Are MOSFETS reliable ?

[Tilman]

Hi,

I might build a small series of amps with DC coupled MOSFET cathode followers as drivers for the output tubes. Crossover distortion and blocking distortion has to be avoided.
I know MOSFETs tend to blow for no reason in power amplifier output stages.
Tubes stop working for no reason too but it seems to me that they are somehow more reliable than MOSFETs.
Should I invest in a tube to do the cathode follower job or should I trust the MOSFET stuff ?
What do you think ?

Apart from the aesthetic reasons, I don't see the need for a tube at that position except maybe reliability.....

Regards!
Tilman

[Steve Conner]

Semiconductors are always more reliable than tubes, because they don't wear out or go microphonic. If solid-state amps blow up, it's because the maker cut corners to save money. Leaving out protection, making the heatsink too small, and so on.

A MOSFET used as a preamp stage needs the following protection: A gate stopper resistor to prevent oscillations, and a Zener (or back-to-back series pair) from gate to source to ensure the gate can never be overvolted. Some FETs even include the zeners, see the datasheet. You also must do your homework to ensure it doesn't run too hot: a TO92 device like the LND150 can't safely dissipate as much heat as a 12AX7.

With these precautions it will still be going long after we are all dead, unless guitar amps go out of fashion in the 2030s and it gets landfilled.

[R.G.]

Kewl! The MOSFET Follies have made it into the boutique amp market!

[defaced]

Using RG's aforementioned MOSFET follies article, I posted a couple of questions, got some great design help from RG and the guys here, and ultimately posted a complete schematic and layout for doing this. I'm admittedly missing the protection Zeners from the gate to source, but I've had it in an amp for over a year without a problem. For me, the bipolar power supply was the most annoying requirement, but depending on your output stage, that may not be a problem.

[Tilman]

Hi,

as far as I understand it, you guys are more on the MOSFET side of the road for this one.
I think I will try the MOSFET approach first and see how it sounds. As a backup I can have a noval socket mounting hole close to the PI stage .

Thanks!
Tilman

[Steve Conner]

Well, I have a couple of amps full of FETs that have been working for more than 10 years. They show no signs of running out of electrons.

In this case, my main reliability worry would be the MOSFET followers overheating the power tube grids, causing the tubes to be more unreliable. A tube follower has an inherent current limit, but a TO220 size FET can deliver several amps. If you want to ship a commercial product like this, you will need to carefully investigate worst-case grid dissipation, and maybe size the grid stopper resistors to keep it under control. You may be opening a can of worms here, as no current production tube datasheet actually lists a maximum dissipation for g1, and hardly any of the NOS datasheets did either. The implication being that the tubes were never designed to run AB2 at all.

RG, fame at last!

[Tilman]

Hi,

I just come back from reading the thread where defaced designed such a MOSFET thing with some great help from RG.
@DEFACED
Did you actually install your PCB into the KT88 amp and if so - what is your impression - soundwise ?

@Steve
Opening a can of worms was not my intention ;-). At the moment I am starting to like the "inherent current limit" of tubes very much.
On the other hand, if a drain resister is used for the follower and a large enough grid stopper at the output tubes too - things might be OK.

Tilman

[Steve Conner]

Well, the problem is that the tube follower current limits by drawing grid current, which may recreate the blocking distortion you were trying to get rid of.

[Tilman]

True, +1 MOSFETs.

[defaced]

Did you actually install your PCB into the KT88 amp and if so - what is your impression - soundwise ?

Yes, and completely transparent. I have wanted to test to see what's going on during operation (am I getting to AB2, etc), but only recently have I owned a scope. So right now, all I've got is my ears, and they didn't hear a change after I installed it so I'd say it's at least worth prototyping to see what it does in your amp

[R.G.]

RG, fame at last!

It feels great... oh! Wait! I'm starting to feel all hollow inside! Oh NOOOO!!

[redelephant]

It sure would be interesting to find out if the MOSFETs are actually doing anything as you say: getting you to AB2 . If you havent lowered the OT primary impedance it may well be that the tubes swing down to clip before gridcurrent is required.
Or am I way off?

I tried using MOSFETs to drive the grids in my OTL amps, and the clipping I got was pretty nasty. Real SS like. Not b/c the MOSFETs were clipping, they were not, but it seems the output tubes were driven to a rather harsh and complete cut-off which resembled SS clipping. I didn't take the time to investigate too much, simply removed it after concluding it was not worth the extra power.

I think the zener mod might be just as good, unless extra power is the main goal, and not just preventing blocking-dist.

[R.G.]

It sure would be interesting to find out if the MOSFETs are actually doing anything as you say: getting you to AB2 . If you havent lowered the OT primary impedance it may well be that the tubes swing down to clip before gridcurrent is required.
Or am I way off?

I don't know about way off. However, if what you mean is that the tubes can't swing further in to conduction, they can do this with more grid drive. The same tube and OT will swing to higher currents with positive grid drive as long as the driving element can push the current for the grid. In transistor terms, the output tube "saturates" to a lower voltage and higher current.

I tried using MOSFETs to drive the grids in my OTL amps, and the clipping I got was pretty nasty. Real SS like. Not b/c the MOSFETs were clipping, they were not, but it seems the output tubes were driven to a rather harsh and complete cut-off which resembled SS clipping. I didn't take the time to investigate too much, simply removed it after concluding it was not worth the extra power.

The MOSFETs work their stuff on the other end, on the tube that's driven on. In a class AB amp, one tube is (nearly) always off, so driving it off further makes no difference at all. It's the tube that's driven on that gives you the extra power and softer clipping.

A 6L6, for example, normally can't get under 50V (roughly, about) with 0V on the grid. If you can supply current to drive the grid positive, the grid actually helps yank more electrons out of the space charge cloud at the cathode and makes more current flow to the plate. The tube actually conducts more current and "saturates" to a lower voltage. This is what gives the extra power. It uses more of the power supply available to it.

What makes the clipping softer is that instead of the grid simply stopping at a fraction of a volt over the cathode voltage, it can be pushed a certain amount more positive. The grid's being clamped to 0V from the cathode can be set up to be as razor-sharp as any opamp clipping. The guys who used to make logic circuits out of tubes use to tie the grid of a triode to B+ with a 470K to 1M resistor to force the grid to cathode value by grid current. This was referred to as "running in clamp", and produces a razor-sharp entry into distortion.

I think the zener mod might be just as good, unless extra power is the main goal, and not just preventing blocking-dist.

That's probably true. Elimination of blocking with a zener clamp is a real advantage. It's one of the only genuinely new circuit techniques with tubes I've seen to arise since the 60s, as far as I can tell by reading. Some ...wink, wink, nudge, nudge... tube amp makers have crowed about and even patented circuit techniques which were well known back in the Golden Age, but the zener clamp was new, I think.

AB2 was always about more power, back when tube followers were used. As noted, you can still get blocking distortion by overdriving the grid of the follower. However, it is much, much harder to drive a follower into grid conduction just because the cathode follows the grid up.

[Tilman]

The grid's being clamped to 0V from the cathode can be set up to be as razor-sharp as any opamp clipping. The guys who used to make logic circuits out of tubes use to tie the grid of a triode to B+ with a 470K to 1M resistor to force the grid to cathode value by grid current. This was referred to as "running in clamp", and produces a razor-sharp entry into distortion.

Oh, I am confused now. If I read this right, the grid gets clamped to near 0V if we apply a higher voltage to it. As long as the grid is not coupled via a capacitor to the driving stage - I did not see that as a problem.
Now you are saying clipping gets razor sharp than.
Does it get softer or razor sharp ? Confused...sorry!
Tilman

[R.G.]

Oh, I am confused now. If I read this right, the grid gets clamped to near 0V if we apply a higher voltage to it. As long as the grid is not coupled via a capacitor to the driving stage - I did not see that as a problem.
Now you are saying clipping gets razor sharp than.
Does it get softer or razor sharp ? Confused...sorry!

No problem. The issue is not one of voltage, it's the available current.

When a 12AX7 grid (that's the tube I know the approximate impedances for) goes from a bit more negative than the cathode to a bit more positive than the cathode, it begins conducting. Looked at another way, it goes from an almost non-conductive extremely high impedance when it's negative, down to between 4K and 10K ohms depending on circuit conditions and the tube itself when it begins conducting because it's slightly positive with respect to the cathode.

If the grid is driven from a moderate impedance, like the plate impedance of another 12AX7, which is about 62K depending on the actual operating conditions, then the grid follows the driving signal as long as it's in the negative/no-current region because 62K can drive many megohms easily enough. When the 62K tries to pull the grid positive, it can't. This is because the incoming voltage is driving through its 62K source impedance into the (about) 5K load impedance of the now-conducting grid. So the driving signal is attenuated to 5K / (62K+5K) = 0.0746 or 7% of the driving voltage. The grid voltage can't change much because any signal changes are suppressed , so the current to the plate can't change much.

If you drive the grid from a source with a 10K internal impedance, then the change in signal level when it goes positive is only a 1/2 reduction, so the amount of clipping is very much reduced.

If you drive it from a 10 ohm source, the driving signal can pull the (about) 5K grid load wherever it wants it to be, so the grid still influences the plate current and signal still comes out the plate, much less distorted.

If you go the other way and drive the plate from a very high impedance, like maybe 1M, then when the grid starts conducting, the source impedance prevents the grid voltage from following the signal because the signal can't pull the grid voltage around any more. You get heavy distortion.

Did that help? It's the available signal current and source impedance that matters when the grid starts pulling current.

Same thing happens in all tubes, even output pentodes and power beam tubes.