• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Input impedance experimentations, aiming to build an A2 stage

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Credit where due: George (Tubelab_com)
Well, yes, I supposed that. I just meant to say that it was you who recommended me to look for it. In reality, I already knew of this possibility, just I didn't get the occasion to work on it yet. This just accelerated the process :)

I did tried the grid stopper resistor, and, as expected, it worked, but THD climbed from 0.4% to 3%, and that's the reason I wouldn't use it.

Re-reading some books, I recalled the grid-cathode zener, which I definitelly will include in my design. Thanks smoking-amp for pointing it out.

About the drain resistor, as I'm winding my own transformers, I'll just probably make a separate supply for the MOSFET, and only if that is not enough, include the resistor. A thin wire will be enough for the winding (~ 0.2mm) and it won't take too much space in the bobbin.
 
Last edited by a moderator:
...plus a Zener from cathode to grid...
Just a reminder, the zener will need an ordinary silicon diode in series to make sure it doesn't also clamp negative signal swings at (-0.7 V) or so. I got bitten by that once in a different application, many years ago, when I forgot that Zener diodes actually conduct in both directions!

The diode should be fast enough for full-bandwidth audio, and able to cope with the maximum negative peak voltage.

Vishay's 1N4148 datasheet specifies 75 volts max reverse voltage, which may be enough, depending on the drive signal.

-Gnobuddy
 
This just accelerated the process :)
Glad I could help. :)
I'm winding my own transformers
Impressive! As a boy, I've taken apart commercially made chokes (for old-fashioned long fluorescent lights), and in the process, realized I could never pack the laminations as tightly myself when re-assembling. I also lost some blood to those sharp-edged metal laminations. :)

I'll just probably make a separate supply for the MOSFET, and only if that is not enough, include the resistor.
I used a small, off-the-shelf 12.6-0-12.6 V transformer to build a small power supply for the MOSFET drivers. It generates +/- 40 V rails by using the entire 25.2V winding, half-wave rectified for each polarity.

Thinking about this now, maybe I should only use half the transformer winding for the positive rail, which should leave me with something closer to +20 volts / -40 volts.

The output valve I have in mind is a 12AB5 (same as 6V6, but in a smaller 9-pin envelope.) The datasheet shows positive G1 voltages up to +15 volts, so perhaps a +20V rail is really all that's necessary, if the grid stopper resistors are small enough.

The other possibility is to use something like 1k - 1.2k grid stoppers. 15 volts at the grid, around 25 V max across the grid stoppers at 20 - 25 mA grid current. The additional distortion might turn out to be a good thing for my purposes (electric guitar), though I will have to try it and see what it sounds like.

Tubelab George uses +/- 160V rails on his MOSFET driver in one of his current threads. I wonder if he incorporates additional protection against driving too much power into the output valve control grids.

-Gnobuddy
 
I wonder if he incorporates additional protection against driving too much power into the output valve control grids.

Seems not.

From here: Limiting the grid current of PowerDrive

"The PowerDrive circuit (as concieved) can indeed feed the grid of a tube as much current as it can eat. Under normal music conditions transitions into the positive grid current regions should occur only on transient peaks, unless you run your amp into clipping a lot. These conditions should not cause excessive G1 dissipation since the time in the grid current region is very low. It should be noted that it might be possible to damage a tube by long term sine wave testing at or above the clipping level since the tube will see grid current on every cycle. I have not seen this happen on conventional G1 drive"

I just simulated the zener+regular diode circuit.

It makes the mosfet and tube grid happy, but not the diodes! If still using 70V for the V+, you can get 4A current during the overload situation, which seems like it can kill all the zeners I have at hand. A series resistor reduces the effectiveness, so you end up not knowing which device will die :D
 

Attachments

  • a2_nmos_diodes_high_current.png
    a2_nmos_diodes_high_current.png
    40.5 KB · Views: 163
Last edited:
Using a 20 Amp Mosfet is asking for trouble, not just with the grid1 or Zener or diode, but input capacitance too.
No, I'm planning to use a STP11NK40Z, 5.5 A, and less than half the input capacitance (< 1nF) that the one I used in the simulation (just because I wanted a rough approximation, I didn't spend the time to look for a more precise model).

I've got IRFU9310, 1.1A, about 300 pF input capacitance, but it can barely dissiapte 1 W without heatsink. So I decided to use the bigger one.

By the way, how does a smaller mosfet help grid1 or zener or diode?
 
... I've got IRFU9310, 1.1A, about 300 pF input capacitance, but it can barely dissiapte 1 W without heatsink. So I decided to use the bigger one....
I think a 5.5A part is not going to perform as good and using a heatsink is recommended. If a lager part is needed or you want to avoid IRF p-channel kink, try something like FQP2P40 still available from Digikey.
 
It makes the mosfet and tube grid happy, but not the diodes! If still using 70V for the V+, you can get 4A current during the overload situation
Well, sure, you can't just put a 24V Zener across a 70 V DC power supply and expect anything to be happy! A drain or source resistor to limit current flow is necessary. Or an active current limit, like the short-circuit-protection circuits often used in solid-state power amp output stages.

I think it may be much simpler to simply limit the maximum AC signal coming in to the gate of the driver MOSFET. That is a high-impedance location with small signal currents, and easy to limit. By comparison the source of the same MOSFET is a low-impedance node capable of much higher currents - which is exactly why it makes a good grid driver for an output valve in A2 mode!

Just an idea - the attached schematic shows a MOSFET driver with a bipolar junction transistor added as a current limiter, to protect the delicate control grid of the output valve.

In the image, (R3+D1) represents the input of the valve being driven. The 60 V negative supply V3 biases the "grid" to (-60 V). Vgrid (the drive voltage to the grid from the MOSFET) can swing from (-60 V) up to (0 V) without any grid current flow, but if it swings positive, D1 conducts, and grid current starts to flow, and the MOSFET is now driving a load of about 1 kiloohm instead of infinity.

The problem we have been discussing is that, left to itself, the MOSFET can now drive up to nearly 70 mA of current into the grid, and this will quite likely fry the grid. (Assuming the input of the valve acts like roughly a 1k resistance once grid current starts flowing.)

This is where Q1 and R2 come in. R2 is quite small, and unlikely to cause much THD due to non-linear grid current flow through it. However, if enough current flows through it, the voltage drop across R2 will turn Q1 on, clamping the gate voltage of M1 and not allowing it to increase any more.

With values shown, this happens at about 16 mA in the LTSpice simulation, where you see the "knee" in the curve. R2 should probably be a 100 ohm trimpot in practice, to allow setting the maximum grid current. Note that R1 plays a role - as the MOSFET is driven harder and harder, but the gate voltage is clamped by Q1, more and more current will flow through R1. This will then flow through Q1, and into the grid of the output valve.

This is why the plot shows the grid current rising slowly past the 16 mA clamp limit, as the gate of M1 continues to be driven more and more positive, all the way to +70 V in this simulation. However, it doesn't increase beyond 21 mA in the simulation.

Note that the plot shows I(R3) - the current flowing into the grid of the output valve - as a function of V1, the signal driving the MOSFET gate.

In practice, one would turn R2 to a low value, then slowly turn V1 up until the output valve is in full clipping, and then just a smidge more. Now adust R2 so that it limits the grid current, but not enough to bring the valve out of clipping - there is still enough grid current for full output, but not enough to destroy the valve.

Adjusted this way, the current limiting should never become audible, because the output valves will be deep into clipping before Q1 steps in to protect them.

Since R2 is so small (a few tens of ohms), I don't think it will have a significant effect on THD. However, if this is a worry, by adding a second transistor (PNP), R2 can probably be moved to the drain side of M1. R2 will turn on the PNP transistor, which in turn will turn on Q1. The current limiting will still work as described here.

An even simpler solution is to simply wire a Zener diode (with series silicon diode) to limit the maximum positive signal voltage at the gate of M1. Choose the Zener voltage to give full drive to the output valve, without driving the grid so far positive that it is in danger of overdissipation and damage. R1 is still needed, to avoid damage to the Zener diode.

-Gnobuddy
 

Attachments

  • MOSFET_driver_w_current_limit.png
    MOSFET_driver_w_current_limit.png
    49.5 KB · Views: 115
Thanks Gnobuddy, I played around with all your suggestions, and I agree. I just didn't understand how some other's driver circuits just don't have any current limiting mechanism, even though being powered by high dual voltages. So instead of looking too much for elaborate solutions, just to hide my lack of understanding, I tried to understand as much as possible the un-protected circuits.

My conclusion so far is that other simple approaches just rely on the fact that music is not a constant high level input, so the average power in the output stage remains on the safe side. But feeding a sine wave input might be asking for trouble, and so a very high (maybe accidental) input level.

So, I'm building a prototype monoblock for real world testing, and I want to use dedicated +/- supply for the mosfet stage, and a small drain resistor. Do some testing, to see how it works under sine wave conditions, and probably adding zener or bipolar for current limiting afterwards, just because they are cheap.

But the chassis is not a big one. I'm using a 20x25 cm aluminium sheet, to fit inside a wooden structure. I plan to place the mosfet stage on a perforated circuit board in the free area between the noval and octal socket, and the two transformers at the back. They will occupy most of the free area at the back. And I need three rectifiers with their filtering. Suggetions on layout/placement?
 

Attachments

  • mini_amp_a2.jpg
    mini_amp_a2.jpg
    732.8 KB · Views: 86
For this grid current cancellation trick to work, you would actually have to stay in positive grid voltage territory even on negative signal peaks........ George (Tubelab_com) was doing this long before I had the same idea

That's actually where I started sometime around 2004 - 2005. I tried to run an 811A in such a manner that the grid stayed positive all the time. I even called this A3. The driver circuits of the day would exhibit distortion, kinkiness, oscillation or some other anomaly at the point where grid current abruptly increased as the grid transitioned positive. There was a popular schematic circulating on the web and this forum that used a 6V6 as a cathode follower driving the grid of the 811A without a resistor to ground. All the current from the 6V6 flowed into the 811A's grid. This guarantees A3.

I built it, didn't like it, started messing with it. One of the dumb blonde experiments involved replacing the 6V6 with a mosfet. This did improve the sound but the real problem turned out to be the extremely high Rp of the 811A. This could be fixed by piling on the feedback, but then the question became "do you want flabby bass, or dull and lifeless midrange in your SET." I chose neither and tried better output tubes.

I did tried the grid stopper resistor, and, as expected, it worked, but THD climbed
You can easily limit the current that the mosfet can stuff into the grid of the output tube by adding resistance in series with the drain of the mosfet. As the grid draws current it will pull the mosfet's drain down. Size the resistor such that the mosfet saturates (drain voltage = source voltage) when the desired grid current is reached. Since the grid voltage for this current is unknown, and varies from tube to tube, some measurements with a scope will be needed.

With the drain current safely limited by this resistor, a reasonably sized zener diode could be used. Remember that zeners, and all diodes are voltage variable capacitors by nature, and regarded by many as something not really wanted in the audio signal path. The audio signal path here is at a very low impedance, so a few pF shouldn't make much difference.

Using a 20 Amp Mosfet is asking for trouble, not just with the grid1 or Zener or diode, but input capacitance too......By the way, how does a smaller mosfet help grid1 or zener or diode?
It doesn't, much. The Crss, or reverse transfer capacitance of the mosfet is directly in parallel with the driving tube. Excessive capacitance here will kill high frequency response, slew rate, and generally cause problems. We try to pick a fet with a low value of Crss. There are plenty of fets in the under 10 pF range. Adding the drain resistor will help buffer this capacitance since it appears in series, so the resistor has two positive effects.

George posted a while ago about using a pair of 6V6 in push-pull for some 35 watts clean output, using.....class AB2 drive to the output valves.
I had one of my Universal Driver Breadboards hooked up on the bench and a discussion about budget guitar amps prompted me to stuff some small tubes in a breadboard and beat on them with some nearly unbelievable results.

I thought that sucking big power out of tiny tubes was cool, but the added expense of multiple power supplies and negative voltage sources negated the cost savings for budget amps.

I pondered on how this could be done differently, simply, and cheaply, which led me to dig out my Grand Universal Theory breadboard from several years ago. After some head scratching, a dozen or so blown parts including two vacuum tubes that actually exploded (a first for me), I now have a working furball prototype. This is a HiFi design that should work well as a guitar amp, and should scale well from small to big. A PCB for a mid sized flavor of the amp has been made and it is being populated now. Time will tell.

I'm wondering if the same recipe will work for 12AB5s.
I haven't tried them yet, but they are a bit smaller internally than the 6V6, so you might not be able to squeeze them as hard.
 
You can easily limit the current that the mosfet can stuff into the grid of the output tube by adding resistance in series with the drain of the mosfet.
We did discuss this idea on this thread a few posts ago. I am curious about pros and cons of the additional drain resistor v.s. simply lowering the positive rail voltage to the driver MOSFETs, which seems simpler.

zeners, and all diodes are voltage variable capacitors by nature, and regarded by many as something not really wanted in the audio signal path.
A while ago, I experimented with Zeners in a couple of places in a valve guitar amp - to replace a cathode bias resistor in the output stage, and to reduce blocking distortion in the coupling caps from PI to output valves by providing a discharge path for negative signal peaks.

I was surprised to hear some semiconductor-sounding distortion in the amp's sound, which didn't go away until I had removed all these Zeners, one by one.

I didn't have access to a 'scope at the time, so I don't know exactly what was going on. What I was hearing was too dramatic to be caused by tiny varying capacitances in the pico-Farad range. I think it had something to do with non-ideal behaviour of the Zeners - leakage currents and / or a soft turn on "knee".

I thought that sucking big power out of tiny tubes was cool, but the added expense of multiple power supplies and negative voltage sources negated the cost savings for budget amps.
I think your demonstrations were not only very cool, but also increasingly relevant as the cost of replacement tubes keeps going up. Even if the additional power supplies cost more when the amp is first made, the ability to get, say, 30 watts out of a pair of $1 tubes lets you keep operating the amp for years without having to spend an arm and a leg.

Between cheap Ebay / Amazon boost voltage converter modules to provide the screen grid voltage, and cheap switch-mode power supplies to provide the +/- voltage rails for the driver MOSFETs, the cost of those power supplies might also be less of a disadvantage as the years go by.

...now have a working furball prototype...should work well as a guitar amp, and should scale well from small to big.
I will look forward to hearing more about this one day!

I haven't tried them (note: 12AB5 valves) yet, but they are a bit smaller internally than the 6V6, so you might not be able to squeeze them as hard.
I will be cautious. I wonder how the 12AB5 anode size compares with the 50C5s, which you also got 30+ watts out of? I have both types, I should go take a look.

-Gnobuddy
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.