F4 Beast Builders

I sssume that simulation is done with thermistor at exact same temperature as mosfet.
Generg's is mounted on heatsink.
Is it possible to simulate the effect of mounting the thermistor in different locations.

I agree that the thermistor mounting location is important. But the junction to heatsink temperature rise is only about 0.2C/W with the IXSYS SOT-227B packages. That means that at 1.3A and @24V the junction temperature is only about 6.3C hotter than the heatsink. I do not think it is a problem.
 
Member
Joined 2009
Paid Member
solder fumes time

and , if possible , cousin at IXYS , to send multiple freebies , for testing

ha, ha, no cousin... but four P and four N are available at my home.

:D:D

In the Sony II paper Nelson writes "There is one other detail in the output stage regulation; R17/C7 and R18/C8
which bootstrap the Cascode power transistors with about a one-sixth version of
the output waveform so as to get a little more voltage swing at maximum and
bring the output VFETs a little closer to the “sweet spot” of their load line curve."

Of course this detail is crying to be used in the "F4 beast" too.

I am only unsure where to measure the "one-sixth" waveform to get the right R.
And of course "one-sixth" might only be correct for the Sony II.

These are the limits of a box of bricks amp builder....:D:D
 
Member
Joined 2009
Paid Member
.....here cascoded and bootstrapped, by the way it was all done already by lhquam in Villars Amp with some different values.
 

Attachments

  • Cascoded and bootstrapped outputstage.jpg
    Cascoded and bootstrapped outputstage.jpg
    43.7 KB · Views: 1,137
  • distortion cascoded and bootstrapped.jpg
    distortion cascoded and bootstrapped.jpg
    76.2 KB · Views: 1,062
Member
Joined 2009
Paid Member
Meanwhile I think the cascoding of the J-Fets is not done to avoid a high rail voltage but as Pa explained here...
"I named this “Cascode feedback”, and besides lowering the output impedance of the front end, it is convenient and effective for reducing and tweaking the specific distortion character of the circuit, as the positive and negative halves of the circuit can be adjusted independently."

So we can need in the F4 beast the lower output impedance as well as some tweaking of the distortion.
Formerly it was done by a P3 Pot you could adjust the distortion, do your remember? Maybe this CFB method is better.

This is similar to what I have modeled, except for the following:

1. No degeneration resistors anywhere
2. Different "coupling" of the driver to the outputs like in the UGS/UP which avoids capacitors in the signal path, but might make bias touchy
3. No cascode on the JFETs
4. IRF610/9610 MOSFET drivers instead of the Toshiba parts (too afraid to test with my precious Toshibas)

The model, for what its worth, looks fantastic. With luck, I will be able to find time to cobble this together some time in the next month...

Now, the important question for generg: How does it sound? Edit: you were four minutes earlier than me.
 
I must point out that I never actually built the Villars amp, just LTSpice simulations. During that past months since I played around with the Villars design, I have learned a few things about the choice and use of the IXSYS fets and would do things a bit differently.
 
ha, ha, no cousin... but four P and four N are available at my home.

:D:D

In the Sony II paper Nelson writes "There is one other detail in the output stage regulation; R17/C7 and R18/C8
which bootstrap the Cascode power transistors with about a one-sixth version of
the output waveform so as to get a little more voltage swing at maximum and
bring the output VFETs a little closer to the “sweet spot” of their load line curve."

You can cascode if you want, but the main reason papa did so in Sony is because the sony devices produce triode curves.
The IXYS devices are strongly pentode and capacitance isn't that high to lose sleep over, so adding cascode is really a personal choice not a requirement.
 
Meanwhile I think the cascoding of the J-Fets is not done to avoid a high rail voltage but as Pa explained here...
"I named this “Cascode feedback”, and besides lowering the output impedance of the front end, it is convenient and effective for reducing and tweaking the specific distortion character of the circuit, as the positive and negative halves of the circuit can be adjusted independently."

So we can need in the F4 beast the lower output impedance as well as some tweaking of the distortion.
Formerly it was done by a P3 Pot you could adjust the distortion, do your remember? Maybe this CFB method is better.

Good point. I had forgotten about cascode local feedback, which might be a good idea in order to tame the gain of the undegenerated front-end.
 
Last edited: