Thinking if all the driver transistors can be matched and also the output transistors can be matched then will there be any issues with the circuit? Like would be using 5 output transistors and would like to use 5 drivers individually to drive each output transistors? any pros and cons of this method
look at the SOA of the driver and the SOA of the transistors like 2sc5949
lets consider we are driving the transistors at 60V then with following driver
https://www.fairchildsemi.com/datasheets/KS/KSA1381.pdf
The output current of the Driver being .02A
the beta of the output transistors consider fairly being 60 the output current from the output transistor will be about 0.02 x 60 = 1.2A but if you look at the 2sc5949 datasheet SOA its clear that it can deliver more current as much as 3A
since the drive current is less the output current from the 2sc5949 will also be less. The actual problem starts when you have multiple pairs like 3 then in that case since the total current which will drive the output stage will be same but shared by the 3 transistors which in result will deliver the same aggregate current output. Hence I just thought why not use individual driver s rather just one struggling to drive all the three or more.
There's no reason not to use multiple driver transistors. Whether you actually need to do it depends on how beefy the driver transistors are, and how much drive current the output transistors need. Ideally you don't want to work any of the transistors too hard, or their beta will droop generating more distortion than you might like. However at this point the amp is blasting high power and the audibility of this distortion is debatable.
The 2SA1381 etc. are OK for drivers in hifi sized amps. The MJE1503x series of drivers are more powerful and would happily drive quite a few BJTs. If the output stage was too big to be handled by one pair of these, rather than using multiple pairs of drivers I would go to an output triple with another output transistor as a driver.
The 2SA1381 etc. are OK for drivers in hifi sized amps. The MJE1503x series of drivers are more powerful and would happily drive quite a few BJTs. If the output stage was too big to be handled by one pair of these, rather than using multiple pairs of drivers I would go to an output triple with another output transistor as a driver.
Last edited:
I guess cascoding would work, at the expense of complexity and maybe some headroom. The diamond driver gives a similar benefit by offloading most of the voltage and power dissipation off the driver transistors onto the current sources that bias them, but then again it is a Class A circuit, not really suited to driving output transistors directly.
Yes, extra complexity but it does allow the use of faster lower powered transistors which allows for a faster and / or more stable amp. My designs nearly always use TO-92 devices for drivers.
A diamond triple is a nice circuit but I wonder about the stability. Only ever tried it once so not much experience in it.
Complexity isn't necessarily an issue when the compexity doesn't really form part of the main signal path.
A diamond triple is a nice circuit but I wonder about the stability. Only ever tried it once so not much experience in it.
Complexity isn't necessarily an issue when the compexity doesn't really form part of the main signal path.
Hi,
There are better ways to improve an amplifiers output stage.
Here your certainly causing more problems than your fixing.
A Sziklai pair for the drivers springs to mind, and the adventurous
could consider a Sziklai pair for the output devices as well, without
complicating the basic EF biasing arrangement, the latter giving
giving each ouput device effectively its own driver, and you
could do that without changing the "nominal" driver part.
rgds, sreten.
There are better ways to improve an amplifiers output stage.
Here your certainly causing more problems than your fixing.
A Sziklai pair for the drivers springs to mind, and the adventurous
could consider a Sziklai pair for the output devices as well, without
complicating the basic EF biasing arrangement, the latter giving
giving each ouput device effectively its own driver, and you
could do that without changing the "nominal" driver part.
rgds, sreten.
Im need good subtle details hence thinking of this method rather I might have directly gone either with bigger drivers or triple..
In the past have used BC327-40/337-40 but despite having good Ic rating found they were too slow. Now using BC560 / 550. Cascoded with MJE15032/33. One pair per output pair.
P.S. Don't see a problem with sreten's idea. Seems an elegant solution to me. Stability issues are just a challenge, not a problem.
Last edited:
Lets see a few things ...
Ok idea is not new it has been done before in one and only amplifier worldwide the Alchemist forseti . Truly the circuit had a lot of issues and it was a nightmare to stabilize ...It ended up with huge miller caps to solve problems impossible to see in a simulator and killed the sonic signature of a very nice amplifier ...
Le us not forget that the alchemist was CFP or sziklai and that will complicate things ten times more in terms of stability ( also not possible to see in simulation )
If you like please search my articles on that issue and you will find detailed info .
Though i have my ideas about the stability on the forseti circuit and to my opinion it has a lot to do with the 4 drivers spread all over the board and resulting different thermal behavior so each driver actually performed with out any compensation and individually. That will be oscillation reason #1 it could be more ...
Am I spreading too much ? If so ask me i will take it slowly or explain any other questions you might have
To useful things
---1 it is a fact that owners of Alchemist talk about twice as much kick next to normal amplifiers
---2 I also find the idea extremely exciting
---3 circuit like the above cannot be simulated regarding the behavior of thermals or beta changing in the driver area versus temperature
Possible solutions
---A In the alchemist and CFP sziklai circuits drivers must be kept away from main heat sink , thermal compensation takes place on the drivers and not in the output stages . So the idea is closely match them and common heatsink them and all problems solved
( in theory sounds nice never had the time to test though )
---B in EFP circuits you may as well install everything in one heatsink and compensation over all and you might get away with it with less or no miller at all
---C Added to the above study there is got to be the aspect that usually circuits of either kind that are biased high ( high AB or straight A ) suffer less from stability issues like described above
One will have to construct to find out Layout and pcb will be critical and to my understanding you may simulate the audio part but not the thermal ..
Let us know of your progress
If there is anything else i might help please let me know
Ok idea is not new it has been done before in one and only amplifier worldwide the Alchemist forseti . Truly the circuit had a lot of issues and it was a nightmare to stabilize ...It ended up with huge miller caps to solve problems impossible to see in a simulator and killed the sonic signature of a very nice amplifier ...
Le us not forget that the alchemist was CFP or sziklai and that will complicate things ten times more in terms of stability ( also not possible to see in simulation )
If you like please search my articles on that issue and you will find detailed info .
Though i have my ideas about the stability on the forseti circuit and to my opinion it has a lot to do with the 4 drivers spread all over the board and resulting different thermal behavior so each driver actually performed with out any compensation and individually. That will be oscillation reason #1 it could be more ...
Am I spreading too much ? If so ask me i will take it slowly or explain any other questions you might have
To useful things
---1 it is a fact that owners of Alchemist talk about twice as much kick next to normal amplifiers
---2 I also find the idea extremely exciting
---3 circuit like the above cannot be simulated regarding the behavior of thermals or beta changing in the driver area versus temperature
Possible solutions
---A In the alchemist and CFP sziklai circuits drivers must be kept away from main heat sink , thermal compensation takes place on the drivers and not in the output stages . So the idea is closely match them and common heatsink them and all problems solved
( in theory sounds nice never had the time to test though )
---B in EFP circuits you may as well install everything in one heatsink and compensation over all and you might get away with it with less or no miller at all
---C Added to the above study there is got to be the aspect that usually circuits of either kind that are biased high ( high AB or straight A ) suffer less from stability issues like described above
One will have to construct to find out Layout and pcb will be critical and to my understanding you may simulate the audio part but not the thermal ..
Let us know of your progress
If there is anything else i might help please let me know
Last edited:
In the last hifi amp I built, I got better than 0.01% THD at 10kHz and all power levels from 1W to full output, using just the standard double emitter follower output stage. That was built and tested with a THD analyser, not simulated. The THD at 1kHz was too low for my old Tektronix analyser to measure, and below 1W at 10kHz the reading was just noise.
If even a simple output stage gives perfectly good performance, then why complicate it? I don't honestly believe that 0.001% THD at 10kHz, as you might hope to get from the latest Hawksford error correction schemes, sounds any different to 0.01%. And the more complicated the output stage, the more likely it is to suffer from parasitics.
I also don't believe that changing the number of driver transistors without changing the topology will make any audible difference, unless the driver stage was running out of steam previously.
If even a simple output stage gives perfectly good performance, then why complicate it? I don't honestly believe that 0.001% THD at 10kHz, as you might hope to get from the latest Hawksford error correction schemes, sounds any different to 0.01%. And the more complicated the output stage, the more likely it is to suffer from parasitics.
I also don't believe that changing the number of driver transistors without changing the topology will make any audible difference, unless the driver stage was running out of steam previously.
Practice have prove you wrong The alchemist sounds different ( not necessarily better ) and that has a lot to do with the output stage and the rest of the configuration .
Let us also not forget that we talk about a thing that has never been tested Alchemist is a CFP and 30+ years old design so its another fruit .
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.