A lot of small changes. Found some different transistors. Better? Not really sure, but I did drop the noise a few dB. Not terribly happy with my choice of drivers. They will need very careful heat sinking.
Some of the changes are based on what SPICE says, and may not be valid on what we hear. I cut the IPS degeneration down a bit. The sim seemed to prefer global feedback over local. Not looking to open that can of worms, I am just mentioning it.
It seems odd to me the sim suggests very lean bias. As expected, it gets worse fast as it gets further lean, not does not climb fast as it gets fatter. If this is at all believable, then the thermal compensation is not quite as critical as it could seem.
I am using totally different FFT parameters, so comparing my values is moot. This version of parameters comes out as .00035% , 20K 1-9 at rated 60W into 8 Ohms. Seems stable at 2, but the clamps are working by then. under 13uV noise.
For protection, I settled on the big outputs combined with the fast diodes on the bias servo. I am open to suggestions to the best diodes here. From what I think I learned, I want something like a HEXFRED. Fast and soft. I will just use the ESP DC and mute card.
Not making much progress on understanding the Tian probe. I sort of get the idea, but I am not totally convinced in my visual mind. Much more work to do. Build it or not, very educational. This months budget went to four new tires.
Some of the changes are based on what SPICE says, and may not be valid on what we hear. I cut the IPS degeneration down a bit. The sim seemed to prefer global feedback over local. Not looking to open that can of worms, I am just mentioning it.
It seems odd to me the sim suggests very lean bias. As expected, it gets worse fast as it gets further lean, not does not climb fast as it gets fatter. If this is at all believable, then the thermal compensation is not quite as critical as it could seem.
I am using totally different FFT parameters, so comparing my values is moot. This version of parameters comes out as .00035% , 20K 1-9 at rated 60W into 8 Ohms. Seems stable at 2, but the clamps are working by then. under 13uV noise.
For protection, I settled on the big outputs combined with the fast diodes on the bias servo. I am open to suggestions to the best diodes here. From what I think I learned, I want something like a HEXFRED. Fast and soft. I will just use the ESP DC and mute card.
Not making much progress on understanding the Tian probe. I sort of get the idea, but I am not totally convinced in my visual mind. Much more work to do. Build it or not, very educational. This months budget went to four new tires.
Hindsight is better than 20-20. I learned a ton here so I went back to revisit where I left off with my Hafler 120 rebuild. It was most successful with the new cleaner power supply and much cleaner sounding with the Exicon outputs and conversion to Miller compensation. I have been happily listening to in in my office for over a year.
So I took what little I learned here and applied it to the SPICE model as I had left off. By changing to TMC, a few part selections and current tweaks, I can reduce the modeled distortion to under .002, five fold better than the previous simulation, three fold better than the OEM spec. and better than 122 dB SNR, for over a 10 dB gain.
While I am at it, I should add the protection/mute board form ESP as I determined I should do on the "Reasonable" and my old Creek. Erno did a pretty good job ( like that is a surprise) on the internal amp protection so nothing to do there. The only "mistake" left to correct is adjusting the VAS standing current is not a clever way to set offset. I should do better.
One of these days it will warm up and I can get off the computer and into my shop to finish a speaker build.
So I took what little I learned here and applied it to the SPICE model as I had left off. By changing to TMC, a few part selections and current tweaks, I can reduce the modeled distortion to under .002, five fold better than the previous simulation, three fold better than the OEM spec. and better than 122 dB SNR, for over a 10 dB gain.
While I am at it, I should add the protection/mute board form ESP as I determined I should do on the "Reasonable" and my old Creek. Erno did a pretty good job ( like that is a surprise) on the internal amp protection so nothing to do there. The only "mistake" left to correct is adjusting the VAS standing current is not a clever way to set offset. I should do better.
One of these days it will warm up and I can get off the computer and into my shop to finish a speaker build.
As I continue to study and refine things, I am almost ready to switch to V-MOSFET outputs. I used Exicons in my Hafler, but I see ALFET as a good possibility. Of course, totally different protection requirements.
I was also looking at a lot of diode specs, as if they are used as clamps, their performance matters a lot. Even to the extent of their Fv being from .5 to 1.8V before current antd temps weigh in. Not finding a lot of diode models though. BAV21 series for newer standatd switching diodes, but nothing on a BAT41 or HF series.
I was also looking at a lot of diode specs, as if they are used as clamps, their performance matters a lot. Even to the extent of their Fv being from .5 to 1.8V before current antd temps weigh in. Not finding a lot of diode models though. BAV21 series for newer standatd switching diodes, but nothing on a BAT41 or HF series.
Another diode with low leakage: 1n3595
Spice model:
******
* Spice Model
* Item: 1N3595
* Date: 7/21/10
* Revision History: A
* ==========================================================
* This model was developed by:
* Central Semiconductor Corp.
* 145 Adams Avenue
* Hauppauge, NY 11788
*
* These models are subject to change without notice.
* Users may not directly or indirectly re-sell or
* re-distribute this model. This model may not
* be modified, or altered without the consent of Central Semiconductor Corp.
*
* For more information on this model contact
* Central Semiconductor Corp. at:
* (631) 435-1110 or Engineering@centralsemi.com
* Central Semiconductor Corp.
* ==========================================================
******
*SRC=1N3595;1N3595;Diodes;Rectifier;150V 0.150A 3us CS CS
.MODEL 1N3595 D (
+ IS=407.36E-12
+ N=1.7759
+ RS=.28325
+ IKF=9.7936
+ CJO=3.5499E-12
+ M=.29569
+ VJ=.3905
+ ISR=10.010E-21
+ NR=4.9950
+ BV=180.16
+ IBV=.52023
+ TT=144.27E-9)
***********
cheers,
Spice model:
******
* Spice Model
* Item: 1N3595
* Date: 7/21/10
* Revision History: A
* ==========================================================
* This model was developed by:
* Central Semiconductor Corp.
* 145 Adams Avenue
* Hauppauge, NY 11788
*
* These models are subject to change without notice.
* Users may not directly or indirectly re-sell or
* re-distribute this model. This model may not
* be modified, or altered without the consent of Central Semiconductor Corp.
*
* For more information on this model contact
* Central Semiconductor Corp. at:
* (631) 435-1110 or Engineering@centralsemi.com
* Central Semiconductor Corp.
* ==========================================================
******
*SRC=1N3595;1N3595;Diodes;Rectifier;150V 0.150A 3us CS CS
.MODEL 1N3595 D (
+ IS=407.36E-12
+ N=1.7759
+ RS=.28325
+ IKF=9.7936
+ CJO=3.5499E-12
+ M=.29569
+ VJ=.3905
+ ISR=10.010E-21
+ NR=4.9950
+ BV=180.16
+ IBV=.52023
+ TT=144.27E-9)
***********
cheers,
I guess you would, seeing ALFET and EXICON are actually identical spec. LATFET brands from the same SEMELAB source. Don't be fooled by maximum v continuous current ratings represented by the part numbers. AFAIK, only Renesas and Semelab manufacture power LATFETS.
- Status
- Not open for further replies.