Capacitor distortion

[fpitas]

Honestly if these subjectivist internet mavens ever chose a reasonably priced cap, especially from (say) an industrial offering like Panasonic has, I'd faint.

 

[Bob from Florida]

Most of the polarized capacitors that you find in audio equipment are for power supply filtering (not in the signal path).

The final capacitor in the power supply - think B+ rails - is absolutely in the audio signal path.

 

[fpitas]

The final capacitor in the power supply - think B+ rails - is absolutely in the audio signal path.

Sure. But even if it's an electrolytic it's heavily polarized by the applied voltage, so negligible distortion. Also, if there's feedback in the amplifier that would tend to reduce any distortion from that cap.

 

[Bob from Florida]

Sure. But even if it's an electrolytic it's heavily polarized by the applied voltage, so negligible distortion.

The point is that cap has an effect on the linearity of the amp due to esr at a given frequency. It pays to look at the manufacturer spec sheets as esr is not always decreasing with frequency. Then verify that by testing your samples.

 

[fpitas]

The point is that cap has an effect on the linearity of the amp due to esr at a given frequency. It pays to look at the manufacturer spec sheets as esr is not always decreasing with frequency. Then verify that by testing your samples.

That's true; that is all part of designing any amplifier.

 

[syn08]

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 )

I don't have an E1DA to compare, but my ES9822 + ES9038PRO analog loop doesn't show the same behavior as your E1DA+D10 loop.

Sampling is 192KHz, THD at 50Hz are about -125dB, while @20Hz are about -119dB. Level is -6dBFS (2.4Vrms) load is 1.4K, ADC coupling caps are 47uF 16V Nichicon MUSE ES (bipolar). Note: due to the ADC input stage inverting configuration, the caps are natively biased at the ADC common mode voltage (+2.25V). Distortions are almost entirely 3rd harmonic over the entire frequency range.

Spectrum for 50Hz -6dbFS (2.4Vrms) @44.1KHz sampling

While I certainly disagree with using $0.05 electrolyics at the E1DA input (in particular in the context of the outlandish E1DA performance numbers quoted by @IVX), I doubt they have the amount of distortions you are showing (-102dB @20Hz). I would rather investigate your DAC distortions, -102dB could be confirmed even with a modest sound card.

I'm still waiting a justification for your statement

I would never use and electrolyte as a coupling cap.

The world is full of junk capacitors, but your statement reminds me of John Curl sweeping generalization about ceramic capacitors in audio being bad.

 

[syn08]

The point is that cap has an effect on the linearity of the amp due to esr at a given frequency. It pays to look at the manufacturer spec sheets as esr is not always decreasing with frequency. Then verify that by testing your samples.

ESR itself does't create harmonic distortions, it's voltage coefficient may.

 

[Bob from Florida]

ESR itself does't create harmonic distortions, it's voltage coefficient may.

I did not mention distortion.

 

[syn08]

I did not mention distortion.

The point is that cap has an effect on the linearity of the amp due to esr at a given frequency.

Huh? Linearity (or non, thereof) is anything but distortions effect?

 

[Bob from Florida]

Huh? Linearity (or non, thereof) is anything but distortions effect?

Linearity as in frequency response. I was not speaking to harmonic distortion

 

[syn08]

Linearity as in frequency response. I was not speaking to harmonic distortion

Ok, but I am pretty confident that extra few ohms ESR has virtually zero effect on the frequency response (or "linearity", as you put it) for any decent input impedance.

Example:

1k (which is already very low) input impedance with 100uF coupling capacitor has a 3dB corner frequency of about 1.6Hz. Adding 4ohm ESR *) moves the 3dB corner with -0.00634Hz and the attenuation at 20Hz changes from -0.0275452dB to -0.02732688db, a whopping -0.000218dB

If this is not laughable, I do not know what else is

*) Values are for the worst 6.3V electrolytic cap available on DigiKey with tan delta = 0.32 at 120Hz and 20 degrees Celsius. This tan delta value leads to ESR=Xc * tan delta = 4.2ohm

 

[fpitas]

If we're talking about power supply ripple, I'd better explain something about practical amplifier design. Your capacitors in the supply will have quite measurable amounts of ripple, at the 120Hz rate and from the audio, mostly the bass since that not only takes all the power but is low frequency. A decent amplifier has extensive supply filtering for the input stages like the differential pair and voltage amps. That's done with zeners or capacitance multiplier transistor regulators, or sometimes a regulator IC. The point being, you don't depend on a pristine supply, because It ain't happening.

 

[Bob from Florida]

Ok, but I am pretty confident that extra few ohms ESR has virtually zero effect on the frequency response (or "linearity", as you put it) for any decent input impedance.

Example:

1k (which is already very low) input impedance with 100uF coupling capacitor has a 3dB corner frequency of about 1.6Hz. Adding 4ohm ESR *) moves the 3dB corner with -0.00634Hz and the attenuation at 20Hz changes from -0.0275452dB to -0.02732688db, a whopping -0.000218dB

If this is not laughable, I do not know what else is

*) Values are for the worst 6.3V electrolytic cap available on DigiKey with tan delta = 0.32 at 120Hz and 20 degrees Celsius. This tan delta value leads to ESR=Xc * tan delta = 4.2ohm

My original reply was stating that the final cap on the B+ rails for a power amp output stage is in the signal path. The AC signal path is through the output devices in and out of your speakers and back through the last cap in your B+ supply. I am not talking about input capacitors. Depending on your output power and speaker inefficiency the AC current through that cap can be high at higher listening levels. Depending on your power supply design there also can be high ripple current at 120 HZ off the FWB. Using a dedicated film cap rated for long life at high ripple current from the FWB then use a regulator or choke between that cap and the final power supply cap will eliminate high ripple current on the final cap. Poor choice of that final cap affects frequency response and reliability.

 

[syn08]

Ok, have it your way. Call frequency response "linearity", then shift the focus to tube amplifier power supply design. Flashnews, the filter cap ESR is a non issue for any competent designed power supply, except perhaps SMPS's (which are not about LF ripple).

If ESR is a problem, then it is about an amplifier with zero or very low PSRR, quite typical in the high end audio realm.

 

[Bob from Florida]

Ok, have it your way. Call frequency response "linearity", then shift the focus to tube amplifier power supply design. Flashnews, the filter cap ESR is a non issue for any competent designed power supply, except perhaps SMPS's (which are not about LF ripple).

If ESR is a problem, then it is about an amplifier with zero or very low PSRR, quite typical in the high end audio realm.

Where did you get the idea I was talking about tube amps? CLC filters can occur in any power supply. Schiit Tyr uses CLC. Do you not understand that all of the music signal goes through the last cap in the supply?

 

[syn08]

Forgot to post the THD vs. level loop results, @50Hz. For comparison purposes, same scale and units are used. 44.1KHz sampling rate

Lesson learned: your loop THD and THD+N vs. level @50Hz are about 6-15db worse compared to my ES98220+ES9038PRO loop. I would certainly suspect the D10 distortion performance, not the $0.05 E1DA input caps. Another good clue is the THD saturation, at high levels, in your setup. Meaning that the D10 DAC doesn't take very good the rather low E1DA input impedance.

 

[syn08]

Where did you get the idea I was talking about tube amps? CLC filters can occur in any power supply. Schiit Tyr uses CLC. Do you not understand that all of the music signal goes through the last cap in the supply?

Because you mentioned above B+. Last time I've checked, this is the common denomination for tube amplifier power supply voltage.

I understand perfectly, and I maintain the effect of the last cap in the supply is virtually zero (except in pathologically bad designs). I don't give a Schiit if any manufacturer decides to use CLC in the SS amp power supply, to me that's a waste of money, space and weight. In particular thinking of the currents a SS amp is drawing a full power.

Come up with numbers if you want to bring anything relevant to the table. Otherwise, this discussion is over.

P.S. From the Schiit Tyr web site: 55 lbs. 12 lb choke. And all the trouble for the results below:

Insanity now has a name.

 

[Bob from Florida]

Because you mentioned above B+. Last time I've checked, this is the common denomination for tube amplifier power supply voltage.

I understand perfectly, and I maintain the effect of the last cap in the supply is virtually zero (except in pathologically bad designs). I don't give a Schiit if any manufacturer decides to use CLC in the SS amp power supply, to me that's a waste of money, space and weight. In particular thinking of the currents a SS amp is drawing a full power.

Come up with numbers if you want to bring anything relevant to the table. Otherwise, this discussion is over.

P.S. From the Schiit web site: 55 lbs. 12 lb choke. Insanity now has a name.

Based on the number of assumptions you have made, I wonder if you actually have any respectful discussions with anyone.

 

[syn08]

Based on the number of assumptions you have made, I wonder if you actually have any respectful discussions with anyone.

New deflection. Because you mentioned it, respect must be earned. So far, you failed miserably.

 

[restorer-john]

Come on boys. Let's get back to the topic at hand.