-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Capacitor distortion
- Thread starter pma
- Start date
solderdude
Grand Contributor
Input coupling caps usually are located before the feedback loop so these are not compensated. Usually low values and any competent designer will use film caps there.
Output caps can be inside the feedback loop and be compensated for.
Also sometimes I see polar caps in the feedback loop (to ground) in order to get a lower DC output voltage which then can introduce distortion that also will not be corrected by the feedback loop.
Except... measurements show they don't. And lots of competent designers absolutely use electrolytics there- for example, in the case of the feedback loop cap to increase DC feedback and reduce offset, I have never ever seen a 100uF (or thereabouts) film cap used, and it would be a lousy choice because of its physical size and consequent noise pickup. In the solid state linear amps I have on hand (Adcom, Bryston), the input caps are all electrolytics.
In both cases, there's negligible signal voltage across the cap at audio frequencies.
solderdude
Grand Contributor
Yes, electrolytics used in the feedback path is actually quite common. It is indeed about physical size. Usage of bipolar would be preferred there but these are usually larger in size.
Large enough values, high enough voltage ratings and is usually enough to get the contribution to the distortion low enough and less than the active components introduce.
Add to that the added distortion (as with the Geshelli) is low enough not to be audible anyway but may be measurable.
I see it as a solved issue as long as the designers knows what caps not to use (easily seen in measurements during development) all is fine and it is not an audible issue.
Quite sure @pma knows all of this given the designs he does.
That said the posts sometimes are a bit quickbaty and might give the impression caps are always an issue where that actually depends on the correct implemetation and component choice. Also sure he knows this very well.
Large enough values, high enough voltage ratings and is usually enough to get the contribution to the distortion low enough and less than the active components introduce.
Add to that the added distortion (as with the Geshelli) is low enough not to be audible anyway but may be measurable.
I see it as a solved issue as long as the designers knows what caps not to use (easily seen in measurements during development) all is fine and it is not an audible issue.
Quite sure @pma knows all of this given the designs he does.
That said the posts sometimes are a bit quickbaty and might give the impression caps are always an issue where that actually depends on the correct implemetation and component choice. Also sure he knows this very well.
You are right; current would have to flow for migration to occur.
syn08
Senior Member
Here's some more data to show the importance of "properly chosen".
Two tantalum electrolytics were measured.
a) Kyocera AVX 10uF/25V "orange drop" type. A few years ago, I purchased from DigiKey a bunch of these for recapping an old piece of ATE that was emitting the magic smoke when powered on. They were used exclusively for power supply decoupling, I think I replaced about 200 of them.
b) Vishay Sprague 22uF/25V axial type. These are NOT wet tantalum, but regular solid tantalum caps. Purchased a few from DigiKey for a sample&hold application. These are x10 more expensive compared to the "orange drops".
I've added the distortion profiles to the previous results chart in #96. The "orange drop" cap distortions is in white, the axial Sprague cap is in yellow. Everything else was identical. Distortions are still mostly 3rd harmonic.
As expected, the "orange drop" tantalum cap is very bad. The bad part extends well into hundreds of Hz, being almost certainly audible.
But look at the Sprague axial tantalum cap. It is actually slightly better than the electrolytic measured in the previous post, although is only 22uF/25V. But since we are clearly in the measurement uncertainty domain I would not declare it "better" but "equivalent"..
Lesson learned:
1. Tantalum caps are NOT intrinsically bad. It all depends on the internal structure/construction and intended purpose. If one lame designer chose these orange drops tantalum caps for coupling, then he is paying the price of ignorance. They do exactly what they are intended for (decoupling), period. They are not doing what the lame designer hopes, period.
2. “properly chosen” electrolytic coupling capacitor is NOT a myth, but a biting reality.
Dedicated to @pma, although I can't hope it will change his beliefs a iota. So be it, reality and facts are more stubborn than humans.
P.S. After a few iterations, I finally found an electrolytic that has significant and reliable measurable THD distortions (some 10dB over the loop @20Hz). It is a very small size (4mm dia 5mm height) NOS 22uF/16V in SMD, of unknown origin (it looks like United Chemi-con though). I've added the distortions to the above graph (in green). Still -114db THD at 20Hz, that's nothing scary about. Except for scare mongers, of course
. Anyway, it is clear that bad electrolytics do exist, what a surprise .The little dip in the THD at LF is right at the current mains frequency (62Hz). I don't know the exact cause, but I suppose it's an interaction with the spectrum bins distribution (128k FFT points).
I'm done here.
Last edited:
- Thread Starter
- #107
For those who might be interested, I have created a web page with summary of all my measurements included in this thread, with some conclusion:
Hi all. Attached is an interesting article by Zak Kaye of Texas Instruments.
One can also download from https://www.ti.com/lit/pdf/slyt796
One can also download from https://www.ti.com/lit/pdf/slyt796
Uh huh.
Please elaborate. I don't remember testing impulse response and recovery when evaluating caps, so I must have missed out on something important.
DonR
Major Contributor
Something beyond its reactance?
Philbo King
Major Contributor
- Joined
- May 30, 2022
- Messages
- 1,037
- Likes
- 1,521
Re: reversed tantalums
I designed a front end PCB for an autopilot mode panel in the 90s. Since it had more than 1000 ICs, the project manager insisted that the factory build the prototype rather than pay me to build it by hand. The factory loaded the reel of tantalum decoupling caps backwards in the pick & place machine. So when I got the new board and powered it up every tantalum cap blew up. The board looked like a moonscape, little holes everywhere. The holes went down 5 to 7 layers (it was a 20 layer board). Fun times in the lab!
I designed a front end PCB for an autopilot mode panel in the 90s. Since it had more than 1000 ICs, the project manager insisted that the factory build the prototype rather than pay me to build it by hand. The factory loaded the reel of tantalum decoupling caps backwards in the pick & place machine. So when I got the new board and powered it up every tantalum cap blew up. The board looked like a moonscape, little holes everywhere. The holes went down 5 to 7 layers (it was a 20 layer board). Fun times in the lab!
First, those terms have to be specifically defined. Which the poster doesn't bother to do. Since caps happily pass VHF, I can't imagine any definition of "impulse response" that makes sense. Recovery? No idea what he means, but if it's related to dissipation factor, that makes even less sense.
I think someone read something somewhere on the internet and adopted the slogans.
theREALdotnet
Major Contributor
- Joined
- Mar 11, 2022
- Messages
- 1,748
- Likes
- 3,126
I remember them fondly for their entertainment value.
Hi Ixnay.
It is very clear you have no idea what distortion really is, so I will help you.
Distortion is the deviation from an original state, whether that be a piece
of steel or an audio signal. It matters not how the distortion was generated.
In my 40 years employment as an electronics engineer I have never heard
the term 'Impulse' being applied to capacitors, or any other electronic device
I can think of. Maybe you had Star Trek weighing heavily on your mind when
you wrote the reply. Presumably your next type of distortion will be warp drive.
Perhaps you as a low level amateur are thinking of rise/fall times, with the
formula of converting rise time to bandwidth being-:
Bandwidth (Hz) = 0.35/Rise Time.
Recovery is also a term I have never heard used in the context of distortion.
I would guess you are referring to the discharge of a capacitor time constant
which would reduce bandwidth. Recovery is something you will need after
reading my scathing reply.
The article I posted was by Zak Kaye of Texas Instruments. Which I case you do
not know make, in my opinion (as an amateur of course), the finest operational
amplifiers.
Choose those you wish to denigrate with the term amateur more carefully in
the future.
Keith
Hi Keith, Impulse Capacitors are really available for very high current load in some circuits (I can't see really an audio application. May be for PS) but they have their market.
See here some from a no "snake oil "company. https://www.wima.de/en/our-product-range/pulse-capacitors/.
Best
Hi BR52.
Wow. Never heard of those before.
Thank you for the information.
Hope not knowing of Impulse capacitors
does not make me look like a dick head.
I was most insulted by Ixnay referring to
me as an amateur.
Thank you very much,
Keith
Wow. Never heard of those before.
Thank you for the information.
Hope not knowing of Impulse capacitors
does not make me look like a dick head.
I was most insulted by Ixnay referring to
me as an amateur.
Thank you very much,
Keith
PlasticDoc
Active Member
Being generic, I don't know if the following is of interest, but since this is a "Capacitor distortion" thread, I'll post it.
Capacitor Performance:
Source: pag. 129, Subjective and objective evaluation of distortion in analogue electronics: capacitors and operational amplifiers, Gaskell, 2016, McGill University
(ESR=Equivalent Series Resistance and Inductance, DA=Dielectric Absorption, Leakage Current)
Capacitor Performance:
Source: pag. 129, Subjective and objective evaluation of distortion in analogue electronics: capacitors and operational amplifiers, Gaskell, 2016, McGill University
(ESR=Equivalent Series Resistance and Inductance, DA=Dielectric Absorption, Leakage Current)
Speedskater
Major Contributor
One of the skills of a good circuit design engineer, is knowing which capacitor technology to use at each point in the circuit.
Similar threads
