Fet Push Pull Mini Amp

[mac]

I was breadboarding a Deacy a few days ago when I decided to do something different:
all small signal fets, even in the power stage. After all it's a 1 watt amp.

I quickly breadboarded a kind of Vox amp, but I did not like the driver stage. It was distorting badly, too many parts, and a matched pair of fets was needed.
So I decided to use a driver transformer. Very simple, as the gates can be coupled to the secondaries with the CT to ground.

Here is the schematic,



I used 2SK117 Toshiba high gain, low vgs fets in the preamp, and two matched MPF102 in the power stage.
Oh, and 18v.
I set the shared emitter resistor near 100 ohm about 6ma per fet. That's 108mW at 18v.
You can go even further, just for fun
R3 could be a trimmer to set the drain near 9v to 12v.

Transformers are from Maplin, UK, LT44 driver (20k:1k2+1k2) and LT700 output (1k2+1k2:3.2) IIRC.
I bet Mouser has something similar.

Although the LT700 drops a lot of voltage, it drives a 12" 8 ohm speaker well.
Try with small speakers, 4 ohm better.

I did not include in the schematic a line out.
I tried it, not for my taste, but you might like it.
You could try a 50k-A pot, and a 3.3 to 4.7 ohm resistor if you use heaphones.
Enjoy,

mac

[robthequiet]

Do the MPF102s get hot enough to need a heatsink?

[mac]

Do the MPF102s get hot enough to need a heatsink?

No, they are cold.

mac

[BubbaFet]

Does it do Metal?

[mac]

Does it do Metal?

It has two irons...

mac

[amptramp]

Would it be better to connect the R6 pot so C1 goes to the top of the pot and the gate of Q2 goes to the slider?  The way it is connected, low frequency turnover goes up as volume goes down.

Have you measured the quiescent current through Q3 and Q4?  It is interesting that with the variation in MPF102 specs that you can get satisfactory balance.

[Transmogrifox]

If the LT44 is an ideal transformer then that gain stage off Q2 is infinite gain.  Regardless, it's really really high gain.

How this sounds good, I don't know.  You're driving Q3 and Q4 as diodes and your "real" power transistor is Q2.  Maybe it commutates quickly enough that you don't hear the crossover distortion.

Try connecting the speaker directly to LT44 and you will probably find you have about the same amount of power output or possibly more.

Might be the LT44 driving the output transformer gives you a sweet impedance matching operating point.

Either way those FET's aren't doing anything in the output stage except acting like diodes in series with the LT44.

[mac]

Try connecting the speaker directly to LT44 and you will probably find you have about the same amount of power output or possibly more.

Actually there is less power that way 
LT700 >> LT44 power.

Would it be better to connect the R6 pot so C1 goes to the top of the pot and the gate of Q2 goes to the slider?  The way it is connected, low frequency turnover goes up as volume goes down.

I'm still tweaking this one.
Today I changed the preamp fets, Q1 BF245A (vgs=0.8v) and Q2 BF245C (vgs=4.5v)
And connected the Pot in the correct way, as Ron noted. My mistake.
Now I have a better transition from clean to distortion.

Have you measured the quiescent current through Q3 and Q4?  It is interesting that with the variation in MPF102 specs that you can get satisfactory balance.

I changed Q3,4 for higher vgs MPF102s. A bit more power now.
I set current at 5-10ma across the shared resistors. Even with R8 at zero, there are no more than 20-25ma across R9.
Drains at 17.90v.
I use this currents noting that (18 v)/(1k2+1k2 ohm) = 7.5ma.
Fets are well matched, I was going to use them in a P90.

Those changes have increased sound quality a bit.
It seems that vgs above 3v is better for Q2,3,4.

Not bad for... 3,29 pounds each iron??!! WTF, I paid 4,98 pounds for the pair in 2013!!!

mac

[bool]

Have you thought about ditching the MPF's and going the mosfet route in the pushpull section (with a xfmr center-tap rebias a-la bjt version)?

Imho even the small BS170's would increase the oomph at your (10mA-ish) bias currents?

But for the proper AB-cls pushpull sag action you should add a cap in parallel to the source bias trimmer to form the time constant.

[anotherjim]

Q3 & 4 look like they could bias ok to me. Drains are supplied from B+ (18v?) via the LT700 centre tap, sources are somewhere above ground and the gates ground referenced from the LT44 secondary centre tap.
Would you need to select MPF102's for high Idss?

[Transmogrifox]

Try connecting the speaker directly to LT44 and you will probably find you have about the same amount of power output or possibly more.

Actually there is less power that way 
LT700 >> LT44 power.

Hmmm...probably because there is DC bias and it's being shut off, so power FETs do something, but it certainly isn't linear...but if it sounds good then who cares if would look ugly on a scope.

Anyway if you put a 10k resistor across the coil of the transformer you will have a known amount of gain on that stage.  It might not sound better.  It might be sounding better just because the transformer is hard-switching the output FETs instead of operating them in their linear region.

[mac]

I recorded two clips with my mobile.

There are some minors changes,
I decided to use BF245A (Vgsoff about a volt) in the preamp,
and two matched 2SK117 (Vgsoff about 0.5v) high gain in the amp stage.
The driver pot is set at around 1k,
and there is just a 1 ohm shared resistor at the 117s sources. I'm reading 25ma idle, 12.5ma per fet.
They get a little warm now, not a big deal.

Would you need to select MPF102's for high Idss?

Imho even the small BS170's would increase the oomph at your (10mA-ish) bias currents?

MPF102s sound cleaner, but a booster is needed to get a decent amount of distortion.
That's why I changed to 2SK117. Their gain is a lot higher, closer to mosfets.

Clean

Dist (clean, dist, Rangemaster)

Now going for a depletion lateral mosfet high voltage amp. This was a low voltage test

mac