Power tube OD in Push Pull

[MWJB]

At clean W RMS rating the Fender schem shows 16.7VAC at the 6V6 grids...the amp might make up to 18W or so under full distortion, if we scale that up, it represents 12VAC at the OT secondary & 20VAC at the 6V6 grid.

As the tubes try to draw more current, the B+ will sag & the 6V6s will head farther from saturation, more towards cut-off.

Also, the rest of the circuit is not really condusive to headroom - relatively low preamp voltages, cathodyne PI, lack of NFB loop, significant drop between plate & screen voltages...stages prior to the output are undoubtedly having a significant effect too.

[lowell]

What about the clipping point of the 6V6 tubes? Also you say that as the tube draws more current it goes towards cut off. I think this is backwards. As tubes draw more current the Va reduces and the tube goes towards saturation.

[MWJB]

"Also you say that as the tube draws more current it goes towards cut off. I think this is backwards. As tubes draw more current the Va reduces and the tube goes towards saturation." Initially, maybe, but after a point (depending on plate voltage & bias conditions) the tubes won't draw more current (unlike in fixed bias), the plate voltage will sag instead...bias point will head towards cut-off, at low idle currents/voltages crossover distortion can be a significant issue with cathode biased amps.

The clipping point of the tubes themselves...a pair of 6V6 in push pull can kick out 25W, so you can only consider that in the context of the circuit that they are in. Whatever limits clean output is the lowest common denominator, in the case of a stock 5E3 it's not the tubes themselves.

A class A amp should be halfway between cut off & saturation at clean power ratings, it will most likely drop in to AB under distortion (away from saturation), a class AB amp (typical 5E3) will head towards crossover distortion under distortion (again, away from saturation).

In short I think what I'm saying is that in the 5E3 the grids 6V6 are more likely to be where the limiting factor is, rather than saturate 6V6s.

[kg]

the reason i asked you to define what you consider saturation is, is because it is really not a clearly defined point.

as vg1 exceeds vk, it collects massive amounts of electrons which tend to reduce its potential.

if the g1 driving circuit is high impedance, then those electrons will remain on the g1 node, which means effectively that g1 will not exceed vk.

this is known as "grid clamping," in most circles.

but here's the key point: NOTHING is being saturated in the tube! it is MORE than capable of passing a greater amount of plate current. all you have to do is increase the potential on g1, which in most cases requires both a lower driving impedance and a lower DCR path back to ground. satisfy those requirements and you now have greatly improved the ability of the tube to "turn on" and pull its cathode and plate together.

the only hard(ish) limit i can think of is that of the cathode emission. even then, overheat your cathodes with a couple extra volts of heater and that gets bumped up. of course, your tubes will die prematurely from outgassing, overcurrent, and cathode poisoning, but you CAN squeeze a bit more current out of them.

so you see i'm not having a go at you--it's just not a simple answer! the only sharp discontinuity you can expect is when the tube completely shuts off (ie grid negative) and the plate current goes to zero, but there really isn't a comparable effect on the grid positive phase.

[Chuck H]

In short I think what I'm saying is that in the 5E3 the grids 6V6 are more likely to be where the limiting factor is, rather than saturate 6V6s.

Really?

The 5e3 does have a low headroom preamp and a high impedance PI with large coupling caps. This would contribute to grid blocking, but ONLY when the 6v6's are in a state that they are trying to draw grid current. And the 6v6's won't try to draw grid current before the plates begin to satutate. So the plates and grids of the 6v6's are at least equally responsible as limiting factors.

Steve Conner is big on recognizing how power tubes do shift operating class depending on the grid conditions. That is, if you hit the grid hard enough that the plates want to overdissapate the tube will go into cutoff unless current is available from that grid (class C?). If current is available from that grid the plate will continue to produce more power (class B?). The bias condition isn't the only determining factor in class of operation. How the grid is driven is also important. Once an AB amp starts to clip and is trying to draw grid current, is it still AB?

A power tubes max diss, the plate voltage applied and idle state are the limiting factors in how soon a given tube will saturate (or cutoff) for a given grid voltage. This is how power scaling takes advantage of lower plate voltages to cleate a lower headroom condition. If you reduce plate voltage you also reduce the amount of grid voltage needed to drive the tube into saturation (or cutoff)

A lucid understanding of saturation (which I can't provide) should then be an understanding of bias condition, plate characteristics at different voltages, any factor that changes the bias condition (cathode R voltage rise, power supply sag, etc.) and just basically what's happening to the grid plate relationship relative to these things. Saturation isn't it's own entity, as in "this circuit will make the plate saturate". Saturation is a consequense (as it were) of other operating conditions.

I really doubt that a 5e3 clips because the 6v6 grids are blocking. The plates are being driven into saturation and cutoff just like any other class A pp or class AB guitar amp.

[TransLucid]

Interesting thread...

A couple of comments:

1. As noted earlier, when the tube is in cutoff there is no plate current. So, it seems that this discussion is focusing on the other half of the waveform (where the grid voltage is positive).
2. I'm not sure there is a firm definition for "saturation" in this situation. The best way to see what's going on is to plot the load line and look at what happens as the grid voltage goes more and more positive. You will see that the slopes of the characteristic curves increase, which means that for a given change in grid voltage the plate current changes by less and less (also, the distortion begins to rise). This produces a kind of compression. As the grid voltage goes way positive, the grid begins to conduct current and this has a couple of effects: it robs current from the plate-cathode path, it can charge up the coupling cap and cause blocking distortion, and it can cook the grid. Blocking distortion, also known as farting out, can be controlled by proper design. That leaves grid conduction, which, as noted earlier, depends on the PI design.
3. My opinion is that the sweet spot for output stage distortion is right at the edge of grid conduction, that is, where the distortion begins to rise and compression begins.
4. Choice of bias operating point will affect when cutoff starts with respect to grid conduction. A lower quiescent voltage on the grid will favor cutoff; a higher one will favor grid conduction.
5. Crossover distortion is separate and unrelated. It is caused when one tube cuts off and the other starts conducting, so it occurs in the middle of the load line, not at either end. It is also affected by choice of bias operating point.

[MWJB]

"Once an AB amp starts to clip and is trying to draw grid current, is it still AB?" I thought so, I thought that in a class B amp, there was no current at idle without audio, current only flows under signal? If you bias an AB tube amp to 4mA, you still have current flowing at idle, bias it to 0mA and you get no audio?

5E3 are normally class AB, even if Class A at clean rating, they won't stay there under distortion, dropping into AB, doesn't this suggest that they are moving away from saturation? I'm making these numbers up for illustration, but just say current peaks at 50mA & the B+ drops 50vdc, how can this be heading more towards, or past class A?

[TransLucid]

Class A means both tubes are conducting current all the time. Class AB means one or the other is in cutoff part of the time. So the answer to your first question is yes, it's still AB.

Don't confuse DC current with AC current - the bias current is DC, the signal current is AC. If you bias an AB amp to 0 ma, it is operating in class B.

In class AB, both tubes conduct current part of the time (at low signal levels). This can be thought of as a kind of class A operation, although it makes things more confusing. The hotter you bias, the greater amount of time both tubes are conducting.

Saturation, at least as used in this thread, occurs at the extremes, so one tube will be in cutoff and the other in saturation. So it is definitely not class A.

[MWJB]

"If you bias an AB amp to 0 ma, it is operating in class B." If you bias a P-P AB tube amp to 0mA, it typically does not operate at all.

[kg]

generally the presence of control grid current is noted by a "2" suffix. ie, A2, AB2, B2.

technically a "1" indicates no current but it's often omitted as it's understood to be implied.

[Chuck H]

Aside from idle conditions I would think operating class should be dictated by what an amplifier is doing. How many "class A" push/pull guitar amps are there that advertise as such, and idle that way (remember, an idling amp makes no music) but are actually class AB1 (albiet hot) when operating (thanks kg). Also, AC drive requirements change with operating class so it seems possible to change a tubes operating characteristics (regardless of it's stated class of operation) to something more like a different class of operation by changing the AC at the grid. We can get all semantic about it and start quoting text books but it won't change these things, these exceptions to the "class" boxes.

Rather than get into another discussion about bias, which seems rather pedantic since guitar amps are all over the map, we should simply consider what the tube is doing. Is it conducting during half the cycle, less than/more than half? Is it trying to draw grid current causing conduction due to a high impedance grid circuit? Is there grid current limiting due to other parameters like cathode bias? Can a typical guitar amp PI allow for a little grid current? How much? How to dial out all the unwanted artifacts that come from the compromises?

Plenty to discuss without quoting operation class descriptions... Again. Not that I shouldn't sharpen my understanding but it probably won't get us any closer to resolving "saturation".

[kg]

well, i can say that g1 current flows following a 3/2 power law as soon as its voltage exceeds the cathode's. it's basically a diode with no forward voltage drop.

the impedance of the majority of grid driving circuits is fairly high--usually not a problem in terms of frequency response since the outputs are generally multigrid tubes with low grid capacitance.

also, perhaps more importantly their current sourcing capacity is severely limited, and they're cap coupled with high value grid leak DCR.

so if you looked at the g1 node as you push those output harder, then you'll see the positive excursions get flat topped as the grids conduct.

this is not necessarily a bad thing! it'll give you a little grunge and higher order distortion. it'll also effectively compress the signal by reducing the peaks and raising the averages.

then there is the dynamic effect of the bias shift. as those grids conduct, the high DCR to ground will cause them to drift lower in potential, decreasing the average current through the tube and effectively biasing it colder.

but in turn, this will actually give the grids MORE headroom before they conduct again on the next positive excursion. assuming a steady state input signal, they'll tend to settle at such an average/dc level that the tops of the waveform just brush 0Vgk. i've seen the exact same thing occur when i was fooling around with diodes in the preamp to clip the signal... if you use only one diode to ground, as it conducts it forces the node voltage to offset from 0vdc.

remove the signal and the grids will slowly discharge through the leak resistors, the time constant dictated by their value along with the upstream coupling caps. tuning this time constant will have a large effect on how the amp responds to transients vs. extended hammering. of course you can't arbitrarily assign cap and leak values, as they will directly effect the frequency response of the driver circuit.

it's a complex system with a lot of stuff going on, that's for sure.

[Chuck H]

El84's seem epecially prone to undesireable artifacts of grid (loading, conduction, clamping, whatever). In all the el84 amps I've built the grids start loading the moment the positive swing starts to compress. On a scope you can watch as the signal peaks flattening and the crossover distortion notch appear and grow simultaneously regardless of idle current. Not that crossover distortion is all bad. With the right OT and speaker using el34's it contributes to that desireable "swirl" effect. But in el84 amps it tends to sound less like swirl and more like BWIZZZERREEE. Agan, by virtue of OT and speaker rare examples of the infamous "18 watt" amps or the many offshoots don't buzz like a bee. The venerable AC30 can even sound buzzy when used with speakers other than bulldogs ( or alnico blue's). And the Orange TT is actually a pretty dark sounding amp so perhaps this keeps the fizz to an acceptible level. Anyhow, the el84 seems to be a real offender with regard to loading the grids.

That's why I love the Paul Ruby mod for these amps. It's just a zener valued just above the bias voltage in series with an opposing polarity 1n4007 strapped across each bias feed resistor. So simple and brilliant. It provides a low impedance path to ground for voltages above bias only while the tube for that bias feed R is in cutoff. So the only audible effect is much less grid loading. I also use another diode across the cathode R (most el84 amps are cathode biased) to clamp the voltage there when the tubes start to clip. Stopping the voltage rise (and bias shift) helps reduce the crossover notch too and gives a fixed voltage to set the strapped zeners. The combination works great and and really polishes the performance of the el84. But I digress...

I've used alternate PI circuits with lower output impedance to good effect too. MOSFET cathode followers after the PI have also been reported here many times. Some guys don't seem to mind the crossover distortion but any more than a little drives me batty.

[kg]

El84's seem epecially prone to undesireable artifacts of grid (loading, conduction, clamping, whatever). In all the el84 amps I've built the grids start loading the moment the positive swing starts to compress. On a scope you can watch as the signal peaks flattening and the crossover distortion notch appear and grow simultaneously regardless of idle current. Not that crossover distortion is all bad. With the right OT and speaker using el34's it contributes to that desireable "swirl" effect. But in el84 amps it tends to sound less like swirl and more like BWIZZZERREEE. Agan, by virtue of OT and speaker rare examples of the infamous "18 watt" amps or the many offshoots don't buzz like a bee. The venerable AC30 can even sound buzzy when used with speakers other than bulldogs ( or alnico blue's). And the Orange TT is actually a pretty dark sounding amp so perhaps this keeps the fizz to an acceptible level. Anyhow, the el84 seems to be a real offender with regard to loading the grids.

well, i can tell you why the el84 seems most prone to this: it's a high gm tube, which means you end up with very low bias voltages. what, usually like 10-20v max? normally, high gm is a "good thing" as it means the tube is easy to drive and is working well as a transconductance device. but when we're deliberately trying to overdrive stuff, that low bias voltage can be problematic.

That's why I love the Paul Ruby mod for these amps. It's just a zener valued just above the bias voltage in series with an opposing polarity 1n4007 strapped across each bias feed resistor. So simple and brilliant. It provides a low impedance path to ground for voltages above bias only while the tube for that bias feed R is in cutoff. So the only audible effect is much less grid loading. I also use another diode across the cathode R (most el84 amps are cathode biased) to clamp the voltage there when the tubes start to clip. Stopping the voltage rise (and bias shift) helps reduce the crossover notch too and gives a fixed voltage to set the strapped zeners. The combination works great and and really polishes the performance of the el84. But I digress...

i like both of those mods (at least in theory--haven't tried them first hand myself yet), for the same reasons you do... keeping things (operating points) during heavy overdrive more similar to those during quiescent conditions.

I've used alternate PI circuits with lower output impedance to good effect too. MOSFET cathode followers after the PI have also been reported here many times. Some guys don't seem to mind the crossover distortion but any more than a little drives me batty.

i agree--i'd much rather have heavy top/bottom hard clipping than crossover distortion myself.

[kg]

i found the chuck H mod discussion:

? - Paul Ruby Zener Diode mod