mosfet split load PI?

[imaradiostar]

Has anyone built a split load PI using an LND150 or IRF820 or similar?

I would imagine the tubes on either side of it would "hit the rails" before the mosfet leaving it pretty much transparent- power and preamp tone would prevail...right?

thoughts?

jamie

[imaradiostar]

so I'll be breadboarding this soon...I'll let you know.

jamie

[Gregg]

R.G did experiments with MOSFETs as PI. There's a thread on that in the forum.
What I remember is you can connect the guitar directly to the PI and drive the amp into clipping.

[Steve Conner]

I've never tried it, but can't see any reason why it wouldn't work. A MOSFET conducts right down to almost 0V across it, so you can get the maximum theoretical swing from a split-load PI. If I remember right, you bias it so the gate/source sits at one-quarter of your B+. Then the drain sits at three-quarters, and the maximum unclipped swing for both outputs is half of B+.

[imaradiostar]

Originally Posted by Gregg

R.G did experiments with MOSFETs as PI. There's a thread on that in the forum.
What I remember is you can connect the guitar directly to the PI and drive the amp into clipping.

Yeah, I recall that post. Pretty sure that was a LTPI.

I think the split load PI would be good because the theoretical swing would be darn close to 1/2 of B+. It's feasible to completely remove the PI from the equation as a tonal contributor given a high supply voltage. The question remains- would this be a good sound or a bad sound. I thought it would lend itself well to an EL84 based amp because they require so little drive.

Steve, I think you're right- essentially I want the resting value of the mosfet to equal the sum of the drain and source resistors for max drive and symmetry. I'll probably connect a 2m2 resistor from gate stopper to B+ and a 1 meg pot from gate stopper to ground. This way I can dial in the right value then replace it with a fixed resistor. I doubt it has to be super accurate so any close value should work. Mosfets can be widely variable but a mosfet biasing 3 volts one way or the other (from device to device) doesn't mean much in a circuit that'll swing 50+- volts.

jamie

[R.G.]

Originally Posted by imaradiostar

Yeah, I recall that post. Pretty sure that was a LTPI.

I think the split load PI would be good because the theoretical swing would be darn close to 1/2 of B+. It's feasible to completely remove the PI from the equation as a tonal contributor given a high supply voltage. The question remains- would this be a good sound or a bad sound. I thought it would lend itself well to an EL84 based amp because they require so little drive.

It was an LTPI. The split load is an obvious extension of the cathode follower, and does not violate my advice in the MOSFET Follies to not use MOSFETs for voltage gain. The gain of a split load is the same as the gain of a follower, which it arguably is; it's unity or negative one on the drain side.

If what you liked about an amp was the distortion of the LTPI, and some people do, or if you were using the LTPI for a clipper/limiter in front of the power stage, the MOSFET will not act the same as a tube. But a split load shouldn't be what is distorting.

It will work OK as long as you have power supply headroom to get the full grid swing needed out of both source and drain sides. This means needing a power supply of four times the cutoff-to_"saturation" grid drive of the outputs, twice on each of the load resistors and two times on the MOSFET. This is because, unlike the LTPI, the split load must swing together in the middle. So to get full swing on each side, the power supply must be 4x plus whatever housekeeping voltage the active device needs. Fortunately, for a MOSFET this can be small.

For EL84s, this is easy, as they only need about a 24V peak to peak swing on the grid. For 6L6's this is more like a 100V swing, so a split load for 6L6's is going to need well over 200V supplied to it, and 300V is a better choice to avoid cutoff and saturation effects.

3V or so of uncertainty is not going to be an issue in the split load MOSFET bias. But you need to put the gate fairly accurately about 1/4 of the power supply. At the minimum 200V power supply for 6L6s, that's 53V. This gets less critical as you get some more power supply to work with, so at 300V, you can have some slop: somewhere between 60 and 75V is probably OK.

The other thing that a split load does is force you to feed it a drive signal as big as you will drive to the outputs, as the gain is unity on both sides. LTPIs have gain and can make do with much smaller driver signal.

[imaradiostar]

This plays well into what I was thinking- EL84's require little voltage so the mosfet PI might work well. The lnd150 can handle 400 volts with no trouble so I don't mind keeping PI supply voltage high with bigger tubes- maybe with a little extra filtering since there won't be as big a drop before the PI node.

Like I said, need to breadboard it and see how it sounds.

jamie

[Gregg]

Speaking of MOSFETs i would like to ask you to comment on this MOSFET based gain stage and FX loop and whether it will be able to provide clean FX Return gain to grive the PI and what output impedance to expect from cascode stage:

http://www.blueguitar.org/new/articl...er/hi_v_ss.pdf

[pdf64]

Should R4 = ~10k, rather than 1k?

[R.G.]

Originally Posted by Gregg

Speaking of MOSFETs i would like to ask you to comment on this MOSFET based gain stage and FX loop and whether it will be able to provide clean FX Return gain to grive the PI and what output impedance to expect from cascode stage:

The cathode follower output will indeed give you clean output drive.

The cascode will also do a good job of clean input gain, at least potentially.

There is an issue with the circuit as shown. The gate of the JFET which provides the transconductance for the circuit must be pulled to ground to work right/reliably. I would put a 1M resistor to ground after C8 and before R10 to fix this.

The raw output impedance of the circuit is the channel resistance of the MOSFET paralleled with the 220K of R6; call it 220K since MOSFET channel resistance is quite large. However, the circuit is operated with negative feedback through R9 and R10. It's set up for a gain of about 20 (R9=1M) to 4 (R9=220K) roughly. That feedback will lower the output impedance by a significant factor. To figure out how much, you need to find the characteristics of the JFET and MOSFET and do the math for those devices in cascode into a 220K to get the open loop voltage gain, then the feedback factor. The output impedance is divided by the ratio of the open loop to closed loop gain. It will be significantly lower than 220K. How much lower depends on the devices which make the open loop gain.

And frequency, because of the caps at C2, C4, and C5.

[pdf64]

If R4 is intended to be 1k, then C7 needs to be increased in value. As is, you'll be losing low end, as they form a high pass filter, corner freq of 130Hz. Peter.

[Gregg]

If R4 is intended to be 1k, then C7 needs to be increased in value. As is, you'll be losing low end, as they form a high pass filter, corner freq of 130Hz. Peter.

You're right and 10k seems to be a reasonable value considering that output from the tone stack can reach 30V or so.

[imaradiostar]

Just a quick reply to say I tried this and thus far I like it a lot!

I had an old stereo chassis with 1 12ax7, 4 cathode biased 6v6's and a 5u4. It has a pair of decent output transformers but I never did much with it because I didn't like the sound of the mediocre phase inverter allowed by a single 12ax7 for both sides. I wired up an LND150 for each side as the split load inverter.

Now I have the ax7 wired as a gain stage (100k plate, 1k5 cathode, unbypassed ala tweed super, allowing for NFB if wanted) into a 22nF cap into a 100k gate stopper resistor- trying a page out of the Valve Wizard book here!

I set the gate voltage with a 3m3/1m combo before the stopper and used 100k resistors at the source and drain because they're what I have laying around. I used 100nF into 1k5 grid stoppers on the 6v6's with 220k to ground for grid leaks.

The cathodes of the 6v6's are connected to ground with 220R per pair and a 40uf cap. Plate voltage is around 340 and screens and PI are above and below 300 respectively. I haven't wired up any NFB yet but I may after using it for guitar.

It sounds pretty great, even heavily overdriven. Thus far I've blasted a few weezer albums through it into a pair of open back 12's. I'm gonna hook up a guitar preamp in a little while and see how that sounds, maybe use my TC electronic chorus in stereo as a preamp.

Just for kicks I threw in a quad of 6l6's and it still sounds pretty great- more bass and tighter sounding. I'm currently letting it run to see how hot the PT gets.

jamie

[imaradiostar]

Just as a quick test I hooked up the tc stereo chorus with a few rca to 1/4" cables and cranked up the gain on the pedal- there was enough to overdrive the output tubes with a strat. Very musical sounding, not the raspy sound I usually associate with an overdriven split load PI. It's probably a combo of the solid state PI and the huge gate stopper. I say it's a winner!

jamie