Single BJT inverting buffer stage

[pacealot]

So I need to add an inverting buffer stage to the end of another circuit (a fairly ordinary wah pedal in this instance), and I'm assuming (thereby potentially making a donkey out of us all) that the best choice available to me — given the need to make it out of a common-or-garden NPN BJT rather than, say, an op amp — would be to use the same architecture as a phase splitter stage, and simply taking the output only from the collector. But am I missing anything more elegant or "correct"? I've yet to see any other BJT buffer topography that addresses inversion besides the phase splitter, so it looks like that's the best option, but I'm open to ideas and/or correction. Doing some (very rough) math (assuming BJT hFE of roughly 250) has led me to this configuration as my starting point:



I've also noted the phase splitters in other circuits (such as the Fender Blender and the Univox Super Fuzz, for two) and seen that their resistor values are scaled up a bit from this, and again have assumed (running the same risk as previously) that the fact that they follow relatively high gain stages explains a need for a larger voltage swing and thus the recalibration. I don't think that the wah's output necessitates more than a 1V swing for my application, but please disabuse me of my various ignorances if I'm missing something...

[antonis]

The lower the Collector (and Emitter) resistor value(s) the more "stiff"(due to higher current capability) the "inverting buffer" output..

Also, Collector/Emitter resistors low value (hence large Collector current) results into more close to unity voltage gain..
(output impedance more "immune" to next stage impedance loading AND intrinsice Emitter resistor much lower..)
e.g. R_C/R_E = 1k, Collector biased at 7.5V and next stage impedance 50k (say), r_e=10Ω and voltage gain is 1k//50k / (1k + 10) = 0.97..    

P.S.
Drawback is BJT low input impedance (h_FE x Emitter resistor) but here is already dominated by voltage divider equivalent resistance..
(unless you bootstrap it..)

[Rob Strand]

If you don't need the swing or the low output impedance you might be better off sticking with the phase-splitter style design.

If you lower the collector resistor the current draw does increase.  It can increase the noise if the source feeding the inverting stage has a significant impedance.

An inverting BJT stage with one transistor always present number of trade-offs.  Even opamp inverting stages present issues with extra noise.

Just for fun here's some different options without deviating into complex designs.    FWIW, the MXR phase 90 has an inverting and mixing output stage and it uses something along these lines but with poorer swing.

Lower output impedance and more swing over phase-splitter form:


Bit of a hack to show noise can be reduced by removing the 10k.  I didn't re-tweak the gain for 0dB.


You can do an in-between design which uses a much larger value resistor in the place of the 10k but uses the feedback network of the second case.  The raised 10k value prevents the noise increasing and by not completely removing the 10k the bias can be made a little less dependent on hFE.

For completeness, a better balance of trade-offs,

[antonis]

It can increase the noise if the source feeding the inverting stage has a significant impedance.

In such a case, noise shouldn't be the sole issue..
A CE amp with significant(*) gain should be needed..
(*)proportional to (Zsource + Zin) / Zin ratio

[Rob Strand]

In such a case, noise shouldn't be the sole issue..
A CE amp with significant(*) gain should be needed..
(*)proportional to (Zsource + Zin) / Zin ratio

With a low input impedance operating at higher currents reduces the noise voltage.  The noise currents don't contribute.    However when the have a source impedance the noise currents then have an effect, and because the operating current is higher the noise current is higher.

You can see the effect in the three cases below:
Rc=3k3  11nV/rtHz,   Rc=470 ohm low source impedance 4nV/rtHz but with 4k7 source impedance the noise rises to 18nV/rtHz
(The addition of the 4k7 does add a bit of noise in itself, about 9nV/rtHz.   Also the 4k7 forms a divider with the bias resistors - another disadvantage of low output impedance.  We could raise the bias resistor a bit.)

What's pretty clear is this form has less noise than the inverting amplifier forms.  However, the swing is much less than inverting amplifier form.


[Sorry for the typo's. Don't have the energy to fix ATM.]

[antonis]

[Sorry for the typo's. Don't have the energy to fix ATM.]

No offense..

BTW, I was talking about signal amplitude recovery due to Zin/Zsource voltage dividing effect..

[pacealot]

Thank you very much, both of you — this is exactly what I needed to chew upon. I was a bit surprised that there's not been more coverage of these configurations, even in the textbooks from the halcyon days of silicon. But this gives me several things to try without having to order more components. Sláinte, gentlemen!

[Rob Strand]

BTW, I was talking about signal amplitude recovery due to Zin/Zsource voltage dividing effect..

OK got it.  The phase-splitter form is better for higher input impedance but pushing the output impedance down puts a limit on how low you can make it (roughly limited by hFE*Re = hFE*Rc).   The feedback types are going to be too noisy for high impedance inputs.

Thank you very much, both of you — this is exactly what I needed to chew upon. I was a bit surprised that there's not been more coverage of these configurations, even in the textbooks from the halcyon days of silicon. But this gives me several things to try without having to order more components

The gain = -1 case certainly doesn't get a lot of air time.   Occasionally I ponder how to do better but even with opamps the gain = -1 feedback case has noise issues.

[PRR]

The gain = -1 case certainly doesn't get a lot of air time.

It is awkward. As a cathodyne you give up a lot of potential output swing, aside from the high plate/collector output impedance. The OTHER general way is the op-amp inverter, which is better Zout but worse Zin plus an awkward bias compromise. I believe I have done this in a 17 cent audiobox (cheaper than a 19 cent chip) and got paid for the gig. Single transistor stages are generally low accuracy (as you see in the wonky part values and the waveform error).

[pacealot]

I suppose gain = -1 remains somewhat wonky until you get into the world of balanced lines and interstage coupling trafos as found in professional mixing desks!

I was casually musing over all this earlier and wondered briefly if there were any advantages to adding either an additional input or output "proper" buffer stage (standard emitter follower, or FET if needed), but quickly concluded that any advantage would probably be offset by greater disadvantages (most likely even more noise) — plus, that also defeats the purpose of keeping the parts count low and the signal/current path sensible. At that point it becomes time for a dual opamp and then I'm just reinventing a not-very-well-rounded wheel.

[Rob Strand]

I was casually musing over all this earlier and wondered briefly if there were any advantages to adding either an additional input or output "proper" buffer stage (standard emitter follower, or FET if needed), but quickly concluded that any advantage would probably be offset by greater disadvantages (most likely even more noise) — plus, that also defeats the purpose of keeping the parts count low and the signal/current path sensible. At that point it becomes time for a dual opamp and then I'm just reinventing a not-very-well-rounded wheel.

The buffer lets you get a low output impedance and run the inverting stage at a lower collector current.   That helps the knock down the noise of inverting form a bit.  The phase-splitter form doesn't have to run at such low currents to get good noise.  The buffer still doesn't fix the swing problem of the phase-splitter.   You can only fix swing with the inverting form.   You get stuck in a loop toggling between the two 

It's not a bad thing to re-invent the wheel.    You can put a lot of thought and time into problems like this and get nowhere.   Once in a blue moon you might actually find a new path.   You don't get there unless you try and it's really cool when you find something hiding in the cracks!

Like the recent discovery of the monotile.   The guy that discovered it appears near the end of the video, he's quite a humble dude.
https://www.youtube.com/watch?v=A1BhOVW8qZU

[Rob Strand]

Playing with trade-offs.  Trade some swing for noise.   Noise is better than an opamp by 2dB and better than previous trade-off circuit by 4dB.  Simpler circuit.   Less dependency of bias on hFE.    You can see it is quite literally a morph of the inverting circuit and the phase-splitter.

[pacealot]

Nearly two months later, but I can now officially confirm that the above iteration worked an absolute treat in my creaky old wah pedal, now enabling my weird and mildly irritating signal-splitting fuzz/wah scenario for bass to be perfectly in phase in all configurations (less on that anon, as it would bore everyone senseless, especially me). Quiet as a church mouse as well. I ended up using an old BC184L that I had to hand, as it had a nice high hFE (well over 600!) and the ECB pinout made for an easier layout somehow. I'm sure the ubiquitous 2N5088 would've been ideal as well.

Thanks once again, Rob. This one goes into the "keep this, it's terribly useful" file...

[Rob Strand]

Nearly two months later, but I can now officially confirm that the above iteration worked an absolute treat in my creaky old wah pedal, now enabling my weird and mildly irritating signal-splitting fuzz/wah scenario for bass to be perfectly in phase in all configurations (less on that anon, as it would bore everyone senseless, especially me). Quiet as a church mouse as well. I ended up using an old BC184L that I had to hand, as it had a nice high hFE (well over 600!) and the ECB pinout made for an easier layout somehow. I'm sure the ubiquitous 2N5088 would've been ideal as well.

Thanks for the feedback.

I'm embarrassed to say the project I was working on then has been moved down the list two projects

[pacealot]

I'm embarrassed to say the project I was working on then has been moved down the list two projects

I am reliably informed that that can have a distinct tendency to occur... 

[Rob Strand]

am reliably informed that that can have a distinct tendency to occur... 

I was wondering if we can step this circuit up one level and use it as a mixer.   Used as an alternative to an opamp, or inverting transistor, or passive mixer.