Inter-stage dividers in high gain circuits

[35L6]

Getting back to basics , I always thought of a voltage divider as a volume pot with a fixed wiper . Typical values being 1M or 470k . The total value of the divider being the load for the previous stage .Following grid resister is in the formula for both gain and coupling cap corner frequency , so is important . Volume pots are typically wired with wiper to the following grid . What about the case of the 'backwards' pot like in a 5E3 , where the signal enters at the wiper instead of the to of the pot ? It seams I've heard that the attenuation is not as good or a different taper or something like that . Also , the top half is now a grid stopper , so does the Miller capacitance thing and reduces treble . how does all that work and could it be used for tone shaping .

[Helmholtz]

Maybe this helps:

A basic voltage divider consists of 2 resistors wired in series and connected to a voltage source.
It has an input (across both resistors) and an output (across lower resistor) terminal.
The voltage applied to the input terminal causes a current of V/Rtot through the voltage divider, where Rtot is the sum of the upper and the lower resistances.
The output is taken from the upper end of the lower resistor, which can be considered a tap of the total resistance.
So if the lower (or shunt resistance) is e.g. 10% of the total resistance, the current produces a voltage drop of 10% of the input voltage across it.
That's the ouput voltage.

When input and output terminals are interchanged, it no longer works as a voltage divider.

The vol. pots in a 5E3 are not wired as voltage dividers but work by loading down the source (input tube).
And yes, the high "output" resistance at low vol. settings and the MIller effect cause a treble cut.

A voltage divider works best if connected to a source having an output impedance less than Rtot and a load having an impedance higher than Rtot.

[Chuck H]

The total value of the divider being the load for the previous stage.

This is a good observation so I wanted to quote it in a post for repetition. In this thread we may have, or may yet refer to the shunt resistor for a voltage divider as a load resistor (probably more accurate so say shunt as Helmholtz did and I sometimes do). But, indeed, the load seen by the previous stage is influenced by the total voltage divider circuit resistance. This is a good distinction to point out.

[Helmholtz]

All the previous stage or signal source "sees" is the load current and this is mainly defined by the total voltage divider resistance.
And it's the load current which causes a significant drop of source voltage (across the internal source impedance) if the total voltage divider resistance is too low (an effect often described as down-loading).

I don't think one can really understand voltage dividers and their secondary effects without considering the current.

[35L6]

So , in the "backwards" (5E3) volume control , does it simply reduce the gain or is it making a voltage divider with the output impedance of the preceding stage or is that just two ways of saying the same thing ? Also , how come a resistor a position B causes the Miller effect and one at position G doesn't ? or does it ?

[Helmholtz]

So , in the "backwards" (5E3) volume control , does it simply reduce the gain..

That.
No voltage divider.
A "reverse connected voltage divider" is no voltage divider, period.

how come a resistor a position B causes the Miller effect and one at position G doesn't ? or does it ?

The Miller effect increases the effective grid to ground capacitance (typical values for a bypassed 12AX7 gain stage being 100pF to 150pF).
It doesn't depend on the B resistor.
But the larger the grid series resistance (or grid stopper) the lower the treble cutoff frequency (=> RC low pass filter).

[Gtr0]

A voltage divider works best if connected to a source having an output impedance less than Rtot and a load having an impedance higher than Rtot.

Nice nugget of info there!

[pdf64]

That.
No voltage divider.
A "reverse connected voltage divider" is no voltage divider, period.

When considered as a circuit feature, away from anything else, I agree. But as I see it, with reference to Chuck’s schematic in post #14, in eg a 5E3, the variable resistance load (ie between the vol control wiper and 0V common) uses the output impedance of the preceding stage as its ‘G’ resistor, so forming an effective voltage divider.

An appreciation of the various output impedances of the various stage types, and how they interact with the loading on them, seems crucial to getting an insight into the mechanisms behind valve guitar amp operation.

… the larger the grid series resistance (or grid stopper) the lower the treble cutoff frequency (=> RC low pass filter).

Just to expand on that, as I understand it, the ‘grid series resistance’ will also include the non obvious effective impedance forming the source impedance that’s feeding the grid. eg with a typical 470k G and 470k C voltage divider resistors, in the context of noise and the Miller Effect, an effective B resistance will be at least 235k will be formed, even if there is no actual B resistor (the output impedance of the preceding stage will make it a little higher than 235k).

[Helmholtz]

.. in eg a 5E3, the variable resistance load (ie between the vol control wiper and 0V common) uses the output impedance of the preceding stage as its ‘G’ resistor, so forming an effective voltage divider.

Right, it's what Enzo once called a hidden voltage divider.
Actually many resistors in an amp are involved in some non-obvious voltage dividing. Still no-one would call those resistors voltage dividers.

The gain reduction can also be explained from the AC loadline of the input stage: Lowering load resistance rotates the loadline to a more vertical orientation and thus lowers gain.

Just to expand on that, as I understand it, the ‘grid series resistance’ will also include the non obvious effective impedance forming the source impedance that’s feeding the grid. eg with a typical 470k G and 470k C voltage divider resistors, in the context of noise and the Miller Effect, an effective B resistance will be at least 235k will be formed, even if there is no actual B resistor (the output impedance of the preceding stage will make it a little higher than 235k).

Good point.

Yes, it's the effective source (AC) resistance (Thevenin equivalent) as seen from the grid which matters regarding thermal noise and treble loss.
A good reason to not use higher than necessary divider resistor values.

OTOH, the contribution of noise from the second stage will be much lower than noise from the first stage.

Regarding treble loss by Miller effect and grid series resistance, the 235K would result in a -3dB frequency of 4.5kHz with the typical inpiut capacitance of around 150pF of a bypassed 12AX7gain stage.
This effect can be neutralized by wiring a suitable treble cap in parallel with the upper voltage divider resistor.
E.g. with 2 x 470R resistors, a 150pF cap across the upper resistor would essentially compensate the effect of the Miller capaitance

[SoulFetish]

I don't buy the "sweet spot" -theory.

For one, typical tube gain stage has gain ranging from 50 to 60, and overdrives at input levels of maybe 1 to 3 volts peak-to-peak... yet, interstage dividers are often only 1/2th ratio and maybe 1 to 10th in extreme cases - and that not even including "treble peaking" bypass, which reduces the ratio even more. No 1 to 50th ratios in sight. If there's a "sweet spot" the division ratios vs. stage input sensitivity ratings do not correspond to it.

Secondly, even most "high gain" amplifiers generate the distortion in just select few of the preamp stages. Carvin's eleven cascaded stages is an isolated incident amongst typical designs and they intentionally introduce as much as DC offset shifting as possible instead of little bit of clipping in each stage. The DC shifting is really one of the key aspects of having several cascaded stages that distort.

I regard interstage attenuation as a relic and often there's more effect to tone from the aforementioned and associated "treble peaking" than from the attenuation of the divider. So, more important to attenuate select frequencies than to just attenuate everything.

^^^ This friggin' guy

Teemuk, every time you post here, you bring something to the conversation and raise the level of discourse. Don't be a stranger, dude.

Carvin's eleven cascaded stages is an isolated incident amongst typical designs and they intentionally introduce as much as DC offset shifting as possible instead of little bit of clipping in each stage. The DC shifting is really one of the key aspects of having several cascaded stages that distort.

Back to the DC offset shifting; is this the reasoning behind the Transtube clipping/biasing circuit? I imagine the reason for biasing the base at a designated +1.5V reference supply allowed for a larger emitter resistor, and provides some bias excursion to control the level and way the stage clips. Is this correct? If not, then that leaves me with – it's incorrect. (<--clearly, you can see I'm a bit of a math wiz)
figured you'd be the cat to ask, being as you wrote the book on solid state guitar amps.

[Jag]

Okay, haven't been here in quite a while but have a question sort of on this same topic.

Let's say I have an amp that has four stages and far too much gain. Let's also say that I have found that a divider with a 1M series and 220K shunt, after the first gain stage, brings the level down to where the following stages have the kind of sound I want. So I'm attenuating the signal by about 80% before going into the second stage. So far so good, all safe and no problem.

My question is, what is the impact, besides extra noise, if I replace the 1M-220K divider with a 4.7M-1M divider?

[Helmholtz]

My question is, what is the impact, besides extra noise, if I replace the 1M-220K divider with a 4.7M-1M divider?

Both dividers approximately attenuate the input signal by a factor of 0.18.

The larger resistance divider has a higher input resistance and a higher output (or source ) resistance.

A larger input resistance means less loading of the preceding stage.
If the divider is fed via a coupling cap, the larger resistance helps to preserve bass content.

A higher output resistance may result in a treble loss, as the output resistance and the input capacitance of the following stage constitute a low pass filter.

It depends on the circuit if these effect are audible.

[Jag]

Thanks Helmholtz! That's pretty much what I thought. Wanted to make sure I wasn't breaking a forgotten law of physics.

[teemuk]

Back to the DC offset shifting; is this the reasoning behind the Transtube clipping/biasing circuit?

Yes. If you read the patent it discusses about overall asymmetry of the clipped output signal, which without the shunt clipping diode at base is actually too symmetric in comparison to an actual triode stage. Patent discusses about certain ratio of asymmetry that is sufficient enough (at least according to Peavey) to mimic behaviour of a typical tube gain stage.

Greater asymmetry will naturally lead to greater amplitude of even order distortion and in cascaded stages to signal offset shifting and dynamically varying harmonic content. I'm not sure if Peavey mentions the latter in the patent but the previous simple, symmetric shunt diode clipping of the Super Saturation family of amps was obviously abandoned. (Pre and post distortion voicing of them wasn't remarkably different than that of TransTube preamps).

I imagine the reason for biasing the base at a designated +1.5V reference supply allowed for a larger emitter resistor, and provides some bias excursion to control the level and way the stage clips.

Maybe. What is actually implied in the patent is that they chose darlington circuit due to its higher gain that aids in increasing the effective input Z and thus helps in mimicking the "voltage drive" characteristic of tubes. There is a transtube patent that covers the use of FET devices but obviously it was easier for Peavey to meet all goals of preferred operation with a discrete darlington circuitry instead.

I would also like to point out that the overall gain staging / frequency response modifications of the transtube preamp are based on their well established all tube amps such as the Ultra series. So they had a recipe that already worked, didn't have to start from scratch and mainly had to figure out how to mimic (in reasonable accuracy) clipping of a common cathode stage. It's not a 100% realistic tube emulation but reaches the point of being sufficiently accurate. IMO, when people have an opinion they are not great amps it tells more of their dislike about that Peavey voicing than about the quality of the implemented tube emulations.

[teemuk]

I reverted back to reading the patent to refresh my memory and the closing paragraphs really stuck me. The main idea of the invention is really summarized there. Anyone can asymetrically diode clip a signal, and there's nothing new in that, but in order to simulate a vacuum tube (preamp) circuit utmost importance is to asymmetrically clip HIGH amplitude signal (e.g. 15V peak) with diode's LOW threshold (ca. 500 mV) so that clipping asymmetry and resulting interstage bias voltage shifts are maximized.

We don't get similar magnitude of asymmetry and bias shifting if clipping lower amplitude signals (e.g. 2 volts peak). This even more counteracts the popular "sweet spot" -theory of clipping each tube stage just a little bit: If we only clip only a little bit there will be very little of asymmetry and interstage bias shifting, IOW "tube sound".