MOSFET Regulator Weirdness

[defaced]

So I am working on a preamp build that is more/less the SLO overdrive circuit (6 12AX7 triodes used) with a crude brute force B+ regulator circuit using a series MOSFET with a zener shunt regulator setting the gate voltage (schematic of the regulation circuit attached). I'm observing something kinda odd though and I'm not sure I understand what's happening.

Symptoms:
When I remove all of the tubes, the regulation circuit puts out the expected 400v DC (this is not completely unloaded, there is still a MOSFET buffer circuit that is drawing about 2 ma). However, as I add tubes, the gate voltage of the MOSFET decreases until it's many 10s of volts low - like 50 to 100v low, which subsequently drops the source voltage. This drop is not coming from the raw B+ off the first filter cap (~470v), that sags maybe 10v from unloaded to fully loaded. I'm using a Allied Electronics PT 6K3VG, 325-0-325, 40ma DC: Allied - 6K3VG - Power Products - Transformers - Allied Electronics

My understanding:
The gate is an open circuit to DC. So the only current path should be through the zener current limiting resistor, through the zeners, then onto ground. Since the resistor was selected to exceed the datasheet "test current" of the zeners, the zener voltage should be stable across any load being driven by the MOSFET.

Thoughts:
I thought maybe I had a marginally operating MOSFET with damaged gate insulation, but I replaced the MOSFET with another and got the same results. I would not rule out both MOSFETs being toast, but I'd like to make sure I'm understanding the operation of the circuit correctly before I start blaming parts for doing exactly what I'm telling them to do.

Thoughts?

[R.G.]

What you're describing is what would happen if the MOSFET was open - not passing any current. As the load on the output increases, current comes in through the gate-protection zener.Is is possible the drain and/or source wiring is open somehow? That would do it, and all MOSFETs would act the same way, as would no MOSFET at all. I would check continuity from the MOSFET device lead to where I think they're supposed to go. Anything over an ohm is likely to be your problem.

[mozwell]

OK, so your B+ starts at 470V and drops down to only 460V from no load to full load. Is there an AC ripple voltage on this ?
What does the voltage on the gate of the mosfet to gnd measure from no load to full load.
What does the voltage on the source of the mosfet to gnd measure from no load to full load.
I would add a 100 to 470 ohm resistor from the bottom of the 27k resistor, to the gate of the mosfet. 1/2W is fine here.
I would also add (say) a 2M2 resistor from the source of the mosfet down to gnd, so that there is always some load. You can remove this after your testing is done, but it wont hurt to leave it in.

I would expect the gate of the mosfet to be at 400V, from your string of 4 x 100V zeners. The source of the mosfet should be around 4V below this, with 4V being the threshold voltage of the mosfet.
Im not sure if the 12AX7 are ok running with a 400V B+, i would have thought a 275V B+ is more normal, many valve data sheets list a max anode voltage of 300V. Before the heater is on, there will be no current flow through the valve, so its anode voltage will be your full 400V B+.

I would suggest a resistor & capacitor in the standard power supply configuration, as this will isolate any power amp ripple from your main B+ to the preamp power supply node. I would then run your regulator from this. I would use 3x 100V zeners, and change the 27k 1W to a 100K 1W. This will give your preamp power supply a B+ of 300V, which shoudl be fine for all you need.
Definitely HEATSINK your mosfet, the heatsink will be live, at your full 475V B+, so dont ground it. The heatsink needs to be floating.
Use star grounding for all preamp gnd connections.

[loudthud]

With everthing disconnected from the FET the gate to source should measure infinite resistance with your DMM. Check that the tip of your soldering iron is grounded. I had an iron that went open after about 20 years. It killed a couple of FETs before I found it.

[defaced]

R.G., I'll check the layout when I get home. I bet I took a bottom view pinout on the datasheet and drew a top view layout with it. I've done sillier things.

Mozwell, thanks for your reply. This is a preamp build, so no power amp. Also it is a copy of a known working circuit - 400v B+ is common in cascaded high gain amps. Also, being a circuit snippet it doesn't show the MOSFET buffer that provides a constant 2 ma load, though that was noted in my OP. I'm going to do the "idiot check" and verify the pinout, but if that doesn't pan out, I'll post a table of B+ voltages for each node under different loading conditions.

[Enzo]

Hey, LT got a new avatar.

[Steve Conner]

Try adding a gate stopper resistor to the MOSFET as suggested above. It might be oscillating when you start to draw load current. The voltage should only droop about 1 volt across the load range.

I've built many circuits like this and they all worked first time, but I never tried without a gate resistor. 100 ohms or so mounted right on the MOSFET's gate pin.

[R.G.]

Could be oscillation. But I'd expect that the output voltage sensed by a meter would drop suddenly and dramatically when the oscillation started, not going down as the load was increased. Could be, maybe.

Easy to find out. Stick in a 100-1K stopper.

To me, the description sounded more like the MOSFET was either not conducting or not conducting enough to pull the source up to near the gate voltage as load increased.

[bob p]

Hey, LT got a new avatar.

I noticed that too.

[defaced]

Alright, back from some work travel (oh the fun) and have some answers to everyone's suggestions and a couple "oh by the way" things I measured. I've also included a complete schematic (less some signal switching stuff which isn't relevant to the regulator issues I'm having), layout, pictures (post troubleshooting butchering) and a table of voltages under different loading conditions (1 to 3 preamp tubes in single tube increments). I was a little off on the voltages in my initial description, but the trend is the same. Updated and correct voltages in the attached table.

• MOSFET pinout is correct (unless I'm triple blind which I'll have to laugh at)
• MOSFET legs do have continuity to points of use
• Added 100R resistor between MOSFET gate and protection zener/resistor/regulator zener node with no change in behavior. I know this is an approximation, but figured it'd be a viable test to see if it's indeed oscillation. I plan to add this resistor anyway, but I have to hack more into the board and I'd like to do that after I get this working (unless that's what I must do to get it to work).
• With the MOSFET pulled from the circuit, I get the following resistance readings:
  • MOSFET gate reads open to both source and drain
  • MOSFET source (meter black lead)/drain (meter red lead) resistance measures open
  • MOSFET drain (meter black lead)/source (meter red lead) resistance measures like a diode
  
  
• Soldering iron tip did not have continuity to ground lead of power cord, but other metallic parts touching the tip did. Surface oxidization at the tip/retainer interface is impeding continuity, so that will be addressed.
• I tacked in a cap across the zeners to see what effect it'd have on ripple (I was curious) and it reduced it from ~4v to ~20mv (assuming my memory is right, those are ROM numbers), so once this is sorted I'll end up keeping that cap. So currently ripple voltage is in the tens of mv range.

The only thing I'm kinda put off by is the source/drain voltage reading open. Not sure if that's normal with the MOSFET in an unconnected state.

[nickb]

In your readings table, the gate and the source are the same? That's not right. I couldn't see a short in your pictures - check with a meter. D6 & D7 could be s/c.

[defaced]

Thanks Nick, that ended up being the issue. Reminds me of being in a forest but only seeing trees. I'll replace them and see if they fail again. I don't suspect they were bad, but I'll test the replacements and see what transpires. If they pop, I'll be back for suggestions. Thanks all for the help and ideas!