Kenpeter: I couldn't get the 10-transistor ladder to stabilize easily - it required base stoppers, Miller caps, and a lot of tweaking. OTOH, the 8-transistor version (without Q3/Q8) stabilized with fairly minor compensation, and it gets to about -120 dB H2 at 1V peak Vout, which is good enough. I also found the 3-transistor Wilson CM to be good enough in this circuit for almost all practical purposes.
I also have a low-power headphone amp circuit coming along that uses my original super-pair with a 3-transistor Wilson CM for about 20 mW into 24 ohms (~ 1V amplitude). Phase margin is still stuck at about 40 degrees, but I'll figure out some compensation schema that improves it eventually, maybe with some THD20 tradeoff.
cbdb: The structure is a wide-band structure - essentially a two-stage voltage follower. With the right kind of transistors, it can probably be made to work from DC to 1 GHz or higher. In audio applications, one might typically close the loop with an op-amp and roll off the closed-loop gain after a hundred kHz or so.
I also have a low-power headphone amp circuit coming along that uses my original super-pair with a 3-transistor Wilson CM for about 20 mW into 24 ohms (~ 1V amplitude). Phase margin is still stuck at about 40 degrees, but I'll figure out some compensation schema that improves it eventually, maybe with some THD20 tradeoff.
cbdb: The structure is a wide-band structure - essentially a two-stage voltage follower. With the right kind of transistors, it can probably be made to work from DC to 1 GHz or higher. In audio applications, one might typically close the loop with an op-amp and roll off the closed-loop gain after a hundred kHz or so.
Last edited:
The circuit (and variations of it) posted at the following link is one that I have used off and on over the years.
http://www.diyaudio.com/forums/solid-state/135637-discrete-ad797-up-running-6.html#post1743249
Again inspired by Pioneer's Super Linear Circuit topologies.
regards, jonathan carr
http://www.diyaudio.com/forums/solid-state/135637-discrete-ad797-up-running-6.html#post1743249
Again inspired by Pioneer's Super Linear Circuit topologies.
regards, jonathan carr
Hi,
Love this circuit. I'm unable to reproduce your results with the models
I have for the 2sc2705/2sa1145. Can you post them?
I was able to get similiar results using 2n3015/2n2905a models that I have.
When I tried kenpeter's circuit, it also oscillated for me with small step
sizes.
I tried 2n3904/2n3906 and 2n4124/2n4126 but these didn't give
near as good results. Tried changing operating points (resistors/voltages)
but to no avail.
Circuit probably model sensitive and also has a sort of sweet spot.
Mike
jcarr: Yes, the idea is similar (with one minor wrinkle) to that in your link. D2 and D4 in your schematic seem to be level-shifters that will interfere with the Vbe cancellation effect between the NPN/PNP pairs Q1/Q3 and Q20/Q2. With ideal voltage sources instead of D2/D4, the circuits will be similar.
In fact, in the symmetric push-pull form (with cross-coupled collectors of Q1 and Q20) shown, you could probably get away with resistors with the same value as R17 and R2 instead of D2 and D4. That little bit of emitter degeneration will also help with stability.
mfc: Here are my 2sa1145/2sc2705 LTSpice models. The comments seem to indicate that I picked them up from these forums earlier. If you've got better (i.e. more realistic) models of the 2sa1145/2sc2705, please post them here.
I'm using the Alternate solver in LTSpice, but AFAIK it only affects the DC operating point analysis.
In fact, in the symmetric push-pull form (with cross-coupled collectors of Q1 and Q20) shown, you could probably get away with resistors with the same value as R17 and R2 instead of D2 and D4. That little bit of emitter degeneration will also help with stability.
mfc: Here are my 2sa1145/2sc2705 LTSpice models. The comments seem to indicate that I picked them up from these forums earlier. If you've got better (i.e. more realistic) models of the 2sa1145/2sc2705, please post them here.
I'm using the Alternate solver in LTSpice, but AFAIK it only affects the DC operating point analysis.
Hi,
Don't think these are any better, but here is what I have for models.
.MODEL 2SA1145 PNP (IS=10F BF=160 VAF=200 IKF=109.96M ISE=221.874F NE=1.66575 BR=70 IKR=880.176M ISC=187.58P NC=1.90472 RE=1 RC=15.5104 CJE=20P MJE=330M CJC=6.24728P VJC=692.028M MJC=340.013M TF=1.08385N XTF=16.9293 VTF=9.36211 ITF=670.025M TR=10N)
.MODEL 2SC2705 NPN (IS=10F BF=160 VAF=100 IKF=42.2761M ISE=1.00598F NE=1.13357 BR=7.54484 IKR=16.2061M ISC=679.914P NC=1.79907 RE=1 RC=706.713M CJE=22P MJE=500M CJC=4.21828P VJC=856.36M MJC=300M TF=1N XTF=2.06578 VTF=9.68976 ITF=384.418M TR=28N)
Possible point of origin might be determined by doing a google on the following string:
2sc2705 "bf=160"
Using your models, I was able to get very similiar results to what you posted, but with these models I wasn't able to.
Mike
Don't think these are any better, but here is what I have for models.
.MODEL 2SA1145 PNP (IS=10F BF=160 VAF=200 IKF=109.96M ISE=221.874F NE=1.66575 BR=70 IKR=880.176M ISC=187.58P NC=1.90472 RE=1 RC=15.5104 CJE=20P MJE=330M CJC=6.24728P VJC=692.028M MJC=340.013M TF=1.08385N XTF=16.9293 VTF=9.36211 ITF=670.025M TR=10N)
.MODEL 2SC2705 NPN (IS=10F BF=160 VAF=100 IKF=42.2761M ISE=1.00598F NE=1.13357 BR=7.54484 IKR=16.2061M ISC=679.914P NC=1.79907 RE=1 RC=706.713M CJE=22P MJE=500M CJC=4.21828P VJC=856.36M MJC=300M TF=1N XTF=2.06578 VTF=9.68976 ITF=384.418M TR=28N)
Possible point of origin might be determined by doing a google on the following string:
2sc2705 "bf=160"
Using your models, I was able to get very similiar results to what you posted, but with these models I wasn't able to.
Mike
mfc: Thanks for your models - your 2sc2705 model looks more realistic. My 2sc2705 model (downloaded from this forum) is most probably broken; Bf=2k is utterly unrealistic - Bf=160 seems entirely reasonable. I used your models and got ~20 dB higher H2 in the 3rd circuit. However, the 3rd circuit still remains better than the other two, and the ranking of the circuits in the order of decreasing THD20 remains unaltered.
Your 2sa1145 model looks similar to mine except for minor variations.
Your 2sa1145 model looks similar to mine except for minor variations.
This thread continues here under the Analog Line-Level sub-forum:
http://www.diyaudio.com/forums/anal...-class-line-headphone-buffer.html#post2361810
http://www.diyaudio.com/forums/anal...-class-line-headphone-buffer.html#post2361810
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.