High Gain Phase Inverter - MOSFET Follies 8-)

[defaced]

Unfortunately, that circuit won't work as shown. the ZVN0545 MOSFETs are enhancement mode MOSFETs. You have to do something to their gate to get them to conduct at all. This is unlike tubes, which are depletion mode devices. Depletion mode devices conduct heavily all the time unless you do something to the gates/grids to turn them off. This circuit has the gates tied through 1M resistors to the sources, and so there is no forward bias at all, and hence no conduction at all.

What you'd have to do is to tie the gates to some source of positive voltage, and let the sources drop below that voltage by the Vth of the MOSFET, about 1-3V for the ZVN0545 acccording to the datasheet. The gate voltage sets the operating point, and the resistor from the source to ground sets the operating current.

Yes, you probably can put negative feedback into it.

Sorry I haven't updated this - I get pulled in a lot of directions, and don't have time to complete a lot of things I'd like to.

Cool, I can provide positive bias no problem. Thank you for the explanation. And don't appoligize for not finishing it. You give enough information to get people like me going, which is perfect.

[R.G.]

Does the high gain necessitate zener regulated bias? Or would a simple divider off the B+ suffice?

I don't think that the high gain would dictate zener regulation. Probably a divider is fine. If you're trying to duplicate the conditions of a tube PI, you'd figure the rough cathode voltage, and make the gate divider be that. The sources will be one to three volts lower. This will wiggle around with the power supply sag, but it probably won't make much difference. Replacing the lower resistor in the divider with zener works too, and is almost as simple.

Back at the "Yes, you can probably put negative feedback into it." I realized I owe any experimenters a warning. One sure way to make an otherwise stable negative feedback device oscillate is to keep turning up the open loop gain. Feedback from the speaker output of a tube amp necessarily includes the output transformer, and that includes two out of the three phase shift components needed for oscillation all by itself. The capacitor coupling to the output tube grids is #3. The high gain of this PI needs to be tamed with local feedback or other degeneration. Otherwise, dropping it into a negative feedback tube power stage probably will oscillate. No (intentional) negative feedback could work, but would be very sensitive to wire placement.

[defaced]

Got this prototyped tonight. Works great, very transparent, schematic/rough board layout attached. Took three days of simulation and re-reading the verbal description, and looking at schematics online to see how to bias the damn thing. But once I got that figured out, it was probably 6 hours to get it made and installed.

So now I'm running DC coupled MOSFET power tube buffers and the MOSFET LTPI. The overall goal is a very tight metal amp that uses tubes for the output and the high gain preamp (clean channel may end up being SS), and MOSFETS/other SS devices for the PI, FX loop, and power tube buffers. Almost there...

I'm still working on getting my NFB to work completely (the NFB was designed for a cathodyne phase inverter). I can hear the NFB working (it dampens the output, cleans up the highs a bit), but the knobs don't do anything. Suggestions?

I did this on strip board, so the board layout is not etch ready, but it does translate to strip nicely.

Next up is playing with a cathodyne phase inverter. The inverter part is easy, but does anyone have advice on a gain stage to precede it? I've easily got 450v to play with.

Thanks for looking into this originally RG. I would have never thought to do this, nor been able to figure it out originally on my own without alot more work.

[R.G.]

Hey! Nice work!

As to advice... hmmm... I'd use a stock triode preamp, possibly with a follower output. You have enough MOSFET drive between the PI and the followers to drive the guts out of the output tubes. The finesse on this one would be to ensure you don't abuse the output tubes too much. Something like melting the signal grids with too much current flow is a possibility.

Watch carefully for oscillation. MOSFETs can oscillate so high that a 20MHz scope thinks it's looking at DC. At any unexplained power loss or hissy audio, shut it down and look for the problem. MOSFETs share with tubes that the diffamp in that PI will be unbalanced because the signal transferred from the first device through the souce to the second device source is degraded by the resistances to ground from the sources. So the second MOSFET has less gain than the first one. This is why you often see unbalanced plate resistors - to get equal voltage swing from each half of the PI. I don't know exactly where this will hit you with followers, but watch for it.

Again, nice work!

[defaced]

Thank you for the compliments R.G. Figuring this all out has taught me alot about using MOSFETs.

I knew there was a reason I have been looking at 100MHz scopes, I just didn't know what it was yet. If high frequency oscillation is found, I guess something like extra capacitance or gate stoppers are in order?

Soooooo, got lazy (translation: I'm failing miserably at translating manufacturer SPICE models into working models in both Multisim and LTSpice so I said to hell with simulating) and decided I'd just shove some high voltage depletion mode MOSFETs (LND150) into my cathodyne phase inverter's tube socket and find out what happens. Well I think too much signal might be where I'm at because I can hear something crying uncle. I need to put a scope on it to see what's happening, but it's more together than the tube cathodyne by first impressions. Again very transparent and at full volume I didn't hear any harsh clipping, probably just AB2 kicking in. NFB works too.

Photo and schematic attached. 18g solid wire is great for impromptu tube pins.

A question. Is there any rough translation/comparison between transconductance in an SS device and gain in a tube? I realize they work on different principals, but is there any "all things considered equal" scale to say a 12AX7 with a gain of 100 is about equal to X device with a transconductance of Y? I'm trying to get my head wrapped around the MOSFET data sheets I'm looking at.

[R.G.]

If high frequency oscillation is found, I guess something like extra capacitance or gate stoppers are in order?

Gate stoppers generally work. They form a first order lowpass with the gate-source capacitance, and the gate-drain capacitance if there is gain. This gets the frequency response back down out of the bringing-down-aircraft-radio range.

Well I think too much signal might be where I'm at because I can hear something crying uncle.

That's likely. Did I mention that these things have a lot of gain compared to a tube?

A question. Is there any rough translation/comparison between transconductance in an SS device and gain in a tube?

Transconductance is current out divided by voltage in. Same for tubes and SS devices. A 12AX7 has a transconductance ("mu") of 1200 to 1650 micro-mhos in "typical" A1 circuits. A "mho" is an amp per volt, so the 12AX7 does, shoot, call it one milliamp per volt.

A BS170 small signal MOSFET (yeah, not what you're using, just handy) has a stated forward transconductance from the datasheet of 320mS. The "mS" is a milli-Siemen, which is the formal name they gave to the "mho", so the BS170 does 320mA per volt.

It has over three HUNDRED times the transconductance of a 12AX7.

Voltage gain is (roughly, lots of other stuff to consider; theoretically) transconductance times load resistance. So if you drive a 100K with a 12AX7 and witha MOSFET, dinking in the biasing and currents to get linear operation, a BS170 will have 320 times the voltage gain. Getting the load resistances to be the same is tricky, but voltage gain is transconductance times load resistor.

I realize they work on different principals, but is there any "all things considered equal" scale to say a 12AX7 with a gain of 100 is about equal to X device with a transconductance of Y?

Not all that different. Triodes and to some extent pentodes have a varying mu depending on operating conditions. MOSFETs have a more static Mu. But they're directly comparable. If you look at the plate curves versus the drain curves, MOSFETs are more like pentodes than triodes. Pentodes have a lot more gain than triodes too.

Using a pair of MOSFETs in a PI is closer to subbing in a pair of pentodes.

[NateS]

I got this working in sim, but I was a little puzzled by the changes I had to make to get it working. First, I ran the numbers using Duncan's EL34 model, and for 450V plate/screen I get cutoff about -50V, and 5mA grid conduction at +20v. So I started aiming for 80V swing out of the PI. Then, I took my 2204 preamp sim and reamed the living daylights out of the input stage to make sure the preamp was clipping heavily, and with all the controls dimed, I saw about 60V swing at the input to the PI. That sets the bar pretty low for gain required. I ended up with feedback resistors at 15k, but that's not surprising given the modest boost I'm asking of it. I never did get it working right with a tail resistor, but using a CCS and gobs of local feedback was required to keep it from starving. It wasn't simply clipping, even with a pretty generous idle current and low input signal, with the resistor tail it kept shutting off one side or the other on swings.

So I ended up using a 2mA CCS, bias @ 50V, 15k source resistors, and 47k drain resistors, and it appears to be working. I thought about beefing up the idle current to be able to supply my 5mA grid current, and DC coupling to the power tube grids, but then I'd have to give up independent biasing.

[NateS]

I'm always amazed at how my ADHD brain, which is usually <SQUIRREL!!> more prone to distraction, can bulldog on thought processes that should be met with 'Well Duh!' It needs to swing 2x V(grid limit, bias), not v(grid limit, cutoff), or about 130V.