Transformer-driven SEPP output stage.

[Circlotron]

For the sake of simplicity the attached cct does not show bias stuff etc. It appears to me that both upper and lower fet would have the gain of a source follower because of the negative feedback applied to the lower end of the transformer input winding. There is no unequal gain and output impedance problem like some othe SEPP's. I haven't tried it out myself yet but I think I might in the next fey days. If anyone is tempted to ask why don't I just use a complementary pair, well I just like using N-channels exclsively because you can get BIG ones.

Has anyone ever taken this aproach commercially, or experimented with it? More likely in tube days I expect.

[Sch3mat1c]

Ah, a transistorized Futterman!

TIm

[dkemppai]

Originally posted by Circlotron
There is no unequal gain and output impedance problem like some othe SEPP's. I haven't tried it out myself yet but I think I might in the next fey days. If anyone is tempted to ask why don't I just use a complementary pair, well I just like using N-channels exclsively because you can get BIG ones.

Has anyone ever taken this aproach commercially, or experimented with it? More likely in tube days I expect.

hmmm, similar designs are used for RF amplifiers, where harmonic distortion is just filtered out... ...not sure if it'll be any better than complete solid state designs. My guess is probably not. I don't believe that transformers are the most linear devices that can be had.

That said, my dad did build one like this as a PA amplfier once, seems to me that it didn't sound all that good. (Of course, biasing was over simplistic on that amplifier).

I'd be interested in hearing what you think of the amp, if you build it.

-Dan



Seems to me the big problem may be the low frequency signals... ...you may need a rather large transformer.

[Nelson Pass]

The input impedance of the Mosfets is very high, so
a small transformer will be up to the job - it will just
need the appropriate # of turns.

To the extent that you match and/or degenerate the
Mosfets, this circuit will work very well; basically it's
a classic.

[dkemppai]

Originally posted by Nelson Pass
The input impedance of the Mosfets is very high, so
a small transformer will be up to the job - it will just
need the appropriate # of turns.

Yes, it is true that the transformer need not supply a lot of power.

Maybe I'm wrong, but I thought that the transformer must be designed so that it cannot saturate at low frequencies. Assuming a small core you'd need a lot of turns on primary to get in the range of 20Hz. I guess this is what you mean by "appropraite # of turns"? (Something like a million turns of #80 copper )

Also, won't you want even more turns than the minimum needed to keep core in a more linear region of the BH loop?

-Dan

[Sch3mat1c]

Depending on the impedance, you'd only need 500-1000 turns I'm guessing. Fine guage such as #36 would probably be adequate. Wouldn't the HF response be reduced by DCR as well as leakage inductance?

Tim

[Nelson Pass]

You need about 100 ohms or so to keep the Mosfet
stable from parasitic oscillation, so I can't imagine this
as a problem.

[dkemppai]

Originally posted by Sch3mat1c
Depending on the impedance, you'd only need 500-1000 turns I'm guessing. Fine guage such as #36 would probably be adequate. Wouldn't the HF response be reduced by DCR as well as leakage inductance?

Tim

Yeah, you'd probably want to go with a ferrite core to get any HF response from the thing.... ...lets play the number game on this problem... (This is coming from my head, so bear with any mistakes I may make

Given we can calculate the number of turns for the primary of a given transformer as: Np = e*10^8/(B*f*Ac*4.44)
Where Np = Number of turns Primary
B = flux density in Gauss
f= frequency in Hz
e = input voltage in volts.
Ac = core ares in cm squared.

Assume that the transformer is a 1:1 ratio. 10 volts to drive the mosfets would be needed, so the primary voltage would also need to be 10 volts. If using ferrite, assume a Bmax of around 2000Gauss (to be safe, of course). Also, assume that the transformer is small, like a small audio matching transformer, with an Ac of about .2cm squared (I just measured a small audio transformer I had laying around).

If we say that we need to go down to 20Hz, with the above transformer and voltage and Bmax we would need 28150 turns! Not a small number! Then add the two secondaries of the same 28 thousand turns each, ouch!

Ofcourse, we could give up a little top end (maybe not) and go with a permalloy core... ...giving us an allowable 5000 Gauss (Again, staying in the linear region). That would reduce the number of turns to around 11250, a little bit better.

Now, if we used... ..say, a permalloy core, and increased the size of the transformer to around 1.25cm Squared (Similar to PC power supply switching transformer in size), we could get down to around 1800 turns at 20Hz. That's a little more reasonable. If we quadrupled the size of that tranformer (Ac = 5cm squared), we could get away with 450 turns. That's a even better yet!

Hey this was fun! Maybe I should build one now! Actually I've been thinking of building one for 145Mhz...

-Dan

[Sch3mat1c]

I don't see inductance in that equation.

L would matter much more than Bmax. At 10V input, you'll practically never see it, unless you push a real low inductance. The key parameters would be capacitance (of the windings) and inductance, primarily controlled by turns and mu of the core. I don't like the idea of ferrites.

Tim

[Circlotron]

I have an ETD49 style N87 core right in front of me and it has a centre pole diameter of... lets see... 16mm. How many turns would it take to get down to 30Hz at least?

/Circlotron - can wind really good SMPS transformers but can't do the hard sums.