I decided to simulate the SWTPC Universal Tiger using newer high speed, high beta output devices. Here is an old thread with the schematic: http://www.diyaudio.com/forums/showthread.php?threadid=41926
It does not simulate well with the original output devices, lots of crossover distortion and so on. However, I don't trust the On-Semi models for the old devices. The newer devices make a dramatic improvement. I was only able to do this using Andy C's improved, actually degbugged, models for the output devices, thank you AndyC: http://www.diyaudio.com/forums/showthread.php?s=&threadid=20460&highlight=
I don't know if I'll finish this, but attached is the LT-SPICE model for the improved output stage. I doubled up on output devices so that it will be safe driving 2 ohm loads, and increased the driver bias current. I suggest rail fuses for protection and of course a front end is required. I don't trust the output stage for thermal stability, it should be analyzed. I thought people might be curious to see it in simulation.
Andy's models must be added to the LT-SPICE library, as well as these below. Anyone checked the mje15034 and mje15035? :
.MODEL mje15034 npn IS=3.92866e-12 BF=260.938 NF=1.02215 VAF=15.3399 IKF=0.160087 ISE=1e-08
NE=2.54491 BR=26.0938 NR=1.10885 VAR=153.399 IKR=1.60087 ISC=1e-08 NC=1.89024 RB=0.41209 IRB=0.1
RBM=0.41209 RE=0.0001 RC=0.208002 XTB=0.897431 XTI=1.39234 EG=1.206 CJE=1.61534e-09 VJE=0.698417
MJE=0.382854 TF=1.03079e-09 XTF=1000 VTF=100000 ITF=42.9041 CJC=1.04458e-10 VJC=0.441587 MJC=0.23
XCJC=1 FC=0.8 CJS=0 VJS=0.75 MJS=0.5 TR=1e-07 PTF=0 KF=0 AF=1
.MODEL mje15035 pnp IS=5.81508e-15 BF=313.373 NF=0.85 VAF=40.5017 IKF=0.897023 ISE=6.74258e-16
NE=1.04249 BR=0.958017 NR=0.894461 VAR=148.639 IKR=7.05393 ISC=6.74258e-16 NC=2.84461 RB=3.62039
IRB=0.1 RBM=0.1 RE=0.000923293 RC=0.233799 XTB=2.92628 XTI=1.01325 EG=1.17461 CJE=1.5597e-09 VJE=0.99
MJE=0.554057 TF=1.35882e-09 XTF=1000 VTF=467.207 ITF=58.3338 CJC=1.58888e-10 VJC=0.4 MJC=0.23
XCJC=0.786287 FC=0.8 CJS=0 VJS=0.75 MJS=0.5 TR=1e-07 PTF=0 KF=0 AF=1
Pete B.
It does not simulate well with the original output devices, lots of crossover distortion and so on. However, I don't trust the On-Semi models for the old devices. The newer devices make a dramatic improvement. I was only able to do this using Andy C's improved, actually degbugged, models for the output devices, thank you AndyC: http://www.diyaudio.com/forums/showthread.php?s=&threadid=20460&highlight=
I don't know if I'll finish this, but attached is the LT-SPICE model for the improved output stage. I doubled up on output devices so that it will be safe driving 2 ohm loads, and increased the driver bias current. I suggest rail fuses for protection and of course a front end is required. I don't trust the output stage for thermal stability, it should be analyzed. I thought people might be curious to see it in simulation.
Andy's models must be added to the LT-SPICE library, as well as these below. Anyone checked the mje15034 and mje15035? :
.MODEL mje15034 npn IS=3.92866e-12 BF=260.938 NF=1.02215 VAF=15.3399 IKF=0.160087 ISE=1e-08
NE=2.54491 BR=26.0938 NR=1.10885 VAR=153.399 IKR=1.60087 ISC=1e-08 NC=1.89024 RB=0.41209 IRB=0.1
RBM=0.41209 RE=0.0001 RC=0.208002 XTB=0.897431 XTI=1.39234 EG=1.206 CJE=1.61534e-09 VJE=0.698417
MJE=0.382854 TF=1.03079e-09 XTF=1000 VTF=100000 ITF=42.9041 CJC=1.04458e-10 VJC=0.441587 MJC=0.23
XCJC=1 FC=0.8 CJS=0 VJS=0.75 MJS=0.5 TR=1e-07 PTF=0 KF=0 AF=1
.MODEL mje15035 pnp IS=5.81508e-15 BF=313.373 NF=0.85 VAF=40.5017 IKF=0.897023 ISE=6.74258e-16
NE=1.04249 BR=0.958017 NR=0.894461 VAR=148.639 IKR=7.05393 ISC=6.74258e-16 NC=2.84461 RB=3.62039
IRB=0.1 RBM=0.1 RE=0.000923293 RC=0.233799 XTB=2.92628 XTI=1.01325 EG=1.17461 CJE=1.5597e-09 VJE=0.99
MJE=0.554057 TF=1.35882e-09 XTF=1000 VTF=467.207 ITF=58.3338 CJC=1.58888e-10 VJC=0.4 MJC=0.23
XCJC=0.786287 FC=0.8 CJS=0 VJS=0.75 MJS=0.5 TR=1e-07 PTF=0 KF=0 AF=1
Pete B.
Simulation Of The Original Design
Attached is the LT-SPICE file for the original design. I used On-Semi models for output devices, and substituted diodes as required to match the library.
SWTPC upgraded the design using 2N5210s in the front end, so that is what I used.
I could not find models for the old 40409/40410 but they were similar to the 2N3053 and I found a model for it at Duncan's site. I made an exact compliment to do the other part, here's what I used. If anyone has better models please speak up:
.MODEL 2N3053 NPN (IS=71.1F NF=1 BF=260 VAF=113 IKF=.42 ISE=27.3P NE=2 BR=4 NR=1 VAR=20 IKR=.63
RE=73.6M RB=.294 RC=29.4M XTB=1.5 CJE=79.2P VJE=1.1 MJE=.5 CJC=25.5P VJC=.3 MJC=.3 TF=858P TR=596N
Vceo=40 Icrating=700 mfg=DUNCAN)
.MODEL COMPN3053 PNP (IS=71.1F NF=1 BF=260 VAF=113 IKF=.42 ISE=27.3P NE=2 BR=4 NR=1 VAR=20 IKR=.63
RE=73.6M RB=.294 RC=29.4M XTB=1.5 CJE=79.2P VJE=1.1 MJE=.5 CJC=25.5P VJC=.3 MJC=.3 TF=858P TR=596N
Vceo=40 Icrating=700 mfg=DUNCAN)
Pete B.
Attached is the LT-SPICE file for the original design. I used On-Semi models for output devices, and substituted diodes as required to match the library.
SWTPC upgraded the design using 2N5210s in the front end, so that is what I used.
I could not find models for the old 40409/40410 but they were similar to the 2N3053 and I found a model for it at Duncan's site. I made an exact compliment to do the other part, here's what I used. If anyone has better models please speak up:
.MODEL 2N3053 NPN (IS=71.1F NF=1 BF=260 VAF=113 IKF=.42 ISE=27.3P NE=2 BR=4 NR=1 VAR=20 IKR=.63
RE=73.6M RB=.294 RC=29.4M XTB=1.5 CJE=79.2P VJE=1.1 MJE=.5 CJC=25.5P VJC=.3 MJC=.3 TF=858P TR=596N
Vceo=40 Icrating=700 mfg=DUNCAN)
.MODEL COMPN3053 PNP (IS=71.1F NF=1 BF=260 VAF=113 IKF=.42 ISE=27.3P NE=2 BR=4 NR=1 VAR=20 IKR=.63
RE=73.6M RB=.294 RC=29.4M XTB=1.5 CJE=79.2P VJE=1.1 MJE=.5 CJC=25.5P VJC=.3 MJC=.3 TF=858P TR=596N
Vceo=40 Icrating=700 mfg=DUNCAN)
Pete B.
I would use the Jensen 990 up to the VAS as an improved front end for this design, and run it on +/- 30 V derived from the main supply. I'd remove the diff pair emitter inductors, and increase the degeneration. Also decrease the VAS degeneration, and redo the HF compensation. Compensation should not be as critical since it does not have to be unity gain stable, and the open loop gain can be lower.
I welcome constructive comments/brainstorming.
I'll use SANKEN 2SC3284/2SA1303 for outputs if I build the improved version which are 125W and this is another reason for doubling up or even more outputs.
Pete B.
I welcome constructive comments/brainstorming.
I'll use SANKEN 2SC3284/2SA1303 for outputs if I build the improved version which are 125W and this is another reason for doubling up or even more outputs.
Pete B.
try using winzip to zip them....... i am using xp on this computer, but i haven't yet tried downloading them to my work computer, which uses win98. winxp's zip routines might do weird things to the header, and so make the zips compatible only to winxp (txn again b.g. for another "undocumented feature" of your software.....
).
i was able to unzip them, but people using other OS's might not be able to.
).i was able to unzip them, but people using other OS's might not be able to.
OK well don't like this incompatibility issue so here it is zipped with WINRAR. Made a few minor mostly cosmetic changes.
It is interesting to run the SIM of the original amp, and look at V(out), not too bad looking, however, then look at V(a) before the output filter. Think that's bad, next look at the VAS collector voltage. It is also interesting to look at the driver and output collector currents. This is of course the amplifier being overdriven by a 10 kHz sinewave.
The input diff pair and VAS are saturating due to clipping and the entire loop has a hard time coming out of saturation.
EDIT:
I must be half asleep here. I included TIGER_AMP2.asc in this most recent, which is the original with a miller cap on the VAS but of such a small value that it does nothing, change it if interested, and a Baker clamp that is also ineffective because it is a Silicon diode and not a Schottky type. It also has base stopper resistors on the outputs but also set to a very low value to represent the original design. Clip them out if you want the original design or I can send TIGER_AMP.asc which was the one I intended.
Pete B.
It is interesting to run the SIM of the original amp, and look at V(out), not too bad looking, however, then look at V(a) before the output filter. Think that's bad, next look at the VAS collector voltage. It is also interesting to look at the driver and output collector currents. This is of course the amplifier being overdriven by a 10 kHz sinewave.
The input diff pair and VAS are saturating due to clipping and the entire loop has a hard time coming out of saturation.
EDIT:
I must be half asleep here. I included TIGER_AMP2.asc in this most recent, which is the original with a miller cap on the VAS but of such a small value that it does nothing, change it if interested, and a Baker clamp that is also ineffective because it is a Silicon diode and not a Schottky type. It also has base stopper resistors on the outputs but also set to a very low value to represent the original design. Clip them out if you want the original design or I can send TIGER_AMP.asc which was the one I intended.
Pete B.
SkyChu said:
Yes, that is the Cdom or Miller capacitor and the original design did not use one, so that low value does virtually nothing. You can increase it to see the effect it has on the circuit. I used this second version, TIGER_AMP2, to experiment with some changes.
Attached is TIGER_AMP.asc which is the original circuit to avoid confusion.
Pete B.
Tried MJL21195/21196 - Model Looks Defective
I tried MJL21195/21196 using the On-Semi models in the original design and the results do not look reasonable. Similar problem of low beta and poor switching characteristics as with the MJL3281A and MJL1302A that we discussed here:
http://www.diyaudio.com/forums/showthread.php?postid=1099958#post1099958
The PNP part looks OK from a quick sanity check, the NPN (MJL21196) seems to be a defective model.
These would probably make excellent substitutes, but it is difficult to work with them when the model is defective.
Pete B.
I tried MJL21195/21196 using the On-Semi models in the original design and the results do not look reasonable. Similar problem of low beta and poor switching characteristics as with the MJL3281A and MJL1302A that we discussed here:
http://www.diyaudio.com/forums/showthread.php?postid=1099958#post1099958
The PNP part looks OK from a quick sanity check, the NPN (MJL21196) seems to be a defective model.
These would probably make excellent substitutes, but it is difficult to work with them when the model is defective.
Pete B.
I've redone the simulation including Bob Cordell's SPICE models, then replaced the semis with reasonable selections from Bob's list. I used NJLs for outputs and they show some significant cross conduction, interesting that the MJLs do not seem to, yet the NJLs should be faster - doesn't make any sense. I also noticed that one of Bob's MJLs did not seem to go as close to the rail as I'm used to seeing in real life, probably a remaining model bug.
The sim should run simply by unzipping the file, Bob's model file is included so it should just run with a standard install of LtSpice.
The sim should run simply by unzipping the file, Bob's model file is included so it should just run with a standard install of LtSpice.
UNIVERSAL TIGER FUNDAMENTAL BUG
Hi David,
No, I don't plan to, it also has the output stage with gain so any issues with the Universal Tiger should also apply to it. Also, since there are issues with the output devices, I'd expect that the stacked outputs might not behave very well. You could always do the front end and then assume high voltage output devices and not do the stacked output.
OK, well I think I've found the fundamental bug in the Universal Tiger (UT). When I run the sim and probe the Vbe of the drivers (Q5, and Q6) I saw a reverse Vbe of 7.5 volts without clipping, and about 18 volts in clipping. The max rating for most transistors is about 5 to 10V, I did not see a spec for the original devices used in the UT but the MJE243 for example is rated for a maximum of 7V. Exposure to this voltage will damage the B-E junction and with enough reverse current burn it out.
Perhaps if someone else runs the sim they could confirm this finding.
Hi David,
No, I don't plan to, it also has the output stage with gain so any issues with the Universal Tiger should also apply to it. Also, since there are issues with the output devices, I'd expect that the stacked outputs might not behave very well. You could always do the front end and then assume high voltage output devices and not do the stacked output.
OK, well I think I've found the fundamental bug in the Universal Tiger (UT). When I run the sim and probe the Vbe of the drivers (Q5, and Q6) I saw a reverse Vbe of 7.5 volts without clipping, and about 18 volts in clipping. The max rating for most transistors is about 5 to 10V, I did not see a spec for the original devices used in the UT but the MJE243 for example is rated for a maximum of 7V. Exposure to this voltage will damage the B-E junction and with enough reverse current burn it out.
Perhaps if someone else runs the sim they could confirm this finding.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.