Hybrid Circlotron ready to build


2007-05-30 4:12 pm
After having tried an awful lot of versions, this is what will be built the upcoming weeks. Instead of building a 10W Class-A, this one is working up to 10W in Class-A and 25W in Class-AB, biased at 1.5A. My opinion is that I rather have a little more AB headroom than a clipping 10W pure Class-A.

With the help of the ThermalTrak diodes and a pair of 1N4002 (mounted away from the heat) a trimmable slope temperature compensation is constructed.

As there is no negative feedback the distortion will rise considerably when going from Class-A to AB. Simmed results are below .1% at 10k below 10W. At 25W the simmed result is just over .3%. It will probably be higher IRL. Output impedance is indicated to be in the region of .3ohm.

12V for the heaters is a switched "notebook-adapter" and the highvoltage goes from there via a switched 5W Recom 12/210V DC-DC module.

After building this little critter I will also try one with a Lundahl LL1635PP (1+1:1+1).

In my opinion the simmed figures are a little optimistic. But the interesting thing will be how it performs when listened at.

Schematic: http://www.eflatjump.se/LTP/LTPsch.JPG

Frequency-plot: http://www.eflatjump.se/LTP/frekv.JPG

Waveforms-A: http://www.eflatjump.se/LTP/A.JPG

Waveform-ABclip: http://www.eflatjump.se/LTP/clip.JPG

(Current plots are taken at the .33 "emitter-resistors")

Temp-plot: http://www.eflatjump.se/LTP/temp.JPG
A few quick notes:
--C4 and C5 may not be necessary. Try the circuit without before committing. The MOSFET Gate capacitance will tend to roll things off on its own and may be sufficient to keep the circuit out of trouble. You can always add them later if you need them.
--R8 and R10 can probably be reduced.
--You might consider a pot to fine-tune the bias for M1 vs. M2, even if you intend to match the devices (which I would advise doing). Unfortunately, MOSFETs don't have a predictable Vgs and even matched ones tend to be off by a bit. Given your bias scheme, you're going to run smack into the Vgs problem. The circuit may work, but you'll end up with unpredictable performance and higher than necessary distortion.
I had something else in mind, but it's slipped away from me and I'm out of time.



2007-05-30 4:12 pm
Hi Grey,

Thanks for your comments!

I also got the indication that the depletion fets might vary considerably. I also did some simming of what might happen. By changing the drain-resistor R6 to 27ohms, I ran 25 resp 38mA through them instead of 38mA through both. Even if this is not exactly the same as two FET´s with different characteristics the following happened:

DC-offset went from 0 to 70mV at the output terminal.
Bias rose with ca 100mA.
Distortion rose insignificantly.

My friend who uses the DN2540s for CCSs has ca 20 items from the same batch so some initial matching won´t hurt.

About C4 and and C5 I chickened out :eek: when I noticed -3dB was at 3MHz.

About the gate-stoppers I was not so serious when choosing them, but rumours tell me these devices have a tendency to oscillate in the MHz area. But maybe 200-300ohms is sufficient?

Anyway, I had to delete the caps and lower the stoppers (to 330 ohm) to see what happened:

With only the caps removed there was a hump at 1MHz at +1dB.
Together with the lowered stoppers the hump moved up to 1,5MHz and was a little higher at +1,5dB.

With the original resistor values 100p was exatly what was needed to get rid of the rise. The -3dB lowered to 1MHz.

So I might keep them until I have checked the real life situation.

About the temp-comp I do not not have the knowledge at all :confused:, just thought the negative compensation was too steep. It would be appreciated if you came up with a smarter solution.
There are people for whom simulation approaches religion; they Believe in it, even in the face of overwhelming evidence to the contrary I can pretty well promise you that a real amplifier built with real parts will not come within a country mile of 3MHz. Simulation folks will come crowding in wanting to know what simulator you're using, what models...blah, blah, blah...just build the confounded thing and you'll know for sure, whereas those who want to simulate will have you tied in knots for the next two weeks trying different models and arguing over little whatnots and at the end of that time you won't really know any more than you do today.
In other words, they'll waste your time.
Build it. If don't like the results, then you can modify the circuit as needed. If you do like it...you're done. What could be simpler?



2007-05-30 4:12 pm
I am fully aware of the spice religion, so we should not waste our valuable time to argue about it. I am not a member of the cult but think it is a good way for the lazy man to try a novel idea when/if he gets one. You can use it to get indications in what direction changing of a parameter leads.

I am in the tube guitaramp business ( www.revintage.se )and you could not make the amps I build based on Spice. Still I had great help of it when I developed the patentable inputstage I use. But I am new(or have forgetten since I graduated -72) to FETs and BJTs so in this area Spice has enlightened me.

Think this sidetracked a little as for me the only interesting thing is if it ends up in enjoyable listening( and thermal stability ;) )whatever method you use to get the results..............

So the for/against Spice discussion maybe could be held elsewhere as my standpoint is not so far from yours.

Back to the circlotron-specific technical issues:

I would not be suprised if the FET/BJT-combination (sourcefollower?) have a little too high upper frequency limit even if it in reality will be far less than the 3MHz. I have not used small depletion mode MOSFETs before but I suspect that these go up higher than a typical enhancement power MOSFET. The datasheet from Supertex indicates constant capacitances over ca 20V but this is probably to good to be true.

About thermal compensation I have not yet found any information about how the optimal curve should look? It is not the most important matter as the amp is intended to be used in its Class-A region. After rethinking I think the "slope" pot is overkill and two resistors will do the job.


2005-07-06 1:13 am
North East
But, then, there are people who enjoy the use of such tools. It does not waste anymore time than this forum :D I just wanted to say That I very often get 1.5mHz response out of my circuits and one of them did about 4.5mHz. Most recently my ZV9 based circuit would do 1mHz at 20Vp-p! I also very often get a "bump" in the response but ussually around 750kHz??? It seems as I recal this has been more the balanced stuff??? I have improved the amount of bump but never really determined the source? I have rebuilt the circuits with better parts and what I thought was a better layout only to have the "bump" move up to 1.5mHz??? I ussually end up limiting the circuit to 4 or 500kHz so the problem is not an issue :xeye:
Very interesting project revintage :D I like the circlotron, and Hybrids!


2007-05-30 4:12 pm
Aha, it exists IRL! The bump is interesting. Probably it has something to do with the source impedance of the stage before?

Had to recheck my 3Mhz figures. These where for the output stage alone! With the triode-drivers, 1k stoppers and 100p caps involved a more realistic -3dB at 500kHz is indicated. No bump when removing the caps!!!!


2005-07-06 1:13 am
North East
These are single stage circuits :) I'm ussually driving it with a 50 ohm source. These are really a little excessively high freqs to be allowed to pass through an audio circuit. Espeacially if there is some gain going on :D I just like to start off with as wide open a circuit as I can.
I spent some time working on Circlotrons using 6AS7s for outputs. I came back to that topology in the solid state realm when I was developing the New-Tron because the LU1014D only came in the N-ch flavor and I just couldn't get excited about doing a quasi-complementary circuit. Mind you, I have nothing against a quasi-complementary output...it's the front ends I don't like.
As for elevated frequency response...I wouldn't lose a lot of sleep over it until I saw it in an actual circuit. I pay more attention to the leading edge of a 10kHz square wave. If the front corner shows a noticeable overshoot followed by more than half to one full cycle or so of ripple, then I'd consider sticking compensation caps in there. This is the sort of thing that can mysteriously vanish when you actually build the circuit, due to stray capacitance. Given the unpredictability of the capacitance of any given layout (particularly if you're using point-to-point, then a breadboard, then a PCB), you're going to drive yourself crazy trying to finalize the capacitance at any point short of your final PCB.
I happen to like bandwidth. I regard 200-250kHz as my minimum acceptable bandwidth. The circuit I'm working on now has around 500kHz open loop, so I'm fairly content for the time being.


P.S.: One milliHertz? Wow! Now that's bandwidth!
(Okay, I'll shut up now.)
New review of Hybrid Circlotron

Hi el`Ol,

your posted circuit is the one of the Thorens TEM 3200 which has been the reference amp of german audio magazin "Stereoplay" for more than 2 years. There is a new and quite large review available at 6moons:


This review includes much more details about the technical backround. The listening tests are interesting as well, because the reviewer (Srajan Ebaen) made some "First-Watt" tests using some exotic high-eff speakers as well as "normal" speakers. The described performance of the TEM3200 is just fabulous. All in all the non-german reviewer confirmed the test results of the german "Stereoplay" magazin.