Cathode Follower......When to use.......

[Jack_Hester]

I've been following Tubenit's Carolina Songbird and more recently, his mod to the TOS.  It's been a slow struggle re-learning tube electronics, but I'm making progress.  Found some really good, but brief discussions on cathode followers.  Seems most say that they are good to precede a tone stack, as that seems to load up the signal coming from the pre-amp.  However, I've seen it in other places, or not at all, when it would appear to be a good circuit to use one.  So, how does one go about deciding when to include it in a build, or mod an amp to include one?  Is economics a major factor in deciding?  I know the '59 Bassman has one just before the tone stack.  But, why would it not precede the tone stacks in the AB763, if a tone stack loads the signal?  Thanks for all help.

Jack

[phsyconoodler]

Because a plate driven tone stack,although it has less gain due to losses,makes the tone stack more responsive.
  I build amps with both styles and it really depends on what you are going for in terms of tones.
 I always seem to go back to the plate driven stack because I like the way it responds and shapes the tone.
   Cathode followers have other uses in multi-channel amps for switching matrix designs too.
I'm no authority,just know what I like.

[jjasilli]

When to use?  Some say it's essential to the early Bassman - Marshal sound.  It does require an extra preamp stage as the follower, and is "wasteful" in that sense.  But we'd gladly engage in that for the tone we want.

The tone stack always puts a load on signal = insertion loss. The load is primarily set by the value of the slope resistor, in parallel with its source impedance, which is its driving stage.  Note that after a cathode follower the slope resistor is usually 33K - 56K; plate driven usually about 100K - 150K.  This approximately equalizes the plate driven and cathode driven designs. But the cathode driven tonestack tends to have less insertion loss and this happens more in the signal's midrange.  Hence Marshal tone seems darker and gainier than BF/SF Fender amps.

[LooseChange]

IMO.... Never. (Except in a loop)

[Geezer]

IMO.... Never. (Except in a loop)

What experiences or knowledge has shaped your opinion on this? 

I have built circuits with & without, & seem to lean towards circuits that employ CF's......those circuits tend to produce the sounds my ears like.

G

[jojokeo]

The effect caused due the biasing of a standard cathode follower yields a compressed signal of the positive half cycles but not the negative halves which yields 2nd order harmonic distortion to be generated.

Aside from the other comments IME a simple description of what I hear is when the stack is in between two gain stages you get more of "bluesy gain" and when the stack's after a CF stage you get more of a "crunchy gain".

[tubeswell]

I like DC coupled pairs (with an inverting stage and a CF stage) in an amp especially if there is another gain stage in front of the DC couped pair. The reason is that the bigger the signal you put into the DC coupled CF pair, then the bigger is the effect of the non-linear output. Those tweed bassman and marshall amps sound best for geetar when they are cranked.

The logical place to put a CF is in front of a lossy tone stack because of the advantages it offers in driving the stack, so the topology would be: input stage (could be parallel?)/vol control/DC coupled pair with CF/TS/PI...
JM2CW

[HotBluePlates]

The effect caused due the biasing of a standard cathode follower yields a compressed signal of the positive half cycles but not the negative halves which yields 2nd order harmonic distortion to be generated.

I would respectfully disagree with this concept.

First "proof" to think about: how significant could the distortion be when a cathode follower has 100% negative feedback? We can discuss further, because overlooking the feedback involved in a cathode follower will lead to well-reasoned, but wrong, conclusions.

... I've seen it in other places, or not at all, when it would appear to be a good circuit to use one.  So, how does one go about deciding when to include it in a build, or mod an amp to include one?  Is economics a major factor in deciding?

Probably.

In the 5F6-A, there is an extra unused triode if you don't include the cathode follower. In AB763 amps (with reverb and tremolo), you already have 6 preamp tubes. We'd need to add yet another to include a cathode follower for the two tonestacks.

In general, there are only a few reasons to use a cathode follower:

1. You need a low output impedance.
2. You want to buffer a stage from the circuit which follows it (perhaps cable capacitance from the output of a standalone unit).

If a circuit you're driving has a low input impedance compared to the stage ahead of it, the heavy load can drag down the gain of the previous stage. In a regular plate-loaded stage, we never quite get a gain as high as the tube's amplification factor, because of voltage division between the plate load and the internal resistance of the tube. So we use a plate load several times as big as the internal resistance. In the same way, the following stage's input impedance should be several times the plate load resistance, to avoid further reducing gain. If we need to drive a relatively low impedance (say, a smallish grid reference resistor dictated by a tube data sheet and the use of paralleled tubes) we might ease the design of the driving stage by placing a cathode follower between the driver and the following circuit.

You might be tempted to think a cathode follower would be a good way to maximize the gain of a preceeding stage. While we could use a bootstrapped cathode follower to present an extremely high input impedance, we'd still have less than the theoretical maximum gain from the preceeding stage due to the interaction of the plate load resistance with the tube's internal resistance. I could show you that with a mathematical example, but consider also that we'd waste a triode to try to get a gain of less than 100 (with a 12AX7), where we could use 2 12AX7 triodes with a gain of 50 each in cascade to get a total gain of 2500.

Anyway, I think it was the Marshall Silver Jubilee (or maybe the Peavey 5150) that went crazy with cathode followers after every gain stage. And you can get the same gain with somewhat less noise and for a whole lot less cost without all the cathode followers.

[VMS]

Here is one analysis that says that gain stage with direct coupled cathode follower reduces second-order harmonics:

http://ampbooks.com/home/classic-circuits/soldano-slo-preamp-4/

[Jack_Hester]

Fine business!  Mighty fine!  I am overwhelmed by the responses.  But, in a good way.  Think I'll print this out this evening and begin the digestion process.  I couldn't have asked for a better lesson.  Just wish the brain was capable of soaking it all up.  I sure am glad you guys are this passionate about tube amps.  I'm having to take it in nibbles.  Thanks a bunch.

Jack

[jojokeo]

Here is one analysis that says that gain stage with direct coupled cathode follower reduces second-order harmonics:
http://ampbooks.com/home/classic-circuits/soldano-slo-preamp-4/

Here's what Merlin says w/ scope info to support his conclusions:
http://www.freewebs.com/valvewizard1/dccf.htm

[VMS]

Here is one analysis that says that gain stage with direct coupled cathode follower reduces second-order harmonics:
http://ampbooks.com/home/classic-circuits/soldano-slo-preamp-4/

Here's what Merlin says w/ scope info to support his conclusions:
http://www.freewebs.com/valvewizard1/dccf.htm

Ah, my second go to page for tube amp info. I don't know enough to start debating which conclusion is the right one but maybe Merlin will chime in.

It would be nice to see scope picture of the output without the CF.

[tubeswell]

If you have a DC coupled pair with an inverting stage and a CF handy, measure the idle voltages including HT, the plate on the inverting stage, the cathode of the inverting stage, and the cathode of the CF stage. You will see that there is less current going through the cathode of the inverting stage than there is through the plate of the inverting stage (because some of the inverting stage's plate current is 'stolen' by the CF stage as 'grid current', which also affects the bias of the CF stage).

[HotBluePlates]

Thanks for pointing me to that! I hadn't realized the cathode follower was biased so close to the 0v grid line.

But the example is also a little bit out of context. Let's place it in the context of a 5F6-A Bassman.

Due to the nature of a cathode follower, you need to ignore the size of the input signal somewhat when you look at the follower's loadline. Instead, know that the gain will be somewhat less than 1, so focus instead on the output signal, as implied by the plate voltage axis. For Merlin's example, as we move along the loadline to 0v on the grid, the plate voltage (and as he says, really the cathode-to-ground voltage) indicated is ~205v. The idle point appears to be at about 195v, so the cathode follower can output a 10v peak signal (or 20v peak-to-peak, as Merlin uses for his test).

Merlin's example assumed a 300v supply and 195v grid and cathode-to-ground. But the 5F6-A Bassman schematic indicates a 325v supply and 180v cathode-to-ground. This means for us that the operating point is shifted down and to the right along the loadline Merlin presented, which allows for more headroom from the follower.

I won't redraw these lines here, because Richard Kuehnel has published a fairly exhaustive analysis of the 5F6-A Bassman, and I will make use of the conclusions he presented. The follower's operating point is still pretty close to the 0v grid line, but his analysis showed a headroom of 37.2v peak (as opposed to Merlin's example with 10v peak headroom).

Now, let's step backward from the output tubes. The schematic shows the bias is -48v. In the same way that we get compression and distortion as we approach and exceed the 0v grid line on the cathode follower's loadline, we will distort the output tubes heavily if we drive them beyond 0v on the grid. Since the bias is -48v, we can apply a positive 48v peak signal to drive the output tube grids momentarily to 0v. That will be our reference point as we work backwards through the amp.

The long-tail inverter is a handful to analyze. A simple rule-of-thumb for figuring on a napkin is that the gain to each output of the long-tail will be about 1/2 the gain one would expect from the tube used as a typical common-cathode gain stage. So if we might normally expect a gain of 52 from a 12AX7 stage, we could expect a gain of roughly 52/2 = 26 from a single output. In truth, the feedback loop reduces this gain somewhat. The gain from each half of the long-tail is also slightly different. The math is a handful to say the least, but Kuehnel found that the gains to be 25 and 26.6 before feedback, and 21.9 and 22.6 after feedback.

From this, an input of 48/21.9 = 2.19v peak will drive the output tubes to significant distortion. I used the lower gain number for the inverter, as it requires more signal from the cathode follower.

The tone stack is an issue. The only time you get a "flat" response (where loss is equal at all frequencies) is when the Mid pot is at maximum, and the other pots are at a minimum. So we oughta select something reasonable as a reference, and realize we will get more and less signal throughput at other settings. Let's work from all controls at half. I used Duncan's tonestack calculator, and plugged in Kuehnel's findings for cathode follower output impedance (531 ohms) and long-tail input impedance (1.9M, worst case), as well as 5F6-A parts values. The midrange dip represents the most loss under these conditions, and therefore the point at which the cathode follower has to supply the most signal to still drive the output tubes to 0v. The loss at the mid-dip with all controls at half is almost -13dB.

[HotBluePlates]

We want a ratio of gains instead of number of decibels. For voltage, that's -13dB = 20 log (A). Rearranging, -0.65 = log (A), so A = 0.22. That's a loss of a little less than 1/5th. Therefore, 2.19v * 5 = 10.95v peak into the tonestack drives our output tubes to 0v.

Merlin's condition suggested 10v peak input was the point at which the his follower example had significant distortion. Kuehnel showed in his book that given 5F6-A schematic conditions, we have the same response at 37.2v peak. This is well above the ~11v peak we need to drive the output tubes to distortion. So we've found that even if the cathode follower is poorly designed, the output tubes are already giving up and distorting well before the cathode follower. As it turns out, the "poor design" is as good as it needs to be.

It might be worthwhile to investigate when the phase inverter runs out of steam. A good amp design has the output tubes run into distortion first, so they are the limiting factor; it might be good to have the phase inverter and then preceeding stages start distorting soon afterwards.

Which brings me back to our original discussion. The follower has 100% feedback. We know the gain is something slightly less than 1. If it takes a 10v peak output (Merlin's condition) to drive the follower to distortion, then it also takes slightly more than 10v peak input to reach that signal level. Look at the curves for a 12AX7; there is not even a -10v gridline. So the signal level the follower can handle on its input will make a normal 12AX7 stage puke, and the distortion shown in Merlins o'scope traces is very much less than a typical 12AX7 stage with the same input signal. So distortion is very much reduced by the feedback inherent in the follower.

So, it is important to place the circuit in a context in order to evaluate it. Merlin also highlights that we can make probably anything perform poorly, and that could be a result of how we use it rather than the circuit itself. I just didn't want you to generalize and make a false assumption based on one example.

[jojokeo]

Thanks for alll that info HBP, I'll have to re-check my Keuhnel book's info too. It's been a couple of years. I wish they made that book better as many pages are falling out.