Rüdiger, what is the reason for that bias arangement ? I see that you shunt the noise with a 100uF cap. Do you want a more stable voltage for the cascode ?
I could show you a way to speed up the current mirror and make it lower in distortion. A simulation does not take into acount the early effect well so it seems that the current mirror does not add distortion but it does in reality.
Ticknpop, there are a lot of prerunners of the circuit i posted here but i made clear why i prefer not to discuss it any further.
Dayton Wright was ahead of it´s time. I have a friend that has the electrostats working and after all this time the heavy gas has not even evaporated. I have heard that they use this gas also in high voltage transformer houses.
I could show you a way to speed up the current mirror and make it lower in distortion. A simulation does not take into acount the early effect well so it seems that the current mirror does not add distortion but it does in reality.
Ticknpop, there are a lot of prerunners of the circuit i posted here but i made clear why i prefer not to discuss it any further.
Dayton Wright was ahead of it´s time. I have a friend that has the electrostats working and after all this time the heavy gas has not even evaporated. I have heard that they use this gas also in high voltage transformer houses.
Hi Joachim,
the very reason for the cascode refs is, a few years ago I made listening tests using different kinds of generating this reference. In a first stage of a phonopre, as well. I found, surprisingly, receivable differences. This particular phono used NimHBlocks for it. I tried resistive, used a TL431, a LM317, and went so far to check how a super reg would perform. This tests weren't wohlgeformt, though. Anyway, back then, I found the TL431 best sonically, but too noisy. So now, a few years later, I come back with this.
I guess it boils down to -lownoise, -no interfering with the rails, -low impedance (always risky (oscillation), but obviously working)
The whole things does sound very well, yeah
Yes, the circuit sims exceptionally well. If you'd give me a tip to improve the design, I'm all ears!!
Rüdiger
the very reason for the cascode refs is, a few years ago I made listening tests using different kinds of generating this reference. In a first stage of a phonopre, as well. I found, surprisingly, receivable differences. This particular phono used NimHBlocks for it. I tried resistive, used a TL431, a LM317, and went so far to check how a super reg would perform. This tests weren't wohlgeformt, though. Anyway, back then, I found the TL431 best sonically, but too noisy. So now, a few years later, I come back with this.
I guess it boils down to -lownoise, -no interfering with the rails, -low impedance (always risky (oscillation), but obviously working)
The whole things does sound very well, yeah
Yes, the circuit sims exceptionally well. If you'd give me a tip to improve the design, I'm all ears!!
Rüdiger
I understand. A TL431 refference can be made low noise too but you found a good solution, fine.
My sugestions and they sound fine too: First use an enhanced Wilson Mirror instead of a Cascode Mirror. That speeds up the mirror by 4 times. In simulation it gives an overshot but in reality the output capacitance of your BD139, 140 are the limit. The reason that the mirror is faster is, that the enhanced Wilson is based on feedback. Then i whould rase the emitter resistors in the mirror to something over 100 Ohm. I can not read it well but when they are 39 Ohm as in your schematic then you loose around 1.5dB in S/N. What do you think is the next improvement ?
My sugestions and they sound fine too: First use an enhanced Wilson Mirror instead of a Cascode Mirror. That speeds up the mirror by 4 times. In simulation it gives an overshot but in reality the output capacitance of your BD139, 140 are the limit. The reason that the mirror is faster is, that the enhanced Wilson is based on feedback. Then i whould rase the emitter resistors in the mirror to something over 100 Ohm. I can not read it well but when they are 39 Ohm as in your schematic then you loose around 1.5dB in S/N. What do you think is the next improvement ?
Another question. You use different source resistors for the N and P channel J-Fets. In my book the P-channel Fets have higher Gm so from intuition i whould do it the other way around and degenerate the P-channels more.
Well, maybe in your design you get less distortion because the N-channels benefit from degeneration. I my simulations it did not make a big difference anyway but i did not degenerate them as asymetrically as you did so i may have missed the optimum.
Thanks Joachim, for spotting that mistake! Yes, the p-breed would need a bit more degeneration!
I sim'd both cascode and enhanced wilson and found no difference, but wasn't aware of the fact, that the early effect shouldn't be implemented in spice. That would be a serious drawback, then!
Unfortunatley, I see no quick way to modify my boards, and those peculiar boards hold my last IDSS-matched 2sk389/2sj109 octets.
Edit: the CM-resistors are 86R, just because I have so many and need them seldom. For lowest output offset I had to tweak a parallel 10k trim pot across some of the output related resistors. I hope that minimizes distortion as well. I loosy hfe-matched the CM-transistors as well (loosly = DVM).
Rüdiger
I sim'd both cascode and enhanced wilson and found no difference, but wasn't aware of the fact, that the early effect shouldn't be implemented in spice. That would be a serious drawback, then!
Unfortunatley, I see no quick way to modify my boards, and those peculiar boards hold my last IDSS-matched 2sk389/2sj109 octets.
Edit: the CM-resistors are 86R, just because I have so many and need them seldom. For lowest output offset I had to tweak a parallel 10k trim pot across some of the output related resistors. I hope that minimizes distortion as well. I loosy hfe-matched the CM-transistors as well (loosly = DVM).
Rüdiger
Last edited:
That 86R resistos are fine. Do you have some more boards ? I could implement the improvements there.
Hi Joachim,
no, I have only the two I need for myself. It would be possible to update the files and make some more boards, of course.
Rüdiger
no, I have only the two I need for myself. It would be possible to update the files and make some more boards, of course.
Rüdiger
If that makes sense to you, you could make some boards with my updates that fit the more common 2SK170, 2SJ74. The enhanced Wilson is only the first step. There is much more to it.
What bells & whistles do you have in mind, Joachim?
When I would improve my boards, I would add space for parallel trimpots for R5, R6, R10, R15 in case they are needed. What more would be needed?
Rüdiger
When I would improve my boards, I would add space for parallel trimpots for R5, R6, R10, R15 in case they are needed. What more would be needed?
Rüdiger
Well, i will do a schematic. Then it is easier to discuss. Actually the changes cost virtually nothing. Just a few more parts to make the current mirror faster, more balanced and lower in distortion.
Rüdiger, this are the improvements i made : i put in two helper resistors. Because they would starve a bit only on the base current i feed them with aproximately 2mA of iddle current. I put in R3. When the same current flows in the left and right side of the mirror no current flows in R3. If there is a mismatch the resistor comes into action and provides differential feedback. i set it arbitrary to 100kOm. In praxis it has to be checked if that value is low enough. It reduces the impedance of the mirror some but it also provides matching and distortion cancelation so i think it is an improvement. I have seen parts of this circuit elsewhere but not implemented in total so i call it a Gerhard Mirror. I have drawn only one side and you have to scroll down a bit to see it. This circuit is much faster and lower in distortion then the simple cascoded mirror and gives an amasing gain in dynamics and slamm.
Last edited by a moderator:
