where are decoupling capacitors required

paulstam

Member
2012-05-30 10:34 pm
I'm building an active amplifier with a DAC, pre-amp, active crossover and six power amp's. Each of these components have their own decoupling capacitor on the input (not the digital input off course) and on the output.

Question, can I bypass the input capacitor from the power amp's, because the active crossover already have capacitors on the output. (between the output and potentiometer)

There is a lot of information about decoupling, and it seems that the quality of these capacitors are important (read expensive). Because they are also in the signal path I like to skip them where possible.
 
As has been said series coupling capacitors are there to block DC between stages and to form high pass elements.

If you have circuits that operate at unity gain and are all ground centred, ie bi-polar, then coupling caps aren't really necessary. Where you get issues is in circuits that have gain (power amplifiers) or between two circuits that operate at different bias points.

In a unity configuration no gain is applied, so if a small amount of offset is present on the output of one stage feeding another, then that offset will be apparent on the output of the final stage, ignoring that stages own DC offsets. If you've got a gain element within the second stage then the offset applied to it will be amplified by the gain factor. This could become extremely significant and cause massive issues. This is why power amplifiers use DC coupling on their inputs. Some do not but these use active circuitry to compensate for any DC that may appear on the output.

In other cases you will have DACs or (usually) single supply circuits that require a DC bias be applied to work properly with audio signals. Typically audio can be represented as a sine wave, that is a signal that swings positive and negative about a common zero point in the middle. Ideally circuits approach this by having both a positive and negative supply, like so they can swing both above and below the common ground reference point (zero volts) and viola we have audio. But if your circuit has to be powered from a single supply, it cannot swing both directions relative to zero. The way these device get around this is by adding a DC bias to their circuitry. This is usually set to the supply voltage divided by 2, or something very similar. This bias is now used to define a new operating point around which the sine wave can function. Zero has now been redefined as supply/2 so the signal can now effectively swing between zero and supply, resting in the middle at supply/2.

Obviously the DC bias applied to the circuit (whatever supply/2 is) cannot be applied to anything before or after said circuit so DC coupling caps will always be used to ensure that the inputs and outputs have zero volts present on them.

They are called coupling caps because they are used to couple one stage to another. Technically they do decouple different DC operating points between stages, but decoupling caps are used to describe caps that sit between a power rail and ground as a way of reducing voltage ripple from the power rail, provide a limited low impedance source of energy for said circuit and to help keep the power rail free from noise.
 

paulstam

Member
2012-05-30 10:34 pm
From my experience any coupling cap should be one only of the best quality you can use.
If you bypass a coupling cap with a smaller one you create smearing at the overlap frequency of the two caps.
Power supply decoupling caps can be bypassed with better quality smaller caps.

Cheers George
It is clear to me now I meant coupling caps instead of decoupling. What I didn't know that it is not recommended to bypass theme with a smaller (with better specs) to enhance the quality.

Thank you for that information.
 
I recommend AC coupling for the Power Amplifier.
I also recommend that the DC blocking capacitor has an optional bypass setting.
This could be a switch, or a pair of pins to be shorted, or a separate phono input.

I recommend the same optional AC/DC coupling at the output of your source equipment.

That way you can choose which DC blocking capacitor gets used for best sound quality and you don't get forced into using a series pair of DC blocking caps, which could be the wrong value and/or an unsuitable type.
 
georgehifi said:
If you bypass a coupling cap with a smaller one you create smearing at the overlap frequency of the two caps.
Power supply decoupling caps can be bypassed with better quality smaller caps.
I'm not sure I understand what is meant by "overlap frequency". In general, caps (whether coupling or decoupling or smoothing) do not need and should not have 'bypass' caps added. In some particular circumstances (e.g. the bias network for an RF PA) extra bypassing is used, but then it is being used by an RF designer who presumably knows what he is doing.

If you use the correct value for a coupling cap then it doesn't matter too much what type of cap it is.
 

paulstam

Member
2012-05-30 10:34 pm
Maybe that's the best answer to this problem, just keep options open. If there is a gain component involved or there is a single power supply in one of the stages, you need at least one coupling capacitor to block the DC component between the output and the input. So far I know that must be a good quality MKP capacitor. Its value depend on the lowest frequency and impedance of the circuit behind it.
 
Maybe that's the best answer to this problem, just keep options open. If there is a gain component involved or there is a single power supply in one of the stages, you need at least one coupling capacitor to block the DC component between the output and the input. So far I know that must be a good quality MKP capacitor. Its value depend on the lowest frequency and impedance of the circuit behind it.
They don't have to be MKP or better.

The reports we have seen have great difficulty measuring any distortion when the capacitor is used as a filter. They have to increase the AC voltage across the capacitor to considerably more than we usually see in capacitor filters for audio just to be able to measure the differences in distortion for the different types.
But the expertise they bring to the problem is welcome because it shows that even types that should be bad are not that bad and when these electrloytics are used in back to back arrangements the distortions do reduce.
I seem to recall that a pair of back to back bi-polars are almost as good as the very good MKT and that puts the dual bi-polar almost upto the quality of the MKP.

Now we come to the crux: what AC voltage is across our audio coupling capacitors?
It is almost none !
and that means the distortion of that almost zero audio signal must also be almost none.

A properly sized capacitor of an appropriately selected type can act as a VERY good coupling capacitor if you ensure it never acts as a filter of the audio signal.
Select a good commercial quality MKT with a passband at least a couple of octaves wider than the audio band you need and you can ignore capacitor distortions.
 
Last edited:
I'm not sure I understand what is meant by "overlap frequency".

We all agree that different coupling caps sound different from one another.

The bypass cap being small can't do the same frequencies as the larger one can. Individually one can sound faster/warmer/whatever than the other, add them together at the same frequency and you have smearing.

But in decoupling caps there's no passing of audio so it doesn't matter if you bypass.

Cheers George
 

paulstam

Member
2012-05-30 10:34 pm
They don't have to be MKP or better.

The reports we have seen have great difficulty measuring any distortion when the capacitor is used as a filter. They have to increase the AC voltage across the capacitor to considerably more than we usually see in capacitor filters for audio just to be able to measure the differences in distortion for the different types.
But the expertise they bring to the problem is welcome because it shows that even types that should be bad are not that bad and when these electrloytics are used in back to back arrangements the distortions do reduce.
I seem to recall that a pair of back to back bi-polars are almost as good as the very good MKT and that puts the dual bi-polar almost upto the quality of the MKP.

Now we come to the crux: what AC voltage is across our audio coupling capacitors?
It is almost none !
and that means the distortion of that almost zero audio signal must also be almost none.

A properly sized capacitor of an appropriately selected type can act as a VERY good coupling capacitor if you ensure it never acts as a filter of the audio signal.
Select a good commercial quality MKT with a passband at least a couple of octaves wider than the audio band you need and you can ignore capacitor distortions.

I used for the calculation the on-line calculator from V-cap.
Coupling Capacitor Calculator by V-Cap

In the design of my active crossover there are two bi-polair capacitor (47uF) in the output (bass). I have the intention to replace them by one MKP of 22uF. Although the 22uF seems to me a very high value here, it seem a good idea....or not.
 

Attachments

  • Crossover-1.JPG
    Crossover-1.JPG
    66 KB · Views: 169
If you want the AC coupling capacitor to pass all of the audio band then you must select a VALUE that gives a low enough frequency that allows all the audio band to pass.
The general recommendation for this is to allow at least two octaves or even a decade lower than our 20Hz limit.
That would put the F-3dB of the coupling capacitor filter at somewhere around 2Hz to 5Hz.
Some would choose to go even lower.

But you need to choose your F-3dB frequency and for that you need to know what filter values to use.
Just specifying 22uF tells you and us nothing about the final filtering F-3dB.
You need C and R values !
 
georgehifi said:
We all agree that different coupling caps sound different from one another.
Do we?

The bypass cap being small can't do the same frequencies as the larger one can. Individually one can sound faster/warmer/whatever than the other, add them together at the same frequency and you have smearing.
Nonsense.

But in decoupling caps there's no passing of audio so it doesn't matter if you bypass.
If the decoupling cap isn't passing audio (e.g. from a supply rail to ground) then it isn't doing decoupling so you can omit it from the circuit.