Questions about decoupling/coupling

[ubersam]

I've been reading some threads about decoupling/coupling an opamp from the power supply using a small ceramic cap between the positive and negative supply pins of an opamp (if using a single power supply). How would this method of decoupling (or is it coupling?) apply to transistors? (Does it apply to transistors?) Is it just a matter of placing a small cap between the collector/drain and ground? How about audio signal ground, can it be coupled/decoupled from power supply ground (0v) using the same method (small cap)? Are there any benefits to doing this?

[R.G.]

The point of using a cap from + to - on a power supply is to (a) act as a local reservoir of charge that supports the local circuit when sudden power supply needs come up and (b) to make the power supply look more like a very low AC impedance so that different stages can't cross-interfere through the power supply.

Let's look at that word - DE-coupling. If we have a perfect power supply, a voltage source that is 9.00000000V and can supply over a million amperes if needed, then we could run two amplifier circuits from it. Each amplifier would pull pulses of current from the power supply in synchronism with the signal it produces at its output. Because the power supply can supply so much current without changing voltage, the two amplifiers cannot affect one another through the power supply because the power supply voltage does not change no matter how much current it provides.

So far, so good. Now let's put a 1K resistor in series with that power supply, and connect both amplifiers to the end of the 1K. We have forced the power supply to look to the amplifiers like it has an internal impedance equal to that 1K resistor. What that means is that for every milliampere of current drawn from it, the power supply sags by 1 volt.

So if both amplifiers draw as much as 1ma of current on high output signals, they will intermittently cause the power supply to be between 9V (when they are drawing very little current) and 7V ( when they are drawing 1ma each). If we run a signal into only one of the amplifiers, the other amplifier will see its power supply wiggle around by a volt as the driven amplifier pulls pulses of current.

So the amp with no signal will be coupled to the driven amplifier through the power supply impedance.

Circuits vary in how well they ignore noise and other junk on their power supplies. Opamps are usually pretty good. Simple transistor circuits are not.

So we could help things a bit by putting a big, fat capacitor from from the end of the 1K resistor to ground out at the amplifier circuits. Now when each amplifier pulls current, it comes out of the capacitor first, and only then out of the 1K resistor. The capacitor has helped DEcouple the two stages by lowering the power supply impedance.

The value of the decoupling cap has to be low for the frequencies of interest. The idea of putting small caps in comes from two motivations. (1) We don't want to amplify RF, so we want caps that will have a low impedance at RF. Small ceramic caps do that well. (2) Large value electrolytic caps are not perfect, and at higher frequencies their impedance actually starts going back up, so we sometimes parallel them with smaller caps that have better high frequency response.

All multistage circuits need the stages to see a clean, low noise, low impedance power supply. There are several ways to do this:
(a) with a tightly regulated power supply
(b) with DE coupling caps
(c) in hard cases, both a and b, as well as series resistors between stages.

So yes, all circuits need decoupling caps to some extent. Sometimes we can get away without it, but it is very much getting away with a non-recommended practice. Simple transistor circuits need it more than opamps for decoupling reasons. Opamps need it more than transistor circuits because their high feedback makes them prone to oscillation.

You would not want to couple an audio signal to ground except when you are trying to reduce its level, perhaps in a lowpass filter or the inter-stage decoupling that I explained above.

Did that help?

[ubersam]

Thank you very much, that helps. So if I understand this correctly, in the example below (lets say that the transistor is acting as an amplifier): Ca is providing the decoupling for the amplifiers/devices connected to the +9v supply, and Cb is providing the decoupling for the amplifiers/components connected to the +4.5v supply. If we choose to do so, a small cap could be paralled with each of the BFCs. Would this be the correct conlusion? Or does the decoupling have to be at the amplifier/component level, i.e. a BFC from collector to ground on the transistor, then another BFC from the Opamp's pin 8 to ground? Essentially a BFC to ground at each amplifier's junction to the power supply. If I can add a small ceramic cap between pins 4 & 8 of the opamp, can I also add a small ceramic cap between the junction of the collector/source and the +9v supply, to ground?



In my original question pertaining to audio signal ground, I may have used the wrong terminology. What I was referring to is audio input and output jacks' ring connection (audio signal's negative return?). Is there a point or benefit in lifting this from the power supply ground? And would this be lifted using a small cap?

Thanks again.

[ubersam]

bump...

[R.G.]

So if I understand this correctly, in the example below (lets say that the transistor is acting as an amplifier): Ca is providing the decoupling for the amplifiers/devices connected to the +9v supply, and Cb is providing the decoupling for the amplifiers/components connected to the +4.5v supply. If we choose to do so, a small cap could be paralled with each of the BFCs. Would this be the correct conlusion?

Yes. Ca and Cb are decoupling their respective supplies.

Or does the decoupling have to be at the amplifier/component level, i.e. a BFC from collector to ground on the transistor, then another BFC from the Opamp's pin 8 to ground?

It is hard to tell, because the schematic leaves out part of what decoupling is all about - the resistance and inductance of the wires and PCB traces which connect the parts together. Whether you need to decouple right at the component or not depends on the wiring conductor size, it's shape, and the path it takes. If in this case the transistor was right next to the opamp, and the Ca capacitor was near (say within 1" of wire or trace) both of them, then they would probably be OK. If the Ca was six inches of wire conductor away, there is a lot more copper resistance and inductance between the cap and the parts to be decoupled. So it's not as effective. If the transistor was on one side of the board and the opamp on the other, say 6" away, they properly ought to have a local 0.1 to 0.01 cap right at their local pins, and then the large Ca is only responsible for the low frequency decoupling, and it can do that from across the board. For high speed stuff, RF, video and high speed logic, you typically have to put a 0.01uF every package. Opamp manufacturers will flatly tell you to put a 0.01uF ceramic cap as close to the device pins as you can get it, on every opamp. They will also tell you to make your power traces wide (0.05"or more) and straight (lower inductance).

The thing about effects is that we often get away with less than ideal conditions because the circuits are so small and simple compared to the industry. So whenever I suggest good practice, there is always somebody who pops up and says "Well, I never did none of that and mine always worked." And they're right - they got away with it. It's important to keep proper practice in mind so that when you bend the rules and get caught, you at least know what you have to fix.

Essentially a BFC to ground at each amplifier's junction to the power supply. If I can add a small ceramic cap between pins 4 & 8 of the opamp, can I also add a small ceramic cap between the junction of the collector/source and the +9v supply, to ground?

Yes. The idea is that the capacitor is a small local power supply that the circuit can draw on when it needs a quick fix of current. Ground traces are important here too. If there is ground resistance and inductance between your circuit and the official ground on the board, then every current spike causes a voltage shift as seen by the local circuit. Local decoupling caps help prevent this by being a local reservoir. You want the local decoupling cap connected to the local ground, NOT to a remote ground. Otherwise, the local IC still sees its ground move when it pulls current. In this situation , the cap may make you feel virtuous, but it's little help to the IC.

[Sir H C]

Here is a great article by one of the giants of IC design, Paul Brokaw with Analog Devices.

http://www.analog.com/UploadedFiles/Application_Notes/135208865AN-202.pdf

He has circuits named after him, he rules.

[ubersam]

Thanks... these are all great info...