Runoffgroove UBE Screamer CD4049UBE mod - why is the clean signal path inverted?

[MikeA]

I recently put together a classic Runoffgroove UBE Screamer CD 4049 UBE TS variant.  I noticed that the clean signal coming into the mix point at U1E from U1A was inverted, while the effected signal coming in from U1D was not inverted.  I'm not sure why it was designed that way, but mixing these two has a significant impact at low gain levels.  The original design partially cancels the effected signal with the opposite-polarity clean signal at low gain, and can distort the waveform and make a significant cut in level (up to 12 dB depending on the value of the clean signal mix resistor Rx.)  At higher gain, the effect is smaller and more subtle, more like a 2 dB cut, since the effected signal swamps the clean signal (again depending on the value of Rx.)   I tried inverting the clean signal by using the vacant sixth inverter on the CD4049 to make it the same polarity as the effected signal at the mix point, and I think it's an improvement.  This mod is shown on the schematic in the red circle, and is independent of other useful mods like making Rx variable and/or making the 330R supply resistor variable.  It also allows rethinking the value of Rx, I believe it could be larger than 470k or smaller than 100k, depending on how much clean signal you want to blend in.  I haven't tried that yet, although Rx does sound good to me around 250k - 300k with underwound humbuckers as the original schematic notes suggest.  You could also vary the feedback resistor on U1F (the clean inverter stage) to change the clean signal level, instead of varying Rx.  Or get ambitious and make Gain a dual-gang pot controlling both the effected and clean signal path levels via the feedback resistors for U1B and U1F.  At any rate, it's not a difficult mod if you want to hear it and decide for yourself the next time you make one of these.   

[teemuk]

Quite frankly put, great number of Runoffgroove "designs" often give me an impression of "painting-by-numbers" or "cutting and pasteing", where they have created something merely by copying parts from other circuits with only vague impression of how tgose things work. Seldom there seems to be any more designing involved than that, in real sense of the word. The end result is more or less random due to that.

That they fail to consider signal polarity doesn't surprise me at all. Many of their circuits are copied in a way that obfuscates the original circuit ideas and operation. There are trends like cloning triode circuits with no other modifications but FETs replacing the tubes, substituting opamps with linear MOSFET inverters, and misproper use of feedback loops. Sometimes this works decently by accident, sometimes it works less decently. The design goal is often accomplished when the circuit simply emits sound.

Who knows, they might have glanced at a similar circuit that does the signal mixing differentially with an opamp with differential inputs.

[anotherjim]

In the regular TS, the clean signal always gets through x1 plus from the addition of gainy/clipped signal. So the clean is affected by the tone control.
It might work to just place the 100n cap and mix resistor from the input stage pin6 direct to the junction of the 1n5 in the tone section. The final inverter can be deleted. Only x4 inverters are required.

[teemuk]

In the regular TS, the clean signal always gets through x1 plus from the addition of gainy/clipped signal. So the clean is affected by the tone control.

In TS the FET switches mute clean (bypassed) signal path when the distortion signal path is selected.

If you talk about clipping diodes in non-inverting amp feedback loop you have:

1) clean signal below clipping diode threshold (at very low input signal amplitudes)

2) distorted signal, that is gain compressed (or "soft clipped") at amplitudes above the diode threshold and below a certain threshold of input signal amplitude

3) distorted signal, now distorted by crossover distortion, as increased input amplitude starts to extend gain compression from signal peaks to ever larger portion of the signal. Now relatively very high gain at small signal input levels distorts the waveform.

4) above plus peak clipping to opamp rails

Generally we wish to confine operation to modes 1. and 2. Extensive shifting to mode 3. no longer serves the purpose of soft peak compression. Clean signal we get through only in mode 1.

In general, whenever we add harmonics of distortion to fundamental frequency/ies we can no longer have "clean" signal with just mere fundamental frequencies in it. Clean and clean+distortion are therefore mutually exclusive.

[anotherjim]

Have to disagree Teemuk. A non-inverting opamp always has x1 clean present on the output. If you accept that the 4.5v DC reference applied to the opamp +input produces +4.5 x1 at the output, then, when the input signal modulates the DC reference the signal must be included x1 at the output. The same applies to any non-inverting single opamp circuit.
The JFET switching in the TS is only doing bypass work and not clean blending.
The Tone section has one preceding fixed low-pass at 720Hz which will affect all signal from the clipper but the Tone opamp will pass the result at x1 too no matter what the Tone pot is doing. Probably not worth replicating this in the inverter version

Voltage gain=(Rf/Rin)+1. The +1 is the +input signal contribution.

[soggybag]

No that you point it I have been hearing the out of phase sound. I built a couple UBE Screamers and always noticed something about the sound. Now you point it out, it's out of phase mix that I was hearing!

[soggybag]

I had a few boards had made for the UBE Screamer. I built another up with the non-inverting mod. Here is a demo of both versions. I added a trimmer for the 100k - 500K mix resistor. It set to about half way in the first pedal. The other was already built I can't remember what the resistor was there.

You can hear a difference. It's almost like an amp/cab sim. The inverting signal is almost like a feature more than bug.

Also it doesn't sound so pronounced, I'm guessing because there is some phase offset going on in the clipping and tone stage.

[soggybag]

Here's the schematic I used:

[Vivek]

Quite frankly put, great number of Runoffgroove "designs" often give me an impression of "painting-by-numbers" or "cutting and pasteing"

My comments on Runoffgroove MOCKMAN and some comments by Teemu K

https://www.diystompboxes.com/smfforum/index.php?topic=126904.0

[Ben N]

I recently put together a classic Runoffgroove UBE Screamer CD 4049 UBE TS variant.  I noticed that the clean signal coming into the mix point at U1E from U1A was inverted, while the effected signal coming in from U1D was not inverted.  I'm not sure why it was designed that way, but mixing these two has a significant impact at low gain levels. 

Isn't that what a Tubescreamer or other non-inverting opamp soft-clipper does? Clipping takes place in the feedback loop which is fed through the (-) input, i.e. out of phase with the clean signal through the (+) input. Come to think of it, at minimum gain, where the clippers are essentially bypassed, you would think there would be some phase cancellation, although, because it is a non-inverting configuration, it can't go below unity. As the gain goes up, more of the signal goes through the soft clippers into the (-) input, but that doesn't cancel because it is distorted. i.e. different from the clean signal.
Or am I completely garbling everything up?

[ElectricDruid]

Clipping takes place in the feedback loop which is fed through the (-) input, i.e. out of phase with the clean signal through the (+) input.

I'm not sure about this bit. The way I think about it is:

A basic rule of op-amps is that they hold their two inputs at the same level. Thus if we have a clean signal on the +ve input, and we have 100% feedback from output to -ve input, the op-amp will drive the output so that the -ve input is the same as the positive - e.g. it will be a unity-gain copy of the input signal. It's in-phase, as you'd expect for a non-inverting amp.

If you now think about what happens when there's a voltage divider hanging off the output (the usual gain network) you can see where the gain comes from. Again, the op-amp will drive the output such that the two inputs are identical. If the voltage divider is 10K+10K (for example) we only get half the output signal going back to the -ve input, so the op-amp must drive the output twice as loud to make the two input equal. E.g. now we have gain = x2.

Hope this helps. It's extremely easy to get tangled up in feedback paths, since everything tends to be back-to-front, but this way of thinking about it helped me get it clear in my head. Perhaps it will help you too.

[Ben N]

Still not quite there, Tom. It is a negative feedback loop, right? Like all negative feedback loops, it reduces gain by mixing an inverted copy of the output with the input, right? To put it another way: It's a differential input--which means it subtracts, right?

[ElectricDruid]

Still not quite there, Tom. It is a negative feedback loop, right? Like all negative feedback loops, it reduces gain by mixing an inverted copy of the output with the input, right? To put it another way: It's a differential input--which means it subtracts, right?

That's why I was avoiding talking about whether it's negative/positive/whatever, and the differential inputs and everything else. When I try to think about it like that, I get lost, or come out with the result inverted from what it should be. The advantage of the "the two inputs must be the same" method of thinking about it (for me, at least) is that it avoids me having to think about those things. If I stick to that one rule, I can work out what's going on without getting tangled up and without getting it wrong.

That said, there's lots of ways you *can* think about how these things work (and none is more "right" than another - they're just different ways of viewing the situation) so it's a question of finding one that makes sense to you.

Let's have a look at it from the "differential inputs / negative feedback" point of view and see where we get...

So we stick an input signal into the +ve input, so we get a non-inverted output coming out. Now, some of that signal is fed back to the -ve input. Since the output is in phase with the +ve input, the signal at the -ve input is as well. That *reduces* the differential input signal (that's your negative feedback action) and hence limits the gain.

That's pretty hand-wavy, but does that help at all? The feedback is in phase, but because the inputs are differential, in-phase feedback to an inverting input is negative feedback.

[soggybag]

The original question was about the clean signal coming from the first inverter stage being inverted with a non-inverted distorted signal coming from the clipping and tone stages.

I built both circuits and tested them. I think I hear some of the phase cancel happening. but to my ear it's not strong. I'm guessing because there is some phase shift happening in the clipping and tone stages.

I'm wondering if people think the inverted signal mixed with the non inverted is a problem or a feature?

[MikeA]

I think it sounds better with the mod, IMHO.  The difference is much more apparent at low gain levels, in your video you were running gain @ 70% and the difference is relatively small up there.  Try it with the gain knob @ 9:00 or so.   

[ElectricDruid]

+1 agree with Mike.

The effect will be worse at low distortion levels, where you're basically cancelling most of the "body" of the signal. Adding more distortion will mask that, but it looks like a problem not a feature to me.

[Ben N]

Let's have a look at it from the "differential inputs / negative feedback" point of view and see where we get...

So we stick an input signal into the +ve input, so we get a non-inverted output coming out. Now, some of that signal is fed back to the -ve input. Since the output is in phase with the +ve input, the signal at the -ve input is as well. That *reduces* the differential input signal (that's your negative feedback action) and hence limits the gain.

That's pretty hand-wavy, but does that help at all? The feedback is in phase, but because the inputs are differential, in-phase feedback to an inverting input is negative feedback.

Right, ok. The signal at the (-) input is the same phase as the (+) input. But the differential input inverts it. In the simplest case, if there were no clippers or bypass cap, decreasing the resistance inside the loop would increase the signal at the (-) input, thereby bringing the gain down to unity (in the inverting configuration, bringing the gain down to zero). So, yes , the feedback signal isn't inverted until it gets inside the opamp, but it does get inverted.

That said, I agree with Mike that it sounds better with the mod, at least on the basis of Soggybag's video. But I don't think ROG was just "painting by numbers" here -- it looks to me like feedback loop clippers work by subtraction, just that subtracting a clipped signal from a clean signal doesn't work the same was as subtracting an identical clean signal, and that's what they were trying to replicate with the non-inverted clean signal.

[teemuk]

The effect will be worse at low distortion levels, where you're basically cancelling most of the "body" of the signal.

Yes. Below clipping (distortion) threshold it's basically two clean opposite phase signals cancelling each other.

Adding more distortion will mask that, but it looks like a problem not a feature to me.

The more there is distortion then naturally the more there are harmonics that won't be "cancelled" thus. And yes, IMO, it's an amateurish design and could sound better. But hey, it does produce some distortion so objective achieved.

[Vivek]

The effect will be worse at low distortion levels, where you're basically cancelling most of the "body" of the signal.

Yes. Below clipping (distortion) threshold it's basically two clean opposite phase signals cancelling each other.

Adding more distortion will mask that, but it looks like a problem not a feature to me.

The more there is distortion then naturally the more there are harmonics that won't be "cancelled" thus. And yes, IMO, it's an amateurish design and could sound better. But hey, it does produce some distortion so objective achieved.

Below clipping

I suppose that the signals from the two routes won't be exactly out of phase

I would expect a comb filter rather than exact total cancellation

[Radical CJ]

I had a few boards had made for the UBE Screamer. I built another up with the non-inverting mod. Here is a demo of both versions. I added a trimmer for the 100k - 500K mix resistor. It set to about half way in the first pedal. The other was already built I can't remember what the resistor was there.

You can hear a difference. It's almost like an amp/cab sim. The inverting signal is almost like a feature more than bug.

Also it doesn't sound so pronounced, I'm guessing because there is some phase offset going on in the clipping and tone stage.

It's definitely an audible difference in your video. But not unpleasant sounding. In fact, pretty good for indi rock, or to get that out of phase Fender Mustang sound. The fact it isn't mentioned in the ROG write up about the design though makes me think it was unintentional.