Dear folks,
just recently I got another small number of nice LU1014 power jfets which immediately broke down all my constraints of messing up other people's perfectly nice designs. In this case I had another look at the Aleph J, F3 (Zen#9) and especially at the ideas of Patrick (EUVL). So in short that's not my design - that's the reason it's not completely dangerous. It's basically Patrick's, however he never published a full schematic. However if you find severe design faults, I fully claim responsibility.
For the technically minded:
As Grey put it, it's a Zen#9 with a front end. So you get the joy of a jfet doing the work (a la Zen#9 with loadline modulation), together with the Aleph current source and a jfet frontend.
the good news: though I had my ideas I finally stayed as close to the original AlephJ as possible, that means amongst others feedback, bias current and AC current gain are identical (AC current gain may be slightly different). I used the Threshold/Zen#9-way of referencing the cascode to the output to maximize voltage swing. However the simulation gives IMHO wrong results about the maximum possible voltage swing, so I guess with regard to the Zen#9 that one looses about 8 V (instead of 3V of the original Aleph J), so wattage is reduced. I increased therefore rails to +-27V to get roughly close numbers (about 20W into 8R, about 30W into 4R).
To get spec wise (damping factor, output impedance) and dissipation wise into shallow waters I use a pair of mosfets and powerjfets (instead of only one jfet). To get lowest distortion I recommend using matched mosfets and jfets. Some may object that using 2x LU1014 here is a waste and of course you can use a single jfet there; however you sacrifice output impedance and damping factor. With a pair of jfets you get the Aleph J specs (damping about 20), with a single Zen#9/F3 specs (damping about 8).
The bad news: the cascoded jfets introduce - at least in simulation - a nasty, sharp resonance at about 300kHz, which is not there in the original design. Increasing the lag cap was no fun. Again Patrick gave me the hint to use a Zobel at the output, the values you see in the schematic give perfectly nice frequency response (in simulation that is) that is a tad extended compared to the original.
More bad news: to get rid of the DC across the LTP load resistor, I chose the cap-variant. I hope I didn't lose the last reader by mentioning a cap right now. Honestly I don't see any advantage by using a level shifter (more active parts), which further means additional rails. Actually I thought of a poor man's level shifter without additional rails, but in the end it brings small disadvantages so I did not consider it seriously.
building hints: you will want to use a pot instead of R27 or better a combination of a resistor and a pot. So you can tune the 3V VDS the jfet sees (remember Mosfets have strongly varying VGS so unless you know the VGS of the fet you're going to use, you have to tune afterwards). 2nd, the bipolars. Nelson uses ztx450/ztx550 because he has a huge bag of these. I use the 2SC2705/2SA1145 because I have a bag of these. Use whatever you feel good with, they're not really critical parts. Oh and don' use R32 (the resistor across the cap), that one is only there to make the simulation going.
DC-stability: as I do not have yet build that amp, I don't have own experiences. However Patrick - he build his own variant of the topic - mentioned in the AlephJ-thread that he didn't have any. So I think tight thermal coupling (as usual) seems to be sufficient. However since I do not know yet for sure, I repeat Patricks warning:
Though nothing new under the sun, I mailed it Nelson and contrary to his friendly custom to reply quickly, I did not got a reply. I freely assume his rule is coming into effect
So that's it from my side, if you have any input I'm happy to receive it. Every critique or tips for improvement greatly appreciated.
Of course I appologize for posting a design that uses obsolete parts. Sorry folks.
Thank you Nelson for sharing your ideas with us, without your kind support all of this would not be possible. Also thank you Patrick for sharing your ideas.
All the best, Hannes
PS: as far as I know www.audioprojekte.de still has some LU1014.
PPS: will I build it? That depends how the AlephJ-LUX turns out
just recently I got another small number of nice LU1014 power jfets which immediately broke down all my constraints of messing up other people's perfectly nice designs. In this case I had another look at the Aleph J, F3 (Zen#9) and especially at the ideas of Patrick (EUVL). So in short that's not my design - that's the reason it's not completely dangerous. It's basically Patrick's, however he never published a full schematic. However if you find severe design faults, I fully claim responsibility.
For the technically minded:
As Grey put it, it's a Zen#9 with a front end. So you get the joy of a jfet doing the work (a la Zen#9 with loadline modulation), together with the Aleph current source and a jfet frontend.
the good news: though I had my ideas I finally stayed as close to the original AlephJ as possible, that means amongst others feedback, bias current and AC current gain are identical (AC current gain may be slightly different). I used the Threshold/Zen#9-way of referencing the cascode to the output to maximize voltage swing. However the simulation gives IMHO wrong results about the maximum possible voltage swing, so I guess with regard to the Zen#9 that one looses about 8 V (instead of 3V of the original Aleph J), so wattage is reduced. I increased therefore rails to +-27V to get roughly close numbers (about 20W into 8R, about 30W into 4R).
To get spec wise (damping factor, output impedance) and dissipation wise into shallow waters I use a pair of mosfets and powerjfets (instead of only one jfet). To get lowest distortion I recommend using matched mosfets and jfets. Some may object that using 2x LU1014 here is a waste and of course you can use a single jfet there; however you sacrifice output impedance and damping factor. With a pair of jfets you get the Aleph J specs (damping about 20), with a single Zen#9/F3 specs (damping about 8).
The bad news: the cascoded jfets introduce - at least in simulation - a nasty, sharp resonance at about 300kHz, which is not there in the original design. Increasing the lag cap was no fun. Again Patrick gave me the hint to use a Zobel at the output, the values you see in the schematic give perfectly nice frequency response (in simulation that is) that is a tad extended compared to the original.
More bad news: to get rid of the DC across the LTP load resistor, I chose the cap-variant. I hope I didn't lose the last reader by mentioning a cap right now. Honestly I don't see any advantage by using a level shifter (more active parts), which further means additional rails. Actually I thought of a poor man's level shifter without additional rails, but in the end it brings small disadvantages so I did not consider it seriously.
building hints: you will want to use a pot instead of R27 or better a combination of a resistor and a pot. So you can tune the 3V VDS the jfet sees (remember Mosfets have strongly varying VGS so unless you know the VGS of the fet you're going to use, you have to tune afterwards). 2nd, the bipolars. Nelson uses ztx450/ztx550 because he has a huge bag of these. I use the 2SC2705/2SA1145 because I have a bag of these. Use whatever you feel good with, they're not really critical parts. Oh and don' use R32 (the resistor across the cap), that one is only there to make the simulation going.
DC-stability: as I do not have yet build that amp, I don't have own experiences. However Patrick - he build his own variant of the topic - mentioned in the AlephJ-thread that he didn't have any. So I think tight thermal coupling (as usual) seems to be sufficient. However since I do not know yet for sure, I repeat Patricks warning:
Though nothing new under the sun, I mailed it Nelson and contrary to his friendly custom to reply quickly, I did not got a reply. I freely assume his rule is coming into effect
So that's it from my side, if you have any input I'm happy to receive it. Every critique or tips for improvement greatly appreciated.
Of course I appologize for posting a design that uses obsolete parts. Sorry folks.
Thank you Nelson for sharing your ideas with us, without your kind support all of this would not be possible. Also thank you Patrick for sharing your ideas.
All the best, Hannes
PS: as far as I know www.audioprojekte.de still has some LU1014.
PPS: will I build it? That depends how the AlephJ-LUX turns out
Attachments
Full marks for not just accepting what is published and trying something new.
I would have done a couple of things differently, though I have no rational arguments for them :
1) Not use C8/R31 and use a current diode in parallel with R2, although the lower end of the diode will have to go to an additional rail at least 8V more negative than the negative rail. The current diode will have the same current as the 2SJ109 bias, thus making bias gate voltage of the LU1014s zero. This has been discussed somewhere before (with Zen Mod). And Nelson came up with the same suggestion by coincidence.
2) I would not hang the top of R28 to the output, but rather to ground. Also I would try to use a current diode (say 10mA) to replace R28, and a Zener to replace R27. See whether you can get rid of the Zobel then, or at least use somewhat larger values for Rz.
3) If you want to X, then you need the top half to drift the same way with the bottom half, which means that you should really use the same devices (cascoded LU1014s) for the Aleph current source as well (see the MOSFET follower thread where I published the DAO follower circuit). But that needs more changes to the Aleph CS circuit then probably you would like. I have not thought about how to do this, so you will have to do some thinking if you want to follow this route.
4) The easiest way to use the LU1014 in a balanced amp is to do circlotron. I have published a schematics for that without values. But the basic idea is there. And yes, it was built. It works. It is DC dead stable, and sounds very nice. Bandwidth over 300kHz, but I was amazed by the tight base with only 2 LU1014s in total.
The rest is just personal choices (BJT types, etc.)
Build something first.
Most people don't bother to look at schematics that's not built.
Best of luck, and have fun,
Patrick
I would have done a couple of things differently, though I have no rational arguments for them :
1) Not use C8/R31 and use a current diode in parallel with R2, although the lower end of the diode will have to go to an additional rail at least 8V more negative than the negative rail. The current diode will have the same current as the 2SJ109 bias, thus making bias gate voltage of the LU1014s zero. This has been discussed somewhere before (with Zen Mod). And Nelson came up with the same suggestion by coincidence.
2) I would not hang the top of R28 to the output, but rather to ground. Also I would try to use a current diode (say 10mA) to replace R28, and a Zener to replace R27. See whether you can get rid of the Zobel then, or at least use somewhat larger values for Rz.
3) If you want to X, then you need the top half to drift the same way with the bottom half, which means that you should really use the same devices (cascoded LU1014s) for the Aleph current source as well (see the MOSFET follower thread where I published the DAO follower circuit). But that needs more changes to the Aleph CS circuit then probably you would like. I have not thought about how to do this, so you will have to do some thinking if you want to follow this route.
4) The easiest way to use the LU1014 in a balanced amp is to do circlotron. I have published a schematics for that without values. But the basic idea is there. And yes, it was built. It works. It is DC dead stable, and sounds very nice. Bandwidth over 300kHz, but I was amazed by the tight base with only 2 LU1014s in total.
The rest is just personal choices (BJT types, etc.)
Build something first.
Most people don't bother to look at schematics that's not built.
Best of luck, and have fun,
Patrick
Thank you very much for your kind words, Patrick!
Of course you go straight for best performance and thus hang R28 to GND or positive rail.
That would avoid the output modulation and very probably improve loadline modulation. I go for more output swing and since Nelson showed in the Zen#9 article - where he did it the same way - that this compromise still performs excellently I'll stick with it.
Thank you very much also for your Zener/current-diode suggestion! I had a few minutes during lunch break to look at it - and indeed it helps with the resonance. But not to the full extend yet. I will look more at this idea!
The SuSy variant is indeed more troublesome it seems to me. You get a hint already by looking at the bandwidth of this AlephJ-LU (being basically a halfed SuSy). Since you sandwitch 2x Alephs together bandwidth roughly multiplies.
So you get a real problem at lower frequencies, since already one Aleph starts to go down below about 10 Hz and this will get worse for the balanced version. Of course this applies to a lesser extent to the upper end as well, but this can be fixed partially by cascoding the front end.
Of course all under the assumption the simulation is correct.
Thank you very much for your comment, Have fun! Hannes
PS: I really hope that I'm going to build that guy. I would say it really performs nicely. The only question is wether some other amp will stop me AGAIN from sending a perfectly ready pcb-layout to the fab (well for the Cello I have the lame excuse that the ordered parts for the VAS turned out to be fakes).
Of course you go straight for best performance and thus hang R28 to GND or positive rail.
That would avoid the output modulation and very probably improve loadline modulation. I go for more output swing and since Nelson showed in the Zen#9 article - where he did it the same way - that this compromise still performs excellently I'll stick with it.
Thank you very much also for your Zener/current-diode suggestion! I had a few minutes during lunch break to look at it - and indeed it helps with the resonance. But not to the full extend yet. I will look more at this idea!
The SuSy variant is indeed more troublesome it seems to me. You get a hint already by looking at the bandwidth of this AlephJ-LU (being basically a halfed SuSy). Since you sandwitch 2x Alephs together bandwidth roughly multiplies.
So you get a real problem at lower frequencies, since already one Aleph starts to go down below about 10 Hz and this will get worse for the balanced version. Of course this applies to a lesser extent to the upper end as well, but this can be fixed partially by cascoding the front end.
Of course all under the assumption the simulation is correct.
Thank you very much for your comment, Have fun! Hannes
PS: I really hope that I'm going to build that guy. I would say it really performs nicely. The only question is wether some other amp will stop me AGAIN from sending a perfectly ready pcb-layout to the fab (well for the Cello I have the lame excuse that the ordered parts for the VAS turned out to be fakes).
Coming back to the 300kHz resonance: I checked now, there is no benefit using a constant current diode and a Zener compared to the resistor network.
What actually influences the resonance is the fet itself; changing to a lateral Mosfet strongly reduced the resonance. So here is some parasitic capacitance oscillating.
So putting the IRFP240 back in I play with the gate-stoppers R29/R30 and indeed one can get a reasonable flat frequency response by omitting the base stoppers completely.
Now I'm a bit in trouble as there's the rule never to use a mosfet without gate stoppers.
Plot is without Zobel, without gate stoppers R29/R30.
Have fun, Hannes
What actually influences the resonance is the fet itself; changing to a lateral Mosfet strongly reduced the resonance. So here is some parasitic capacitance oscillating.
So putting the IRFP240 back in I play with the gate-stoppers R29/R30 and indeed one can get a reasonable flat frequency response by omitting the base stoppers completely.
Now I'm a bit in trouble as there's the rule never to use a mosfet without gate stoppers.
Plot is without Zobel, without gate stoppers R29/R30.
Have fun, Hannes
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.