DIY F2 clone

Hi Paulo, I’m not an electronics pro here, but yes, you should still get the 2nd harmonic characteristic, but with the 2SK1530 it will be lower. My guess is that total THD would drop to around 0.5% at 1 W into 8 Ohm.
Thank you very much for your help Ripson.
Obviously that total THD around half of the original F2 is very welcome, but, I wouldn't want to lose too much of the 2nd harmonic character.
Keep in mind that you’ll probably need to adjust the bias resistors, because the Vgs(on) of the Toshiba FETs is lower than that of the IRFP240 / FQA19N20C.
Observing the datasheets 2SK1530/2SJ201 - IRFP240/IRFP9240. they look so similar.

Paulo
 
It’s interesting I said that, because what I remember from some simulations I did a few years ago is that the IRFP240 showed the highest THD, followed by the 2SK1530, and then the SJEP120R100, it was long time ago... I could never find a model for the FQA19N20C, but I did measure them on a curve tracer... in the transfer curve they match 2sk1530 in terms of linearity. It also depends on which model you use in the simulation.

The topic about "triode like" sound is very subjective and I prefer to avoid it, but consider this: the F2J built with SJEP120R100 JFETs measures even less: 0.34% THD at 1 W into 8 Ohm in my setup. These JFETs are even more linear, but does that sound less "tuby"?

1751534956363.jpeg


If I remember well Nelson Pass said years ago that simple single stage amps are exactly where THD numbers matters, how we actually hear them.

So why not give it a try and experiment yourself? 🙂
 
  • Like
Reactions: Paulo Batista Gomes
…and if you’re looking to mimic a tube sound, maybe some VFET amplifiers would be a better choice? Their characteristics are closer to a triode: the plate voltage doesn’t just carry the signal, it also changes the tube’s own amplification and output resistance as the signal swings. When the signal drives the plate voltage up and down, the tube’s gain and its plate to cathode resistance move together.

But most MOSFETs are more predictable: they follow the square law of Vgs voltage, and their behaviour isn’t much altered by Vds swings.

F2 is a different beast. If you have full range speaker I am sure you would be very happy with F2.

Just a thought.
 
It’s interesting I said that, because what I remember from some simulations I did a few years ago is that the IRFP240 showed the highest THD, followed by the 2SK1530, and then the SJEP120R100, it was long time ago... I could never find a model for the FQA19N20C, but I did measure them on a curve tracer... in the transfer curve they match 2sk1530 in terms of linearity. It also depends on which model you use in the simulation.

The topic about "triode like" sound is very subjective and I prefer to avoid it, but consider this: the F2J built with SJEP120R100 JFETs measures even less: 0.34% THD at 1 W into 8 Ohm in my setup. These JFETs are even more linear, but does that sound less "tuby"?

1751534956363.jpeg


If I remember well Nelson Pass said years ago that simple single stage amps are exactly where THD numbers matters, how we actually hear them.

So why not give it a try and experiment yourself? 🙂
Thank you very much for all your valuable explanations.

What I'm listening to today is a Class A SE valve amplifier.

My interest in the F2 is because it's a transconductance amplifier, which I hope to use with full range 8" Coral speakers, but I wouldn't want to be far from the 2nd harmonic character.
I've already decided to build the F2, but I wasn't sure yet whether to use IRFP240, 2SK1530, or even the Cree component that @twitchie tested some of.

But most MOSFETs are more predictable: they follow the square law of Vgs voltage, and their behaviour isn’t much altered by Vds swings.

F2 is a different beast. If you have full range speaker I am sure you would be very happy with F2.
Now you've helped me a lot, and, noting that your simulations "F2 with MOSFET Toshiba 2SK1530 1W (8 Ohm)" and "F2J with SemiSouth SJEP120R100 JFET 1W (8 Ohm)", were quite similar in numbers, and considering that the SJEP120R100 is almost impossible to have, I'll start with the F2 with 2SK1530/2SJ201.

Paulo
 

Attachments

  • 2sk1530-F2-FFT-1W-8.png
    2sk1530-F2-FFT-1W-8.png
    4.4 KB · Views: 17
  • SJEP120R100-F2J-FFT-1W-8.png
    SJEP120R100-F2J-FFT-1W-8.png
    10.3 KB · Views: 16
Regarding the Cree SiC fets that I experimented with a few years back, I didn't listen to them but others had tried them in other amps like the F6 if I recall correctly. Some people didn't care much for them, others liked them so the only way to know if you'll like them is to try them yourself. They will work but without SOA data, no guarantee they will or will not last. At this point, I would recommend you go ahead with the Toshibas.

I haven't looked recently but odds are that you may be able to find a suitable OnSemi/Fairchild mosfet with high enough forward transconductance (gFS in the 12+ range) and low enough input capacitance (Ciss below 2000ish) but I wouldn't recommend this route if you aren't interested in experimentation and the unnecessary risk without any reward other than the fun and satisfaction of doing it. For reference, the IRFP240 has gFS > 6.9 and Ciss is 1300

Actually, while looking again, looks like there are many more options for SiC devices. Not to say that there is any benefit to using a SiC, actually it sounds like the contrary is true so again, it goes against common sense to pursue that it seems ...

Now after saying all that, I did find this little surprise on Mouser that I'm going to try. DC SOA graph available, power dissipation 145W, gFS - 14, Ciss - 1870 and under $20 CAD ea. PD may be a bit on the low side for some people and I highly discourage anyone from trying this. The curious and capable will disregard. Anyway, for those reading this, you've been warned and I'm absolved with this disclaimer.

Only fools dare tread

My measurement regime and equipment are better than last time I tried all those parts. I'll revisit all of them and redo a part roundup properly sometime soon if the stars align.

Cheers,
Stephen
 
Thank you Stephen.
I haven't looked recently but odds are that you may be able to find a suitable OnSemi/Fairchild mosfet with high enough forward transconductance (gFS in the 12+ range) and low enough input capacitance (Ciss below 2000ish) but I wouldn't recommend this route if you aren't interested in experimentation and the unnecessary risk without any reward other than the fun and satisfaction of doing it. For reference, the IRFP240 has gFS > 6.9 and Ciss is 1300
IRFP240 and Toshiba 2Sk1530 do not meet your gFS forward transconductance recommendation. While Cree SiC fets C3M0065090D and OnSemi NTHL045N065SC1 do. But, I guess it's like in the valve world, we shouldn't just stick to the numbers!
 
From my simulations, I found that to achieve precise current matching/settling in F2, the bottom N-CHANNEL FET needs a high gm. The amplifier’s overall gain is then tamed (local feedback) by the source degeneration resistor Rs=0.666 Ohm (actually six resistors R7-R12 to spread out the heat dissipation).

EDIT: I meant the output current, the current flowing through the speaker’s voice coil. Out of curiosity, I also tried low gain factor FETs (lateral MOSFETs), also VFETs but its square wave current response was not great.

EDIT 2: and when I tried VFET it was without the use of source degeneration resistor, but it does not meant it would sound bad, the response was not as precise as with high gm MOSFETs

EDIT 3: and I know from @Zen Mod that someone built F2 with VFET, no idea how it sounds
 
Last edited:
  • Like
Reactions: Paulo Batista Gomes
This is what I was talking about:

1751981937099.png



Using a low gm VFET lowers the amplifier’s output impedance compared to a stiff current source. The result is a transconductance amplifier that still feels a little bit like a voltage amplifier. When in simulation I drive a small coil (to simulate a speaker and to check the current response), the square wave current is not as perfectly square as with a very high gm FET like the SJEP120R100. However, that might still be sweet sounding amp, especially if we measure and find the sweet linear spot of this Sony VFETs.

This is very similar to a papa's VFET design, but instead of using common drain at the bottom (voltage output), I tried common source resulting in transconductance output, still dominant CCS character. But it requires negative biasing :/

The "voltage gain" is only 2 with 8 Ohms in LTSpice.
 
Last edited:
  • Like
Reactions: Paulo Batista Gomes