More useful distortion from a second stage that is preceded by a gain control

[Mike Sulzer]

Suppose we have an amp with a gain control after the first stage, enabling adjustment of distortion in the second stage from clean and on up. The goal is to make this distortion more useful. I hesitate to use the much overused term "touch sensitive", but it actually is not such a bad description.

I will begin by showing the circuit and describing its differences from the usual one, although you might immediately notice that it draws functionality from some overdrive pedals, most famously the tube screamer. (This tube circuit here seems kind of natural, and I assume that it must have been done before many times; maybe someone has a reference to something very similar.)

The first change from the usual circuit is a tiny coupling capacitor from the first plate to the 500K gain control. This begins cutting gain from about 700 Hz on down. After the second stage, the response below 700 Hz is restored. In between there is distortion. When the gain is adjusted low for linear response, the frequency response of the amp is essentially the same as it would be "unmodified". When the gain is increased, the distortion is very different from the unmodified circuit.

It is useful to look at how a Tube Screamer works. The frequency dependent distortion level is obtained with a single op amp with diodes the feedback path and a series RC in the input path (negative input). This is a very precise circuit operating over a wide range of gains. The tube circuit is similar in effect, but realized in a very different way. There is some variation in the cutting time constant as the gain is changed; this might be thought of as a feature rather than a bug. What does this frequency dependent distortion level (bass cut) do in both the TS, or other similar overdrive pedal, and this tube circuit? When a single string is played, the slower changing parts of the waveform are distorted less than the faster parts, and the higher frequency strings are distorted more and differently from the lower. This means that how you excite the string vibration can have a significant effect on the transient response and the harmonic content of the signal. When multiple strings are played, the intermodulation distortion is very different. The effects of the fundamental and lower harmonics of one string affect another string less than they would otherwise.

In the tube screamer, the filter after the distortion produces mid range emphasis. It is, after all, an overdrive pedal. This tube circuit has a different function, a kind of replacement for turning the amp up and getting output stage distortion, and so the original response is restored.

Why diode distortion in the tube circuit? There are other possibilities. For example, you could just let the tube distort. Or you could use the grid cathode conduction as a diode in one direction, and add silicon in the other. I tried these circuits, and the diode circuit in the schematic is best for me, by a lot. I use two diodes in series to get enough clipped voltage to move the second tube operating point over most of the possible range. This higher level adds more tube harmonics, but without additional clipping.

This circuit by itself does not give huge distortion and sustain. But when that is wanted, a tube screamer in front works very well.

[Helmholtz]

Nice! So we get pre-emphasis before the clipper and de-emphasis after.

I guess this doesn't allow for a clean sound anymore, so I would make the diodes switchable.

[Mike Sulzer]

Nice!

How's the clean sound?

Thanks!

I do not hear any difference in the clean tone. That probably just means that the component values close enough so that the restoration of the frequency response is reasonably close to the original. Of course it would have to be off by quite a bit to be very noticeable.

Gain and master volume levels are somewhat different for the same clean perceived volume level. I have not done careful tests to see if these changes are as one would expect. This change does reduce the available gain somewhat.

[loudthud]

Ya know how the gain equation for a non-inverting opamp is 1 + ( Rf / Ri ). This means that a replica of the input waveform ( the plus one part ) is added to whatever the ( Rf / Ri ) part of the equation does. In the case of the Tube Screamer, when the signal is large, it just looks like you have a gain slightly higher than one with variable crossover distortion. Your circuit doesn't do that, probably a good thing.

The 700Hz EQ thing is an interesting idea, I'll have to try it if you don't post a sound clip.

[Helmholtz]

In the case of the Tube Screamer, when the signal is large, it just looks like you have a gain slightly higher than one with variable crossover distortion.

I do not recall seeing anything that looks like crossover distortion in the output signal of a TS.
Can you provide a scope pic showing this effect?

In the gain equation Rf lowers with increasing signal due to increasing diode conduction, which means signal compression above some threshold.
Result is a more or less flatter top.
Zero-crossing always happens in the linear range.

[loudthud]

I do not recall seeing anything that looks like crossover distortion in the output signal of a TS.
Can you provide a scope pic showing this effect?

In the gain equation Rf lowers with increasing signal due to increasing diode conduction, which means signal compression above some threshold.
Result is more or less flattened tops.

Try increasing the signal level about 6dB. The squarewave part will get a little squarer, but the tops will grow more rounded. If you turn the pot across the diodes all the way down ( Rf = 0 ), the signal still gets through with no distortion. The emitter follower on the input of the TS might limit the signal, I've never looked at one with a scope.

A similar thing in Mike Sulzer's circuit would be to add resistance in series with the diodes. It might be interesting to add resistance to only one side.

[Mike Sulzer]

Ya know how the gain equation for a non-inverting opamp is 1 + ( Rf / Ri ). This means that a replica of the input waveform ( the plus one part ) is added to whatever the ( Rf / Ri ) part of the equation does. In the case of the Tube Screamer, when the signal is large, it just looks like you have a gain slightly higher than one with variable crossover distortion. Your circuit doesn't do that, probably a good thing.

The 700Hz EQ thing is an interesting idea, I'll have to try it if you don't post a sound clip.

Please do try it; I think you will like it, and you could find some improvements! I was aware that the original signal is added to the distorted signal in the TS, but I decided to ignore that. It is difficult to analyze for a harmonic rich input signal, and I suspect it does not do very much to the sound, but I do not really know that. On the other hand, the engineer might have done it for some very good reason. It would have been possible for the emitter follower input buffer to connect to the RC network that goes to the op amp negative input instead. One possible reason for it is that it could make it easier to adjust accurately low distortion levels since a nearly constant signal level is present and you do not have to adjust the output level much if at all in this case while adjusting the distortion.

[Leo_Gnardo]

I used a similar trick way back when modding amps was much more popular than it is these days. For a nice variation try germanium diodes in your clipper. Gives a softer sounding crunch when they start conducting. Might even put in a switch to select silicon or germanium. My suggestion for the day, over & out.

[Helmholtz]

I think when adding the diodes to an existing amp, the diode clipping threshold should not be lower than maybe 1V to still allow for some clean headroom and sufficient output drive level.

[g1]

I do not recall seeing anything that looks like crossover distortion in the output signal of a TS.
Can you provide a scope pic showing this effect?

I think this is what he means, although I would not call it crossover distortion as such. The sides of the waveform become flat. Of course I may be mistaken and he may be talking about something I am not familiar with.

[Chuck H]

Not a discussion on the incorporation of diodes, but possibly still a useful idea...

Something I do quite often when using a lower value capacitor to trim LF behind the earliest clipping stage is to use a dual ganged pot for the gain control to increase circuit impedance in the early, low capacitance stage as gain is decreased while simultaneously cutting gain in a more linear way on the following stage. The result is more LF when gain is lowered and a tighter sound when clipping. This and the careful use of bypass caps on the pot can get pretty close to a do-all circuit but it can be tricky.

[Helmholtz]

I think this is what he means, although I would not call it crossover distortion as such.

Thanks for posting the wavehapes.
Neither looks nor sounds like crossover distortion (crossover relates to zero-crossing).
Real crossover distortion shows its ugly face with small signals and when the signal decays.
A TS cleans up nicely when the sound decays.
After all players don't use a TS for its waveshape but because they like the sound.

I really like Mike Sulzer's idea with the pre-emphasis/de-emphasis thing in a tube amp.
Lowering intermodulation certainly is highly desirable.
Hope a few will try the circuit.

[Mike Sulzer]

I used a similar trick way back when modding amps was much more popular than it is these days. For a nice variation try germanium diodes in your clipper. Gives a softer sounding crunch when they start conducting. Might even put in a switch to select silicon or germanium. My suggestion for the day, over & out.

Good idea! Maybe use enough in series to match in voltage the two silicon diodes in series for a good comparison. I took a look at some of the physics to see why a germanium diode has a more rounded characteristic. It turns out that that is not so simple to understand.

[Helmholtz]

I took a look at some of the physics to see why a germanium diode has a more rounded characteristic. It turns out that that is not so simple to understand.

I think the main reason is that they have a larger intrinsic series resistance (as can be seen from the slope of the characteristic).

[Mike Sulzer]

I think when adding the diodes to an existing amp, the diode clipping threshold should not be lower than maybe 1V to still allow for some clean headroom and sufficient output drive level.

Exactly. If you make the clipping threshold too low it is hard to adjust for clean. And you do not want to lose that gradual transition between clean and dirty.