comining rail to rail and rail to ground decoupling?

[jazz]

Hi,

I've tried a search but didn't come up with a direct answer hence this post.

I'm wondering what would happen if you combine two ways of 'filtering' in a psu. By that i mean combining rail to ground decoupling with rail to rail decoupling in symmetrical supplies.

Let's say we have a + and - 24V supply for an F5 amplifier and we use a crc type filter. The supply is layed out using 22mf - 1R - 22mf for the rail to ground decoupling. What happens if we add another 22mf capacitor both to the first and second stage of the crc filter(s) and connect it rail to rail?

My questions are:
- Am I stupid to have this idea?
- If I am stupid, why?
- If I am not stupid, what do I have to think about?

Regards,
Joris

[indianajo]

Bad nonlinear things happen around the 0 v point on both BJT and FET transistors. Most amplifiers have those bad things happening at the "ground' point, not the rails. That is the gate and bases of the transistors are usually referred to "ground" not the rails, especially in class AB or B. (the cheap ways to operate). So if you buy 4 capacitors instead of two, and hook up as you say, you have improved predictibility of the transistor performance out at the rail, where it is pretty linear already. But you haven't done a thing for performance near "ground" where expensive circuitry is installed to handle the non-linear effects that happen there in class B, AB etc etc. So you have wasted your money. Better to put the additional money regulating the "ground" where bad things happen, or instead operate class A which makes "ground" irrelevant but wastes a lot of electricity and causes heat sinks to be expensive.

[godfrey]

Hi Joris

I suppose rail-to-rail capacitors would provide some improvement, but I would expect more improvement by adding extra capacitance from rail to ground, especially after the "R" in the CRC.

Separate smoothing for each channel would be good too, if you're not already doing that.

Cheers - Godfrey