Capacitor distortion

[pma]

Though I posted this as a reply to @syn08 in another thread (preamp ULTIM2), I think it deserves its own thread.

As an example, 2 electrolytic capacitors were measured:
1) Aluminium 10uF/100V
2) Tantalum 4.7uF/6.3V

they were put in the loop between Topping D10s and E1DA Cosmos ADC. Load impedance is 830 ohm.

Below are the results of distortion measurements at 50Hz. X-axis is in Vrms. Y-axis in dB relative to fundamental (50Hz) level.

Test fixture

Question: Does capacitor distortion exist?
Answer: Depends.

 

[fpitas]

That accords with what I've seen elsewhere. Put much voltage across a polar cap, especially in reverse, and you get non-linearity.

If you have a bipolar electrolytic around that might be an interesting test subject. Nelson Pass showed results testing Nichicon Muse bipolar caps that was quite good.

 

[pma]

That accords with what I've seen elsewhere. Put much voltage across a polar cap, especially in reverse, and you get non-linearity.

If you have a bipolar electrolytic around that might be an interesting test subject. Nelson Pass showed results testing Nichicon Muse bipolar caps that was quite good.

Yes, I know. However I wonder if NP made measurements with similar high resolution. And there is not "much" voltage across the cap. It should be understood that the AC voltage is mostly across the ADC input impedance. Even for the 4.7uF cap and 830 ohm load we have F(-3dB) frequency at 40.8 Hz (high pass). It is good to take a calculator and check the conditions. Input Vrms (plot below) is about 1.7V. And there is an impedance behind the cap, it is not shorted to ground.

 

[fpitas]

Yes, I know. However I wonder if NP made measurements with similar high resolution. And there is not "much" voltage across the cap. It should be understood that the AC voltage is mostly across the ADC input impedance. Even for the 4.7uF cap and 830 ohm load we have F(-3dB) frequency at 40.8 Hz (high pass). It is good to take a calculator and check the conditions. Input Vrms (plot below) is about 1.7V. And there is an impedance behind the cap, it is not shorted to ground.

View attachment 256696

Things might be different if a lot of voltage was across the cap at some frequencies, like in a crossover application.

 

[Bamboszek]

Can you compare much bigger capacitances such as 33uF or even 100uF?
Sufficiently sized electrolytic should have non-measurable distortion as line level coupling capacitor.

 

[solderdude]

A tantalum is a semiconductor capacitor an MUST have a DC voltage across it. It cannot be used with AC signals unless it is biased with a DC voltage higher than the AC.
An electrolytic capacitor also should have a DC on it. When this is not possible they should be used in anti-series which is equivalent to a bi-polar capacitor.
I prefer to use the term non-polar for capacitors without an electrolytic.

In this case (AC coupling) the problem is not the voltage across it (there should not be any) but the current flow (which there is) which is the problem.
Biassing the electrolytic lowers this distortion.
Would be fun to test perhaps using a battery in series with the load resistor for instance.

Tantalum are very poor audio coupling caps but work well as power supply rail caps. Also not very suited for low leakage current applications (timers).

Another option for tantalum is to use 2 in anti-series and bias the middle with an appropriate DC voltage but this better be very low noise

I often see electrolytics used in the feedback path of a power amp (to ground) which also is poor practice. Also sometimes see electrolytics as an input capacitor on amps without any bias voltage. Poor engineering.

 

[fpitas]

A tantalum is a semiconductor capacitor an MUST have a DC voltage across it. It cannot be used with AC signals unless it is biased with a DC voltage higher than the AC.
An electrolytic capacitor also should have a DC on it. When this is not possible they should be used in anti-series which is equivalent to a bi-polar capacitor.
I prefer to use the term non-polar for capacitors without an electrolytic.

Tantalum are very poor audio coupling caps but work well as power supply rail caps. Also not very suited for low leakage current applications (timers).

Another option for tantalum is to use 2 in anti-series and bias the middle with an appropriate DC voltage but this better be very low noise

I often see electrolytics used in the feedback path of a power amp (to ground) which also is poor practice. Also sometimes see electrolytics as an input capacitor on amps without any bias voltage. Poor engineering.

I agree with most of that, but we've had nothing but trouble with tantalums on supply rails. They sometimes just explode, despite being very down-rated in voltage. We stick to big ceramics, which are cheaper and have lower ESR.

 

[DVDdoug]

That accords with what I've seen elsewhere. Put much voltage across a polar cap, especially in reverse, and you get non-linearity.

Yeah, you don't normally use a polarized cap for unbiased AC audio. Normally they would only be used (in the signal path) to filter-out the DC component when you have DC and audio together so the voltage across them is never reversed.

Most of the polarized capacitors that you find in audio equipment are for power supply filtering (not in the signal path). But I have an older amplifier with a single-ended power supply so it has big electrolytics on the speaker outputs.

If you have a bipolar electrolytic around that might be an interesting test subject. Nelson Pass showed results testing Nichicon Muse bipolar caps that was quite good.

The only place I've seen non-polarized electrolytics is in a speaker crossover where you heed a high uF value and other types of caps get expensive. I had the impression it was sort-of a "hack" and I think high-end crossovers avoid electrolytics, but it's not something that I've worried about and my DIY speakers have them.

Electrolytics also age and lose some capacitance. In a DC-blocking application that's not a big deal because you can higher than necessary value to begin with. But with filters (crossovers) you want a more-precise value.


P.S
Reversed tantalums can explode like a firecracker! It's pretty spectacular and it get's everybody's attention when it happens on the production line! I've never seen anyone hurt but it could cause eye-damage if one of those hot little particles happens to hit you in the eye. (There's not enough energy in a line-level audio signal for an explosion.)

I don't think I've ever seen an electrolytic explode accidently but when I was in high school we used to explode them by hooking them up across the power line, when the teacher wasn't around. I have seen black & melted electrolytics on circuit boards, probably shorting-out from aging..

Oh... I did have some in a home built amplifier burn-up and release "toxic" smoke when I turned-on the amplifier after not using it for a couple of years. I think they were on the hairy-edge of their specs, maybe 16V capacitors used in a 16V power supply or something. (They weren't the main power supply capacitors.) I don't remember if there was being an explosion... I don't think so and I couldn't see what what was happening inside the cabinet... but I do remember that I had to get out of the house to breathe! (I wasn't worried about a fire since they were inside a metal cabinet and I had turned it off before the smoke filled-up the room.) The amplifier was OK after replacing the caps.

 

[fpitas]

Yeah, you don't normally use a polarized cap for unbiased AC audio. Normally they would only be used (in the signal path) to filter-out the DC component when you have DC and audio together so the voltage across them is never reversed.

Most of the polarized capacitors that you find in audio equipment are for power supply filtering (not in the signal path). But I have an older amplifier with a single-ended power supply so it has big electrolytics on the speaker outputs.

The only place I've seen non-polarized electrolytics is in a speaker crossover where you heed a high uF value and other types of caps get expensive. I had the impression it was sort-of a "hack" and I think high-end crossovers avoid electrolytics, but it's not something that I've worried about and my DIY speakers have them.

Electrolytics also age and lose some capacitance. In a DC-blocking application that's not a big deal because you can higher than necessary value to begin with. But with filters (crossovers) you want a more-precise value.

A surprising number of popular speakers used (and still use) electrolytics. It is a point of discussion in audiophile land, and I'm sure there's some effect. It may not be the sonic disaster some make it out to be. Speakers designers I know say they avoid electrolytics in midrange and tweeter filters. Of course, that's convenient since the values are dramatically smaller.

 

[DonH56]

A hot topic, nice to see the data. Ceramic caps have all sorts of issues used as coupling caps, fine for decoupling. Film caps exhibit lower distortion and hysteresis though quality of film matters for things like 1/f^N noise. Audibility is always a question.

There are so many papers published on this... A recent one on electrolytic caps is https://www.aes.org/tmpFiles/elib/20230112/20891.pdf (quick Google search). There are zillions, and of course the old Jung and Marsh paper (again from a quick search): https://www.reliablecapacitors.com/information/picking-capacitors.html

 

[solderdude]

I agree with most of that, but we've had nothing but trouble with tantalums on supply rails. They sometimes just explode, despite being very down-rated in voltage. We stick to big ceramics, which are cheaper and have lower ESR.

me too (blowing up spontaneously) but also have seen many over 30 years old still functioning.

 

[fpitas]

me too (blowing up spontaneously) but also have seen many over 30 years old still functioning.

Yeah. Still, it only takes a couple blowing up to get unhappy customers.

 

[somebodyelse]

See also https://linearaudio.nl/cyril-batemans-capacitor-sound-articles

 

[pma]

Please have a look at this set of 3 plots, THD distortion vs. frequency, measured at 1.7Vrms ac input voltage, into 830 ohm load in all cases. We have 3 plots:

1) Topping D10s ==> E1DA Cosmos ADC loopback. Violet plot. I am quite sure that the rise of distortion below 70Hz is a result of distortion in input electrolytic capacitors of E1DA (thoughts for @IVX and @syn08 )

2) 10uF/100V aluminium electrolytic capacitor in the loop as (1). Orange plot, rise of distortion below 100Hz is from this capacitor.

3) 4.7uF/6.3V tantalum electrolytic capacitor in the loop as (1). Green plot, rise of distortion below 300Hz is from this capacitor.

Electrolytic aluminium/tantalum capacitors do increase distortion of the circuit if:
- ac voltage across them is approaching to their rated DC voltage (high pass filters). In the high pass filter, distortion rise may start as soon as below 10 x F(-3dB) filter corner frequency

There are 3 cures:
- use the cap with highest possible rated DC voltage
- use the cap with highest possible capacitance, i.e. do no use el. capacitors in filters, minimize ac voltage across them
- use bipolar connection of el. capacitors

 

[Matias]

Since we are at it, have fun!

Humble Homemade Hifi - Cap Test

Humble Homemade Hifi - Very high quality loudspeaker kits, components, upgrades, modifications and custom solutions.

www.humblehomemadehifi.com

 

[pma]

Since we are at it, have fun!

Humble Homemade Hifi - Cap Test

Humble Homemade Hifi - Very high quality loudspeaker kits, components, upgrades, modifications and custom solutions.

www.humblehomemadehifi.com

Hmmm ... no measurements, right?

 

[BR52]

Hi Pavel, is it possible to do a comparison between an MKP and MKT capacitor with your test setup would be interesting?
Bernd

 

[fpitas]

Hmmm ... no measurements, right?

Measurements would just detract from the spiritual purity of his subjective guesses. I mean, anyone can measure a cap!

 

[BR52]

Hmmm ... no measurements, right?

More precious metal> better.

 

[fpitas]

More precious metal> better.

And more $$$$$$$$$