Page 6 of 6

Re: Dielectric Absorption

Posted: Wed Mar 13, 2019 5:56 am
by terkio
You scared me with dielectric absorption.
I did encounter DA, years ago, while debugging a water pump electric moteur. Checking the motor starting cap ( 20uF 400V AC ). I was amazed to read DC after I had shorted it to make sure it was discharged, and see that DC coming back.
Recently, interested in a preamplifier using a servo, I met again dielectric absorption, reading about it, here, then thinking of the servo integrator capacitor.
After extensive brain warping and research, I finally concluded there is nothing to fear about dielectric absorption.
No, it cannot mess the cap of the integrator. I can add DC at the cap, whatever value, it works just the same.
And this is true of any system I can think of, using caps.
The reason is that in a system, whatever the absorbed voltage there is inside the cap, the cap will charge in a way that exactly cancels out that voltage.
Think of the phantom power isolation caps. With phantom power off, you will measure 0V ( or some voltage if there is a voltage biais at the input of the preamplifier). Let's say, you measure 0V, but who knows the voltage from dielectric absorption, may be 1 Volt canceled by -1V of charge. Now turn phantom power on, the cap will charge, now you will measure 48V at the caps, may be it is 1V of dielectric absorption and 47V of charge.
So I forget DA and I am back at the real problem of capacitors when used in integrators or sample and holds, namely the Isolation Resistance, this is the true problem that makes error.

Re: Dielectric Absorption

Posted: Wed Mar 13, 2019 6:12 am
by mediatechnology

Re: Dielectric Absorption

Posted: Wed Mar 13, 2019 6:45 am
by terkio
In page 7 and 8
Assuming phantom power has once been turned on, dielectric absorption
impacts the circuit’s behavior when phantom power is off for a very long time. In one experi-
ment, after being charged to 48V for a few minutes, we turned off all power for over one
hour. Coming back to the circuit on the bench, with power still off, we measured voltage
across C1 and C2
on the order of 10mV. (One measured about 7mV, the other about 14mV.)
We then shorted out both capacitors for a few seconds. After removing the short, we saw
the voltage across both caps rise over many seconds to nearly the same value as before
shorting out the capacitor.

I say, this is a wrong experiment, giving a wrong conclusion.
If they had turned the power back on, then measured, they would have seen 0mV.
I was not there when they conducted their experiment.

The purpose of their paper was about clicking noise, obviously it has to do with capacitor charges, but nowhere but their wrong experiment, they show how dielectric absorption comes in the clicking. Of course this must not be confused with the clicks and bangs from turning equipments power on and off.

EDIT: The right experiment is:
Measure the caps with the equipement turned on.
Turn phantom power on for minutes.
Turn off phantom power, then measure the caps.

Re: Dielectric Absorption

Posted: Wed Mar 13, 2019 8:01 am
by mediatechnology
I think THAT's point was to remove Ibias from the experiment.

What is relevant is that it clicks regardless of how you run the experiment.

I ran a slightly different experiment where I turned off phantom and kept the circuit powered so that I could check it for clicks over time.
It clicks long after phantom has been turned off due to "soakage."
Electrolytic capacitors do this - film don't.

(Even with nice film caps the Ibias * Rbias voltage term will cause clicks with gain applied.)

(I would never do a servo with an electrolytic capacitor if I didn't absolutely have to.)

I co-wrote DN-140.
I spent a great deal of time explaining to my co-workers that this type of circuit will click.
They, like yourself, didn't believe me until they tried it themselves.

(The same frustration applied to the second Phantom Menace AES paper where I had to prove to the same very reluctant audience that differential fault currents could be in the >>1A range. ... eturns.pdf)

So I'm tired of arguing theory.
Building something and measuring it is where the rubber hits the road.

In both cases I built something, found its fault mechanism, measured it and then forced them to sim what I'd found.
Only when sim'd were my assertions believed by the theoreticians.

When Rosalfonso Bortoni at THAT took the bull by the horns and sim'd the phantom fault I identified he quickly became a believer.
Rosalfonso, some time later, also recreated at THAT HQ my clicky line input.
If it hadn't been for Rosalfonso duplicating my efforts I'd still be arguing theory today.