2SA1294/2SC3263 are the easiest to source.
2SA2223/2SC6145 are the more powerful 160W TO-3P's , same die as the MT-200.
Good .. almost 4.5A /100ms SOA @ 60 V , way better than most TO-3p's.
2SA1859-a/2SC4883-a for drivers.
OS
Last edited:
I still have more to look at but one recommendation already, run the 12 V stuff (servo and level shifters in the front end) off the 30 V low rail, not the 60 V one.
Save power, even if it's not much, runs cooler.
Best wishes
David
I could do that with my planned 10 pin connector between IPS/OPS.
IPS will have +60/+30/PD+/GND/NFB/ND-/-30/-60 .
I want to have my SMD IPS's fully compatable with both this OPS and
the slewmaster OPS's. They just keep on building those Slewmaster EF3's.
But consider that when the greenamp is running short of class H , having
the VAS/IPS running off the 60V rail is like having a fully separate supply for
the IPS. I'm sure PSRR would be superior.
Also , both the VAS and preceding stages benefit from the cap multipliers.
I would need to have redundant multipliers for both 30 and 60V supplies.
BTW - IPS only draws 20ma per rail. Most of that is the Zener regulators and LED's. Very green !
OS
IPS will have +60/+30/PD+/GND/NFB/ND-/-30/-60 .
I want to have my SMD IPS's fully compatable with both this OPS and
the slewmaster OPS's. They just keep on building those Slewmaster EF3's.
But consider that when the greenamp is running short of class H , having
the VAS/IPS running off the 60V rail is like having a fully separate supply for
the IPS. I'm sure PSRR would be superior.
Also , both the VAS and preceding stages benefit from the cap multipliers.
I would need to have redundant multipliers for both 30 and 60V supplies.
BTW - IPS only draws 20ma per rail. Most of that is the Zener regulators and LED's. Very green !
OS
Last edited:
They are very powerful BJTs. I used 2 pairs, with +/-55V, to push 2ohms load, with no problem. Easy to get the from Digikey, of from Profusionplc to Europe (before brexit).
Sajti
I thought you already had my models for the 2SA4883A/C1859A.
Here are some models for the C3264/A1295 I just worked out. They are the MT-200 versions of the 3263/1294, same die different package.
.model 2SC3264-Y_kq npn ; First revision by keantoken - Feb 24 2019
+ Bf=83 Vaf=30 Is=900f Cjc=600p Cje=15n Tf=2.2n
+ Rb=.22 Rbm=0 Irb=200m Re=38m
+ Rc=1m Rco=750m Vo=1000 Gamma=12n
+ Br=8 Ise=8p Ikf=1000 Nk=0.2
+ Itf=10 Xtf=100 Vtf=0.8 Xtb=1.4
+ Vceo=230 Icrating=17A mfg=Sanken
.model 2SA1295-Y_kq pnp ; First revision by keantoken - Feb 24 2019
+ Bf=90 Vaf=150 Is=60f Cjc=1.2n Cje=20n Tf=3.9n
+ Rb=1.4 Rbm=0 Irb=3 Re=33m
+ Rc=8m Rco=380m Vo=1000 Gamma=160p
+ Br=8 Ise=0 Ikf=22 Nk=0.5
+ Itf=10 Xtf=100 Vtf=0.8 Xtb=1.7
+ Vceo=230 Icrating=17A mfg=Sanken
EDIT: added more reasonable capacitance values. The model can be improved if someone measures collector capacitance at 0V. Also helpful would be for someone to measure Vaf, as I had to guess based on pixels in the datasheet.
Here are some models for the C3264/A1295 I just worked out. They are the MT-200 versions of the 3263/1294, same die different package.
.model 2SC3264-Y_kq npn ; First revision by keantoken - Feb 24 2019
+ Bf=83 Vaf=30 Is=900f Cjc=600p Cje=15n Tf=2.2n
+ Rb=.22 Rbm=0 Irb=200m Re=38m
+ Rc=1m Rco=750m Vo=1000 Gamma=12n
+ Br=8 Ise=8p Ikf=1000 Nk=0.2
+ Itf=10 Xtf=100 Vtf=0.8 Xtb=1.4
+ Vceo=230 Icrating=17A mfg=Sanken
.model 2SA1295-Y_kq pnp ; First revision by keantoken - Feb 24 2019
+ Bf=90 Vaf=150 Is=60f Cjc=1.2n Cje=20n Tf=3.9n
+ Rb=1.4 Rbm=0 Irb=3 Re=33m
+ Rc=8m Rco=380m Vo=1000 Gamma=160p
+ Br=8 Ise=0 Ikf=22 Nk=0.5
+ Itf=10 Xtf=100 Vtf=0.8 Xtb=1.7
+ Vceo=230 Icrating=17A mfg=Sanken
EDIT: added more reasonable capacitance values. The model can be improved if someone measures collector capacitance at 0V. Also helpful would be for someone to measure Vaf, as I had to guess based on pixels in the datasheet.
Last edited:
They work , slight retune of currents. Oh , over 10A ... these models go "haywire". 8R load , OK . 4R = "haywire"
2SC3264-Y_kq - this one "acts" wrong. Negative class H works far different
than with FJP5200_Okq .
OS
2SC3264-Y_kq - this one "acts" wrong. Negative class H works far different
than with FJP5200_Okq .
OS
Check that the drivers are not starved at peaks, that is the problem I'm seeing.
The base current does have some spikiness, due to the supply switching (saturation makes base current very sensitive to Vce) and also probably due to the fact that the transition to quasi-saturation in the simulator model is more abrupt than in real life. I fitted the model to the curves given in the fairly good Sanken datasheet; it could be that the misbehavior is realistic. Maybe we can find out one way or another.
In case you missed them, here are my 4883/1859 models:
.MODEL 2SC4883A_kq npn (Bf=140 Ikf=100 Is=600f Vaf=500
+ Rb=1.7 Re=105m RC=0 Rco=6 Ibc=150f Vo=30 Gamma=250n
+ Cje=1.2n Cjc=72p Tf=875p Vtf=1.2 Itf=1 Qco=8p
+ Nk=1.2 Br=2 Var=22.9 Ikr=36 TR=85n
+ Xtb=0.34 Xtf=1.36
+ Vceo=180 Icrating=2A mfg=Sanken)
.MODEL 2SA1859A_kq pnp (Bf=162 IKF=10 Is=500f Vaf=350
+ Rb=0 Re=0 Rc=90m Rco=3.25 Ibc=150f Vo=120 Gamma=22n
+ Cje=1.5n Cjc=72p Tf=1.9n Vtf=1.9 Itf=1 Qco=8p
+ Br=2 Ikr=10 Var=23 Tr=188.8n
+ Xtb=0.138 Xtf=5
+ Vceo=180 Icrating=2A mfg=Sanken)
The base current does have some spikiness, due to the supply switching (saturation makes base current very sensitive to Vce) and also probably due to the fact that the transition to quasi-saturation in the simulator model is more abrupt than in real life. I fitted the model to the curves given in the fairly good Sanken datasheet; it could be that the misbehavior is realistic. Maybe we can find out one way or another.
In case you missed them, here are my 4883/1859 models:
.MODEL 2SC4883A_kq npn (Bf=140 Ikf=100 Is=600f Vaf=500
+ Rb=1.7 Re=105m RC=0 Rco=6 Ibc=150f Vo=30 Gamma=250n
+ Cje=1.2n Cjc=72p Tf=875p Vtf=1.2 Itf=1 Qco=8p
+ Nk=1.2 Br=2 Var=22.9 Ikr=36 TR=85n
+ Xtb=0.34 Xtf=1.36
+ Vceo=180 Icrating=2A mfg=Sanken)
.MODEL 2SA1859A_kq pnp (Bf=162 IKF=10 Is=500f Vaf=350
+ Rb=0 Re=0 Rc=90m Rco=3.25 Ibc=150f Vo=120 Gamma=22n
+ Cje=1.5n Cjc=72p Tf=1.9n Vtf=1.9 Itf=1 Qco=8p
+ Br=2 Ikr=10 Var=23 Tr=188.8n
+ Xtb=0.138 Xtf=5
+ Vceo=180 Icrating=2A mfg=Sanken)
Last edited:
I misunderstood, sound quality difference is noticable, (flac being superior to mp3) I thought you were referring to lag or other network performance issue(s). I'll shut up now.
VAF is usually in the order of the breakdown potential, often somewhat more.
Admittedly Early effect models in base level Spice are simplistic but 30 V seems very implausible for a 230 V transistor.
Bump it up.
Best wishes
David
It's hard to assess your models quickly because the parameter order is somewhat inconsistent.
Could I recommend that you adopt a standard order, with related parameters collected?
Also a tip from my professional experience rather than DIY, I find it useful to document each parameter, whether it's an estimate, or derived from what measured data or what plot on what datasheet.
Last edited:
Vaf was taken from the 20mA Ib trace in the first plot. It makes sense this would be conservative as Vaf is lower at low Vce.
http://www.farnell.com/datasheets/1854823.pdf
Without data I can only bump it up to 150 like the NPN model and hope that is close to realistic. If I had some of these I could do the capacitance and Vaf measurements myself.
http://www.farnell.com/datasheets/1854823.pdf
Without data I can only bump it up to 150 like the NPN model and hope that is close to realistic. If I had some of these I could do the capacitance and Vaf measurements myself.
Yes, Early effect varies but Spice only has one parameter to match (in the basic model at least)
So I would pick it to be more representative of normal use, even if it matches one particular plot less well.
In the first plot the curve hasn't flattened out yet.
150 seems plausible to me, NPN are usually better than the complementary PNP so at least don't have it worse.
Best wishes
David
So I would pick it to be more representative of normal use, even if it matches one particular plot less well.
In the first plot the curve hasn't flattened out yet.
150 seems plausible to me, NPN are usually better than the complementary PNP so at least don't have it worse.
Best wishes
David
Last edited:
I made a thread for the Sanken models so we can take the discussion out of this thread:
https://www.diyaudio.com/forums/sof...-models-sanken-2sc3264-a1295.html#post5710053
https://www.diyaudio.com/forums/sof...-models-sanken-2sc3264-a1295.html#post5710053
Ok, that explains it, not that you just overlooked it, fine.
I continue to play with it, quite educational.
Mainly I want to simplify, did you ever consider a self-biased JFET IPS?
Lets you skip not only the current sources but the level shifter emitter followers on the inputs.
Linear Systems now have the J74 and K170 so we have nice low noise and decent transconductance parts available.
Best wishes
David
Hi - Dave.
Updated asc ...
-Silicon Carbide Schottky's modeled.
-R123/R136 bias adj. can be tweaked from (full switching) class H up to "stacker" mode for the inner/outer devices.
IF they are set at the middle ground between the two , power fully
is shared between diode and class H. Higher power transitions into
(more) transistor only operation. Low power (<.6V input) is all SiC diode I.
One can adjust between fidelity/efficiency.
It is basically 2 control voltages that one could automate with a fancy control circuit based on average musical level.
-MOSFET's seem to work VERY well when replacing Q109/Q114.
Edit - I stay away from hard to find , expensive Jfet's. All will be $.40c cheap mouser BJT's (for 20ppm).
OS
Hi Pete, please stay away from these bastard j74 k170.
We can't spent 10 or 15euro for a small signal fet.not to give money for fake parts.
We can't spent 10 or 15euro for a small signal fet.not to give money for fake parts.
Last edited:
JFET = fire and pitchforks !! 
😱
Many through-hole are going end of life on mouser. I have a mouser account ,
so they are my main source - 2 days from Texas ....
As they go "end of life' many unscrupulous (even some here at DIYA) are buying them up.
I see ksa1381 drop from thousands to 0 in one day.
So , expect these "horders" to fleece you with .80c Mouser parts at a big markup.
$$$ $crew them , will make easy HQ projects with bigger SOT and avoid
"unobtanium" B$.
OS
😱Many through-hole are going end of life on mouser. I have a mouser account ,
so they are my main source - 2 days from Texas ....
As they go "end of life' many unscrupulous (even some here at DIYA) are buying them up.
I see ksa1381 drop from thousands to 0 in one day.
So , expect these "horders" to fleece you with .80c Mouser parts at a big markup.
$$$ $crew them , will make easy HQ projects with bigger SOT and avoid
"unobtanium" B$.
OS
What will you replace the 1381 drivers with to get to the higher than 60 volt rails needed for the 200, 300 or 400 watt versions that you have mentioned before?
Thanks for all your Pitch fork hard work on this and many other projects. I have learned a lot even if I do not understand all that you detail out in your posts. Thanks
Thanks for all your Pitch fork hard work on this and many other projects. I have learned a lot even if I do not understand all that you detail out in your posts. Thanks
ZTX957/857 +/- 300V/85mhz/13pf Cob for SOT223.
For this new age , 10ma VAS's are GONE.
EF3's with 2ma pre-drivers , VAS's with <5ma ....
The ZTX's should give the same performance as the 1381/3503's.
The only through -hole I will use is large resistors , caps ,
TO-220 and TO-3P (sanken/toshiba/ON).
Large DIODES will be 3 pin D-PAK - like a TO-126 that solders on to the
copper.
For small signal they make so many SOT23's , or dual NpN / N/P pairs.
OS
- Status
- Not open for further replies.