Sizing PSU Decoupling R's and C's

can someone point me to a refrence or give an opinion of the practical way to do this?

the way I do it is to pick the R based on the current for an accecptable amount of voltage drop, then 1/2piRC to find the corner freq. with regulated supplies you obviously don't need to shoot for .1 Hz, but Im not really sure how low is necesary in general.

additionally, two issues can crop up.

1) If you are decoupling an output stage and have limited voltage at the rail to start with, the desired voltage drop might be "zero". I tend to go for half a volt, but I picked that number out of the air.

2) If you are trying to minimize supply current variations all the way down to low frequencies with low resistances in series, this leads to 1000000000000uF capacitors. does anyone know of a tricky way to bootstrap this C without negating the current-smoothing effect?

I know, it is a comprimise that varies from case to case. I just wondered what the "guidelines" are for this kind of thing.

mike p

I'm not sure what exactely you are talking about--decoupling opamps (or generally speaking active circuitry)?

If so, I think you might misunderstand the issue a bit--decoupling is not basically about ripple/hum suppression, it's about making things stable by providing a low-impedance supply at high frequencies. Typical opamp-circuits have very high PSRR within the audio range and together with the 60-80 dB ripple suppression of the regulator you get hum way below the noise floor without further precautions (stray injection e.g. from the mains transformer and grounding is much more of a problem).

I'd say for most situations, it's not necessary to include a series resistor, in fact it might rise distortion in certain circumstances. If you have very high gain-bandwidth opamps and perhaps the layout isn't optimum (long and narrow supply tracks), it might help though to dampen the resulting LC filter--10 ohm is a good starting value. For details see 135208865AN-202.pdf.

My usual decoupling strategy for IC-based standard circuitry is to provide a 10 ohm/100 uF (or 220 uF) filter at the entrance of the board for both supply lines, and one 100 nF polyester per opamp. That seems to work pretty well.

Samuel

i use thermal fuses, also called a "pico" fuse or a "resetable" fuse. they reset when they are powered down and the load is taken off. kinda expensive but they don't current limit before shutdown like a resistor will.. and they don't burn up and take part of the PCB with the like a 10r will.

and for decoupling, 1000uf for every amp of current drawn parallel with a smaller capacitor of 10-100nf in normal applications.

[quote author="mediatechnology"]

SSL used carbon comps on standoffs so that they would open. Stinky mess but no board damage when they did. Carbon films will start a small fire enough to burn a hole in a PC board.[/quote]

And contribute to Global Warming too!

Samuel's guidelines are good ones. Also, I would never decouple where there's big power involved, like on the output stage of a power amp, unless it was a ferrite bead to a film cap/'lytic perhaps. As discussed, you generally just want to provide a local low impedance at high frequencies, and the caps alone, with due attention to where they are grounded, usually suffices. Self and others have discussed this issue iirc.

ok, that helps. even though I didn't really give enough information. sam's post is pretty much exactly how I would treat ICs.

I was thinking about a simple discrete circuit with very poor PSRR running on lower than ideal voltage (can't use a shunt regulator). trying to come up with a decoupling scheme to reduce crosstalk between channels.

> Damn! The session arguments in those last two links lead into the next County!

Use the forum's URL syntax.

And the RS link is expired, and does some really screwy stuff. Opened in a new window, it closes that and redirects the parent window.

Mostly, if you find something on a snazzy vendor site, if it doesn't have a short URL, it isn't going to be the same URL 15 minutes later, or for someone else. Give the main site and a part number/name, let us hunt.

I was thinking about a simple discrete circuit with very poor PSRR running on lower than ideal voltage (can't use a shunt regulator). Trying to come up with a decoupling scheme to reduce crosstalk between channels.

Click to expand...

So is there a specific reason why you don't want to make the simple discrete circuit a little less simple for better PSRR? Looks like the easier way to me?

Samuel

There are some charts for selection of power supply components, it's mostly just a bunch of lines with a 1:1 slope, parasllel to each other, tou plug in your VA and some other parameters, and read the va;lues off the chart.
This will change dependingon if you are running choke input, cap input, half wave, full wave, bridge, center tap, solid state or gas rectifier.

Here's a twist:
Warming up your amp with a ringing pwr supply.
http://www.siteswithstyle.com/VoltSecond/CLC_ringing/CLC_ringing.html

One thing you can do that's kind of like "bootstrapping" is to feed the voltage regulator's sensing from a point after the decoupling resistor. This doesn't work for multiple loads each with their own decoupling resistor fed from a single regulator, but as others have suggested you really don't need or want to be doing that anyway. You're trying to keep the supply impedance low, right?
Remember that the output current of an amplifier will modulate the power supply current, which will cause a modulating voltage drop across the supply impedance. In other words, your decoupling resistor can serve to *increase* supply noise. Since this noise signal is generated by the amplifier, the amplifier's finite PSRR will produce a feedback loop that can cause oscillation.