Why can't BJTs be used instead of JFETs in phasers?

[Rob Strand]

I have a Univox U-65 amplifier that uses a phase shift oscillator driving a BJT as a tremolo, working against 68K resistors from the input.  It works and although it is difficult to hear non-linearity in a tremolo, it should be suitable for a phaser.

One aspect of the non-linearity is the distortion, which seems to be like JFETs without linearization.   The other is the transistor no longer behaves like is resistor for one of the polarities.  It kind of 'lets go' which would weaken the effect.   

IIRC the 'letting go'  is soft in that it transitions from about 50mV pk to about 300mV pk.   These are off the top of my head values as different configurations I tried had different limits.

IIRC there's two details which help the phaser:   When you look at two stages at a time the low-frequency inversion of the all-pass helps the cause.  It's possible to get some cancellation of the LFO feed-through.  If the first all-pass 'let's go' due to high signal level the second stage will not.  So it does maintain some phasing effect at higher inputs.

For guitar and bass it looked feasible.  For a high-output keyboard maybe it needs some pre-attentuation and post-amplification (with a small amount of pre-emphasis/de-emphasis in the upper freqs.  No too much for keyboards.)

[Rob Strand]

I've seen those, and those are trace schematics. I had a bit of a hard time following them because of how they're broken up. I was using a factory schematic, which I found had a few little differences compared to Mark's unit, so I penciled them in (so to speak). I also cleaned it up so it's readable (open in a new window to see it all).

Just passing casual eye over that schematic, without writing out equations, to me it looks like the transistors are still being used as controlled resistors.   No different to the method I used, which is shown in the *second* Hoshuyama snippet.    Notice the low-impedance return paths on the emitters to avoid LFO currents getting into the audio.

The thing that is a little weird is the NPNs are low resistance when the LFO is positive and the PNP are low resistance when the LFO is negative.   Each BJT type operates on alternate polarities of the LFO.   What this is doing it making the control the absolute value of the LFO waveform.      Maybe the idea is to approximate the skewed LFO idea used on the second revision Ross phasers.   IMHO it won't achieve a result as good as the Ross.   It will definitely give the LFO a different character - perhaps that's what you don't like.    The other thing is does it doubles the apparent modulation rate.

When I was playing around with mixed NPN and PNP my aim was make the non-linearity more symmetric and to remain in control for larger signals but the Peal is using the two BJT polarities in a different way.

[PRR]

> Univox U-65 amplifier that uses a phase shift oscillator driving a BJT as a tremolo, working against 68K resistors from the input.

An older Univox from the same designer used a 12AU7 twin-triode, back-back, as the VCA, after a gain stage.
https://el34world.com/charts/Schematics/files/Univox/Univox_u_1226_lead_amp.pdf

It was fairly fussy about that tube. Not all of "the right type" worked.

[digi2t]

To be honest, I'll have to plug it back in and listen to it to define what exactly isn't right with it, but IIRC the phasing sounded like shite, and the speed control was way too fast. I'll dig it out tonight and report back.

IC15 looks quite oscillatey, according to the other site schematic, R99 is 470k, which should slow it down quite a bit.

Alrighty-o, finally had time to fiddle with it. As scruf called it, changing R99 to 470K did the trick. Looking at the other schemo, I also noticed that it called R102 as a 470K, but 470K here stalled the LFO completely. Went back to the 47K and all is good with the LFO now.

So...

The only thing left to solve now is the phase sweep. With the manual and depth maxed, with minimum slow speed, the phase seems to go too high and too low. In the low end, it seems to hang for a while before coming back up. Then, when it gets near the top, the audio changes to a hissing sound, which lasts until the phase starts to come back down.

Turning down the manual pot doesn't help, but if I turn down the depth, or speed up the slow LFO, it goes away. It's as if at max slow speed, it surpasses the top of the phase sweep, and goes into a hiss fest. The trimmer does seem to dial it out either.

Now, I have noticed a couple of discrepancy between the factory schemo and the one at the other forum, C36 (0.47uf electro) polarity is flipped. I've tried it both ways, and I don't notice any difference.

Also, R76 on the other schemo sits atop R71, and not where it's shown in the factory schematic. This kind of makes sense to me, since positive voltage to the base would be provided by the RV2 trimmer. I switched the resistor over to the other base, and effectively, I can hear the trimmer adjusting the center of the manual setting.

The hunt continues....

[Scruffie]

So your description of the sweep "hanging" and especially that turning down the depth solves it says to me the sweep is clipping somewhere, not an uncommon problem, question is... where?

The LFO output could be too big (the 470k for R102 would reduce its gain but apparently, too much) it could be the transistors in the phase stages don't meet the necessary spec or it could be in the current control for them.

R76 makes sense to me too above R71, the whole manual/lfo/trim section is various competing offset voltages and I can't see what benefit it would have there as it would just be in parallel with RV2 and the manual control.

Gut shots to check resistor values would be handy, do you have any?

[Mark Hammer]

Turning down the manual pot doesn't help, but if I turn down the depth, or speed up the slow LFO, it goes away.

This is why, historically, we tended not to see many 4-knob phasers and flangers.  It was too easy to "mis-set" them.  Both MXR and EHX would issue pedals that had a single control knob, or else were simple and foolproof enough to operate that end-users wouldn't misinterpret dysfunctional settings for "lousy pedals".

[ElectricDruid]

Does the LFO design produce equal amplitude at all frequencies? Some don't. That can actually be handy for things like chorus where you don't want so much depth as you increase the frequency (since pitch shift is related to rate of change).

[digi2t]

Does the LFO design produce equal amplitude at all frequencies? Some don't. That can actually be handy for things like chorus where you don't want so much depth as you increase the frequency (since pitch shift is related to rate of change).

I'm not sure if that is how it's designed. All I know is that at very slow speed (from 0 to about 2 on the slow pot), the width of the sweep seems to be way too wide. Unfortunately, I don't have Mark's unit on hand to compare, but I do remember that the sweep was symmetric, no matter what the LFO speed was.

[Mark Hammer]

If you can provide me some clues about where to look, I can try and measure/scope it.

[jatalahd]

Related to BJTs as voltage controlled resistors, I recently built the Kay T1 Tremolo on a breadboard to try it out (schematic is easy to find from Internet). In the Kay Tremolo there are 2 NPN BJTs connected together and used under identical control voltage. Directly from that idea I extracted the following voltage divider simulation circuit for testing:

   (I guess you need to click the image to see it better)

The input signal is 1 V sine and the control voltage is DC source that is sweeped from 0 volts to 8 volts. I changed the 68k RB resistor to 1 megohm since that was more useful in this voltage range. The 68k is more sensitive and reaches its limit already around 2 volts. The transistors are basic BC547B, which I also used when successfully building the Kay Tremolo.

The simulation output looks like this:



The transistor combo seems to behave as a relatively linear resistor up to 4 V control voltage in this case (midway of the simulation). After that the resistance drop is not that prominent anymore.

Although it seems quite promising, in the Kay Tremolo it slightly passed the LFO control voltage to the audio path. But other problems I did not really notice.

[Rob Strand]

Although it seems quite promising, in the Kay Tremolo it slightly passed the LFO control voltage to the audio path. But other problems I did not really notice.

The issue with that set-up is the base current flows through the emitter into the audio path.   That make the feedthrough much worse than it can be.

Try this,
- Use a PNP and NPN in parallel with the emitters to ground, collectors together.
  Keep the bases separate with separate resistors, like you already have.
- create a second LFO output which is inverted from the first LFO output
- connect the first LFO to the NPN base via a resistor
- connect the inverted LFO to the PNP base vias a resistor.

You can tweak the PNP base resistor so the NPN and PNP have the same attenuation.

The main point of this configuration is the base current goes to ground through the emitters and not into the audio path.

There's quite a few variations you can play with.   I spent ages tinkering around with these things.