On Jun 30, 2024, at 11:03, sergio_logic via groups.io <sergiu757@...> wrote:

?

I am trying to get away from Electrolytics

What is your motivation? If it's reliability, solid polymer aluminum electrolytic capacitors are a good option, e.g. something like . In short, they won't dry up like regular Al electrolytic ones. Tantalum is also an option, but you should apply 50% voltage derating (e.g. get them rated for 20V if they have 10V across them), and there are ethical regarding tantalum sourcing.

On Jun 30, 2024, at 14:09, Henry Spragens via groups.io <henryspragens@...> wrote:

?What matters with coupling capacitor distortion is the AC signal drop between terminals. DC across the cap doesn't distort the audio. It may change the capacitance a little, but the resting DC doesn't generate distortion. What's the second harmonic, or the third of DC? Zero Hz, right?

If the cap is used for DC blocking as most coupling caps are, it should be large enough that both terminals go up and down together, and there is no AC signal across the cap. The signal is passed undistorted. Incoming signals below the passband DO create a voltage change across the cap, and suffer distortion, but they are being deliberately attenuated, and attenuation is certainly distortion.

What may be a concern is signals below the passband modulating signals within the passband. This sort of thing happens in low and high pass filters, equalizers, and other tone shaping circuits. In these applications, there is an AC voltage between terminals of the cap, and the potential for varying capacitance. The measured capacitance change is a small percentage between zero volts and full rated voltage. One thing you will notice in every article where someone is measuring capacitor distortion in audio amplifiers - the biggest item the author talks about is his test rig, and how hard it is to get a measuring setup clean enough to detect the distortion.

OK, so let's spitball Jules' MLCC output caps. Given they are 22uF 35V rated. They see 22VDC across them in operation. Fine, the DC might have reduced the capacitance from 22uF to 21.9uF, but it isn't changing. No distortion. The mic output is maybe 100mV with LOUD audio, so this is causing the cap to swing from 22V to 22.1V, or under 0.5%. Let's look at worst case, deep bass. Let's see how 40Hz fares. We get 1/2 power, 3dB down, or 70.7mV out at 9Hz. AC drop across the cap is 29.3mV. According to my calculator that makes the load at the preamp 800 ohms. That seems low to me, but let's stick with Jules' numbers. At 40Hz we're dealing with 2.5 mV AC drop and capacitance change between 22V and 22.0025V on the cap's C/V curve is negligible. The delta in capacitance is from 21.9uF to 21.18999...9 uF, which is gonna change the LF rolloff by ... aahhhh, never mind ... that's way off in the weeds. With a higher preamp input impedance, bass rolloff and distortion will be much less. Jules, did you mean 0.9Hz, I find more typical?

Microphonics can be a problem, but not the way you're thinking. High-K dielectric caps are tiny, inefficient ceramic transducers; think of old crystal mics and earphones from the 1950s, or ceramic transducers generally. Flexing the PCB will generate a voltage, and changing the voltage across the cap will cause it to bend. In DC-DC converter applications running at low frequencies, the converter may whistle audibly. Worse, the solder joints may crack in time. But microphones are by definition small signal devices. It's good practice to check a mic after building for handling noises or resonances etc. by disconnecting the capsule and tapping on the body, the PCB, etc. Usually such effects are completely inaudible with capsule connected, and aren't worth worrying about, but sometimes it helps while debugging a new build. You may find a cold solder joint or cracked PC trace. Or maybe that priceless NOS silver and germanium doohickey you built around is kinda crackly when not masked by room noise.

I'm not claiming that all caps sound the same, or that differences don't exist. I'm just saying that the "horrible distortion" claimed by some audio pundits just can't be measured. Expectations from building and listening to amps, speakers, etc. often carry over where they don't apply. It's possible to pick the wrong cap, too, and if you've been building stuff long enough, you've probably done so.

Yeah, Jules, I've built a couple of mics with MLCC output caps, and they sound fine. I just never told anyone because ... you know, ...

On Jun 30, 2024, at 19:12, Goran Finnberg via groups.io <mastering@...> wrote:

?

Jules Ryckebusch:

>I can’t measure a difference and of course with the OPA1642?

>I’m already 100dB in the weeds for THD to begin with.

Jules,

How are you measuring this ?

In my experience unless the input stage differential JFET pair is fully cascoded with very little?

extra nonlinear input capacitance then every opamp I have measured using a capacitor in series?

with the + input of the opamp will show around -70 to -50 dB of distortion due to the nonlinear?

input capacitance.

The input capacitance of the OPA1642 is 8 pF

Question 1:

Did you actually measure the distortion of the OPA1642 using a series capacitor having the actual?

capacitance of a real condenser capsule ?

Question 2:

Using an electret capsule instead of a true capacitor capsule will provide a much lower source impedance?

than a true condenser capsule so the opamp distortion when fed from an electret capsule will be lower due?

to the much lower source impedance of that kind of source.

But in that case the actual distortion will be due to the scrappy single N channel JFET that buffers the electret?

capacitor and the distortion then is much higher than any kind of reasonable good opamp so the opamp distortion?

means nothing at all in that case.

The OPA1642 uses, fig 37, a source resistor of 10 kohm to measure distortion below 0,001 % @ 5V rms in a unity?

gain configuration.

Thats impressive, but I wish to see the distortion using a 22 - 33 pF capacitor so we can see what that means to?

the actual distortion using a true condenser capacitor.

Have you done this Jules ?

Thanks !

-----------

Best regards,

Goran Finnberg
The Mastering Room AB
Goteborg
Sweden

E-mail: mastering@...

Learn from the mistakes of others, you can never live long enough to
make them all yourself.??? -?? John Luther

(\__/)
(='.'=)
(")_(") Pyret, Ranglet, Aron, VovVov, Nero, Smurfen & Pussin:RIP

[Edited Message Follows]
[Reason: Oops! Corrected a typo "21.18999" to "21.8999".]

What matters with coupling capacitor distortion is the AC signal drop between terminals. DC across the cap doesn't distort the audio. It may change the capacitance a little, but the resting DC doesn't generate distortion. What's the second harmonic, or the third of DC? Zero Hz, right?

If the cap is used for DC blocking as most coupling caps are, it should be large enough that both terminals go up and down together, and there is no AC signal across the cap. The signal is passed undistorted. Incoming signals below the passband DO create a voltage change across the cap, and suffer distortion, but they are being deliberately attenuated, and attenuation is certainly distortion.

What may be a concern is signals below the passband modulating signals within the passband. This sort of thing happens in low and high pass filters, equalizers, and other tone shaping circuits. In these applications, there is an AC voltage between terminals of the cap, and the potential for varying capacitance. The measured capacitance change is a small percentage between zero volts and full rated voltage. One thing you will notice in every article where someone is measuring capacitor distortion in audio amplifiers - the biggest item the author talks about is his test rig, and how hard it is to get a measuring setup clean enough to detect the distortion.

OK, so let's spitball Jules' MLCC output caps. Given they are 22uF 35V rated. They see 22VDC across them in operation. Fine, the DC might have reduced the capacitance from 22uF to 21.9uF, but it isn't changing. No distortion. The mic output is maybe 100mV with LOUD audio, so this is causing the cap to swing from 22V to 22.1V, or under 0.5%. Let's look at worst case, deep bass. Let's see how 40Hz fares. We get 1/2 power, 3dB down, or 70.7mV out at 9Hz. AC drop across the cap is 29.3mV. According to my calculator that makes the load at the preamp 800 ohms. That seems low to me, but let's stick with Jules' numbers. At 40Hz we're dealing with 2.5 mV AC drop and capacitance change between 22V and 22.0025V on the cap's C/V curve is negligible. The delta in capacitance is from 21.9uF to 21.8999...9 uF, which is gonna change the LF rolloff by ... aahhhh, never mind ... that's way off in the weeds. With a higher preamp input impedance, bass rolloff and distortion will be much less. Jules, did you mean 0.9Hz, I find more typical?

Microphonics can be a problem, but not the way you're thinking. High-K dielectric caps are tiny, inefficient ceramic transducers; think of old crystal mics and earphones from the 1950s, or ceramic transducers generally. Flexing the PCB will generate a voltage, and changing the voltage across the cap will cause it to bend. In DC-DC converter applications running at low frequencies, the converter may whistle audibly. Worse, the solder joints may crack in time. But microphones are by definition small signal devices. It's good practice to check a mic after building for handling noises or resonances etc. by disconnecting the capsule and tapping on the body, the PCB, etc. Usually such effects are completely inaudible with capsule connected, and aren't worth worrying about, but sometimes it helps while debugging a new build. You may find a cold solder joint or cracked PC trace. Or maybe that priceless NOS silver and germanium doohickey you built around is kinda crackly when not masked by room noise.

I'm not claiming that all caps sound the same, or that differences don't exist. I'm just saying that the "horrible distortion" claimed by some audio pundits just can't be measured. Expectations from building and listening to amps, speakers, etc. often carry over where they don't apply. It's possible to pick the wrong cap, too, and if you've been building stuff long enough, you've probably done so.

Yeah, Jules, I've built a couple of mics with MLCC output caps, and they sound fine. I just never told anyone because ... you know, ...

On Jul 1, 2024, at 15:08, j.postma8 via groups.io <j.postma8@...> wrote:

?Henry, I basically agree with your first three paragraphs. But where you go wrong is with your assumptions and calculations. The capacitance drop is not just 0.1uF at 22V, but it drops to just ~3uF! This results in a clearly audible and measurable decrease in sub-50Hz frequencies. Just tested and verified this with a Murata 22uF 25V X5R size 1210, which decreases to ~4uF. If I would want to keep an effective ~50uF capacitance with MLCCs, I would have needed almost $10 worth of huge 2220 size X7R caps, whilst an 47uF Panasonic EB series elcap would cost me less than $30. For me, X7R, and X5R even more so, remain a no-go.

THD @ 20Hz measured 0.2% at 400mV AC input with the Murata MLCC. Inaudible at this frequency, AFAIK.

As usual, YMMV.

Jan