BMP feedback loop question: That cap in series with the diodes

[midwayfair]

Kitrae's schematic indicates the cap is meant as a high-pass so that bass frequencies are clipped less than higher frequencies. I think the assumption is that since the effect is very bass heavy after the tone control, that maybe this had something to do with it.

I can't come up with how that's true, though. I would think that the treble loss at the tone stack is created by the tone stack values and the fact that there's a 723Hz cut from the 470pF caps in the feedback loops.

Second, my calculator says that the 470K resistor and a 47nF cap in the feedback loops have a cutoff of a mere 7Hz. It would have to get mighty tiny before it starts clipping the bass less, much smaller than what's used in Muffs that are touted as being "bassy". EDIT: I'm talking about C6 and C7 here, which are 47nF in a Green Russian and many other editions (just taking the smallest value I've seen).

From this, all I can see is that it's used to avoid upsetting the bias and to get an AC signal going through two diodes.

Am I missing part of the calculation in figuring out the cutoff frequency?

And if I'm not -- if the diodes really are essentially an all-pass -- is there really much of a benefit to two independent stages with the same diodes and tone shaping compared with the Darlington Bazz Fuss stage in the Wisker Biscuit? Assuming that the builder doesn't want to tinker with the tone shaping and diodes in one or both stages, of course. From what I recall, you can even still put a pF cap across the diode in the Bazz Fuss stage in the middle, but I'm guessing you have to use a much smaller cap since the diode will have a few megs of resistance across it.

[anotherjim]

Sorry, where do you see a 47nF cap? Is that a typo?

The tone stack has permanent treble bleed. Due to the output impedance of the preceding stage. It's collector resistor forms a potential divider at any frequency when C8 and/or C9 have low impedance. Reducing the collector R's wouldn't only reduce gain, but reduce some of the tone stack treble loss too, but at the expense of higher supply current.

Whisker Biscuit is not much different in this regard.

A low output impedance buffer before the tone stack would cure.

[midwayfair]

Sorry, where do you see a 47nF cap?

Just incomplete ... C6 and C7 are 47nF in the Green Russian. I can't see how that makes any audible difference from the 1uF in the Kitrae schematic.

EDIT: Also, I'm not looking for any change in the tone stack and I know why there's treble loss from the loading there. I'm only asking about the feedback loops.

[antonis]

C6 and C7 are 47nF in the Green Russian. I can't see how that makes any audible difference from the 1uF in the Kitrae schematic.

If you take in mind that these caps create a RC filter with the respective Q3 & Q2 biasing resistances (R20 & R16)...

[midwayfair]

C6 and C7 are 47nF in the Green Russian. I can't see how that makes any audible difference from the 1uF in the Kitrae schematic.

If you take in mind that these caps create a RC filter with the respective Q3 & Q2 biasing resistances (R20 & R16)...

So, what, R16 divides against R15, and that's the resistance that ends up in parallel with C7?

[anotherjim]

Sorry, where do you see a 47nF cap?

Just incomplete ... C6 and C7 are 47nF in the Green Russian. I can't see how that makes any audible difference from the 1uF in the Kitrae schematic.

Gotcha, and think I see your point, and it's a pretty good one.
Going down to 47nF seems to make it a rumble booster more than a low end instrument boost.

But, if I look at it as an AC path in parallel with the DC feedback, and compare the significance of the impedances at spot frequencies, I see the cap Z is 1/10th (47k) of the R at 70Hz. The 1uF original is zero ohm at 70Hz. So, it's a significant change to the low end despite the filter cut-off calculation.

Thinking of the high pass filtering by input and stage coupling caps. I make out 4 poles at about 20Hz throughout. There's not going to be much at 0.7Hz to boost! Does the Green Russian change the coupling caps from 0.1uF?

So, there's a hump in low end from 20 to 70Hz before the diodes limit everything upwards from there?

[samhay]

Second, my calculator says that the 470K resistor and a 47nF cap in the feedback loops have a cutoff of a mere 7Hz. It would have to get mighty tiny before it starts clipping the bass less, much smaller than what's used in Muffs that are touted as being "bassy". EDIT: I'm talking about C6 and C7 here, which are 47nF in a Green Russian and many other editions (just taking the smallest value I've seen).

Not quite - the 47n/1u cap only comes into play when the diodes conduct, at which point you have 470k || the diode forward resistance under conduction (i.e. bugger all).

From this, all I can see is that it's used to avoid upsetting the bias and to get an AC signal going through two diodes.

Yup - otherwise you get into Bazz Fuzz territory, with very asymmetric clipping.

[midwayfair]

So, what, once the diodes conduct, the clipped negative feedback forms a low-pass filter with the 100K at the bass, NOT a high-pass with the 470K feedback resistor?

But, if I look at it as an AC path in parallel with the DC feedback, and compare the significance of the impedances at spot frequencies, I see the cap Z is 1/10th (47k) of the R at 70Hz.
...
So, there's a hump in low end from 20 to 70Hz before the diodes limit everything upwards from there?

I don't really see how an entire decade of frequencies could possibly matter. Isn't 70Hz up only 0.3dB in that case compared with the 1uF? I can't imagine that's audible.

There's not going to be much at 0.7Hz to boost! Does the Green Russian change the coupling caps from 0.1uF?

GR's caps are 47nF in the feedback loop; so far as I know, ever version keeps the 100nF coupling caps.

I should say I'm simply passing on what people have said, that the Green Russian has more bass, and extrapolating from the Kitrae schematic. I haven't observed the change myself.

Bonus question to extra-confuse Jon ...

What do I use to calculate these filters if I have, say, my base resistor to a 50K/50K voltage divider for Vb? 

[samhay]

^So, what, once the diodes conduct, the clipped negative feedback forms a low-pass filter with the 100K at the bass, NOT a high-pass with the 470K feedback resistor?

Right - As Jim said, it forms a low fairly aggressive low pass filter. I think you need to factor more than the 100k resistor in to the frequency calculation though.

[anotherjim]

Before the diodes act, or a frequency where the Cz is significant, there is ALL the gain to make up for the cut. The diode cap is making a lpf, but it is only one pole. How much gain is there? Maybe enough to take the transistor itself into clipping and negate some of the cut from that lpf?

Bias resistor is in parallel with source impedance, BUT, we have negative feedback from collector -  it is 180deg out. Input signal is opposed by collector signal. An imperfect (base current) virtual ground at base. So, I say the feedback R and diode paths are terminated at the base and don't see the bias R.

Well, maybe...

[PRR]

I think, in this circuit, if you do not have some capacitor in series with the diodes, the DC bias will be very different from the case where DC bias is set by 470K+100K resistors.

Or was this point already made?

[midwayfair]

I think, in this circuit, if you do not have some capacitor in series with the diodes, the DC bias will be very different from the case where DC bias is set by 470K+100K resistors.

Or was this point already made?

No, I know you need the cap in series with the diodes. I'm just trying to figure out how to calculate the frequencies it passes under various circumstances.

[PRR]

> figure out how to calculate

The impedance of the diodes will swing from >1Meg for very small signal to <300r for very large signal.

And it will do this over every half-cycle.

So there's no "one frequency" it works at. The roll-off varies with signal level.

If the cap is "small" (we don't yet know what "small" might be), then for low signal levels it will flatten bass and treb, for large signal levels only treb.

In fact the diode effect is small until they come down to say 250K (because they shunt 470K). And they probably max-out (least impedance, most clipping) at a few K-Ohm (in this circuit). If we split-the-difference, ratio-wise, between 250K and 2.5K, we have 25K. That may be a fair number for initial estimates.

Assuming 1uFd and 25K, action is uniform above 7Hz. Someone thought we should clip "everything".

Assuming 1uFd and the lower likely diode impedance 2.5K, action is uniform above 70Hz. This is still "everything" for guitar (rarely plays below 82Hz).

If you shorten the cap to say 0.1uFd the action slides up three octaves. Small bass will be clipped, but large bass will hardly be limited here (it may get hammered somewhere else in the system).

There's two such stages, essentially identical. I wonder what the sound is with very different caps at each stage, and what order they should be in. But that is far too many unknowns to grok without lots of experimentation.

> figure out how to calculate

Tip: in musical electronics, especially distortions, the fastest "calculator" is often the soldering iron and your ear.

[nocentelli]

If you shorten the cap to say 0.1uFd the action slides up three octaves. Small bass will be clipped, but large bass will hardly be limited here (it may get hammered somewhere else in the system...

Well, that kind of chimes with the received Internet wisdom that 100n diode cap clips most of the signal, and reducing to 47n clips bass slightly less and so is less mushy.

[jatalahd]

I guess there is also the time constant R18*C6 and R11*C7 affecting the rise time (of the squared signal) when the diodes are significantly clipping. Perhaps the decay time is controlled by hfe*R21*C6 and hfe*R10*C7 ...

I don't see any reason why and how the values of C6 and C7 would have any effect on filtering the signal. Only in the cases where the input signal has a very specified low level or where these caps would be taken in the picoFarad range.