op amp distortion problem

[tdkrause]

Im working on an overdrive circuit using the basic tube screamer topology with a TLO72. I've built the same circuit in the past with no problems. I re-designed the circuit board to accommodate a charge pump section, built one and I'm getting horrible distortion that sounds like the op amp is moving in and out of saturation. On soft notes it is clean but very dull sounding. Then when I play louder it distorts and becomes very loud.

I thought it might be a cold solder joint somewhere, so I built it on a bread board. Same problem. So I went back to a standard 9v power supply section like before. Same problem. So I switched op amps. No more problem with an NE5532, RC4558P, or LF353. An OPA2134 has the same problem on a smaller scale. It's mostly audible as the note decays. I also tried tying the circuit in to the reference voltage through a few different resistors.

So in summary: not a power supply problem, apparently not a biasing problem, same circuit worked just fine before, some op amps don't have a problem. Any thoughts?

[diydidi]

You'll have to post your complete circuit

[olddawg]

Some op amps are internally compensated and some are not.

[pdf64]

It may be that cable capacitence is causing your design to become unstable, dependant on the characteristics of the device used.
Suggest you try a series resistor (~ 2k2) between the output of the circuit and the output socket, to help 'isolate'the device from the cable capacitence, and report back.
Pete

[Mark Hammer]

Sees to me that maybe what you are asking the charge pump to do exceeds what it can do.

Ironically, though, and contrary to my suggestion, a higher-current chip like the NE5532 is not experiencing any problems, which would suggest it's not stemming from the charge-pump's current limitations.

Why a charge-pump anyways, if your goal is to generate clipping? Seems a little antithetical.

One of the reasons why there are so many different dual op-amp chips in the world is because they all have optimal conditions that can be a little different. Sometimes those optimal conditions are very different. And sometimes, particularly in the case of distortion, we avoid those optimal conditions and exploit the sub-optimal.

For instance, the classic MXR Distortion+ uses a pair of 1M resistors to provide Vref, and then feeds that bias voltage through another 1M resistor providing an eensy-weensy bias current to the non-inverting pin of the LM741. In contrast, the DOD250 has pretty mucht he same topology but provides Vref through a pair of 22K resistors, and feeds that to the functionally equivalent pin through a 470k resistor. Many folks who build a Distortion+ (including myself) will tell you that the thing almost never gets clean, even without clipping diodes. Part of that is the 741 itself, but I imagine some of it is the "starved" bias. In effect, the unit is a double clipper, by virtue of coloring the signal in the way it asks the 741 to act, and then by clipping that product with a pair of diodes.

That's the long way of saying you may be asking the TL072 to do something it doesn't like, and asking those other chips to do something they are perfectly at home with. Trouble is, we don't know what thatmight be.

[tdkrause]

I tried a 10k resistor just before the output capacitor. That didn't make any difference. As soon as I get home I'll post a schematic.

[tdkrause]

Yes, I am trying to clip. But at a higher voltage, this circuit has much better clarity. I tried it at 27 volts once and it made a remarkable difference. All the mush is gone. I want all the clipping to come from the diodes. I know that the TLO72 can handle it because I've built a dozen or so of these pedals and never had any problems. This is the exact same circuit I was using before but with a charge pump. At first I thought it was a problem with the new circuit board design, but why then would I suddenly have this problem on the bread board too?

[tdkrause]

You'll have to post your complete circuit

Not sure if I drew the charge pump section correctly, but the rest of the circuit is correct

[diydidi]

It seems your second opamp is missing bias (Vref). Try removing .1uF after first opamp first, and add a coupling Capacitor from its output pin to volume put. If that doesn't work, also add a High value res from Vref to input pin of second opamp.

[tdkrause]

Well, I know what the problem was. Turns out it had nothing to do with the circuit. In the case of the one that I built, there were a couple of solder pads that were barely hanging on. Re-flowing the solder wasn't enough. I guess that's what I get for not enlarging the pads in the cad library and then trying to eyeball the holes on a drill press. As for the other one, the breadboard i used was from radio shack. Enough said. I tried it again on a better bread board and had no problems.

This experience does, however, tell me that i need a better understanding of the causes of these symptoms. Many of your suggestions have pointed me in the right direction. One question that I have is this: how often does the circuit need a bias reference? Is there any advantage to connecting each op amp to a reference voltage if it will work just fine with just a 1 meg to VR at the first stage?

[J M Fahey]

One question that I have is this: how often does the circuit need a bias reference? Is there any advantage to connecting each op amp to a reference voltage if it will work just fine with just a 1 meg to VR at the first stage?

1) you *always* need to provide proper biasing
2) if this means wiring a high value resistor (1 M or whatever) to a suitable point, so be it.
You just can't leave it "open" .

[Steve Conner]

Correct. All op-amps leak a tiny DC current from their inputs, and if you don't provide a path for it, your circuit will behave pretty strangely. Some leak more than others: the NE5532 is about the worst you'll come across, and the TL072 about the best.

When troubleshooting op-amp circuits, always check what DC voltage the outputs are sitting at. They should almost always be at 0v (in a split supply system) or half the supply voltage in a single-supply system.

Pro tip: If you overdrive the inputs of a TL07x hard enough, it can flip phase, with the output popping over to the opposite supply rail from where it should be. I imagine if this ever happened in a distortion pedal, it would sound like crap.

[tdkrause]

p: you overdrive the inputs of a TL07x hard enough, it can flip phase, with the output popping over to the opposite supply rail from where it should be. I imagine if this ever happened in a distortion pedal, it would sound like crap.

So would a starved bias or low supply current cause it to reverse phase like that? I have a feeling that might be what was happening. At times it sounded like a (not sure if I'm naming the right effect) ring modulator. The fundamental was barely Audible. Other times it sounded like splatty, gritty fuzz. I took voltage readings at several points and everything looked ok.

[J M Fahey]

Your voltage measurements are "static" (no signal) or, best case, "average" (small glitches are too fast or short to be shown by meter) and you simply don't "catch" them ... but you sure can hear them.
That's where the scope shows its true value.
You can wire 2 opposite polarity diodes from + to - input, so never there's more than 650mV across them.
It does not affect "normal" signal.
And in the case of your voltage pump, maybe it can't cope with high demands.
I never had great results from them, specially since the basic problem is that the 9V battery does not like to be overtaxed.
I *did* use charge pumps (made with a humble NE555) in my 12V gel battery powered amps, to get a -12V rail, but situation is different, 12V alarm type batteries can supply *Amperes*.

[Mark Hammer]

The problem is that you had a bias voltage, but removed it with the .1uf cap after the first op-amp, which served to block DC. It's not wrong to do so, but you would need to bias the second stage if you aren't going to share the biasing of stage 1.

I've seen various TS-topology circuits, some that provide a 10k biasing resistor going to Vref at the non-inverting pin of stage 2, and some without. The majority, however, have a 10k biasing resistor there.

I'm no expert in such matters, but my sense is that the preference is to avoid sharing bias voltages across stages when the stage you want to share from is noticeably higher than unity gain. You'll often see bias-sharing across stages in things like chorus or flanger pedals, but in the majority of those instances, all those stages are run at essentially unity gain. If a stage has a gain of several hundred (as is true in your drawing), one really does not want to trust whatever DV voltage is coming out of it as an accurate reflection of Vref.

All of which makes me confused about how the standard TS-9/808/SD-1/Timmy/Fulldrive design has no DC-blocking cap on the output of the gain/clipping stage. Does biasing via a lower-value resistor (10k) supercede prior biasing? Someone wipe the sleep from my eyes.