allFET circlotron

Nice work there Kees. That's a very simple design - I like it. Can you show 20kHz square wave and 100kHz square wave?

Also what are harmonic distortion components at 22v? (60 watts into 8ohms) for 1kHz sine input?

Hi X

This is the current section only, need a voltage amp who I still need to do, as you did see I put in 20 volts, making 40 out.

input is high impedance for tube driving.

Here some with square wave. distortion components I can not see with sim.



  • ScreenHunter_1129 May. 26 20.11.jpg
    ScreenHunter_1129 May. 26 20.11.jpg
    101.1 KB · Views: 533
  • ScreenHunter_1130 May. 26 20.11.jpg
    ScreenHunter_1130 May. 26 20.11.jpg
    97.7 KB · Views: 516
  • ScreenHunter_1131 May. 26 20.12.jpg
    ScreenHunter_1131 May. 26 20.12.jpg
    104.4 KB · Views: 513
  • ScreenHunter_1132 May. 26 20.17.jpg
    ScreenHunter_1132 May. 26 20.17.jpg
    132.6 KB · Views: 498
  • ScreenHunter_1133 May. 26 20.33.jpg
    ScreenHunter_1133 May. 26 20.33.jpg
    111.8 KB · Views: 493
Thanks for the sims, it looks very promising. So you say because of the Circlotron topology, 20v gets 40v at driver?

You need to have log scale in vertical to see HD components. It does look low though this could be one clean sounding amp.

Hi X

You now clean sounding amps are also clinical, as this is most of the time the case, I have to say most because of heavy feedback, this balanced one can sound also clinical, because it cancels out harmonics who are even, left oneven dominant.

This link to sound is the new amp I did make some months ago, do work after simming, with just slight modifications, so you see sim thins are quite effective. This amps do not use feedback, do not use cathode followers, and so has no really feedback at all, it contains .8 % distortion al even harmonics.

This is the sound, on just a mid speaker and a tweeter, no bas system at all, hear how dynamic it is, at first it miss some high, but this comes back later, my finger was on mic of mobile.

the amp uses laterals 2sj 62 and 2sk1058.

On the picture I have done distortion, jhowever in sim program something is not oke, I do not see a line at all.


  • ScreenHunter_1144 May. 27 13.52.jpg
    ScreenHunter_1144 May. 27 13.52.jpg
    74.6 KB · Views: 143
Here I did use the voltage amp part with the circolotron, for compensation of dc the input J fets and the above fets needs tight together. for Idle maybe I can use a Vgs multiplyer below the driver stage? as a constant voltage.

I did make a sim with distortion, Thirth are very dominant when use a single feedback path, however using the other do correct this nicely, looks a lot better. But using this extra feedback line make one input have a very low impedance, so this is not the solution, need here a other way to do this.

Something in multisum is not oke, do let see nothing in distortion. also fourier do strange, sometimes oke then not, I think ltspice is better but have no models for 2sk lat en hexfets and dn25 series.

If you have X can you send me this? thanks.



  • ScreenHunter_1143 May. 26 22.05.jpg
    ScreenHunter_1143 May. 26 22.05.jpg
    621.7 KB · Views: 277
  • ScreenHunter_1142 May. 26 22.01.jpg
    ScreenHunter_1142 May. 26 22.01.jpg
    88.1 KB · Views: 242
Last edited:
Hi All

I had work to do so I was away for a while, still have to watch for rain trouble, this evening again havy thunder and rain. When we go to thorium plants?? and rebuild our forrests in the tropic area? because we can not go on this way, the trees who are destroyed are the biggest cause.

Oke, this have said, I have some tests done in simulation who get some nice ideas, people who have something to add, please do, I now the correction for temp and such is not yet done, so run away can happen, but a temp resistor or so can help, or lat fets.

I have not current feedback, I do like this more, is undependent of bandwidth and tim is very low, so l watch the outcomes of the experts.

Pictures are from different setups, like laterals, hexfet or J-fet power and double current feedback.

Distortion photo let see that it is quite a friendly amp, the off harmonics are quite below the even harmonics and wil make this a friendly sounding amp.




  • ScreenHunter_1168 Jun. 03 15.05.jpg
    ScreenHunter_1168 Jun. 03 15.05.jpg
    113.1 KB · Views: 113
  • ScreenHunter_1174 Jun. 03 18.37.jpg
    ScreenHunter_1174 Jun. 03 18.37.jpg
    628.6 KB · Views: 120
  • ScreenHunter_1173 Jun. 03 18.28.jpg
    ScreenHunter_1173 Jun. 03 18.28.jpg
    612.6 KB · Views: 837
  • ScreenHunter_1171 Jun. 03 18.14.jpg
    ScreenHunter_1171 Jun. 03 18.14.jpg
    626.2 KB · Views: 117
  • ScreenHunter_1170 Jun. 03 18.13.jpg
    ScreenHunter_1170 Jun. 03 18.13.jpg
    626.6 KB · Views: 155
  • ScreenHunter_1169 Jun. 03 15.39.jpg
    ScreenHunter_1169 Jun. 03 15.39.jpg
    566.7 KB · Views: 163
Last edited:
And this ones are before the previous post but forgotten to post.

Only here I have trouble with feedback, it is voltage feedback and not double version so one half of the amp is not feedbacked healthy, the speaker is between them.


  • ScreenHunter_1152 May. 28 14.26.jpg
    ScreenHunter_1152 May. 28 14.26.jpg
    620.4 KB · Views: 111
  • ScreenHunter_1153 May. 28 14.27.jpg
    ScreenHunter_1153 May. 28 14.27.jpg
    629.6 KB · Views: 92
  • ScreenHunter_1154 May. 28 14.29.jpg
    ScreenHunter_1154 May. 28 14.29.jpg
    633.2 KB · Views: 89
Last edited:
Founder of XSA-Labs
Joined 2012
Paid Member
Thanks for all the nice designs Kees. It is interesting how the vertical FET IRFP240 has lower THD than the lateral FET 2SK1058.


Is there a way to do this without an expensive input trafo? Is there a unbal to bal line driver made of silicon I can use?

If using a Lundahl, would this be an appropriate model - LL1690?

Why is the speaker load 100R and not 8R or 4R?
Last edited:
Founder of XSA-Labs
Joined 2012
Paid Member
So hexFET is the winner then? Surprising result actually.

So here is the lower THD design:


I would like to get critiques from other members who are experts. We still need some sort of Vgs or Vbe multiplier so that will increase active parts count to 11 - still not bad.

Could a single-can SK389 matched dual JFET be used on the input stage?

Can you walk us through in words what the various stages do and why they are there?

Hi X

The reason that the hexfet do better can be because of model, it has to be accurate and most of the time it is not real accurate.

The amp is in fact two single ended amps, put one away use a cap and it works, that is why also there are two feedback lines, however make it unbalanced do not work, you need a balanced driver for it, for excample two burr brown opamps do, but a transformer is also oke, there are also cheaper ones then the murdorf.

the amp is quite simpel, just a amp and driver amp, both is cascode, I like them because it "give "air" in the sound, cascodes have a large bandwidth and it make it possible to give high supply voltage.

You have to now, explane much about amps such as bootstrapping etc, I just combine and sim it, your question about the be multiplayer I do think about, for me a hexfet Vge looks te direction, because no bipolairs here.

Maybe putting drivers and such on the heatsink works al for a part, and a temp resistor is series with the idle pots who are the R17 and R15 resistors.
Founder of XSA-Labs
Joined 2012
Paid Member
I think the Vge multiplier can be made by substituting the two zener diodes that supply the voltage to the bias control on the driver stage. Put the Vge across both zeners and combine to one Vge multiplier with Q3 and Q5 gate connections. I don't know if it will work but maybe similar to how I did it to turn the FX8 into an FH9:

Look at the block around T7, R15, R18, and P1.

Replace MPSA42 with an appropriate FET mounted to same heatsink. Maybe IRF540?

Btw, I just priced the BOM for this allFET Circlotron amp (hexFET variety) and $42 will be enough transistors to make like 4 amps.
Last edited:
Hi X

The amp has some corrections, the zvn mosfets has quite a lot voltage and are just little plastic ones who can torse 200 mA, I have change the zeners so the voltage is not more then 20 volts and the rest go over the bigger irf9610 fets who are mucho bigger, I did use the small mosfets because of the capacitance who needs to be low, therefore I do uses cascodes, these also half the capacitance on his own.

I think what concerns run aways, maybe we have to build it and test, a temp sensitive resistor can do wonders.

Making a multiplyer on this one, you need a double one, a constant voltage source between the two driver halfs?, but even not needed when using laterals.

the sim does exectly what I want, it do very well react on the adjust pots, but need two idle adjust points because as I say it is a balanced one, using it as class ab will however rise the oneven harmonics, something I gess nobody wants.

Schematics are a little a mess but coming oke later.




  • ScreenHunter_1185 Jun. 03 22.13.jpg
    ScreenHunter_1185 Jun. 03 22.13.jpg
    530.4 KB · Views: 362
Last edited:
When using a vge multiplyer in series with the zener diodes of the bias control of drivers it is I think a problem because of the offset amp, because it are two half single ended amps we need two, always there are some differences, using it between supply and the bias mosfet can do better, for example a diode,, or two or a TH resistor.

But can I sim runaways? I do not think so, maybe we or I have to make a pcb and go test it with the irfp mosfets to see what happens, also I need to put in protection zeners for the gates.
Last edited:
You mean op amp for the single to balanced driver?

The dual balanced probably requires a temp resistor such as a thermistor. In fact I hunk the Circlotron on pass labs paper uses that method. One each in each leg?

Yes a thermistor is the answer for this, need experimenting thought, because of tracking, a parallel resistor and a thermistor parallel.

The opamps can be used for driver the mostfets, I can sim this, and also you can use opamps to go from unbalanced to balanced, use burr brown ones, are the best as my experience was.

Als Pass Labs did making new amps is more a combination of excisting ones, we have not much new ideas anymore, these are shift to digital amps who I do also follow.

I did see Kaneta amp do use also thermistors, and i have a half of that amp used, but more beafy driving setup, Kaneda uses 5 mA not very much..