Post Phase Inverter Cathode Follower - Pointless in anything under 100W?

[roberto]

Thanks R.G.

may I ask you which other methods can be applied for the same purpose?
To correctly dimension the needed current, using logic I would calculate:
(B+plates - V-mosfets) / Rgridstopper.
I would also use V-mosfets <= 2 x V-bias, so at least -100V.
So, considering B+ around 500 V and grid stopper around 5k6, it’s something like 110 mA.

I really would like to know more about the possible designs.
If I’d open a thread about this, would someone more experienced collaborate to explain the design and draw a basic design that can be implemented in existing amps?

[R.G.]

may I ask you which other methods can be applied for the same purpose?

Sure. The three types of devices we have available to us are (1) tubes (2) bipolar transistors and (3) MOSFETs. For bipolars and MOSFETs, you can have either N type or P-type. Tubes only work as N-type devices (that is, plate positive wrt cathode).
You can make simple followers from any of these. It used to be that you couldn't get bipolars capable of withstanding the couple of hundred volts needed for this application, but modern bipolars can do the job now. So for a simple one-device follower, you'd use one of these.

The fourth option is to use more complicated, multiple-device followers. Simple followers are heir to nonlinearities with big signal swings close to their power supplies, especially if they use simple resistor loads, as the resistor does all of the pulling toward the non-device side of the follower's power supply. The first enhancement to a simple follower is to put a constant current load on it. The CC load then takes over the job of pulling down on the load while the main follower pulls in the up (more positive) direction. One alternate is to make a complementary follower from two devices, so one can pull positive, one negative. Complementary followers use one N-type device, and one P-type device. If you did a complementary, you'd probably use an N-type and a P-type. If you did a CC loaded follower, you'd use two N-types.
Another possible "follower" is the complementary or two-transistor compound, set up for unity gain. These things, along with complementary followers, can be d@mn fast and low distortion, but require some skill to set them up correctly.

With a more complex follower, you'd have to decide whether you used the "raw" few-device follower or put an amplifier/driver ahead of it to get predictability, lower distortion, etc. Then the complexity really opens up.

To correctly dimension the needed current, using logic I would calculate:
(B+plates - V-mosfets) / Rgridstopper.
I would also use V-mosfets <= 2 x V-bias, so at least -100V.
So, considering B+ around 500 V and grid stopper around 5k6, it’s something like 110 mA.

Actually, the grid does not change from near-infinite impedance at Vgk<0 to a short circuit when Vgk goes positive. The grid has a fairly high impedance even when it's driven positive and this is usually much higher than the grid stopper's resistance. Also, you'll probably not drive it anywhere near B+. For a power tube with a cutoff voltage in the -50 to -80V range, driving the grid positive by 10 to 20V is likely to be all you can expect. Power tubes will run out of cathode emission before they get too much above Vgk=0. This is heavily dependent on the design of the internals of the tube.

As to how big the current is, I'd guess - and this is a guess! - that it's significantly less than screen current ever gets. Maybe 1/10 or less of screen current, if that. So you'd be good with a current of much less than that 110ma number. Which is good, as that would represent a power of 500V x 0.11A = 55W that would have to be wasted somewhere to do the driving.

[Colossal]

I am very interested in is topic and thanks for getting it started. One request, if possible...while the discussion of MOSFET and other drivers is very useful from the theory perspective, can we please focus on using tubes as the driver for an actual application?

[martin manning]

The GE 6L6GC data sheet (1959) has some Ig1 data. For Vg2 = 400V, and Vg1 = +20, Ig1 begins to shoot up a little below 150V Va. At that point, Ig1 is 26.5 mA, and Ig2 is 57 mA. At a more modest Vg1 = +4, Va = 100V, Ig1 is 6 and Ig2 is 45.

[Tony Bones]

The following I say with respect and an attitude of wanting to learn: I've heard "transparency" or the lack of it as a criticism of practically every audio circuit ever used. What makes me a little nuts is that no one can ever say what this "transparency" they refer to actually is, or how it might be measured in the interest of making things more transparent. About the best working definition I've been able to come up with is that transparency means "it sounds like I expected it to sound", and that lack of transparency means "it doesn't sound like I expected it to sound". I'd love to design for "transparency" if I could figure out what it is and when I'm being successful.

I totally get that, which is to say that I feel the same frustration. Though it's probably a safe bet that hundreds, maybe thousands of engineers have felt it ever since the first one that realized "faithful audio reproduction" isn't the slam-dunk they thought it would be.

But, surely you must be able to recognize something that *you* would describe as transparency. That's where you have to start, isn't it? I didn't really start to trust my own ears until I started spending a lot of time with other people listening critically to the same systems. (Here I'm talking about music playback systems, a.k.a "hi-fi", but the principles are the same.) Not surprisingly, we heard much of the same things and developed a crude vocabulary to describe it. I know, human perceptions are flawed and all that, but you work with what you've got.

I suppose that if I were a chemical engineer and wanted to understand wine connoisseur in order to engineer delicious wine, then I would need to become a wine connoisseur myself, wouldn't I?

[pompeiisneaks]

The following I say with respect and an attitude of wanting to learn: I've heard "transparency" or the lack of it as a criticism of practically every audio circuit ever used. What makes me a little nuts is that no one can ever say what this "transparency" they refer to actually is, or how it might be measured in the interest of making things more transparent. About the best working definition I've been able to come up with is that transparency means "it sounds like I expected it to sound", and that lack of transparency means "it doesn't sound like I expected it to sound". I'd love to design for "transparency" if I could figure out what it is and when I'm being successful.

I totally get that, which is to say that I feel the same frustration. Though it's probably a safe bet that hundreds, maybe thousands of engineers have felt it ever since the first one that realized "faithful audio reproduction" isn't the slam-dunk they thought it would be.

But, surely you must be able to recognize something that *you* would describe as transparency. That's where you have to start, isn't it? I didn't really start to trust my own ears until I started spending a lot of time with other people listening critically to the same systems. (Here I'm talking about music playback systems, a.k.a "hi-fi", but the principles are the same.) Not surprisingly, we heard much of the same things and developed a crude vocabulary to describe it. I know, human perceptions are flawed and all that, but you work with what you've got.

I suppose that if I were a chemical engineer and wanted to understand wine connoisseur in order to engineer delicious wine, then I would need to become a wine connoisseur myself, wouldn't I?

I think the problem with that analogy, or maybe not problem, it may actually tie in tightly, is that wine connoisseur's are often completely bull. There have been multiple blind taste tests with people that have pretty high reputations and end up not being able to distinguish a 100$ bottle from 3 buck chuck. Because it's about sensation (taste or hearing) it is extremely relative, and it's easy to trick ourselves into believing something that isn't real. This is very common in high end audiophoolery. This is an approach that seems to be common here, is that we feel that without solid proof and a clear definition of terms, you can't accurately know or measure something like this. It doesn't mean I don't thinks some things are very clearly noticeable, but defining and measuring them has become painfully unobtanium level stuff.

My 2c anyway.

~Phil

[Tony Bones]

I hear you and don't disagree. Human perception is a notorious liar. But it's all we got.

I *know* that within the morass of unreliable sensations that I experience, some of them are reliable and reproducible. I'm not going to throw those out with the bathwater. They're all I have to work with. What's the point if I choose not to trust any perceptions, mine or yours? Seriously, I should just throw my hands in the air and find something else to do?

People that demand proof are on a fools errand. They're traveling a road that I have no interest in because I know that proof is no more reliable than perception. I know that's a bit philosophical, but think about what "proof" means in the social sciences. It's not like a mathematical proof at all. Further, I understand that progress can be made without knock-down proof. It's a process that many amp designers use successfully. They start by making something that sounds good to them, then they give it to someone else. That person say they want more "xyz". So the designer tries to guess what "xyz" means and experiments until they think they've got it. Then they give it back and ask "is that better?" It's a process that works and nobody gets bogged down with triple blind experiments intended to satisfy some people that are never going to use the gear anyway.

I believe it's impossible to describe the experience of eating pecan pie. All we can do is say "it tastes just like chicken" and hope the other person has eaten chicken at some point in their life.

Measuring and quantifying what makes one piece of pecan pie more delicious than another must start with a personal understanding of what that means. In other words, we have to eat some and agree with others about the sensory difference. We can't depend on others to describe the difference. No great cook ever learned their craft without tasting the food they prepare.

[martin manning]

Well, I don’t think we are going to solve that one here, so getting back to the amps and volts of low impedance drivers...

A recent thread concerning a Musicman amp got me looking at its output section, where the output tube grids are provided with a fixed, low impedance, zener regulated supply voltage, and driven at their cathodes via BJT’s. Seems like this is another way to avoid hard grid clipping.

[gingertube]

NOT POINTLESS EVER

A tube cathode folower or mosfet source follower between phase splitter and output tube grid allows us to not repeat the bad design of most guitar amps, over many years.

What is wrong with those designs - inadequate control of the output tube grid by use of too high a grid leak resistor.

A peek at a KT88 datasheet will show what I mean:

The KT88 is a 42W rated tube.
If run at 42W in Cathode bias then you have a specified maximum grid leak resistance.
If run at 42 W in Fixed bias you have a specified maximum grid leak resistance which is 1/2 or 1/3 of the above.

Any grid current generates a voltage across the grid leak resistance which subtracts from the bias. That increase tube current which increases grid current which subtracts more from the bias, round and round untill boom - the tube melts down.

Hence the maximum grid leak resistance specification and why the max. resistance can be a bit higher for cathode bias where the increase tube current is partially opposed by increased bias generated by increased voltage drop in the cathode bias resistor.

Next:
On that same KT88 Datasheet you will see a second set of max. grid leak resistance ( about double) for when the tube is run at 35W or less dissipation (<= 70% of 42W).

For other tubes datasheets you will generarally just see the maximum grid leak resistances for fixed and cathode bias at their maximum rated power dissipation.

Designers have typically just used a value of twice the maximum allowed value and then set bias for IDLE at 70% of maximum dissipation.
They do this so as to not load the phase inverter too much with low value grid leak resistors.
This makes it all good and reliable at idle, but it is still tube abuse when you start applying signal or, heaven forbid, actually overdriving output tubes. Then bias stability becomes so bad as to contribute the the very high rate of output tube failures that we see in Guitar Amps.

Adding the tube cathode follower or MOSFET Source Follower allows use of a sensible gridleak resistance which then imparts tight control of the output tube grid and good bias stability.

It is good design - always, regardless of output power level and will result in output tube reliability improvements, less burned out screen resistors etc..

Sonically? Well sufice to say that the HiFi tube Amp I designed (called the Baby Huey and posted at DIYAudio) used MOSFET Source folowers to drive the output tubes. Some guys have done PCB layout from my design and have now sold around 300 board sets for stereo amps. No complainers about sonics. (If there was going to be a complainer it might have been me on another issue entirely, I got nothing for my design, not even the offer of a couple of free boards.)

Cheers,
Ian

[roberto]

Hi Ian,

is this one the design you are referring to?

[SoulFetish]

SVT and SSS DBeasley asks about are direct coupled, so there is the possibility of AB2 operation.

SVT is directly coupled, but will not operate in AB2. The 47k grid stoppers will clamp the grids, once the grid impedance starts falling rapidly. I can't imagine any positive signal swing surviving the voltage drop across those, right?

[roberto]

The first enhancement to a simple follower is to put a constant current load on it. The CC load then takes over the job of pulling down on the load while the main follower pulls in the up (more positive) direction.

Clear thanks. Something like the evolution of the Pass amps or the tubecad studies?

Another possible "follower" is the complementary or two-transistor compound, set up for unity gain. These things, along with complementary followers, can be d@mn fast and low distortion, but require some skill to set them up correctly.

Something like this can be in that direction? https://www.diyaudio.com/forums/solid-s ... ost5027612

The grid has a fairly high impedance even when it's driven positive and this is usually much higher than the grid stopper's resistance. Also, you'll probably not drive it anywhere near B+. For a power tube with a cutoff voltage in the -50 to -80V range, driving the grid positive by 10 to 20V is likely to be all you can expect.

Clear, so I could consider to asymmetrically supply the CFs at twice the extremes of the supply, like +40 and -100 to -150V and (following the next quote) let's say 50 mA, so around 10W?

The GE 6L6GC data sheet (1959) has some Ig1 data. For Vg2 = 400V, and Vg1 = +20, Ig1 begins to shoot up a little below 150V Va. At that point, Ig1 is 26.5 mA, and Ig2 is 57 mA. At a more modest Vg1 = +4, Va = 100V, Ig1 is 6 and Ig2 is 45.

[gingertube]

Hi Ian,

is this one the design you are referring to?

Yes that is it.
Current Source loaded Mosfet source followers plus the Baby Huey feedback scheme, balanced shunt feedback from output tube anodes
Cheers
Ian

[roberto]

Yes that is it.
Current Source loaded Mosfet source followers plus the Baby Huey feedback scheme, balanced shunt feedback from output tube anodes
Cheers
Ian

Thanks, so from your design I see that there's no need to double the bias voltage to get linearity, 1.5 times is enough with that design of CCS-loaded source-followers.
And the negative side is also useful to give more swing to the PI.

EDIT
I've two questions more:
- why going to two different CCS designs for the PI and for the source followers?
- have you tried to implement the R3&R4 as shown here: http://www.elektronik-kompendium.de/pub ... rr2pol.htm

[didit]

Hello --

Catching up with the week.

Have I missed something key?

Yep. No better than tubes. Just cheaper, longer lived, and smaller.

OK. What precisely I'd understood. And given only those minor benefits, will stick with 12AU7 or similar. Some in my spares box are a decade plus older than me, and anticipate our children will still be able to find new spares fifty plus years from now. Nothing wrong with silicon. I'd doubtless be unemployed & broke without it.

NOT POINTLESS EVER
[...]

YES, EXACTLY ON POINT. Have rough idea (mixed onto my growing heap of other rough ideas) for a slightly stressed ~15W quad of 6005/6AQ5/EL90 in a Dumble-esque head. My back still hurts from hefting a #102 clone that was delivered recently.

(If there was going to be a complainer it might have been me on another issue entirely, I got nothing for my design, not even the offer of a couple of free boards.)

As a diyaudio aside, have Z565 pair I've decided want to be in a Baby Huey. I would want PCBs when this project gets attention finally. And happily will send you my spares, given lots are 5-10 minimum.

Best .. Ian