Hi,
no, the last comparison has to do with derating the devices to take account of case temperature. Onsemi include a graph on the datasheet showing the maximum power you can extract from their devices at elevated case temperature.
Both To147/264 and To3 start at full power when Tc<=25degC.
But To147/264 are down to zero power when Tc=150degC.
However, To3 can still dissipate more than 28% of maximum power at the same Tc=150degC, due to their maximum junction temperature being 200degC.
This derating has to be applied to all the currents extracted from the SOAR curves. It's this dual derating that reduces the power that you can pull out of the output stage (reliably).
In classAB the device temperatures can be quite low when the amps are not driven at continuous disco levels. However ClassA run hot all the time, they even run slightly cooler when driven beyond ClassAB currents, but you would never do that with a domestic system.
It is at the elevated temperatures of ClassA that the plastic packages fare a little worse than the 200degC To3s. But you can alleviate this by increasing the heatsink and/or more pairs but increasing the cost.
no, the last comparison has to do with derating the devices to take account of case temperature. Onsemi include a graph on the datasheet showing the maximum power you can extract from their devices at elevated case temperature.
Both To147/264 and To3 start at full power when Tc<=25degC.
But To147/264 are down to zero power when Tc=150degC.
However, To3 can still dissipate more than 28% of maximum power at the same Tc=150degC, due to their maximum junction temperature being 200degC.
This derating has to be applied to all the currents extracted from the SOAR curves. It's this dual derating that reduces the power that you can pull out of the output stage (reliably).
In classAB the device temperatures can be quite low when the amps are not driven at continuous disco levels. However ClassA run hot all the time, they even run slightly cooler when driven beyond ClassAB currents, but you would never do that with a domestic system.
It is at the elevated temperatures of ClassA that the plastic packages fare a little worse than the 200degC To3s. But you can alleviate this by increasing the heatsink and/or more pairs but increasing the cost.
Thanks for the clarification.
Seems sensible in that case to buy more than three pairs then I can see what the temperature gets up to and experiment from there.
I am still not clear on what I need to order though. For complimentary pairs do I need half of one type and half the other, i.e. 3 of type MJL4302 & 3 of type MJL4281?
Seems sensible in that case to buy more than three pairs then I can see what the temperature gets up to and experiment from there.
I am still not clear on what I need to order though. For complimentary pairs do I need half of one type and half the other, i.e. 3 of type MJL4302 & 3 of type MJL4281?
Hi,
sorry, I should read your Q.
Yes, our Klone uses complementary pairs of output devices.
If you stick to the standard Klone, you need 3pair of each device per channel. i.e. 6off MJL4302 and 6off MJL4281 for a stereo amp.
If you double up that's a lot of hardware. And it probably won't cool adequately in a stereo chassis. You would need to consider monoblocs.
And Yes, you should spread out the devices along the sinks to the detriment of amplfier performance.
You may have to tailor the stability components to get a clean output under all load conditions when using long leads to and/or from the drivers. You may want to consider doubling up the drivers to copy KSA100. That achieves a slight shortening of lead lengths if laid out carefully.
A slight advantage in buying more pairs, is the ability to select either gain or Vbe for group matching to go into each upper and lower half. I think our school teachers would go mental if I were to suggest that four halves make a whole. 😉
sorry, I should read your Q.
Yes, our Klone uses complementary pairs of output devices.
If you stick to the standard Klone, you need 3pair of each device per channel. i.e. 6off MJL4302 and 6off MJL4281 for a stereo amp.
If you double up that's a lot of hardware. And it probably won't cool adequately in a stereo chassis. You would need to consider monoblocs.
And Yes, you should spread out the devices along the sinks to the detriment of amplfier performance.
You may have to tailor the stability components to get a clean output under all load conditions when using long leads to and/or from the drivers. You may want to consider doubling up the drivers to copy KSA100. That achieves a slight shortening of lead lengths if laid out carefully.
A slight advantage in buying more pairs, is the ability to select either gain or Vbe for group matching to go into each upper and lower half. I think our school teachers would go mental if I were to suggest that four halves make a whole. 😉
Sound advice - thanks.
From what you say, seems the 2SA1943 /2SC5200 that you are using may be a better combination than MJLs I planned.
Would you recommend them in my position, and do you know if there is a big difrerence in the way the two sound?
From what you say, seems the 2SA1943 /2SC5200 that you are using may be a better combination than MJLs I planned.
Would you recommend them in my position, and do you know if there is a big difrerence in the way the two sound?
Output transistors
Hi Bud,
The specific choice of transistors for the output is not without some controversy. Here's my synopsis of the threads 'findings':
mj15003/4: The original amp used a pair each of these, per channel. Very rugged, but sort of slow, technically there are now better transistors. Not available in the plastic packages, so you need to be good at drilling or have pre-drilled heatsinks. With transformers in the 300-400va per channel 2x of each are more or less bulletproof. Bigger transformers would suggest more paralleled transistors, since the rails won't collapse as low impedance loads are driven and the transistors would be at risk...
mj21193/4: The closest 'modern' transistor to the originals, much better gain linearity, more package options, easier to use. In the TO3 form can be substituted pretty much 1-1, in plastic package the ratio is probably better kept at 2-3. Since they are not much faster than the originals instability is not likely any more of a problem.
mj4302/4381: Very much faster, technically a much better transistor, AFAIK not available in TO3, so more are needed for an equally bulletproof output stage, 3 pairs per channel for instance. These transistors are very much faster than the originals and I'd exercise care in the output stage wiring, better to be safe than sorry.
The other transistors you mention (2SA1943 /2SC5200) are more or less unavailable, but there are a lot of fakes on the market, so the consensus seems to be avoid them and use the modern equivalents.
I have tried the mj15003/4 and the 21193/4 and they both sound excellent to me...I think the amp is probably 'better' into low impedance loads with the 21193/4 because of the better gain at high currents loads the drivers less, and it's easy to use more of the plastic outputs.
HTH
Stuart
Hi Bud,
The specific choice of transistors for the output is not without some controversy. Here's my synopsis of the threads 'findings':
mj15003/4: The original amp used a pair each of these, per channel. Very rugged, but sort of slow, technically there are now better transistors. Not available in the plastic packages, so you need to be good at drilling or have pre-drilled heatsinks. With transformers in the 300-400va per channel 2x of each are more or less bulletproof. Bigger transformers would suggest more paralleled transistors, since the rails won't collapse as low impedance loads are driven and the transistors would be at risk...
mj21193/4: The closest 'modern' transistor to the originals, much better gain linearity, more package options, easier to use. In the TO3 form can be substituted pretty much 1-1, in plastic package the ratio is probably better kept at 2-3. Since they are not much faster than the originals instability is not likely any more of a problem.
mj4302/4381: Very much faster, technically a much better transistor, AFAIK not available in TO3, so more are needed for an equally bulletproof output stage, 3 pairs per channel for instance. These transistors are very much faster than the originals and I'd exercise care in the output stage wiring, better to be safe than sorry.
The other transistors you mention (2SA1943 /2SC5200) are more or less unavailable, but there are a lot of fakes on the market, so the consensus seems to be avoid them and use the modern equivalents.
I have tried the mj15003/4 and the 21193/4 and they both sound excellent to me...I think the amp is probably 'better' into low impedance loads with the 21193/4 because of the better gain at high currents loads the drivers less, and it's easy to use more of the plastic outputs.
HTH
Stuart
Thanks for your synopsis Stuart.
There is a place called Profusion that sells 2SA1943 /2SC5200 in the UK for a reasonable price, I have never bought any transistors but they have a good reputation as far as I know.
My question would then be: If the 2SA1943, 2SC5200s are available and genuine, would they be a better buy the MJL4281/4302s?
There is a place called Profusion that sells 2SA1943 /2SC5200 in the UK for a reasonable price, I have never bought any transistors but they have a good reputation as far as I know.
My question would then be: If the 2SA1943, 2SC5200s are available and genuine, would they be a better buy the MJL4281/4302s?
reputation...
Hi Bud,
Several very aware people here, have recently been 'bitten' by fakes of these devices, reputable distributors are normally where they get them, caveat emptor...
That having been said, when you can get the darn things they are apparently very good transistors. I don't have any personal experience using them, but there are lots of people who really like them...
AFAIK the on-semi devices (4281/4302) etc are 'descendants' of the same technology that Motorola bought from Toshiba that was used to create the originals.
Stuart
Hi Bud,
Several very aware people here, have recently been 'bitten' by fakes of these devices, reputable distributors are normally where they get them, caveat emptor...
That having been said, when you can get the darn things they are apparently very good transistors. I don't have any personal experience using them, but there are lots of people who really like them...
AFAIK the on-semi devices (4281/4302) etc are 'descendants' of the same technology that Motorola bought from Toshiba that was used to create the originals.
Stuart
Hi,
2sa1943 have a very poor high voltage ability.
In a Klone running on +-40Vrails this is not an issue.
But 150W per device is an issue.
I had to use either 1943 or 4302 for Leach or Klone.
1943 running on a 3pair Leach clone PCB with +-60Vrails, then high voltage ability is becoming an issue and SOAR id definitely an issue.
Assessing their high voltage, current, power ability and SOAR and the ease of adding devices I went with 1943 to Klone and 4302 to Leach. Others could very easily arrive at a different choice.
2sa1943 have a very poor high voltage ability.
In a Klone running on +-40Vrails this is not an issue.
But 150W per device is an issue.
I had to use either 1943 or 4302 for Leach or Klone.
1943 running on a 3pair Leach clone PCB with +-60Vrails, then high voltage ability is becoming an issue and SOAR id definitely an issue.
Assessing their high voltage, current, power ability and SOAR and the ease of adding devices I went with 1943 to Klone and 4302 to Leach. Others could very easily arrive at a different choice.
Does anyone know how the Sanken 2SA1216 / 2SC2922 combo would perform?
I am thinking of using these. They look like very rugged devices to me, and I like the looks of their MT-200 casing.
Seems not many have used these in their clones, the wiki says only Jacco build one.
What modern equivalent does come close to the 2SA1943 / 2SC5200?
There is much written abouth these sa1943/sc5200, and it seems to me they have some reputation. But compared with modern equivalents, do they keep up their reputation?
I am thinking of using these. They look like very rugged devices to me, and I like the looks of their MT-200 casing.
Seems not many have used these in their clones, the wiki says only Jacco build one.
What modern equivalent does come close to the 2SA1943 / 2SC5200?
There is much written abouth these sa1943/sc5200, and it seems to me they have some reputation. But compared with modern equivalents, do they keep up their reputation?
Hi,
1943 is the modern equivalent.
I believe Toshiba brought it out when they made 1302 obsolete.
This was to try to overcome the fakes problem.
But it has been on the market for a while now and new fakes may be available.
Sanken seem to do good semis and they usually have a higher fT.
I think they are a potential alternative.
Their downside is pricing in the UK.
1943 is the modern equivalent.
I believe Toshiba brought it out when they made 1302 obsolete.
This was to try to overcome the fakes problem.
But it has been on the market for a while now and new fakes may be available.
Sanken seem to do good semis and they usually have a higher fT.
I think they are a potential alternative.
Their downside is pricing in the UK.
dr.strangelove3 said:
I've constructed half a dozen amplifiers with these ring-emitters, not counting RETs made by other Japanese manufacturers.
The 2SA1215 and 2SC2921 are the first series Sanken RETs, and the best, but only good for 150 watts.(see Burmester 828)
No idea how the Sanken version compares to the standard parts setup with Al's boards.
On the same loudspeaker system, comparable source and preamp as i audited the genuine KSA50 with, the KSA clone is certainly worth the additional hardware. (even used the same Stereoplay test-CD i used for comparisons in the 80s)
Al's layout, foil bypass capacitors, bigger and better electrolytics, and higher bias may have a much more important influence than the output device choice, who knows ?
These Sankens are a very good choice for high bias output stages, that i do know. In terms of ruggedness, they outperform TO247s. (and i find ways to buy them cheap)
AndrewT said:
Thanks for clearing that for me. What do you mean by higher fT? (I'm quite new in this) And how does this affect the amp?
The Sanken's aren't cheap here either, but then, they aren't the most expensive parts of an amplifier. So I have no problem spending some more on these parts.
Hi Jacco,
Thanks for your information. Did you run into problems using the Sankens? Or did they swap just in?
And what made you decide to use the Sankens?
Thanks for your information. Did you run into problems using the Sankens? Or did they swap just in?
And what made you decide to use the Sankens?
75 % more case to sink surface.
2-screw versus 1-screw torqueing, much more homogenous
pressure distribution to the heatsink.
ring-emitter output devices have as the name says many
emitters on the die, regular BJTs only 1. An RET is actually a
number of transistors on 1 die, you get a much more even heat distribution across the die as the emitters are spread out and divide the output current.
All these factors make them very rugged and perfect for class A use.
And they are the most linear production output devices that have become easily available, high transition frequency translates to linear behaviour. (and cheap, regular cost was $11 the piece 20 years ago, before that date only producers who ordered series of 1000s could get them overhere)
I'm in the corner of the ones who view that an output stage can not be fast enough. Why i admire the work of Mr Curl(his Parasound designs use the Sanken RETs too, the higher voltage types, see the JC-1s. $200 MSRP per output device, cheap too.)
I've used safe styroflex lead compensation cap values so far to make sure the amplifiers remain stable.(took out the CDE silver micas i used initially)
I had the Sankens and Al was willing to change the spacing of the output boards to use these big devices.
2-screw versus 1-screw torqueing, much more homogenous
pressure distribution to the heatsink.
ring-emitter output devices have as the name says many
emitters on the die, regular BJTs only 1. An RET is actually a
number of transistors on 1 die, you get a much more even heat distribution across the die as the emitters are spread out and divide the output current.
All these factors make them very rugged and perfect for class A use.
And they are the most linear production output devices that have become easily available, high transition frequency translates to linear behaviour. (and cheap, regular cost was $11 the piece 20 years ago, before that date only producers who ordered series of 1000s could get them overhere)
I'm in the corner of the ones who view that an output stage can not be fast enough. Why i admire the work of Mr Curl(his Parasound designs use the Sanken RETs too, the higher voltage types, see the JC-1s. $200 MSRP per output device, cheap too.)
I've used safe styroflex lead compensation cap values so far to make sure the amplifiers remain stable.(took out the CDE silver micas i used initially)
I had the Sankens and Al was willing to change the spacing of the output boards to use these big devices.
Changing the spacing on the output boards so the Sankens would fit was a minor problem, and I was glad to oblige. The topology of the Krell seems very robust, certainly in my experience, and I think the basic boards allow a lot of experimentation with component choice. I think that the outputs are the point where you can make the most difference, and it would be nice to hear some comparisons with the different devices that constructors have used. If only we were all rich, and could jet off to a nice venue to meet and compare our amps! 🙂
fT
Hi Bud,
In a simplified sense the fT of the transistor tells you it's upper frequency limit, it is the gain bandwidth product. The actual limit depends on a whole host of factors, but the bigger this number the better the response of the transistor at high freqencies.
Ordinarily you'd think that bigger must be better, but as the transistor becomes more capable at very high frequencies (more than a few megahertz) the components in the amp designed to control the gain can begin to 'fail', leading to situations where the amp is unstable. This happens because common components don't always behave well at high frequencies, so the amp can oscillate unexpectedly, either from caps whose impedance has begun to rise from the expected low impedance, or wires that couple RF from one part of the amp to another, resistors that look more like inductors etc.
The gist of it is, faster transistors demand more effort elsewhere to control their high gain at high frequencies, failure to take care of this can make the amp destroy transistors and in the worst case speakers etc.
The original mj15003/4 were 2MHz parts, the 21193 are 4MHz, the 4302 gets to 35MHz, while the 2sa1216 hits 40MHz. At frequencies in the low megahertz range the newer transistors will have a gain 17-20 times higher than the originals...and will still have some gain in the low 10's of megahertz...
HTH
Stuart
Hi Bud,
In a simplified sense the fT of the transistor tells you it's upper frequency limit, it is the gain bandwidth product. The actual limit depends on a whole host of factors, but the bigger this number the better the response of the transistor at high freqencies.
Ordinarily you'd think that bigger must be better, but as the transistor becomes more capable at very high frequencies (more than a few megahertz) the components in the amp designed to control the gain can begin to 'fail', leading to situations where the amp is unstable. This happens because common components don't always behave well at high frequencies, so the amp can oscillate unexpectedly, either from caps whose impedance has begun to rise from the expected low impedance, or wires that couple RF from one part of the amp to another, resistors that look more like inductors etc.
The gist of it is, faster transistors demand more effort elsewhere to control their high gain at high frequencies, failure to take care of this can make the amp destroy transistors and in the worst case speakers etc.
The original mj15003/4 were 2MHz parts, the 21193 are 4MHz, the 4302 gets to 35MHz, while the 2sa1216 hits 40MHz. At frequencies in the low megahertz range the newer transistors will have a gain 17-20 times higher than the originals...and will still have some gain in the low 10's of megahertz...
HTH
Stuart
badges/signs/whatever are etched and I should have them this tuesday. I'll try to have them out on wednesday at the latest.
Welcome!
Hi Kmj,
thanks for the update.
No need to rush on my account, my cases are not yet built.
Thursday will do just fine.
thanks for the update.
No need to rush on my account, my cases are not yet built.
Thursday will do just fine.
Something from my corner of half-finished stuff for the motivation then, not the needles but the thingie in the middle.
An externally hosted image should be here but it was not working when we last tested it.
