Coupling Caps, ground leak and series resistors

[Unlikekurt]

Hi all,

I've been looking at schematics with op amps a lot lately and have found enough variation in practices that I'm left with a few questions.
-Why would one not utilize coupling caps between op amp stages?
-Why would one not have coupling caps before the first op amp stage after the  pedal input or a coupling cap after the last op amp stage?  (Dismissing the appearance of caps due to some other circuitry such as passive tone stack etc)
-Why do sometimes coupling caps appear without subsequent pull down leak resistors?  (dismissing if there is a path to ground via some filter, voltage divider, etc.)
-Why are series resistors at op amp outputs sometimes employed and sometimes not?  Aside from as part of LPFs or when current limiting is absolutely essential?  Is it a best practice to always utilize?

I realize that is a lot, but would certainly appreciate some thoughts.
Thanks.

[amptramp]

Hi all,

I've been looking at schematics with op amps a lot lately and have found enough variation in practices that I'm left with a few questions.
-Why would one not utilize coupling caps between op amp stages?

If you want to retain DC coupling or minimize the loss of low frequencies or the layout doesn't tolerate any more space taken up.

-Why would one not have coupling caps before the first op amp stage after the  pedal input or a coupling cap after the last op amp stage?  (Dismissing the appearance of caps due to some other circuitry such as passive tone stack etc)

In some cases where you have complete control over what the previous or subsequent stage is (because it is in the same pedal), you may be able to perform whatever function you need because the input from a previous stage is at the correct voltage to bias the op amp.  If the pedal uses positive and negative supplies, it is quite simple to bias everything at zero volts.  Otherwise, the op amp may have an input dynamic range that includes ground.

-Why do sometimes coupling caps appear without subsequent pull down leak resistors?  (dismissing if there is a path to ground via some filter, voltage divider, etc.)

Sometimes it's just bad design but if the pedal that follows the one with no pulldown resistors is never switched off, you rely on the input pulldown of the next pedal to keep the capacitor voltage from rising at the output.  Sometimes if film capacitors are used, the output leakage is negligible.

-Why are series resistors at op amp outputs sometimes employed and sometimes not?  Aside from as part of LPFs or when current limiting is absolutely essential?  Is it a best practice to always utilize?

If the op amp drives any significant amount of capacitance, you can get a lowpass lag from the output impedance of the op amp and the capacitance of the cable plus that of the subsequent stage that causes a loss of high frequencies in the feedback loop.  This results in a rising closed-loop gain as frequency goes up (because the negative feedback disappears) that can cause oscillation when it meets the falling characteristic of the op amp itself.  It pays to isolate the output of the op amp from the output of the pedal.  It is also used to limit the damage when someone connects a power source to the output (some roadies are dumb).  It is a good practice as is the practice of putting a small capacitor across feedback resistors to keep oscillation at bay.

I realize that is a lot, but would certainly appreciate some thoughts.
Thanks.

[Unlikekurt]

If you want to retain DC coupling or minimize the loss of low frequencies or the layout doesn't tolerate any more space taken up.

In some cases where you have complete control over what the previous or subsequent stage is (because it is in the same pedal), you may be able to perform whatever function you need because the input from a previous stage is at the correct voltage to bias the op amp.  If the pedal uses positive and negative supplies, it is quite simple to bias everything at zero volts.  Otherwise, the op amp may have an input dynamic range that includes ground.

Does that mean when utilizing a bipolar supply and referencing 0v/gnd as bias you don't need to decouple between stages? Isn't there still the possibility of developing an offset at an output?

Sometimes it's just bad design but if the pedal that follows the one with no pulldown resistors is never switched off, you rely on the input pulldown of the next pedal to keep the capacitor voltage from rising at the output.  Sometimes if film capacitors are used, the output leakage is negligible.

That's what my guess had been.  To rely on the next unit in the chain.  Which seems foolish.  But at the same time, let's say the last thing in your pedal is a volume pot and the first thing in the next pedal is a pull down resistor, should you account for that possibility in your design since that pull down would be in parallel with the output of your voltage divider?

-If the op amp drives any significant amount of capacitance, you can get a lowpass lag from the output impedance of the op amp and the capacitance of the cable plus that of the subsequent stage that causes a loss of high frequencies in the feedback loop.  This results in a rising closed-loop gain as frequency goes up (because the negative feedback disappears) that can cause oscillation when it meets the falling characteristic of the op amp itself.  It pays to isolate the output of the op amp from the output of the pedal.  It is also used to limit the damage when someone connects a power source to the output (some roadies are dumb).  It is a good practice as is the practice of putting a small capacitor across feedback resistors to keep oscillation at bay.

Would it be a good practice to simply put a series resistor sized so if the full possible output swing from the data sheet were to be across it the current wouldn't exceed the data sheet max for the op amp?  And is it better to have the series R right at the output or after the coupling cap/shunt resistor?

Thanks!

[ElectricDruid]

Does that mean when utilizing a bipolar supply and referencing 0v/gnd as bias you don't need to decouple between stages?

Not only with bipolar supplies. In any situation where the output from the previous stage has the same bias that the next stage is going to use.
Here's an example:

(From my Hard Bargain distortion: https://electricdruid.net/building-druid-pedals/

The first stage is a simple buffer. It uses R6 to provide bias to the op-amp.The next stage uses the same bias, so there's no point AC-coupling the signal and then having to add another resistor to put back the bias level we just removed.

The tone control stages do the same thing. The buffer after the clipping/lowpass stage sets the bias, and then the two control circuits both use it without needing inter-stage coupling caps:

Isn't there still the possibility of developing an offset at an output?

Yes, there's possibility, but if it's only mV from the op-amp and the following stage doesn't use large gains (which would increase the offset) then all it's doing is eating a few mV of headroom, and since your bias resistors weren't that accurate to begin with, you'll never know.

(In the Hard Bargain example, I didn't care about even the gain because I knew the drive stage was going to slam the signal into the rails anyway....)

That's what my guess had been.  To rely on the next unit in the chain.  Which seems foolish.  But at the same time, let's say the last thing in your pedal is a volume pot and the first thing in the next pedal is a pull down resistor, should you account for that possibility in your design since that pull down would be in parallel with the output of your voltage divider?

Yes, you should make sure not to use a 1M volume pot so that it doesn't matter! What's wrong with 10K?!

Would it be a good practice to simply put a series resistor sized so if the full possible output swing from the data sheet were to be across it the current wouldn't exceed the data sheet max for the op amp?  And is it better to have the series R right at the output or after the coupling cap/shunt resistor?

Yes, this is reasonable. I often use a 560R for this job. Assuming 9V across it, that'd be 16mA - that's not going to kill an op-amp.

Series order of components doesn't matter.

HTH,
Tom

[antonis]

(From my Hard Bargain distortion: https://electricdruid.net/building-druid-pedals/

The first stage is a simple buffer. It uses R6 to provide bias to the op-amp.The next stage uses the same bias, so there's no point AC-coupling the signal and then having to add another resistor to put back the bias level we just removed.

I've just noticed that DC on pin 5 (IC1.2 non-inverting input) should be about 1.58 times Vref, according to 1 + 33k/(10k+47k)..
(is there a missing cap for DC roll-off to unity or is just me missing something..??)

P.S.
Don't bother with a coffee addicted brain, Tom...
(just took Vref as GND point..)

[Unlikekurt]

Thanks guys!
I went back to the breadboard armed with some additional knowledge.  And made some improvements.  I'm still left wondering a little regarding the ground bias reference.  If, when using a bipolar supply the circuit is referenced to ground at the first stage and there is no need to shift that at any subsequent point, is there no need for coupling caps nor bias resistors through the remaining stages until the output?

[amptramp]

If you are cascading linear stages, the only thing you need to worry about is whether the offset voltage of the first and subsequent stage is amplified sufficiently to cause trouble in the subsequent stages.

In the Electric Druid Hard Bargain above, the first stage has a gain of 4.3 and the subsequent stage has a DC gain of 1.  If you have an offset of 10 millivolts in each op amp, you have 43 millivolts at the input to the second stage and 53 millivolts at its output.  This hardly affects your signal range at all and the second amp is AC-coupled to the antiparallel diodes, so the effect of the offset never gets to them.  If the diodes were DC-coupled, the offsets would make the output go to a DC value that was not zero and the clipping would not be repeatable between two samples of the same pedal.

This advice doesn't change whether you are biased to ground with ± supplies or to Vcc/2 with a single-ended supply.

There are also power amp stages where the amplifier is DC-coupled but there is an integrator taking the output and feeding it back into the input to minimize the DC offset.