Differential Fuzz Design

[antonis]

I'd just like to add that there's a difference between "not reading thoroughly" and "asking for clarification when you don't understand something."

True and Correct..!! 

[mdcmdcmdc]

As for that "notorious" CFP" (Sziklai) buffer, make 10k upper voltage divider bias resistor a 82k one (or 100k by also making lower 120k resistor a 150k one) and make 150k bootstraped resistor a 47k  one..

The former for a more healthily biased Emitter follower (symmetrical undistorted swinging ability) and the later for more stiff bias configuration (more current through bootstraped resistor + voltage divider Thevenin equivalent one)

Just going back through these suggested changes and realised that the 10K upper is a typo in the schematic - I actually have a 100K on the breadboard. My mistake—I'll make the other swaps you suggest and see what happens.

To the point about the 100R - I tapped the output after the BPF section as-is, and as y'all kindly pointed out there is indeed some slight distortion happening when I hit a string quite hard. It's maybe not as obvious as I was expecting, but it's definitely there. I'll keep tweaking and see how it goes. Thanks again to the folks chiming in!

[mdcmdcmdc]

Ty to ianpunk and antonis for the help in dialing in the buffer—it's starting to sound pretty ok.
I added some more LPF-ing between stages and made a few other small tweaks.

[antonis]

I added some more LPF-ing between stages and made a few other small tweaks.

Too many H/LPFs..
(many interacting poles..)
e.g. 10nF/220K 72Hz HPF is followed by 100nF/100k 16Hz HPF which has no meaning..
Higher cut-off frequency HPF comes after lower one so the higher filter's slope gets more "stiff" - 12dB per octave due to 2nd order formed filter..

A Tone control should be more easy and effective..

[iainpunk]

i partially agree with antonis, some/most high pass filters are below audible range. on the other hand, low pass filtering before clipping will greatly affect the chaotic brightness of the tone, and a normal tone control after distortion will lessen the amount but not the character of the brightness, however, those 10nF caps after the 100nF caps might not be the best solution. i'd place the 10 nF cap in place of the 1n or 470p caps.
also, tune the low and high pass filters by ear instead of theoretical numbers.

if you want you could try adding feedback to the gain stage to shape its frequency response, for instance use a extra transistor and a bridge tee filter in the feedback loop to determine the gain as a ''bell shape'' bandpass boost. the one here is centered around 500-ish Hz

this is a proposed experiment, not a recommendation, and changing the values can greatly alter the depth and center frequency of the filter.

cheers

[mdcmdcmdc]

I pulled some of the filtering out and to ianpunk's point, I think there were two things going on that were beneficial:

1) LPF-ing before before a gain stage taking away the unpleasant high frequencies, minimizing string wipe, etc
2) some level of volume/voltage attenuation between gain stages

Also, with regard to the HPFs, the only one that was intentional was the first one as part of the BPF after the buffer. Any other HPF-ing in the circuit is totally unintentional.

That said, with the extra caps and resistors pulled, it's a chaotic mess.

Again, clearly I'm a bit out of my depth here but I'm trying to parse this out... If I switch the coupling capacitors to 22n that's going to form a 72Hz HPF with the 100K base-ground bias resistor in the differential stages, correct? If so, I can effectively drop the BPF from the beginning and not really worry about it.

So would something like this be a step in the right direction?

[mdcmdcmdc]

There's also the matter of those two 10nF caps coming off the base to the ground - I have no idea what they're actually doing but they're doing *something* helpful with regard to the high frequency noise. I added them when I was trying to chase down some HF whine and honestly was just randomly bridging points to ground with a cap to see what would happen. I just sort of stumbled on it.

[mdcmdcmdc]

I swapped the final cap for a 1uF and still need to tweak the TBD resistor and cap values, but it seems to be working and sound... pretty good?

EDIT: 100K/1nF for the first filter and 33K/4.7nF for the second is nice, and apparently I'm now live blogging my resistor choices.

[iainpunk]

There's also the matter of those two 10nF caps coming off the base to the ground - I have no idea what they're actually doing but they're doing *something* helpful with regard to the high frequency noise. I added them when I was trying to chase down some HF whine and honestly was just randomly bridging points to ground with a cap to see what would happen. I just sort of stumbled on it.

they should be seen together with the output impedance of your fancy input buffer.
is has about 3.3k output impedance, and with that 10n it forms a low pass with a cutoff at about 5kHz-ish

some components labeled TBD is that ''to be determined''? or is that the company i work for?
(just kidding, it just reminded me of my work, ahah)
but that TBD resistor between the two diff. stages already interacts with the 10nf goint to the 2nd stage's base, so you can leave out the TBD capacitor there.
also, idk if the software you use permits that but if possible, number the components, like R1 R2 C1 C2 Q1 Q2 etc. so talking about specific components will be more easily done.

nothing wrong with live blogging component values.

cheers, Iain

[antonis]

If I switch the coupling capacitors to 22n that's going to form a 72Hz HPF with the 100K base-ground bias resistor in the differential stages, correct?

Partially, yes..
(and partially no..)

22nF cap forms a HPF with 100k//180k//2*h_FE*(25k_set+1k5)

P.S.
Plz delete 10nF caps after 22nF ones.. They serve nothing but (unwelcome..??) capacitive voltage dividers..
(unless you intend  to form LPFs with previous stage output impedance - in such a case you better move them right after respective Collectors..)

[antonis]

they should be seen together with the output impedance of your fancy input buffer.
is has about 3.3k output impedance,

Despite its "fancy" configuration, it is an Emitter follower..
Output impedance is 1/gm = re = 0.026/I_CQ..
(actually, it's a bit lower due to parallel combination of 3k3//150k//100k ..)

[mdcmdcmdc]

Ah, got it - I wasn't sure if the 22uF cap between the resistor and the cap to ground would inhibit the two components from forming a LPF. That's very helpful, thanks.
So it seems like a huge rolloff in the high frequencies really helps before the signal starts getting slammed.

I did some more messing with the filtering and gain staging. I moved the gain/distortion control from a variable resistor on the emitter of the first differential pair to a simple voltage divider after the buffer, it seems to give a wider range and retains a bit more of the high end when it's rolled off (although it still gets pretty wooly at the lower end). I think it's in a good spot for me to leave it for a few days—thanks again for all the help and feedback on this. I've been learning a lot.

[antonis]

Could you plz enlighten me about the purpose of C5/C6 50% voltage divider..??
(don't deal with differential pair bias configuration at the moment..)

[mdcmdcmdc]

Based on Iainpunk's post above, my understanding is that C6 forms an LPF with the output impedance of the buffer.

The circuit doesn't function without C5.

[iainpunk]

Based on Iainpunk's post above, my understanding is that C6 forms an LPF with the output impedance of the buffer.

The circuit doesn't function without C5.

yes but no

those DC blocking caps are to small, that's what i understand from Antonis.
also, the frequency cutoff changes with the gain control, turning darker at lower gain and brighter at higher gain, normally the opposite is desired.

they should be seen together with the output impedance of your fancy input buffer.
is has about 3.3k output impedance,

Despite its "fancy" configuration, it is an Emitter follower..
Output impedance is 1/gm = re = 0.026/I_CQ..
(actually, it's a bit lower due to parallel combination of 3k3//150k//100k ..)

well... when pulling up it has extremely low output impedance, when pulling down, its 3.3k, since those transistors can't take current down. this gives asymmetric drive to capacitive loads like filters, but in practice and my experience, 3.3k gives accurate frequency cutoff points, albeit with a less steep slope than a conventional filter, and some ''odd order'' distortion for frequency's above the cutoff point.
funny, we have had this same squabble 2 times before IIRC

cheers

[mdcmdcmdc]

Based on Iainpunk's post above, my understanding is that C6 forms an LPF with the output impedance of the buffer.
The circuit doesn't function without C5.

yes but no

those DC blocking caps are to small, that's what i understand from Antonis.
also, the frequency cutoff changes with the gain control, turning darker at lower gain and brighter at higher gain, normally the opposite is desired.

That's exactly what's happening - the overall range of gain works well, but the 100K pot is creating a variable LPF with C6. I was getting to the end of my ability to focus at that point, so that was definitely a "look at this with fresh eyes" issue. Is the fix as simple as moving C6 to directly after the buffer, before C4?

As far as the DC blocking caps go, pardon the basic question but... too small to do... what? Block DC? Is this a function of the output impedance of the buffer?

[iainpunk]

to small in relation to the filter cap after it.
if two caps are forming a string like that, and the capacitors are equal, it halves the signal over the full band width.
i'd say place this filter cap parallel to R6, straight form the emitter of the sziklai pair.

also, smaller cap lets through less bass. might lose some fatness there.

cheers

[mdcmdcmdc]

That all makes sense - I moved C6 to the emitter/collector junction of the sziklai pair and it solved the issue with the high end rolling off with the gain being reduced. I'll need to go back and readjust the attenuation between stages now as that accidental voltage divider was actually helping things. Every change cascades...

[iainpunk]

Every change cascades...

that what makes it fun!

cheers

[antonis]

well... when pulling up it has extremely low output impedance, when pulling down, its 3.3k, since those transistors can't take current down. this gives asymmetric drive to capacitive loads like filters, but in practice and my experience, 3.3k gives accurate frequency cutoff points, albeit with a less steep slope than a conventional filter, and some ''odd order'' distortion for frequency's above the cutoff point.

True and Correct for a sigle BJT Emitter follower..
(that's exactly the reason for suggesting (enough times) Emitter to be biased higher than power supply mid rail - although, a current source instead of Emitter resistor should be the ultimate solution..)

Nevertheless, this particular CFP configuration significally improves linearity by increasing negative-feedback factor via increasing the open-loop gain..
(it can even give voltage gain, by splitting 3k3 into 2k2 and 1k, say, and place first BJT Emitter on resistor junction, due to output and feedback point been no longer inherently the same - that's another story which OP might want to implement.. )

P.S.
That's what I'm talking about..