Troubleshooting dull pulse wave from a PNP circuit

[BenJBX]

Hi everyone, it my first post on the forum. Looking forward to contributing!

I've been going through my collection of synths and studio gear this year, selling off quite a bit of stuff and trying to fix things that are broken. I used to get my synths looked after by Andy from EMIS in Bristol but since he retired I've been doing more myself. I've got a degree in electronic engineering from 15 years ago so a lot of it is in my head somewhere but it often takes a good rummage to find it!

I have a Syntecno Teebee, which for those who don't know is a terrific, ballsy 303 clone from the 90s. It has not sounded right for several years, the sound was weak and it wasn't tracking MIDI pitch well. Robbie from Syntecno still answers emails via the website and he pointed me in the direction of a few known and documented issues with transistors in the oscillator circuit. I've fixed those, recapped everything and also replaced the external PSU which was outputting 16VDC not the 12VDC it was rated at.

This has all improved things massively, but the one issue that's still kicking my ar$e is that the pulse wave sounds dull in the lower octaves, so that when you crank the resonance the sound disappears because there's no higher harmonics for the resonant filter to emphasize.

The circuit that generates the pulse from the sawtooth is shown below. It looks simple but there are hidden depths. I'm also finding that most DIYers use op amp comparator circuits to generate pulse waves so there's not much info on PNP circuits..



The story so far ... I have confirmed the rail voltages, replaced the PNP and both capacitors, and checked all the resistor values in situ. The filter is fine on the sawtooth and if I put a raw pulse wave from my modular into the Teebee's external input it sounds great, so it points to the problem being in this circuit. It's not an issue with PNP switching speed, because the higher octaves are OK.

Here's the trace of the pulse wave playing E1...



and here's what the PNP base is doing ..



After thinking it over, my guess is that the slew on the rising edge of the pulse wave is because the PNP is not switching on cleanly. The RC network on the base pin is designed to sink the emitter-base current when the PNP is on, and when it's off the discharging C10 sums with the falling sawtooth wave to create a steep ramp to turn the PNP back on. This works well at higher octaves, but here C10 has discharged fully before the emitter-base voltage is established (the impulse spike in the blue trace above), and the sawtooth ramp is too gentle at this frequency to activate the PNP like a true switch. Hence the slewed waveshape and consequently the attenuated higher harmonics.

I'm confused though because the circuit is working as it was originally intended as far as I can tell, yet something is clearly and audibly not right. Has anyone ever experienced transistor based pulse oscillators behaving like this, or can you point out if I've missed something in my analysis?

I'm also going to go back to Robbie to see what he thinks, but any tips on a way forward would be helpful. Thanks!

 

[Voyager10]

A bit more info on those scope traces would be useful, like horizontal and vertical scale, what the DC levels are, and also what the input to the circuit (R35/C10 junction looks like).

If you have a 2ch scope, pairs of traces would be good.

What I don't get here is that the time constant of R35/C10 is 1ms, which is bang in the middle of the audio range. It is certainly going to change behaviour between say 300Hz and 3KHz.

 

[BenJBX]

A bit more info on those scope traces would be useful, like horizontal and vertical scale, what the DC levels are, and also what the input to the circuit (R35/C10 junction looks like).

If you have a 2ch scope, pairs of traces would be good.

What I don't get here is that the time constant of R35/C10 is 1ms, which is bang in the middle of the audio range. It is certainly going to change behaviour between say 300Hz and 3KHz.

Hey, thanks very much for taking the time to reply, I really appreciate it!

Good points re the missing info on the scope traces and the input to the PNP base RC network. The R35/C10 junction is connected directly to the output of the sawtooth oscillator, but you're correct that I hadn't made that clear in the OP. The base pin trace I took originally had the sawtooth input as a second trace but I removed it to make it clearer. I also trimmed all the scaling info off the screenshot when I cropped it which with hindsight was not the best idea. I will make sure I include more pertinent info when I post traces to the forum in future, so thanks!

I made some progress with this problem last week by recreating the circuit on the breadboard and playing with different component values. I quickly found that increasing the cap value sharpened the square wave rise time a lot. In fact removing the C10-R34 leg altogether was a marked improvement. I got a rise time I was happy with, but the duty cycle had drifted away from 50/50 by this point so I then started working on R45 and R36 values to change the PNP biasing. All of this was pretty iterative with not much in the way of design/analysis, but I eventually came up with these values that fit the bill:

R45 = 10k, R36 = 22k i.e. R45 and R36 swap values
R34 = 22k
C10 = 680nF - a whopping rise this, partially dictated by what was available in the spare caps box. A smaller one would probably have done just as well but I didn't have much available in the 10nF - 500nF range.

The original C10 was nonpolar. The new C10 is polar and worked best with the anode on R35 and the cathode on R34.

For those interested, here's the AC coupled trace showing the ramp of the original circuit (blue) vs the same circuit on the breadboard with the revised component values (yellow).

 

[Voyager10]

So I guess that's a now audible increase in higher harmonics?

The other factor that could affect this is the gain of the transistor. On the rising edge, the dV/dt will be R36 x gain x d(Ib)/dt, where Ib is the base current (determined by the input network and Q8's base-emitter diode). A higher gain transistor would give better rise time, other things being equal.

 

[BenJBX]

So I guess that's a now audible increase in higher harmonics?

The other factor that could affect this is the gain of the transistor. On the rising edge, the dV/dt will be R36 x gain x d(Ib)/dt, where Ib is the base current (determined by the input network and Q8's base-emitter diode). A higher gain transistor would give better rise time, other things being equal.

Sorry, I've been away for a few days so I missed this reply. Yes, the sound is audibly way better because once you start cranking the resonance there are now harmonics present for the filter to emphasise. There's also more 'boom' with the resonance down because the duty cycle is closer to 50/50.

You may well be onto something with the transistor gain. The original circuit specced a 2SA733 PNP which aren't made any more, so I replaced it with a BC212. Published hfe range is 90-600 for the 2SA733 vs 60-400 for the BC212 which fits your theory, and they're big ranges so if you replaced a high gain 2SA733 with a bit of a duff BC212 the difference in gain could be approaching an order of magnitude.

It's definitely a puzzle that the circuit appeared to be functioning as intended before the component value changes but the sound was audibly not right. Another friend suggested checking for ripple on the supply rails but they were clean and stable, so that was a dead end.

I'm just really glad I got it fixed. Any 303 emulator needs to sound good with the resonance way up otherwise what's that point?