The "JCN-JX1" Mosfet amplifier


here I want to introduce my actual project.
To the background: My previous project was a Pass Aleph-JX amplifier which works very file. The next step should be a .5/.8 clone, but for severals reasons I decided to build up an independent project, but with some "Pass-DNA" - the full balanced circuit, low count of amplification stages, and a current source at the outputs to rail voltage.

So after a lot of LTSPice twiddling I came to the following circuit.
The basis is a symmetrical JFET input stage in the "John Curl style", built as folded cascode to avoid a 2nd voltage stage.
Its followed by a Mosfet output stage - since the folded cascode is high impedance the Mosfet output stage (or perhaps a triple) is mandatory for low distortion figures. In order to drive the Mosfets with high speed I also introduced a simple driver stage.

Thats in principle an easy circuit, but in detail it became a bit more complicated to solve all details. Common mode offset and so on.
At the end I think it works very well and during prototyping I never has issues with oscillation although I would call it a quite "fast" circuit, capable of slew rates above 200V/us. With the shown (quite simple) bias spreader and IRFP140/9140 the bias is rock stable. Thermal compensation can be altered easily by chosing annother zener / resistor combination.

Because there is only one voltage amplification stage, the open-loop-gain is somewhat limited to a maximum around 60dB. Therefore the amp is designed for high Class-A bias. But for "casual listening" an optional AB-bias gave also quite good results so it can be used optional.
For this, there a 2 bias pots to set-up different bias values. I have the AB-setting for casual listening (150mA per Fet), watching TV, playing PS etc... and the A (400mA per Fet) for serious listening. At +/-20V Rails this is ~65Watts per channel plus drivers.

Amplifier power is limited by the voltage capability of the input Jfets, the J74 is rated to only 26V. I don't liked do add an "cascode cascode" so the power is limited to ~70V/8R. But it can be increased by adding a zener into the drains of the J74s, found no problem with this so far.

Because of the type of feedback circuit there is a trade-off between gain / input impedance / noise. Without an input buffer by soulution was a 68k/6k8 feedback network which gives 20dB gain and about 9-10k input impedance. With an input buffer the resistors can be lowered, noise decreased and input impedance increased. It depends on personal preference, on my latest board I have the input Jumpered for with / without buffer.

Power supply is a bit special: There is a standard PS for the output stage and boosted rails for input / drivers. In the first stages of development the amp had not the best PSRR, so I introduced a boosted rails filterd by a capacitance multiplier. For simpler built-up I designed a "booster" supply which give doubled rail voltages by adding a Elko and a Diode in the rectifier. So only 1 transformer is neccessary. But of course a normal supply with higher voltage can also be used.

At both outputs (+out and -out, full balanced) I have some CCSs which can be switched of +Ub or -Ub, so there is the possibility to alter the sound a bit by switching the outputs to + or - Ub. But with I=0.4A in the prototype the changed are clearly measureable, but hardly noticeable in the listening test.
In short, both to the same Voltage keeps the H3 characteristic but alters the values, and One to +Ub and one to -Ub gives a H2/H3 characteristic.

I have finished eagle layouts which I would call "fool-proof", but I dont see it as a beginners project.

The big drawback is that it uses the unobtainium K170/J74 combination and there is no way in using other types because their high transconductance is really needed here - and I have 2 in parallel. But only 1 ( => 2 pairs per channel) might also be possible if the drivers used have high beta.
Input stage with BJT is possible, but with a completely redesigned input stage because they need to biased in a different way.

There are severals options: Prototype has 2 Fets in parallel, annother version is with 4 of them with increased rails for >100W/8R and better performance at low impedance loads.
Actually I test at an input stage with full-symmetric reference current source, without servos (I think they are not really neccessary) and a K170/J74 buffer stage in front for lower impedance feedback network = lower noise.

Now some pics and measurement data... and ideas for further improvements are welcome ;-)

Regards, Jean-Claude


  • JCN-JX1-8a.PNG
    77.3 KB · Views: 540
  • Like
Reactions: 1 users
Forgotten the most important...
the sound. Its hard to describe, every time a new self-made amp / loudspeaker etc. is the best of the world. My first impression was "Oh, that sounds expensive" :cool:.
But: I have a audio-fried who told me that he never heard an (affordable) solid state amp he liked. He tested the Aleph-JX and doesn't liked it, But the JCN-JX1 he likes a lot and forced me to sell him my prototype amps. Now I have to build-up annother set :cool:... and also a set of PCBs for annother (tube-loving) friend of him who heard them.
But I would not say it sounds like a tube amp! Sound is dynamic and clean (more than an average tube amp), but never harsh. More dynamic than the Aleph-X / JX, but not so sweet. For me a matter of taste, depending on music I more like the Aleph-JX or the JCN-JX1. For electric bill the latter is far superior since it allows to switch to Class-AB with much less power consumption for watching TV etc.

Edit: I have all my K170/J74 (hundreds of them because the amp will be built-up in dozends by our "amplifier anonymous" group) from BDent, all real toshibas.

Regards, J-C
Last edited:

today evening we have fired up the first Version with 4 parallel Mosfets. Runs well in the first attempt without problems (oscillation etc.). About 250W/4R. Sound is very dynamic as already found in the 2 parallel Mosfet Version.
The Version with the buffer is the next step to try, but at the moment I have no board populated.

Regards J-C
  • Like
Reactions: 1 user
Hello again,
the Prototype-Version of the Version with 16 Mosfets per Channel runs well in pratice. This week we had a session with a Loudspeaker-topic and als switched the amplifiers. To our hearing this version runs very well and kicked the Aleph-JX *** by a remarkable margin.

For me a very good amplifier, a pity that there is no interest here…

Regards J-C
  • Like
Reactions: 1 user
at the moment the prototypes are only a collection of wired PCBs. But I take some pics the next time I visity my friend who at the moment has the first running version. But even for me as the developer its quite a wonder that it runs so well without any problem :cool:
Seems as its much more foregiving since it has only 1 voltage amplification stage with much less phase shift compared to the typical 2 (ore even more) stages.

Regards, Jean-Claude
Thanks ;)
That session was funny: After firing up the amps my friend liked to test the amps upstairs with my speakers… because he dont thrusted them and he dont like to make the first test with his own speakers :p. Additional information: My Speakers = 22W Revelation, his = 22W discovery:sneaky: