Controlling tube gain with logic signal

[Matador]

I've seen digital control of tube gain accomplished with Vactrols, such as in Soldano and Mesa Boogie preamps, where a LDR is used to place elements in parallel (like adding a cap to a cathode resistor, or adding a cap in parallel across a series resistor to make a filter), even in high voltage portions of a circuit. However Vactrols are becoming incredibly expensive ($5-$10 a piece), and vary greatly from unit to unit.

Has anyone experimented with using low cost, high voltage MOSFET's to accomplish something similar? For example, the STN1NK60Z NMOS has a 600V breakdown voltage, a 1V threshold, and has a PMOS equivalent, that run only 30 cents each in bags of 100.

Something like this:



VP and VK are incoming, logic signals (3.3V), that add a plate resistor and cathode resistor respectively. The NMOS MOSFET above is spec'd at 1.2 ohms with 3.3V VGS drive, so when 'on' should place R7 in parallel with R5 when VK is on. Such a scheme could also be used to add a cathode bypass capacitor, or LED, etc. I don't think Q3 needs to be rated for high voltage, as the drain shouldn't ever see another much above 1V.

Q2 needs to be level-shifted, and will only turn on with VGS at 3.3V below VB+, so when Q1 is turned on, it will pull ~1mA through R2 and R3 (if VB+ is about 250V), and will turn on with the drop across R2 thus adding R4 in parallel with R6. With Q1 off, the gate is pulled to VB+ thus turning Q2 off.

 

[ruffrecords]

Both logic signals will cause a change in dc conditions of the tube so may well result in a thump in the output. Maybe ac couple them instead?

Do you just want to switch gain values or vary them smoothly?

Cheers

Ian

 

[Matador]

Do you just want to switch gain values or vary them smoothly?

Just to switch them. I'm thinking of adding another one (NMOS) on the output in parallel with R8 (as a mute), so I can turn that one on first, adjust the others, then turn it off (mute-adjust-unmute).

I tried this on LTSpice, switching R7 for a 10uF bypass capacitor, and see the gain jump at the point where Q3 is switched on which "adds in" the bypass cap (after settling):



The natural question is: why do this? I'm trying to make a test fixture, where I can measure a tube under 4 different biasing conditions, to test suspect tubes to make sure they bias as expected, and I can listen for self-noise by tapping the output at the plate. I already made such a fixture, but it requires I use trim pots and adjust each plate and cathode resistor for each of the conditions, and it takes a long time per tube (plus adjusting a plate resistor when the tube is on is a miserable scratchy mess of noise).

If I use a cheap rotary encoder and a PIC to switch in the circuit elements by just turning a knob it would greatly speed things up and allow me to truly A/B different biasing schemes while listening.

 

[merlin]

The natural question is: why do this? I'm trying to make a test fixture, where I can measure a tube under 4 different biasing conditions, to test suspect tubes to make sure they bias as expected, and I can listen for self-noise by tapping the output at the plate.

Why do you need to change the anode load? Can't you just switch the grid bias voltage through 4 different options?
If you can specify exactly what it is you're expecting to test, there's probably a simpler way to accomplish it than what you're proposing.

 

[Matador]

Why do you need to change the anode load?

Different gain structure. I've found a few tubes that operate "correctly" with a 100K load (e.g. a standard Fender/Marshall/Vox value), but burst into dynamic oscillation with a 220K or 270K anode load (typical on higher gain amps like 5150's, Mesa Boogie Mark's, or SLO's. I typically see this on negative going peaks when measuring the plate with a scope, requiring either a larger grid-stopper or an anode bypass cap.

I'd like to test not only the DC biasing, but also the dynamic operation with an AC input.

 

[Matador]

In case it helps, here's what I've already tried:

I already have a test fixture that does part of this test: it's a standard Fender V1 gain stage (100K plate, 1K5 cathode, and 25uF bypass), with a 0.1uF coupling cap leading into 1M load resistance. The output is tapped from the 1M load resistor, and leads to a dual-BNC where I can connect a scope, and also connect to a power amp. I insert a tube, after warming up, I measure the plate and cathode voltages to see where the tube settled, and observe the plate voltage when rapping on the side of the tube, while also listening to the output through speakers to get an idea of the background noise and how microphonic the tube is under operating conditions.

I expanded the cathode resistance options with SPST switches, so I can disconnect the bypass cap, or raise the cathode resistance up to 2K7. The plate resistor is a 250K pot, so as I said, I can adjust it to other values, however I need to disconnect the speaker as it's a scratchy mess adjusting it due to the DC through the pot.

Lastly, I'll run a 100mV, 440Hz test sine wave into the grid and observe the output signal and make sure the gain is as expected.

What I also have found, is that some tubes appear to operate just find under these pretty standard conditions, some tubes start to misbehave when under higher gain conditions: for example, a Mesa operating point is a 270K plate resistor, 3K3 cathode, bypassed with 0.22uF. In some amps the 270k is bypassed with a 1n. The same tube that works fine in the first situation above will start to oscillate on the negative peaks of the output signal under the higher gain conditions. There's usually three things that will fix it: a) removing the cathode bypass, b) adding in the 1n plate bypass, or c) adding a large series R onto the grid connection (many high gain ams use 220K to 470k in this position). These are all just different ways to reduce AC gain. Many EH tubes run fine under these high gain conditions without adjustments, whereas (new) JJ's I've found need a lot of band-limiting to behave.

Some Marshalls will also squeal like banshees when the 10K cold clipper resistor (10K) is bypassed with a smaller 0.1uF. Other tubes test fine in this condition.

I'd like to be able to quickly run a tube through these various test conditions to make sure it's usable in a given circuit/amp, without having to troubleshoot the amp first if it's misbehaving, without having to a) wire up 4 or 5 separate sockets with each configuration, and b) having to stop and adjust trim pots. I'd like to quickly flip between 4 or 5 configurations, all while running live, so I can quickly check for problems with an unknown tube, and A/B between them. If the elements can be added and removed with logic signals, then it's easy to encode a sequence that I can use a single pot to rotate them through, because at that point it's just basic software to make it work.

I know that SPST switches will work, but I would need to flip them manually to set up each condition. I know that LDR's will work, but they are expensive, and have their own problems (like hugely varying on-resistance from unit to unit, which doesn't matter when in series with a 270K anode resistor, but can make a big difference when compared to an 820 ohm cathode resistor). LDR's really need to be driven from CCS's which also adds more to the cost. I know JFET's and depletion MOSFET's can be made to work, but they need supporting circuitry to make sure they can be turned off around a ground-referenced signals (usually at least a negative rail of -3 to -5V for the gate), and careful attention that the body diodes aren't turned on.

Hence I was wondering if enhancement MOSFET's can be made to work, because they are cheap, tolerant of high voltages, and need only a few resistors to bias correctly, and work well with standard 3.3V logic levels.

 

[ruffrecords]

When you say the output is tapped from the 1M resistor can you explain exactly what you mean by this?

Cheers

Ian

 

[Matador]

When you say the output is tapped from the 1M resistor can you explain exactly what you mean by this?

Taken from the junction of C1 and R8 in the first schematic above.

 

[ruffrecords]

Taken from the junction of C1 and R8 in the first schematic above.

OK, just wondering if scope probe capacitance, amplified by the Miller effect, is going to skew your frequency response and noise results.

Cheers

Ian

 

[merlin]

Hence I was wondering if enhancement MOSFET's can be made to work, because they are cheap, tolerant of high voltages, and need only a few resistors to bias correctly, and work well with standard 3.3V logic levels.

In that case yes, MOSFETs are probably your best option for this application.

 

[zamproject]

I know that SPST switches will work, but I would need to flip them manually to set up each condition. I know that LDR's will work

Have a look at opto-mosfet, you'll get the best of both world.
Low Ron close to mechanical switches, and isolation from V control as a LDR

 

[Matador]

Have a look at opto-mosfet, you'll get the best of both world.

I did take a look at these. For the low side switches, I think these would work fine. And unlike LDR's, they achieve low RDS-on even with small LED forward current (most LDR's are spec'd above 20mA, whereas opto-MOSFET's are 1mA or even less).

The concern is on the plate side: when off, the source of the PMOS sits at B+ (say 250V), however the drain sits at whatever the anode quiescent point is, say 150V. At negative peaks of the output signal, there may be north of 200V across the PMOS. This is 100V-200V across the transistor when it's off. For high voltage PMOS this is no problem, being rated to 400V or more.

However the opto-MOSFET's all seem to be rated at 80V or less, however it's not clear exactly what this voltage is referenced to. If the datasheet is consistent, it appears to be analogous to a source-drain max, so I'm not sure these would work on the plate side.

 

[zamproject]

Nope...
I used various optomos, that can take 400V, yet I just have in mind the G3VM series from Omron, but there is other, in various contact form, single NO or NC, mixed dual etc...

Probably the thing is you often need to search for -mosfet relay- or -solid state relay-
searching with -opto- usually send you around optocouplers (mosfet or darlington) for small signal (4Nxx etc...)

 

[Matador]

Thanks for the pointer! I scoped out the datasheet for the G3VM-351VY1, and it seems to have all of the correct specs: 350V max on the FET side, turns on with only 2mA forward LED current (which means it can easily be driven directly from a GPIO output), rated for 110mA continuous load current, and worst case only adds 50 ohms, and comes in a tiny SOP-4 package (barely larger than a 1W resistor).

I think these would even be usable in some power tube applications, to dynamically change screen resistor values, or lifting cathodes in tube quads for a half-power mode.

They are a tad pricey ($3.50 in quantities of 10) but might clean up the layout enough to be worth it.

 

[merlin]

FYI: For higher voltages you can also roll your opto-MOSFETs using PIV devices like the VOM1271, though they are somewhat more expensive.

Maybe this is a dumb question, but wouldn't a 3P4T rotary switch be just as quick to use as a digital solution, if you only need 4 test configurations? (Or two rotary switches, one for DC conditions, the other for bypass cap options?)

 

[Matt Syson]

Real relays could be your friend in this application with a major plus that they will probably survive a dud tube (valve) without being destroyed. Coil voltages whatever you want so 5 Volt logic chips possibly with a 'driver' chip if you need it. Sometimes the 'old ways' really are the best. Making a digital computer using EX Post Office relays is a bit tredious though!