaking you need to increase the resistance but that will increase the HF impedance.

So another way is to add resistance between the regulator output and the cap.   That increases the low frequency impedance but doesn't affect the high frequency impedance.

What you can do is set the resistance in series with the regulator output equal to the caps series resistance.   That produces a flat impedance curve.    However unless the the total resistance is large enough you will still get peaking  you then need to increase both resistances to flatten the peaking.

All I'm doing here is highlighting the tradeoff between HF and LF impedance and how to play to game of reduce the peak.  A 100% technical exercise.

I'm not advocating this is a good thing.   Unless the peaking is particularly high it's not worth worring about.    Also the LF impedance would typically be more important than the HF impedance.  So not using resistors in series with the regulator seems wise and choosing a common garden electrolytic will have enough ESR to keep the impedance peak down.     Keeping the impedance low at high frequencies is best done at the amplifier with another electrolytic cap.  *if* you are concerned with the peaking then only then would you add resistance between the regulator cap and the amplifier cap.    You might find the PCB track resistance magically does this for you.

[Bunkey]

Thanks Rob.

I meant it was the problem in that the slightly higher ESR values required to reduce peaking are hard to come by when going on datasheet values in order to pick out a capacitor, as manufacturers don't seem to publish values in the 200-300mOhm range I'd liike - That's why I'm wondering what the mystery values of the audio caps are and whether they're suitable stand-ins here, as opposed to just picking out a cheap cap and hoping its ESR happens to be right.

The slightly higher ESR to dampen the resonance is exactly what I'm trying to do.

Here's one for you to go look at: the datasheet on the LM386 says that if you put a 10uF cap as close to it as you can reasonably put it, the chip has a 50db power supply rejection ratio from few hundred hertz on up.

As for a 10uF cap at the amp:-

The Ti datasheet I'm looking at suggests putting a bypass cap on pin 7. Checking this against the schematic on electrosmash shows that would form an RC filter with the 15k resistor R7 giving a cutoff @ 1hz, so yeah that makes sense and looks pretty decent.

It's nice when the solutions begin to fall into my own scope of understanding 

[Rob Strand]

Thanks Rob.

I meant it was the problem in that the slightly higher ESR values required to reduce peaking are hard to come by when going on datasheet values in order to pick out a capacitor, as manufacturers don't seem to publish values in that range - That's why I'm wondering what the mystery values of the audio caps are and whether they're suitable stand-ins here, as opposed to just picking out a cheap cap and hoping its ESR happens to be right.

The slightly higher ESR to dampen the resonance is exactly what I'm trying to do.

Ooops, sorry for the misunderstanding.

I did a quick sim of the double cap idea.  You can see how this in itself dampens the peak.
Both caps are identical and the ESR is step 10m, 100m, 1000m.




Maybe should explain
- the "I" source is just so VZ plots in units of impedance.
- LM317 output impedance is: 3.2uH + 10m
- we are interested in the impedance looking into the supply

[Rob Strand]

Here's one where you use a common garden cap at the regulator and the one at the amp varies the ESR.




EDIT:

FYI, a way of getting higher ESR is to use smaller value caps.

[Bunkey]

Thanks for running these man, I'm gonna have to take another look tomorrow as it's just gone 3am here and all those lines are beginning morph into one.

You might find some of the graphs on that linked tnt-audio page interesting, especially regarding the cap values.

Another rule of thumb for higher ESR seems to be the biggest voltage rating in the smallest physical package, so that could always be something to fall back on, and I would speculate the Fine Gold audio caps may have a higher ESR than the larger KZ Muse caps for exactly that reason... It's obviously way passed my bedtime

Cheers

[PRR]

Bypassing pin 7 should be considered "mandatory". Without it all the power crap reflects over to the output at half level, and the signal gain function is whacky.

Why do you think you want zero or near-zero source impedance?? Everything is relative! Sure I'd like it if my house lights didn't dim when the well pump came on, which means I would like <0.1 Ohms in the feeder line. But my 0.4r feeder cost a lot and a 0.1r feeder would be 3.5X more. I can live with the flicker.

A power amp driving 8 Ohms looks-like about 50 Ohms to the power supply. So you want less than 5 Ohms for low sag, and many favored-flavor guitar amps allow 20% sag and that's over 10 Ohms. Third-Ohm is better than you need. And sure not worth the brain-pain here.

[R.G.]

Here's one for you to go look at: the datasheet on the LM386 says that if you put a 10uF cap as close to it as you can reasonably put it, the chip has a 50db power supply rejection ratio from few hundred hertz on up.

As for a 10uF cap at the amp:-

The Ti datasheet I'm looking at suggests putting a bypass cap on pin 7.

That's good too, but you also need a local Vcc-to-ground bypass cap right at the chip; generally a 10uF or so and a 100nF will do it.

Whether a chip's datasheet says so or not, it's good practice to put a power supply bypass cap as close to the chip as you can get it. Kemet had a great paper on capacitor basics, titled "capacitors_basics_applications.pdf" if you can find it.