Op-amp clipping way too early?

[edvard]

I have a circuit I'm working on; a high-gain CMOS circuit using NOR gates, and I need a buffer with a little "goose" in the front end.  So for simplicity's sake, I have an op-amp front and center.  My problem is that I want the output of the op-amp to be squeaky-clean, but I can't get much more than a gain of 5 before I get clipping when strumming hard.  "Welcome to the Real World, Ed" I hear you say, "Deal with it, Ed" I hear you say.  OK, but I have a question:

My guitar is a home-built and it's my #1.  It's got cheap chinese mini-humbuckers that don't output more than a ~350 mV (my analog meter died, so I'm probably missing some fast transients that may reach 1 volt or more), but they sound damn good.  My op-amp is running on ±12V, so even with a gain of 10,  I should not be reaching anywhere near the clipping threshold, even with a LM741, right?  I've tried LM741, TL071, OP27, one side of an RC4558, and LF353 (all from my junkbox, I don't have any others), and they all exhibit the same behavior.  I've gone over the datasheets, re-read Walt Jung's "Op-Amp Applications Handbook", Ron Mancini's "Op-Amps For Everyone", and I just can't find what may be at the root of this.  I found on the TL071 datasheet the "Common-Mode input voltage range" and it's ±11V, so you'd think I'd be safe, right?  Nope, still clips if I slam it at gain of 10, which would put my signal at ±3.5V.

Is there something I'm missing?  Is it really just transients?  Am I asking too much?

Circuit below:

[PRR]

Few meters will correctly read musical transients. When we used VU meters we designed-in 10dB to 12dB of "lead"-- peaks at 3X to 4X over what the needle said. Or in your assumptions, 10.5V to 14V peaks.

What do you need more than 3.5V peaks for??

[edvard]

Well, technically I don't NEED 3.5V peaks, just looking to shove a little 'oomph' into the CMOS stages and drive a tone stack when running clean, and I was curious as to why I got clipping at such (relatively) low inputs.  I suppose one clean-channel solution would be to split the signal to another op-amp configured for lower gain, but I also want to avoid clipping in the op-amp itself to keep the dirty channel gain 'cleaner' (yep, oxymoron...). 

See, I can't stand that "silicon chainsaw" buzz that you can get from overdriving an op-amp to the rails, or (curiously) from CMOS circuits using very high feedback resistance for gain.  That sort of thing is OK in a Fuzz, but it's not the sound I want.  I've discovered (and as others here have advised), that chaining multiple fairly low-gain CMOS stages creates a smoother distortion that doesn't do that, even at Heavy Metal gain extremes, and is not as prone to the noise and hiss that commonly plagues many CMOS designs.  I'm not necessarily trying to make Tubes-in-a-box either, but I have discovered that chained-low-gain CMOS circuits as previously described can get that ragged-but-beefy grind that I've only encountered previously in some Tube amps.

So you're saying it IS transients?  So I need to design a bit of soft-limiting into my booster stage?  Not a bad idea, but I don't like unnecessary complexity.  Maybe run it at a higher voltage (±15V) for the extra headroom?  And yes, sans proper measuring equipment, those peaks are quite the assumption on my part.

[amptramp]

I think what you are getting is oscillation.  Put a small capacitor (100 pF would be OK) from op amp output to inverting input and see if it changes.  If you have a pot in the feedback loop, it may have enough capacitance to ground that you have a phase lag in the feedback that causes a rising response at the output.  You may be getting oscillations far above audio frequencies into the RF range but these will ride on top of the signal and when the sum of oscillation and signal hit the clipping threshold, it will clip even though the audio portion of the signal is below the threshold.

[edvard]

I think what you are getting is oscillation.  Put a small capacitor (100 pF would be OK) from op amp output to inverting input and see if it changes.  If you have a pot in the feedback loop, it may have enough capacitance to ground that you have a phase lag in the feedback that causes a rising response at the output.  You may be getting oscillations far above audio frequencies into the RF range but these will ride on top of the signal and when the sum of oscillation and signal hit the clipping threshold, it will clip even though the audio portion of the signal is below the threshold.

Just tried your suggestion, no change.  Tried replacing the Pot with a 100k resistor, same thing.  I tried a 1n cap on the input to cut out just a little more bass and it was a bit better, but still get crackles on the peaks.  I went and looked up some op-amp-based booster schematics, and most boosters were only wired for a gain of 2-4 max, unless it was also an overdrive, so the 5-10x I'm looking for is kinda pushing it. 

EDIT: Many of those boosters are also running off a split 9V or two 9V, so maybe I'm not pushing it running at ±12V.... Anyway, I'm seriously thinking of running the op-amp off of ±15v just for the extra headroom. I mean, I intend this to be for a complete amplifier or rack unit...

[amptramp]

What do you have in the way of power supply bypass capacitors?  You don't have to have long leads running to the supply from the op amp power supply terminals, some will oscillate with as little as an inch of unbypassed power supply leads.  Also in this circuit, C2 is going to be reverse biased half of the time and some electrolytics fail under this condition.  If you are using breadboard construction, oscillation is an almost unavoidable possibility.

Are you located close to any transmitters that could be injecting RF power into the circuit?  I restore antique radios and the noise and hash is so great that the lower range of the AM band is almost unusable.

[composition4]

An oscilloscope would confirm Ron's suspicion, if you have one. I'd be more inclined to believe it's a power supply headroom issue. If it is, increasing your rails from +-12v to +-15v probably won't make much of an audible difference, if any.

[edvard]

What do you have in the way of power supply bypass capacitors?  You don't have to have long leads running to the supply from the op amp power supply terminals, some will oscillate with as little as an inch of unbypassed power supply leads.

I have two 500µf caps, one for each of the power rails, and I have the power at the op-amp bypassed with a 0.01µf cap.  I looked up power bypassing on the 'net and apparently it's good advice to have two when running on a bipolar supply; one for each power rail, because  a spike on both rails would be seen as no voltage by a single capacitor and therefore won't be bypassed.  It got crowded, but I re-configured for that scheme.  No change.

Also in this circuit, C2 is going to be reverse biased half of the time and some electrolytics fail under this condition.

Noted.  Since it is a bipolar supply, there's no need for that to be AC grounded.  Cap removed, no change in sound.  I believe I was experimenting with using that cap for setting the High-Pass cutoff, which it did, but not as effective as the input cap. 

If you are using breadboard construction, oscillation is an almost unavoidable possibility.

Yep, I'm aware of that, which is why I use those honkin' caps for the power rails, and bypass the power at the ICs, just in case.

Are you located close to any transmitters that could be injecting RF power into the circuit?  I restore antique radios and the noise and hash is so great that the lower range of the AM band is almost unusable.

I am in the same room as our WiFi router, but that's never given me any trouble so far.  Dodgy cables and running a large fan directly upstairs has more of an effect, even more than the fluorescent lights in my 'shop'.
I turned the WiFi router off while my wife was at the store, and no change.

[edvard]

An oscilloscope would confirm Ron's suspicion, if you have one.

Nope, don't have one.  I should get one, though I'd need to save my pennies, even for one of those cheap Asian jobs that are all over eBay.

I'd be more inclined to believe it's a power supply headroom issue. If it is, increasing your rails from +-12v to +-15v probably won't make much of an audible difference, if any.

Eh, maybe, maybe not.  I'm inclined to do it anyway and see what happens.

[iainpunk]

how about you replace the opamp with more CMOS gain, accept the clipping and call it a day...

cheers, Iain

[dschwartz]

How are you testing the circuit?
My guess are two causes:
If the output of the opamp is directly connected to the cmos input, the clipping from the cmos could be affecting the signal.. disconnect the cmos stage and retest..

Also, capacitive loads and/or low resistive loads (under 10k)will reduce the opamp headroom heavily.

[bean]

I don't know if this is helpful at all but, maybe doing two lower gain clean-ish stages in series (a dual op-amp) then just follow that with a voltage divider type gain control to your higher gain CMOS stages?

[edvard]

how about you replace the opamp with more CMOS gain, accept the clipping and call it a day...

cheers, Iain

HA!  I knew someone would say it.  At least it would be softer than op-amp rail clipping... 
I've tried wiring the op-amp as a unity buffer, and using four stages of the CMOS, but that last one just puts everything way over the top 
As I stated before, I'm going for high gain, not Fuzz, and that's what I get (and a bunch of noise) with four stages running, even at lower gains.  3 seems to be a magic number; nice grind-y distortion without a bunch of fizz but with an acceptable noise floor.

[edvard]

How are you testing the circuit?
My guess are two causes:
If the output of the opamp is directly connected to the cmos input, the clipping from the cmos could be affecting the signal.. disconnect the cmos stage and retest..

Also, capacitive loads and/or low resistive loads (under 10k)will reduce the opamp headroom heavily.

I discovered that pretty early on that the distorted signal will "leak", so I have always tested the op-amp either alone connected to my high-impedance computer interface, or through a tone stack, which simulations say has ~100k to 360k input impedance, depending on where the knobs are.

[edvard]

I don't know if this is helpful at all but, maybe doing two lower gain clean-ish stages in series (a dual op-amp) then just follow that with a voltage divider type gain control to your higher gain CMOS stages?

Not sure how that would help, but I can give it a shot.  I've tried the plain ol' voltage divider (like a volume control) between the op-amp and the CMOS input, and I like the response of the pot-in-the-feedback-loop scheme much better.

[Fancy Lime]

For a non-inverting opamps stage followed by a CMOS inverter, I like a "bluesbreaker style" gain control. Gives a wide gain range while keeping the maximum gain of the opamp stage within reasonable bounds.

Andy

[iainpunk]

As I stated before, I'm going for high gain, not Fuzz, and that's what I get (and a bunch of noise) with four stages running, even at lower gains.  3 seems to be a magic number; nice grind-y distortion without a bunch of fizz but with an acceptable noise floor.

if you think the distortion is to fuzz like but still want more gain, just take out some bass. the recommended frequency is generally thought off as 700Hz, but its a starting point, not a fixed rule to use that frequency.
a simple first order filter will do the trick

cheers, Iain

[aron]

Yes, nice tip Iain!

[Mark Hammer]

As I stated before, I'm going for high gain, not Fuzz, and that's what I get (and a bunch of noise) with four stages running, even at lower gains.  3 seems to be a magic number; nice grind-y distortion without a bunch of fizz but with an acceptable noise floor.

if you think the distortion is to fuzz like but still want more gain, just take out some bass. the recommended frequency is generally thought off as 700Hz, but its a starting point, not a fixed rule to use that frequency.
a simple first order filter will do the trick

cheers, Iain

I like to say that most of the signal, or at least a lot of it, "lives in the basement".

[edvard]

For a non-inverting opamps stage followed by a CMOS inverter, I like a "bluesbreaker style" gain control. Gives a wide gain range while keeping the maximum gain of the opamp stage within reasonable bounds.

Andy

That's exactly the scheme I'm using, though the Bluesbreaker has some additional filtering going on in the inverting connection to ground.