AC Coupling and OP Amp Stages

[eev_carl]

Hi,

I have a single supply.  If I'm connecting several op amps together for an audio guitar effects pedal, should I AC couple each stage?  I have an input buffer, clipping stage, filter, and output buffer implemented with a 4-op amp IC.  I remember reading somewhere that if you don't AC couple everything, you'll get more noise as the when DC is passed along and amplified.  Do only gain-producing stages need this treatment?

Thanks,
Carl

[Benta]

There's no reason to AC couple each stage, input and output is enough.
With a single supply, you have the issue of setting a noise-free DC bias for the first stage. With a dual supply you have a nice quiet ground as bias.

[eev_carl]

I was breadboarding the attached circuit but don't get similar results unless I put a non-polarized coupling capacitor at the NPN's emitter.  I re-implemented this with a pair of op amps on a 1458 with the first op amp serving as the input buffer and got the same results on the scope.

Are there any situations where I'd need to AC couple op amp stages?  I'm planning on chaining together an input buffer > clipping circuit > filter circuit (tone control) -> output buffer with a single supply.  I've seen coupling and de-coupling between stages in some schematics I've been reviewing but don't have a feel for the rule of thumb.

Thanks,
Carl

[MiDi]

AC coupling is used if there is the need for a specific high pass filter and/or if the DC part is out of common mode input range of amplifier (e.g. high voltage gain stages suffer from input offset voltage)

[eev_carl]

AC coupling is used if there is the need for a specific high pass filter and/or if the DC part is out of common mode input range of amplifier (e.g. high voltage gain stages suffer from input offset voltage)

But my circuit doesn't meet either criteria (audio frequencies, low input levels, and modest gain (x5)). 

If I had to guess, I think I need the caps for the same reason I put a cap between the source voltage and the circuit.  DC is flowing back to the input and interfering with the preceding signal.  At least that's what it seems from my A/B testing of the cap and cap-less circuits.

[Benta]

The problem is, you have no control over the DC level of the emitter follower.

Bigger question: why do you need this at all?

I'd simply build it as a non-inverting amp and eliminate the emitter follower completely.
Input signal capacitively coupled to the non-inverting input, with a 100 kohm resistor to Vref to set bias.
R5 to Vref.
Done.

Note that both the input cap and the cap at Vref influence the frequency response.

An even better solution would be to buffer Vref, either with an opamp voltage follower, or... your emitter follower 

[eev_carl]

Bigger question: why do you need this at all?

I'm implementing a larger circuit and am breaking it down into subcircuits because my first try didn't work and I wasn't able to troubleshoot it.

There's no practical purpose to buffer the high impedance second stage.  In the real circuit, a JFET switch is taking a clean (but buffered) input from the emitter follower and bypassing the clipping and tone control stages.

[MiDi]

You asked for op amp and not for your input stage with EF these are different things...

And in addition to that, for audio you only want to have ~20-20.000Hz bandwidth.
Speakers or Headphones do not like dc voltages (output AC coupling needed, if dc voltage is more than a few mV)
It is a good advise to reject any dc voltage on input as you never know what is connected...

[eev_carl]

I get similar results with a unity gain op amp circuit feeding an inverting amp on a 1458.  No AC coupling cap b/t Vout1 and V-2 --> a noisy 2V DC signal.  Cap --> an amplified source wave.

[eev_carl]

Here's some more background to my question.  I'm wondering why there's a coupling cap between the Input/Clipping stages, the Tone/Output stages, but not the Clipping/Tone stages.

https://www.electrosmash.com/tube-screamer-analysis

Is there a rule-of-thumb about when one is needed?

[Benta]

Is there a rule-of-thumb about when one is needed?

Not a rule-of-thumb, but a rule: whenever the DC levels change, you need a coupling cap. And this is exactly what happens at the points in your link.
Example: the opamp in the volume stage has an output cap. Why? Because the volume pot is ground referenced. If it was referenced to 4.5 V it wouldn't be necessary.
And so on.

[floobydust]

It's not an issue over coupling capacitors. Your circuit is wrong, the single 500k resistor base-bias resistor to Vcc (not 1/2Vcc).

I dislike emitter followers in audio circuits as I find (at high levels) they don't have symmetrical input/output impedances. Your EF can source many mA but sink almost nothing with the 10k pulldown. The op-amp stage has a low 1k input impedance. The result is some rectification of the signal and a DC average that moves around, and it does sound not so great with the high distortion.

[eev_carl]

Your circuit is wrong

Thanks.  Here's the circuit with the correct connection to Vcc/2.

[Benta]

Why do you insist on the emitter follower? Start from scratch again and stop leaning on the "Tube Screamer" schematic. Either copy it 100% or do your own thing. A mix won't work.

Read my reply #5 again.

[eev_carl]

I'm at scratch.  I've made and tested each stage individually, but still am missing something when biasing two or more stages.  I plan on using a 4 op amp chip and was getting similar results when I made the unity gain amp from an op amp (rather than the NPN).  I can post an op amp-only LT Spice example later. 

[floobydust]

I think there might be a misunderstanding about the op-amp's operation when it is level-shifted.
Usually textbooks have the (+) input grounded at 0V for the inverting configuration, with perfect world +/- rails.

The op-amp stage has a voltage gain of around -5 in comparing the difference between the (+) input which is level-shifted about the Vref center point at 4.5V, and the (-) input before the 1k resistor.

What would be expected as an output when the emitter-follower is providing say 4.5V? 4.5-4.5=0V difference and the op-amp outputs 4.5VDC which is fine.
But with 4V out of the emitter-follower, the op-amp's output is shifted up near the rail at 7V, as it tries to maintain 4.5V at the (-) input with a single 9V supply. So any DC offset away from the Vref point gets amplified which can cause saturation, so an input capacitor is always used to block DC.

I'm not the greatest at explaining this, some people like math, others get a feel for it. I don't know your perspective.

[eev_carl]

Does a set of op amp stages that are level shifted need coupling caps between each stage then?  We're blocking the DC that was added in the + terminal.  Is this positive feedback?

[Benta]

Does a set of op amp stages that are level shifted need coupling caps between each stage then?

You've now asked this question in different ways several times in this thread. And you've been given the answer every time.

What you need to do is to set the DC bias voltage in the first amp correctly. Then the following amps will also sit at this DC bias point without coupling capacitors.

Your input emitter follower brings nothing, remove it, and run the opamp as a non-inverting amplifier as previously described.

[eev_carl]

Does this look correct then?  I ditched the emitter follower.  The role of the first op amp is to provide a buffered signal that will bypass the second (and third) op amps.  A JFET switch will toggle between this clean tone and the processed tone which is clipped and filtered.

[Benta]

Should work perfectly. You now have a circuit with a gain of appr. -5 and a DC bias at V+/2. You can add following stages without coupling caps.

One comment: I don't know the LT1001, but 500 kohm is pretty high for a bias resistor. It depends on the opamp input bias current.

[floobydust]

Generally speaking, once you have level-shifted there is no need for blocking capacitors downstream.
That circuit looks fine.

But...

It's trickier for guitar effects because they are usually overdriven so the textbook no longer applies and there are other things to look at.

When overloaded, some op-amps will phase-invert and blast out frequency-doubled (2nd harmonic). You can also get a DC-average (sub-harmonic and rectification) that can make one stage upset the bias of following stages, without coupling caps.

Example is you overload the input stage and it has a net DC center or operating point move up a volt or two of the 1/2Vcc point.
Without blocking caps, this now propagates to the amplification stages downstream. So stage two will now get saturated and cut out.
If you have a diode-clipper, it is expecting a symmetrical waveform but will instead be getting a waveform centered up, so it would clip one side only.

The art here is to have things sound reasonably like a guitar when overloaded. The "crunch" and "overdrive" sound is a lot of work to do after the basic clean signal path design is done.
I'm saying technically you don't need interstage coupling capacitors but they can help a stage stay centered and in balance, instead of just saturating.

[David Hess]

AC coupling is only required to the extent that DC offsets need to be controlled including the DC level shifting offsets in direct coupled amplifiers which can be quite high.  Operational amplifiers typically have such low offset voltages that AC coupling is not required between stages.

[Zero999]

Your circuit is wrong

Thanks.  Here's the circuit with the correct connection to Vcc/2.

The problem with that common emitter configuration is the bias point is highly dependant on the transistor's Hfe. Try changing the transistor model in LTSpice, run the DC operating point simulation and note how VBuffer changes. It will also change depending on the temperature.

This is why a proper potential divider is used to bias a BJT circuit. The downside is it lowers the input impedance, but that can be overcome with bootstrapping.