Feedback diode clipping mystery..to invert or not to invert?

[dschwartz]

Hi guys..
I was working on a new design that starts with an opamp od stage a la tube screamer (non inverting, feedback clipping diodes) but i boosted the gain a lot (1M gain pot with 470ohm shunt resistor..lots of gain)..it seemed to sound very nice, until i noticed a very ugly "blurry" distortion on hard attacks, specially on single notes on the 3rd string...

So i took my "solid state guitar amplifiers" book by teemu ktala (a must read for anyone on guitar electronics) and he wrote that feedback diode clipping should be implemented on INVERTING configuration..because on non inverting clipping, there is crossover distortion produced ...

I simulated the circuit with ltspice and i couldnt see a difference in clipping , but on "everycircuit" (a mobile app i use for quick sim of ideas) it clearly shows that the non inverting configuration produces crossover distortion as the diodes start to conduct..also the ouput amplitude is not really limited by the diodes...

I'm a little confused..anyone have an idea what I'm talking about?  I makes sense to me since the tube screamer never sounded like an overdriven amp, instead it has this "clean bleed" and the clipping indeed sounds like crossover distortion that fades away as the note decays...

Have you experimented with soft clipping on both configurations?  Im intrigued

[PRR]

Say you set non-limit gain to 10, inverting or non-inverting.

In inverting connection, when the diodes limit, they limit. Output does not increase.

In non-inverting connection, when the diodes limit, they only cut the gain down to *unity*. Output still increases, just not as fast.

How it *sounds* with high-high gain and complex guitar signals may be another matter.

[Quackzed]

       i'd think if it was usual crossover distortion, it'd be MORE apparent with low level signals than high level signals... like sending a signal through antiparallel diodes in the signal line? otherwise it might be that the signal is so hot, anything under the diode clamp level is basically vertical and the unity signal riding on top is big enough to hit the rails and possibly get inverted by the poor tortured opamp? some opamps do weird inverted things when really slammed?
      you could always use some led diode combo's before the opamp/ after the input tranny (if theres one)  to keep the signal flat juuuust before the point that it'll do this to the opamp? after the opamp might work too, with 2 btb leds maybee depending on what the issue is? just a few thoughts..
      i'd record a sample and then zoom into the waveform on the recording to see whats going on... if theres any weird inverting going on etc...

[Voltron]

Hey would you post a little more details, I mean, from what you said Is your non inverting amp trying to go about 2k gain?. I have to check the Teemu pages but usually happens what PRR said. Diodes on non inverting have somewhat like a "compression" effect. I never had problems on ltspice related to crossover, maybe you can share a file?

Saludos!

[dschwartz]

Yeah maybe I'm  hitting the rails with this amount of gain..the effect is different than harsh clipping, its more like an intermodulation , out of tune wobble i get at loud attacks..with a strat it sounds awful..

[anotherjim]

Maybe the amp in not stable with such high gain. Have you tried the usual cures? Cap across diodes, series resistance in output...
If it sounds like IM distortion, maybe it is, which could mean there's another frequency present.

[ashcat_lt]

TubeScreamer distortion is crossover distortion, and it is dependent on the signal level.  There is something important in the particular gain structure in that circuit that it is about just right with typical guitar inputs to stay in the crossover region most of the time so that the peaks look squashed in comparison.  With larger signals or more gain, though, the effect changes drastically.  The peaks if the waves look relatively normal, while the areas around the zero crossings look strange by comparison.  A TS has at least a couple of volts between the diode turnover and as close to the rail as it will swing, a reasonable amount of room to clean up some. 

It's interesting to remember that the whole treble boost thing that the TS does only really happens in the crossover region also.  When the diodes are "open", the frequency response is (relatively) flat.

[dschwartz]

Yes i ended up lowering the gain with a 2.2k /47n shunt and the weird clipping went away..i use this to boost 2 inverters with relatively low gain, it does the ts into driven amp thingy pretty well..

[PBE6]

Interesting, I wonder if the 470 ohm load on the opamp was too much and was causing the chip to get stuck?

[dschwartz]

Maybe..but ive used that value on other experiments with no issue..the difference is that i was driving a jfet muamp, and this time i was driving an inverter...impedance matching maybe?

[Voltron]

SOrry but I'm still confused with this.

"1M with a 470ohm shunt"
You trying to do a TS with x2000 gain? Shouldnt this be more an oscilator instead?

[ashcat_lt]

SOrry but I'm still confused with this.

"1M with a 470ohm shunt"
You trying to do a TS with x2000 gain? Shouldnt this be more an oscilator instead?

Nah.  It's just a lot of gain.  A Rat has 100K over 47R, which is the same ratio.  Mine actually goes to like 180K with the gain all the way up and it still doesn't actually oscillate.  The caps in the circuit help to make sure that no frequency ever really gets that full 2000x gain, and GWP starts to come into the picture at these levels also, but basically it's not going to oscillate unless you give it positive feedback - intentional or otherwise.

[dschwartz]

Actually i ended up with a 1M/2.2k  it's  more natural soundind and not so squashed and weird