Polypropylene vs. Silver-mica for Cdom.

[Edmond Stuart]

This appears to be a language difficulty problem, rather than anything else.

Apparently.

I used the word 'multiply' but that meant 'scaling' or something like that, rather than making the percentage of the distortion in the cap higher. The main thing is that a small cap can dominate the transfer function of an entire amplifier.

Agreed.

Therefore, if it is non-linear, then the amp can be 'non-linear' too.

Agreed.

 

[prairiemystic]

who makes teflon caps

I see Cdom as about 5-20pF at 100V and haven't seen any ptfe offerings. Cornell Dubilier has 7-9pf max. on their MIN02 parts, for 300V RF apps. Any others?

This is a good Bob Pease article comparing dielectric absorption (Fig. 7- teflon, mylar, PS, PP, ceramic etc.) Understand Capacitor Soakage to Optimize Analog Systems

 

[Bob Cordell]

Thanks to everyone who puts links up to read. I have found them very informative.
Getting back to the original reason for this thread, Cap compensation, I remember using ceramic caps 40 years ago, with impunity for caps up to 500pf or so, for different things, especially comp caps for circuits. Years later, I did some researching as to whether these caps, commonly found at the local electronics supply store were NPO or a slightly worse ceramic. You see, they were available in about the same package, with just a few cents of difference per part. Well, that is what I try to warn against. Even Mark Levinson used these ceramics in our early modules, before I was shown the potential problem at Tektronix, back in 1974. Then, we switched to a film cap of some type.
It seems to me that many comp caps rely on Miller multiplication to work or at least driven from an extremely high impedance current source. Now, IF this comp cap changes its characteristics, through microphonics, voltage change, or DA, it is probably amplified 1000's of times compared to measuring the cap and its characteristics, alone driven by a conventional generator, etc. This is why I use polystyrene comp caps when I can.

Hi John,

I don't think it is quite right to say that the effect of a change in Miller capacitor characteristics is multiplied by thousands of times (as by the Miller multiplication factor).

In the ideal case of a simple amplifier, the gain at frequencies above the open-loop pole is approximately equal to the product of the input stage transconductance and the impedance of Cdom. The range of frequencies over which this applies is often from frequencies well below 1 kHz to above the gain crossover frequency (for those amplifiers that have high feedback factors and low open-loop bandwidth).

If the impedance of the capacitor changes by 1% for whatever reason, the magnitude of the open-loop gain will change by about 1%.

Similarly, if a microphonic effect causes a 1 mV voltage change across the capacitor, the voltage at the output of the VAS will change by 1 mV. This is simply a result of the tight, wideband shunt feedback around the VAS formed by the Miller compensation loop. Ultimately, of course, the global negative feedback will reduce this step offset to a very small value, with an exponential decay on the order of the time constant corresponding to the closed-loop bandwidth of the amplifier.

In no way is what I am saying an endosement of using crappy ceramic capacitors for Cdom. X5R, X7R, Z5U and most all other dielectrics for ceramics, other than COG, are completely unsuitable for use as Cdom. Obtaining really good film capacitors in small values less than 30 pF appears difficult, so sometimes other choices are necessary. I believe that silvered Mica is the safest choice.

Cheers,
Bob

 

[Conrad Hoffman]

I haven't done it for audio, but for other instrumentation applications I've just twisted a couple wires together to get small capacitance values. Now, if those wires happened to be Teflon insulated (or PP or your choice), the capacitor will be pretty decent.

 

[wrenchone]

I've used "gimmick" capacitors twisted from teflon wire where I needed small capacitance values. One frill that may be necessary is some heat shrink so that the wires stay twisted as tightly as when originally done. I start with a twist longer than necessary, and chop it to length as needed. It's not "mass manufacturable", but this is diy, after all. Does anyone make a trimmer cap with teflon dielectric? I suspect some of the small ceramic trimmers are made with NPO dielectric, but it would be nice to have some confirmation.

 

[ostripper]

Hi John,

I don't think it is quite right to say that the effect of a change in Miller capacitor characteristics is multiplied by thousands of times (as by the Miller multiplication factor).

In the ideal case of a simple amplifier, the gain at frequencies above the open-loop pole is approximately equal to the product of the input stage transconductance and the impedance of Cdom. The range of frequencies over which this applies is often from frequencies well below 1 kHz to above the gain crossover frequency (for those amplifiers that have high feedback factors and low open-loop bandwidth).

If the impedance of the capacitor changes by 1% for whatever reason, the magnitude of the open-loop gain will change by about 1%.

Similarly, if a microphonic effect causes a 1 mV voltage change across the capacitor, the voltage at the output of the VAS will change by 1 mV. This is simply a result of the tight, wideband shunt feedback around the VAS formed by the Miller compensation loop. Ultimately, of course, the global negative feedback will reduce this step offset to a very small value, with an exponential decay on the order of the time constant corresponding to the closed-loop bandwidth of the amplifier.

In no way is what I am saying an endosement of using crappy ceramic capacitors for Cdom. X5R, X7R, Z5U and most all other dielectrics for ceramics, other than COG, are completely unsuitable for use as Cdom. Obtaining really good film capacitors in small values less than 30 pF appears difficult, so sometimes other choices are necessary. I believe that silvered Mica is the safest choice.

Cheers,
Bob

WOW !, an excellent explanation that was easily digested. I was so concerned by JC's statement that I took the amp off the speaker 🙄 (burning amp by micro-phonic miller feedback .. yikes 😱 ).
OS

 

[davidsrsb]

It is very hard to stop any board movement from putting stresses on the leads of the semiconductor device that the Cdom cap is wrapped around. This is likely to have a bigger effect than a NPO ceramic.

 

[PA0SU]

no sound from a capacitor!

In general: if the capacitor produces sound if a 'large' audio-signal has been put on it (in combination with DC or not), the capacitor is bad for audio applications. Such capacitors are microphonic!!
Ceramic caps sometimes act as tweeters.....

To me, styroflex and mica sounds best in filter applications because they do not store charge. Anyway, always choose the value of the caps large enough (See Self: Small signal audio design, Figure 2.17 on page 60) if used for coupling amps or decoupling power supplies.

 

[balaboo]

I agree completely with you - this has been my own experience too.

 

[michaelkiwanuka]

In theory, by the way, the Miller compensated stage without a resistor in series with the cap can actually introduce a right-half-plane zero in the open-loop transfer function, which is a bad thing.

The right half plane zero is only an issue if you use a MOSFET second stage.

 

[Bob Cordell]

The right half plane zero is only an issue if you use a MOSFET second stage.

Hi Mike,

You're probably referring to the fact that the gm of the MOSFET VAS will typically be lower than that of a BJT, causing the right-half-plane zero to occur at a lower frequency for the MOSFET design. It is true that the RHP zero is pretty far out for a BJT VAS with high transconductance, but a BJT VAS with 10:1 emitter degeneration has about the same transconductance as an undegenerated MOSFET VAS if they are both operating at the same current.

In any case, I agree that the RHP zero is often not an issue, especially given that some designers will put a resistor in series with Cdom that is big enough to bring the zero into the left half plane to gain some phase margin.

Cheers,
Bob