Give me Some General Tips on Caps and Resistors

[The New Steve H]

I guess I should be thinking about the volume pots and the resistors in the preamp, then.

I was too cheap to get plastic pots.

It doesn't matter; the 5f6a sounds great. Just fun to fiddle with it.

[thejaf]

In my personal limited experience, layout, lead dress, and solder technique takes care of 90% of noise issues. I doubt there'd be any audible difference is noise using metal film vs carbon film resistors if layout, lead dress, and soldering aren't at or above par.

[Structo]

Can I assume, then, that changing the 4.7K to a big metal film might reduce hiss? I would think it would have an effect on the output plates.

The benefits of certain types of resistors in the resistor/hiss equation manifests itself more greatly where resistors have a direct effect on the AC in the signal path, i.e.: plate resistors, cathode resistors, grid leak resistors. Changing to different type of resistors in the power supply may alleviate a teency weency portion of hiss, but in the overall scheme of things won't reduce hiss that much, because the supply resistors are decoupled from the signal path by the filter caps.

But having said that, there are extraneous noises/signals in the power supply (that one tries to eliminate with RC* decoupling, and electrolytic filter caps are far from perfect at decoupling AC), which can benefit from HF decoupling with smaller high voltage film caps (e.g. 10-100nF) in parallel with the filter caps. Usually one (high voltage) 10nF or 100nF film cap in parallel with the reservoir cap is sufficient to get rid of these more unstable high frequency elements in a B+ supply.

* the particular combination of the values of 'R' and 'C' produce different points of dB reduction at particular frequencies, hence when you decrease a supply resistor, you really (technically speaking) should increase the filter cap at that node by a corresponding inverse ratio, in order to keep the freq roll-off of extraneous noises in the power supply at the same point.

I have read of doing this quite a few times, adding a poly cap on an electrolytic.
Say on a Dumble ODS type amp, 100w I have two large caps in series for a total of 160uf for the power tube plates, a choke (screens come after the choke), then a 22uf for each preamp node. (4 x 22uf)

Which node or caps would benefit the most from the parallel poly cap treatment.
My amp sometimes has this weird high frequency thing going on I can't seem to track down.

[Cliff Schecht]

Tom the idea behind the smaller capacitances is that the effectiveness of the large electrolytic bypass becomes less and less as you increase in frequency (i.e. the impedance decreases with frequency). The smaller bypass caps can be chosen to overlap where the larger caps fall off. What I expect to see is for something at say 100uF would be decoupled with a film cap (or other type that is good at higher frequencies) of 0.1uF or so. You can also strap a 100pF cap across the 0.1uF cap to make the equivalent network effective over a large range of frequencies. This is more common in super low noise INA's or in applications where you are mixing analog and digital signals and need the two types to not crosstalk. It also knocks down on broadband noise and increases power supply rejection.

Here is a neat little graph that shows different cap values vs. impedance. Notice that all of the caps eventually "fall off" and rise again - this is the point where the capacitors start looking inductive and are completely useless as caps.
[img:378:297]http://www.analog.com/library/analogDia ... 909_01.gif[/img]

[Structo]

Thanks Cliff that is helpful.
The stuff I have forgotten over the years......

So what node of the power supply is most beneficial to bypass with a poly?
The first preamp tube?

[Cliff Schecht]

Yeah anything that is sensitive to noise. The PI and powerstage are usually push pull circuits and have enough built in PSR to not need extraneous filtering usually.

[Aurora]

Also remember the basic noise model for all amplifiers-the noise product of each amplifer stage gets amplified by all the succesive stages.
Thus the input stage(s) will always be the largest noise producer....

[martin manning]

Interesting plot Cliff, but it looks like this is only important for RF or at least ultrasonic noise though, right? For audible noise we are only interested in frequencies up to 0.02 mHz at most, which would appear to be clear of the inductive break point even for a very large filter cap value. What kind(s) of caps does this chart represent?