How would a cathode follower output section sound for guitar?

[printer2]

I thought the NFB in a cathode follower and, a mosfet source follower stays clean also but it has a wider voltage swing. So that stage stays clean and is transparent (more or less) until clipping. So any distortion has to come from the tube the cf/sf is helping out. Now when we overdrive a pentode in a preamp circuit it sure does not sound like a triode (just generalizing here, sure you could find an example of one sounding like the other or even like a SS device).

So if we took a small power tube, say a 6AK6, run into a CF into a OT what would be the result? The 6AK6 would be running into a high impedance, pretty much the same thing we try to give our preamp tubes to run into. You never hear complaints that an EF86 is being run into too high of a impedance rather than too low of one. Or maybe even use a EF86 as the 'output tube'. Just trying to think a little outside the box. Also if you have a CF on the 'output tube' you may not have to retube the output of the amp any more than you would the input. Still want to retain the quirky OT though. Well at least for now.

[Steve Conner]

That seems right. But if you admit that a power tube used as a CF is transparent and the tone comes from the stage before it, then you might as well replace the tube with a power MOSFET source follower. It will still be transparent, and won't need retubing, or a floating heater winding.

[J M Fahey]

Fully agree and add:
1) yes, plate curves must *still* be used, as they are the ones which describe how much current flows for a given plate to cathode voltage and grid to catode (and even screen to cathode)
Why do I seem to insist so much on "to cathode"?
Because it's implicit (thus not explicitly mentioned) in "normal" designs, but here you may be mistaken by thinking "hey! ... on a cathode follower plate voltage does not vary, so let's take it out of the equation" ... which is not true, in this case cathode voltage follows signal across the load so the *difference* between them still rules.
2) when I said "transistor-y" I didn't specifically refer to the output transformer (although yes, it's still *a* factor) but to the very different operating condition:
a) as Steve mentioned, it operated unity gain, with 100% local feedback, *ver* different
b) there's no grid rectification and heavy bias point shifting (there may be a little, ut much less important) , so:
c) no curve kink, no gain modulation, no important input signal clipping because of that, etc.
In sum, a very different way of operation.
Not surprisingly, the way *most* SS power stages operate.

[Alan0354]

Fully agree and add:
1) yes, plate curves must *still* be used, as they are the ones which describe how much current flows for a given plate to cathode voltage and grid to catode (and even screen to cathode)
Why do I seem to insist so much on "to cathode"?

Hi J M Fahey

I don't understand why Vpk matter. For one, pentodes and beam power tubes has much flatter curves where Ip essential stay constant. In fact, in RDH4 page 34 in chapter 2, it stated the cathode current is constant at various plate voltage and is equal to plate current plus screen current. The reason of the plate current change and the calculation of rp=Vp/Ip is only because of the increase of screen current when Vp is lower. So the cathode current is totally uneffected by the plate voltage. So the common plate characteristics should not be a factor for cathode follower within the limits. From the surface, these tubes are even better than FETs where the source current actually change with drain current as it affect the channel pinching width.

I understand that is a simplistic view as RDH4 is only a cook book...a very good cook book, but that at least tell me that the plate characteristics are not that important. Tell me what am I missing, I am still learning.

[J M Fahey]

To say it in as few words as possible: we are designing a *power* amp, not a *current* amp.
If I had to swing a meter needle, from, say, 50 to 300 mA yes, plate voltage becomes somewhat irrelevant.
As RDH4 correctly says, a power pentode can swing from 50 to 300mA (a made up example but with reasonable real world values) with 100Vpk or with 500Vpk just as easily, provided we supply it with proper screen voltage (say, 400Vsk) and apply needed grid (Vgk) voltage as shown by tube curves.
That's what you are saying and within those limitations I agree.
BUT we are designing a power amp, and we need to apply to that transformer primary , besides that 250mA swing, *also* a, say, 500V to 100V swing , or 400V total.
We are after power after all, and that involves voltage swing times current swing, we are driving a certain impedance (which in this case would be 400V/250mA=1600 ohms).
Here we can not ignore plate voltage (and plate voltage variation) as you seem to understand.
Please remember that tube curves are drawn in a curve tracer *which is not an amplifier* and applies voltages and measures currents *straight to the raw tube, no resistances or loads involved* , it's a Lab apparatus, not any real world application.
Which is fine, it shows *universal* data, you then draw your own load lines or whatever on them.

[printer2]

That seems right. But if you admit that a power tube used as a CF is transparent and the tone comes from the stage before it, then you might as well replace the tube with a power MOSFET source follower. It will still be transparent, and won't need retubing, or a floating heater winding.

And that was my original thought.

[printer2]

Fully agree and add:
1) yes, plate curves must *still* be used, as they are the ones which describe how much current flows for a given plate to cathode voltage and grid to catode (and even screen to cathode)
Why do I seem to insist so much on "to cathode"?
Because it's implicit (thus not explicitly mentioned) in "normal" designs, but here you may be mistaken by thinking "hey! ... on a cathode follower plate voltage does not vary, so let's take it out of the equation" ... which is not true, in this case cathode voltage follows signal across the load so the *difference* between them still rules.
2) when I said "transistor-y" I didn't specifically refer to the output transformer (although yes, it's still *a* factor) but to the very different operating condition:
a) as Steve mentioned, it operated unity gain, with 100% local feedback, *ver* different
b) there's no grid rectification and heavy bias point shifting (there may be a little, ut much less important) , so:
c) no curve kink, no gain modulation, no important input signal clipping because of that, etc.
In sum, a very different way of operation.
Not surprisingly, the way *most* SS power stages operate.

Excuse my lack of knowledge if this seems dumb, but could we not have the screen supply of the pentode really saggy causing compression when we push the amp?

[J M Fahey]

In a conventional common cathode amplifier (like 99% of them out there) the output pentode has gain, which can be roughly defined as transconductance x load impedance.
Transconductance depends on many parameters, but an important one is screen voltage.
When you lower such voltage, transconductance lowers.
SO, when you push an amp, supplies sag (meaning they lose voltage), and transconductance lowers= less gain at higher power= compression. Simple as that.
This is *one* of the factors explaining power tube compression (there are more).
NOW, in a cathode follower amp (what we are discussing), 100% internal Negative Feedback means gain fixed at roughly 1 (really a little less, say 0.9 or thereabouts), so what we described earlier simply does not happen.
Unity gain is unity gain !!!
So, no compression on cathode follower amps, no matter what the sag.
Whose only influence is limiting maximum power available, but that's another topic.

I'm sorry but these and most other questions about tube amps, design or behaviour come because not many study tubes from the ground up, from absolute zero to whatever's needed, and everything in a solid functional block.
Not many have the full Map open before their eyes.
So there are all over the Net zillions of ready-made answers , such as "sag in screens compresses the signal" , which *IS* correct, *but* in certain circumstances, which are usually not defined.
That's why if you search a little , so many apparent contradictions appear.
And I'm talking about "correct" answers!!.
Forget about the actual cr*p and b*llsh*t !!

[Steve Conner]

I'd press "Like" on that one 10 times if it would let me.

Needless to say, I don't think the proposed topology, with a preamp tube driving a cathode follower output stage, would sound all that great. Many hybrid bass amps use a slightly more refined version, with a SRPP tube circuit driving a push-pull pair of MOSFET followers that drive the speaker directly, and bassists seem to like them. But that's kind of my point, the topology is better for bass or hi-fi IMO. The interaction between the power tubes and the speaker is part of the fun in guitar amps.

I built a hybrid guitar amp that runs the output transistors as current sources, not followers, to give a high output impedance. It sounds great. IMO.

[Alan0354]

I'd press "Like" on that one 10 times if it would let me.

Needless to say, I don't think the proposed topology, with a preamp tube driving a cathode follower output stage, would sound all that great. Many hybrid bass amps use a slightly more refined version, with a SRPP tube circuit driving a push-pull pair of MOSFET followers that drive the speaker directly, and bassists seem to like them. But that's kind of my point, the topology is better for bass or hi-fi IMO. The interaction between the power tubes and the speaker is part of the fun in guitar amps.

I built a hybrid guitar amp that runs the output transistors as current sources, not followers, to give a high output impedance. It sounds great. IMO.

Do you mean you use HV N channel MOSFET in common source to substitude the beam power tubes? If you look at Class A1 or AB1 where there will be no grid current drawn, the plate curves looks awfully like the FET drain curves. I was just thinking about using MOSFET and OT. It is easy to buy HV MOSFET. I have been using 1000V MOSFET a lot before.

Hey J M Fahey

I have no come back on your comment as I am still studying the more basic stuffs. I'll think about this when time comes.

[Steve Conner]

Using two N-channel MOSFETs with an OT works. The Russian guy KMG has done it. He used a low supply voltage and a custom wound OT with a low ratio, so he could use lateral MOSFETs (they only come in 160 or 250V) for easy biasing.

I went for a different approach, a complementary pair of BJTs driving the speaker directly. It looks the same as a regular SS amp output stage, but the base drive is by a floating transformer winding referenced to the emitters, not ground, so it ends up as a controlled current source rather than a voltage follower.

[printer2]

In a conventional common cathode amplifier (like 99% of them out there) the output pentode has gain, which can be roughly defined as transconductance x load impedance.
Transconductance depends on many parameters, but an important one is screen voltage.
When you lower such voltage, transconductance lowers.
SO, when you push an amp, supplies sag (meaning they lose voltage), and transconductance lowers= less gain at higher power= compression. Simple as that.
This is *one* of the factors explaining power tube compression (there are more).
NOW, in a cathode follower amp (what we are discussing), 100% internal Negative Feedback means gain fixed at roughly 1 (really a little less, say 0.9 or thereabouts), so what we described earlier simply does not happen.
Unity gain is unity gain !!!
So, no compression on cathode follower amps, no matter what the sag.
Whose only influence is limiting maximum power available, but that's another topic.

I'm sorry but these and most other questions about tube amps, design or behaviour come because not many study tubes from the ground up, from absolute zero to whatever's needed, and everything in a solid functional block.
Not many have the full Map open before their eyes.
So there are all over the Net zillions of ready-made answers , such as "sag in screens compresses the signal" , which *IS* correct, *but* in certain circumstances, which are usually not defined.
That's why if you search a little , so many apparent contradictions appear.
And I'm talking about "correct" answers!!.
Forget about the actual cr*p and b*llsh*t !!

Thanks for you explanation, hopefully this winter when I get more time I will be able to spend some time learning the intricacies of tubes, already have a lot of reading material waiting for me. So I am a little embarrassed to say, I don't get it (your response that is). I am assuming that you are saying that the pentode has a gain of 1 and no compression due to all the NFB that goes with the circuit.

Now just to be clear I will say we are using a mosfet as a source follower instead of a pentode (in order not to get confused). And the mosfet has a gain of 1 (actually a little less). Now this mosfet is following a pentode operating with gain (hopefully, this is where I am getting confused) and its screen is being fed by a saggy supply, when the source follower pulls a lot of current the voltage drops. I would think this tube would compress if the screen voltage sags. But then if the pentode does not have gain due to the source follower delivering the current behind it, well it does not matter if the screen sags.

[J M Fahey]

@printer 2
You are mixing two good individually answers into a not so good one.
1)

I am assuming that you are saying that the pentode has a gain of 1 and no compression due to all the NFB that goes with the circuit.

No, it's not "the pentode" but "a pentode connected as cathode follower" and in that case the gain and compression comments apply.
2) if a pentode is operated with gain and its screen voltage sags, yes, it will lose gain.
a) If that sag comes because the cathode follower output stage is driving a load , fine, you have compression.
b) you're having your compression in the last pentode with gain, and once a signal is compressed, it remains compressed (as long as the next stage is linear) so if it drives cathode followers or source followers or even bipolar *emitter* followers, you will still have your compressed signal.

@ alan354

If you look at Class A1 or AB1 where there will be no grid current drawn, the plate curves looks awfully like the FET drain curves.

Yes, I consider MosFets (bad) Pentodes, with one important difference: they behave *as if* they had screens (fine), BUT that virtual "screen voltage" is FIXED, meaning it does not sag, meaning it does not compress ... so the "magic" we were after, is not there any more.
I *have* built transformer output stages with Mosfets and the results, to put it mildly, have gone from horrible (with VMOS) to acceptable (with LMOS) but "hard" sounding even in the best case.
FWIW I have also made output stages with Darlington Bipolars (TIP 142) and results were not bad, but not that good either.

3) 3 guys I know have succeeded, so far:
a) Steve Conner, (yes, the guy upstairs), who transformer drives honest Bipolars with a wonderfully crunchy pentode, and
b) KMG, who needs no pentodes, but has studied his stuff *very* well, and by combining Fets, Mosfets and even Germanium diodes has achieved very good results.
c) SSGuitar's PHATT, who preprocesses signal, drives a small power pentode balls to the wall, reprocesses it and drives an SS power amp. Excellent sound.
So it *is* possible, but so far it means a lot of tweaking, experimenting, smashing everything against the wall (literally, no kidding), starting anew, and so on.
But it is definitely FUN.

[printer2]

OK got it, thank you. Originally I envisioned a small cheap pentode like the 6AK6 I mentioned and then into a mosfet into a transformer in order to get a tube sound but without the high cost of replacing the output at regular intervals. More of a design exercise than anything else as I went a little overboard on the tube thing and bought more than I probably will need in my guitar playing lifetime (can't resist a good deal).

[Austin]

OK got it, thank you. Originally I envisioned a small cheap pentode like the 6AK6 I mentioned and then into a mosfet into a transformer in order to get a tube sound but without the high cost of replacing the output at regular intervals. More of a design exercise than anything else as I went a little overboard on the tube thing and bought more than I probably will need in my guitar playing lifetime (can't resist a good deal).

This may be better posted in the "why not power pentodes for drivers" but would somebody help me understand why a 6ak6 or other smal signal pentode might be better than a el84 or a 6v6 or even larer tubes for a driver if money was no object? Is it the "inter-electrode capacitance" that might be a low pass filter with the grid stopper? Thanks