Good question, what to look for is important.
In my experience, for emitter followers (normally in the output stage), where input and output are separated by Vbe, you need to look for the way in which Vbe changes with current, AND with Vce. This latter is often missing from the datasheet; in which case go for reasonably flat hfe with increasing collector current, and Vbe/Ic linearity for constant Vce.
For common emitter (input stage and VAS of audio amps) you should look for low Cob, less than 3pF is good (generally at Vce of 10V), speed is significant, 100MHz or greater, and I prefer lower peak collector currents, particularly in the VAS, no more than about 100mA. For diff pairs, Vbe with current should be as linear as you can find.
Not too much interest in common base for an audio amp, unless you are using cascodes.
The gnfb loop picks up all these non-linearities, but if you make each stage as linear as possible to begin with, then the distortion will be reduced before feedback is applied. This is always good.
Hugh
In my experience, for emitter followers (normally in the output stage), where input and output are separated by Vbe, you need to look for the way in which Vbe changes with current, AND with Vce. This latter is often missing from the datasheet; in which case go for reasonably flat hfe with increasing collector current, and Vbe/Ic linearity for constant Vce.
For common emitter (input stage and VAS of audio amps) you should look for low Cob, less than 3pF is good (generally at Vce of 10V), speed is significant, 100MHz or greater, and I prefer lower peak collector currents, particularly in the VAS, no more than about 100mA. For diff pairs, Vbe with current should be as linear as you can find.
Not too much interest in common base for an audio amp, unless you are using cascodes.
The gnfb loop picks up all these non-linearities, but if you make each stage as linear as possible to begin with, then the distortion will be reduced before feedback is applied. This is always good.
Hugh
You must be kidding, as a sub for the old 2sc3281/2sa1302
in all those old JVC/ sanyo /etc. receivers 1-they run cooler
(same factory bias) 2-sound better (opinion) 3- Only 3$
,cheap.I busted one apart and its die is bigger than the 2SC's.
With 2X the SOA and Vceo the mjl4281/4302 is just
great for an "abuse" mode amp.
I'm building a full EF "blameless" design now and my choice
is the MJL's even as I could buy the 2sc's for much less.
definitely the best cost/performance ratio out there.
OS
Homemodder -
I think you are referreing to the Thermaltrack NJL BJTs that have the built-in diode,
but I am not referrring to those, but to the rather new
NJW types. NJW 0302/0281
OnSemi has continued to improve on the 3281/1302 and 0302/0281 types and changed lettering from MJL to MJW to NJW.
I do like these but if there are any better ones out there, I would not hesitate to try them out. You mentioned that Toshiba has released new drivers and replacements for the 1943/5200. Would you please let us know what these are?
Sigurd
I think you are referreing to the Thermaltrack NJL BJTs that have the built-in diode,
but I am not referrring to those, but to the rather new
NJW types. NJW 0302/0281
OnSemi has continued to improve on the 3281/1302 and 0302/0281 types and changed lettering from MJL to MJW to NJW.
I do like these but if there are any better ones out there, I would not hesitate to try them out. You mentioned that Toshiba has released new drivers and replacements for the 1943/5200. Would you please let us know what these are?
Sigurd
homemodder said:
Sigurd, those devices look good maybe they can now compete with the japanese, I think Ill drop them a email and ask for a sample pair. I have the latest Toshiba product manual somewhere on my harddrive in pdf, Ill just need some time to locate it again, but I think I mentioned them in some thread here before.
Ill get back to you soon. The dealer I get parts told me at the time he would not be able to get any for some months though.
Ill get back to you soon. The dealer I get parts told me at the time he would not be able to get any for some months though.
JLH 2005
Korkotam
I don't see how you didn't recognize the parts on the schematic. Are
you referring to the parts I used on mine?
If so, here we go:
Q4, 5 , 6 - 2SA 970
Q7 - 2SA970
Q3 - 2SA1358
Q8 - 2SC3421
All the outputs are 2SC3857
I hope this gives you an idea of how I built the JLH this time around.
BTW I'm using 28.5 V, plus and minus.
Cheers
Korkotam
I don't see how you didn't recognize the parts on the schematic. Are
you referring to the parts I used on mine?
If so, here we go:
Q4, 5 , 6 - 2SA 970
Q7 - 2SA970
Q3 - 2SA1358
Q8 - 2SC3421
All the outputs are 2SC3857
I hope this gives you an idea of how I built the JLH this time around.
BTW I'm using 28.5 V, plus and minus.
Cheers
There is nothing new with regard to the lastest toshibas, the driver and power seems to be very much the same as the parts they supposed to replace.
The Onsemi have have made a giant leap forward though and spec sheet wise is now in sanken territory, warrent enough for me to order some samples.
The Onsemi have have made a giant leap forward though and spec sheet wise is now in sanken territory, warrent enough for me to order some samples.
After seeing this datasheet:
http://www.semicon.toshiba.co.jp/docs/datasheet/en/Transistor/2SC5948_en_datasheet_061116.pdf
I'm starting to believe the time axis on the tranisent thermal plots of the Fairchild FJL4x15 and FJA4x13 are off by a factor 10 or so. There can't be that big difference in die size. It would also explain why you can use 4 pairs of this size transistor in a 700W class AB or 1200W class H amp without problems. (Yes that's QSC RMX)
http://www.semicon.toshiba.co.jp/docs/datasheet/en/Transistor/2SC5948_en_datasheet_061116.pdf
I'm starting to believe the time axis on the tranisent thermal plots of the Fairchild FJL4x15 and FJA4x13 are off by a factor 10 or so. There can't be that big difference in die size. It would also explain why you can use 4 pairs of this size transistor in a 700W class AB or 1200W class H amp without problems. (Yes that's QSC RMX)
Same as what? It seems all four might be wrong.
Either that or else the Fairchild versions must be weaker.
I'm thinking like this: The 2SC5948 has the same pulse SOA as 2SC5200. So 2SC5200 should have a similar thermal plot, at least for times shorter than 100mS or so. 2SC5200 should be equivalent to FJL4315.
If you compare diagrams the new Toshiba has the same transient thermal impedance at 100ms as the Fairchild is claimed to have at 10ms.
It's the difference between having to derate for peak power or average power at bass frequencies.
It also seems strange that the package has no time constants longer than 10ms or so... So I'm pretty sure the fairchild diagrams are wrong (in a good way!). I'll have to measure them to be sure though.
Either that or else the Fairchild versions must be weaker.
I'm thinking like this: The 2SC5948 has the same pulse SOA as 2SC5200. So 2SC5200 should have a similar thermal plot, at least for times shorter than 100mS or so. 2SC5200 should be equivalent to FJL4315.
If you compare diagrams the new Toshiba has the same transient thermal impedance at 100ms as the Fairchild is claimed to have at 10ms.
It's the difference between having to derate for peak power or average power at bass frequencies.
It also seems strange that the package has no time constants longer than 10ms or so... So I'm pretty sure the fairchild diagrams are wrong (in a good way!). I'll have to measure them to be sure though.
And that's why I believe the transient thermal impedance plots are wrong for the FJx:s as the SOA is claimed to be the same as the Toshiba It's not the pulse SOA I believe might be wrong, it's the transient thermal impedance plot.
The Toshiba 2SC5200 does not have a transient thermal impedance plot in it's datasheet but the new 2SC5948 has one which is quite different from the FJx:s plot. The pulse SOA is claimed to be the same though.
Sanken publish transient thermal impedance plots too, and these are quite similar to the plot for the 2SC5948.
It's not the pulse SOA I believe might be wrong, it's the transient thermal impedance plot. The Toshiba 2SC5200 does not have a transient thermal impedance plot in it's datasheet but the new 2SC5948 has one which is quite different from the FJx:s plot. The pulse SOA is claimed to be the same though.
Sanken publish transient thermal impedance plots too, and these are quite similar to the plot for the 2SC5948.
homemodder said:
A Darlington conected VAS buffer is fine if it is cascoded. That allows you to use a low voltage transistor with high beta for the VAS buffer as well. Linearity is also aided as the majority of the VAS transistor base current is summed with its collector current instead of going to AC ground. I don't think D.Self has come up with anything new here.
And you are still ignoring the fact that the Stoc. fast amp has a cascoded VAS!
Cheers,
Glen
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.