Time constants for bias shift effects due to grid conduction

[phasetrans]

All, me again. I'm the guy that did the DC coupled optocoupler based clipper thingy a bit ago.

I'm working on my second distortion related solid state thing (also DC coupled for now). But as I look at Merlin's book, it strikes me that the bias shifting that happens due to the combination of DC blocking cap after a GCA, and the variable input impedance of the subsequent tube entering grid conduction, is certainly part of the sound, and perceived compression.

I have plenty of experience with useful attack and decay times for compression AFTER an amp, but not within it. And the spice tube models I have don't include the change in grid input Z as it enters conduction.

Does anyone know what these time constants are, typically, and how dramatically they change (shorten) with input level as the follow on valve enters grid conduction?

Thanks!

[merlinb]

You can model grid conduction well enough for this application by adding a diode plus a 1k to 2k resistor between grid and cathode of the tube (i.e. diode anode goes to the to grid, and the resistor is in series with the diode). The rest you can copy from popular amp circuits. 

[phasetrans]

You can model grid conduction well enough for this application by adding a diode plus a 1k to 2k resistor between grid and cathode of the tube (i.e. diode anode goes to the to grid, and the resistor is in series with the diode). The rest you can copy from popular amp circuits. 

I should clarify that I'm not trying to work with tubes at the moment, only hoping that the tube models would characterize the input Z behavior.

So just assume the input Z collapses to a couple kilo ohms, and the diode knee behavior is similar to the cathode electron cloud behavior? Interesting.

The specific circuit that I wanted to get square on the shifting is for the Soldano-style cold biased GCA with degeneration followed immediately by DC coupled CF. And specifically how the bias of the follower is changing when interplayed with the cutoff behavior of the GCA.

For context, I was moonlighting for a production company in college, in the state of Florida, in the thick of the nu metal era. So I have intellectual curiousity about this particular triode pair, but no interest in repeating it! 

[PRR]

You will not go wrong STEALING time-effects from classic Fender and Marshall. Leo had a good ear and Jim had a different time/taste; between them they covered the waterfront.

Guitar Amplifier Overdrive, Neumann and Irving, is a study of various dynamic overdrives in guitar work. More theory than guide. Deserves a follow-up with some variant experiments and some musical observations. And not the cheapest book in the store.

Tubes or crystals is no real different, amplifiers are amplifiers and a Guitar Amplifier has to do specific things. You can emulate directly with JFET approximations, or massive BJT analog computer, or dump into a DSP and program a sound, but the User should not have to care.

Agree that a sim model with no grid diode is *defective*, a diode and resistor is a fair approximation for modeling overdrive. Neumann&Irving give some data on grid I/V curves. It varies between brands; since no brand has an overwhelming "sound" advantage I suspect the variation is not critical, meaning model values don't have to be exact. A few-tenths Volt change of diode threshold is mild compared to many-Volt overdrive.

[phasetrans]

Tubes or crystals is no real different, amplifiers are amplifiers and a Guitar Amplifier has to do specific things. You can emulate directly with JFET approximations, or massive BJT analog computer, or dump into a DSP and program a sound, but the User should not have to care.

I'm a materials engineer by education. Long ago I worked in a CMOS Fab for one of the big brand firms. And in graduate school I worked on work function lowering materials for thermionic emission. So I've seen the physics of both. And I agree that it should be transparent to the user, but certainly they are different beasts to each other.

I'm trying to stay away from history, or at least be mostly ignorant. There's enough copying afoot already.

But the bias shift thing, especially with a number of stages in series, would be something I'd like to have a clear sense of the behavior. E.g. does only the shift in the final, most overdriven stage(s) dominate all the effects.

A few-tenths Volt change of diode threshold is mild compared to many-Volt overdrive.

Great point.

[amptramp]

Don't forget the millions of radios made with a 12AT6/12AV6/6AT6/6AV6/12SQ7/6SQ7 first audio stage where the grid bias came from a 10 megohm resistor to ground.  Contact potential (the impingement of electrons on the grid) was quite capable of biasing the grid at -0.8 to -1.0 volts with no signal applied.  The 6AV6 triode was one half of a 12AX7, so you could pull the same stunt with one of those.  Your model of grid conduction would have to include the diode and resistor as outlined above with a battery of about -0.8 to -1.0 volt in series as long as you had 10 megohms to ground on the grid.  For lower values, you would get lower voltages.

[Eb7+9]

Does anyone know what these time constants are, typically, and how dramatically they change (shorten) with input level as the follow on valve enters grid conduction?

The dominant time constant in triode gain stages comes via the cathode bypass cap (when present) and the non-linear grid current might rise to maybe around 1uA from its nominal sub nA value at idle ... these are only rough estimates based on what is necessary in grid biased circuits via the typical 10meg grid resistor value used in the circuits to produce the required 1volt back bias ... the dynamic grid current is so small that it seems to play a near negligible role in all cases // whether in linear mode or in all out cutoff/saturation mode

You can check out my 12ax7 model as it contains a scaled down expo function derived from published pentode grid source functions ( see Glass Audio 2/98)

As for pentode circuits, especially cathode biased pp circuits it's difficult to say again ... personally I cannot trustany modeling of pentode devices on account of the screen current function which is typically ignored ... and the bias shift seen at thecathodebypass cap is due primarily to the non-linearity of the plate current function

In both cases, triode and pentode ckts, it's difficult to pin a time constant to half-cycle non-linearcurrent draw ... partly because of its non-linear action on surrounding capacitors (when present) and I the more dominant non-linear currents

The only reliable sim i've been able for to perform (seen in my tube amp book) is of the Fender Champ psu where i model the 5Y3 rectifier tube and show in basic first order terms how a sudden transient load gives rise to attack/recovery times (total 5x Tau) of about 100ms

Otherwise, all other time manifestations in tube circuits are either too small to measure or too non-linear in nature to equate to the linear concept of time constant

[merlinb]

the dynamic grid current is so small that it seems to play a near negligible role in all cases // whether in linear mode or in all out cutoff/saturation mode

Um, I think you may have mistyped something there. Grid current is the important factor in 'saturation' [sic] mode.

[highwater]

Rather off-topic, but there's been an idea kicking-round my head for a while, and this is the first time grid conduction has come-up since I thought of it.

Would it be possible to parallel a JFET with a BJT as a sort-of über-Fetzer, such that the JFET covers the negative-grid range of input and the BJT covers the positive-grid range?

The biasing would be a nightmare, of course, and SPICE wouldn't help at all...

...but if the idea has any merit whatsoever, it'll go straight to the top of my breadboarding list.

[Eb7+9]

Grid current is the important factor in 'saturation' [sic] mode.

I think you’re mixing two things up here Merlin ... what you’re saying is true only in the more extreme scenario. where things go into hard gating ... but not really until then, or not in the same way we see in other pre-saturation tube compression (ie,. clean compression)

I’m being careful to distinguish between compression generated by bias shifting (as seen more evidently in  the output stage of SE amps - and what I meant in relation to cc triode gain stages) and the gating caused by increasing grid currents of one stage starving the previous stage running out of current as it enters its highest voltage range - why we typically see large valued grid blockers in high gain tube preamp designs ... if you lump the two mechanisms into one umbrella concept I can see grounds for misconstruing what I wrote above

Grid currents in a 12ax7 tube cannot really exceed tens of micro amps (because the driver stage driving it is simultaneously shutting off) while a saturated plate/cathode current reaches around 1~2 mA at most simultaneously - swamping the grid current levels ... and so the bias shift caused by cathode voltage bump is in reality dominated by the much larger plate/cathode current / the gating side of things is another story altogether

This is why single ended Amps, like a Fender Champ say, will typically exhibit a strongly perceivable degree of compression well before it enters its clipping zone (why they’re used in the studio a lot) ... contrast this against a preamp stage built out of a pair of 12ax7’s (ie., say the normal channel if a BF reverb amp or simple standalone preamp circuit) and said compression is completely or near completely absent ... it’s way more subtle with triodes alone, I would say virtually non existent with typical circuit values (Fender/RCA)

Stick a scope on the cathode of a Fender Champ and see how much the bias shifts when playing clean ... then ask the question “what causes the shift” ... you will find that the difference comes from one half cycle current being appreciably larger than the same for the next half cycle (something btw that does not take place in PP output stages) ... the difference is tantamount to an additional per cycle dc current // hence the bias shift ,... hence the compression effect perceived strongly in SE amps

Of course keep moving in that direction from that point on and stronger grid current on that positive half cycle starts adding to that picture, entering what I would refer to as the over-compression zone (slamming or gating) ... some SE Amps have it so bad at it that they are practically un-usable ... something I’ve seen when modding tape deck amps

All this to say that it is not really appropriate to attach a time constant to either the former nor the later as they are products of dynamic effects - my main point here ... it is really impossible to do so primarily because the resultant dc shift is a product of dynamics also and isn’t a constant or fixed value ... I fought up my psi sim because it is only there that s time constant can be attached ... of course, in a real amp like the Champ all three mechanisms start kicking in various degrees, forming a descriptive basis for the dynamic complexity that a Champ offers

[merlinb]

Would it be possible to parallel a JFET with a BJT as a sort-of über-Fetzer, such that the JFET covers the negative-grid range of input and the BJT covers the positive-grid range?

Yes, but why? The FET already behaves like a tube; it's gate can be driven into conduction.

[merlinb]

I think you're mixing two things up here Merlin ... what you're saying is true only in the more extreme scenario. where things go into hard gating ... but not really until then, or not in the same way we see in other pre-saturation tube compression (ie,. clean compression)

Hmm, I find it hard to follow what you're saying as it sounds like you're using technical terms in an imprecise way.
Saturation is maximum possible cathode current; this cannot occur in a conventional tube amp unless something has gone horribly wrong. I think you're using saturation to mean something else.
Hard gating? Is that your term for blocking distortion? i.e. caused by the negative shift in grid voltage due to coupling cap charging when the grid conducts?
Tube compression? Are you referring to soft clipping or to changes in overall gain like in an audio compressor?
Clean compression? You have lost me.
Clipping zone?

There are two very important time constants created when the grid conducts or does not conduct. Conduction leads to coupling cap charging and hence to bias shift at the grid, and these time constants determine the 'attack' and 'decay' times of this grid shift. This is the most important factor controlling the dynamic distortion tone in a preamp. And you're right, it only takes microamps (steady state) to bring this shift about.
Shift in the cathode voltage due to changes in the average anode current are a secondary and lesser effect, in preamp stages at least. Average anode current is, after all, a slave to changes in the grid voltage.


I can recommend this book if you want to get really deep into the subject! 
http://www.lulu.com/shop/ulrich-neumann-and-malcolm-irving/guitar-amplifier-overdrive/paperback/product-22447917.html

[highwater]

Would it be possible to parallel a JFET with a BJT as a sort-of über-Fetzer, such that the JFET covers the negative-grid range of input and the BJT covers the positive-grid range?

Yes, but why? The FET already behaves like a tube; it's gate can be driven into conduction.

I was under the impression that when a tube grid is driven positive, the plate current continues to increase... whereas with a JFET, only the source current increases.

Perhaps the better question for me to ask would be: is that increase in plate current anywhere near large enough to be audible (or even measurable)?

[Eb7+9]

Hmm, I find it hard to follow what you're saying ...

In using the term “compression” you’re lumping extreme clipping caused by one sided gating (when the tube goes into hard clipping and gets stuck - as you point out - hard on one side per cycle) with gain reduction caused by bias shifting (drop in trans-conductance) which takes place well before the stage goes into any kind of clipping ... ie,. producing clean gain reduction

grid current plays a dominant role in the former hard-clipped/gating case, while plate current non-linearity plays a dominant role in the later clean compression case

basic stuff ...

[teemuk]

But the bias shift thing, especially with a number of stages in series, would be something I'd like to have a clear sense of the behavior. E.g. does only the shift in the final, most overdriven stage(s) dominate all the effects.

This is my rough estimation:
The bias shift will most prominently effect 1) stage gain and 2) overall thresholds of clipping (distortion), as bias determines plate voltage and therefore "headroom" of voltage swing to either direction.

I would be inclined to think that regarding "audibility" the effect of 1) is largely negligible. You will not likely hear any effects of minor bias shifts (and gain change) as long as the output waveform remains moderately distortionless.
Regarding audibility 2) is likely more easy to perceive so the effect is largely created in the stages in which clipping is interacting with bias shifts. Generic PP output stage is one of such stages (but only when the amp is cranked to very high output power and loudness levels) but it is customary for "high gain" preamp designs to also feature preamp stages where bias shifts due to asymmetric output from previous stage, or due to effects such as grid conduction.

Without any further details given of the design one pretty much has to ponder case-by-case, which stages in the circuit contribute to such effects and at which magnitude. For example, what applies to, say, Fender 5E3 circuit will not apply to, say, Soldano SLO 100 circuit because of entirely different architectures.

[amptramp]

Back in the old days when pterodactyls plied the skies, some radios were built using the 6N6G output tube.  It was effectively a Darlington tube with the cathode of the input triode direct connected to the grid of the output triode.  There was no need for a cathode resistor and the grid resistor to the input was grounded, so it was considered easy to design with and was used by Crossley among others.

Another interesting tube was the 6AC5GT which was always direct driven from plate of another tube, in the case of a Silvertone set I have, a 6P5GT.  It was always run in class B but with grid current flowing.  Internally, the tube has two grids that are connected together.  The 6AC5 is said to have a µ of 125 or 135 depending on whose tube characteristics you read.