Most of you know that i am an Sziklai lover !
In many of my posts you will find plenty of reference in various sziklai attempts that usually include some historical data .
The subject here is the Alchemist Forseti that is the most unique sziklai that ever designed since it features double sziklai pairs in the output .
Other than the output stage design the Alchemist Forseti is a well designed machine that has a reputation of sweet and round sound and plenty of kick since its a dual mono and features plenty of capacitance and good drive ability from the extra pair of outputs and drivers .
The only problem of the Alchemist APD 15 was the sonic performance that was limited by huge 680 pf miller caps on both the drivers and the outputs to keep stable with double an sziklai pair for an output stage .
-------------------------------------------------------------------------------------
so please ladies and gents allow me to quote these questions from my P3A thread in order to get some answers
---APD 15 senses the idle of the amplifier in the wrong place as we now know .
--- It senses over the outputs and not over the drivers , Plus that drivers all 4 of them share no heatsink at all ( may be not needed) but no heatsink at all means also no common behavior to the 4 drivers since these are behaving on their own when it comes to temperature
---Then APD 15 features huge miller caps of the size 680pf to keep stable in the output these may prevent oscillation but will do nothing to thermal stability .
Result :
Is there a chance that the designer tried to cure stability issues with the wrong way ? it is common practice that miller caps put a brake in oscillation issues but will only make the circuit worst if the source of the problem is thermal stability .
Why something is telling me that this amplifier may also work without the sonic killers miler caps on the outputs or have them replaced with something a lot lot smaller that will work as an oscillation stopper without killing the sonics ?
The only thing you have to do is redesign the thermals in the drivers and stability will come to visit you ....
Think please ...
kind regards
Sakis
In many of my posts you will find plenty of reference in various sziklai attempts that usually include some historical data .
The subject here is the Alchemist Forseti that is the most unique sziklai that ever designed since it features double sziklai pairs in the output .
Other than the output stage design the Alchemist Forseti is a well designed machine that has a reputation of sweet and round sound and plenty of kick since its a dual mono and features plenty of capacitance and good drive ability from the extra pair of outputs and drivers .
The only problem of the Alchemist APD 15 was the sonic performance that was limited by huge 680 pf miller caps on both the drivers and the outputs to keep stable with double an sziklai pair for an output stage .
-------------------------------------------------------------------------------------
so please ladies and gents allow me to quote these questions from my P3A thread in order to get some answers
---APD 15 senses the idle of the amplifier in the wrong place as we now know .
--- It senses over the outputs and not over the drivers , Plus that drivers all 4 of them share no heatsink at all ( may be not needed) but no heatsink at all means also no common behavior to the 4 drivers since these are behaving on their own when it comes to temperature
---Then APD 15 features huge miller caps of the size 680pf to keep stable in the output these may prevent oscillation but will do nothing to thermal stability .
Result :
Is there a chance that the designer tried to cure stability issues with the wrong way ? it is common practice that miller caps put a brake in oscillation issues but will only make the circuit worst if the source of the problem is thermal stability .
Why something is telling me that this amplifier may also work without the sonic killers miler caps on the outputs or have them replaced with something a lot lot smaller that will work as an oscillation stopper without killing the sonics ?
The only thing you have to do is redesign the thermals in the drivers and stability will come to visit you ....
Think please ...
kind regards
Sakis
Schematic
Here is the schematic.
Sakis, I also like Sziklai OPS, however I don't like its implementation in ADP15. To be honest, I don't like many things in this design
Examples of good designs with Sziklai - Audiolabs (8000A, 8000P, 8200M). They use Sankens without any additional compensation in OPS and the circuit is rock stable.
For this particular one I don't have any good idea so far, though I will think a bit more...
Cheers,
Valery
Here is the schematic.
Sakis, I also like Sziklai OPS, however I don't like its implementation in ADP15. To be honest, I don't like many things in this design
Examples of good designs with Sziklai - Audiolabs (8000A, 8000P, 8200M). They use Sankens without any additional compensation in OPS and the circuit is rock stable.For this particular one I don't have any good idea so far, though I will think a bit more...
Cheers,
Valery
Attachments
Member
Joined 2009
Paid Member
OK - as an idea - I would do the following:
1) get rid of one of the drivers (source both OPs' bases from one driver);
2) add resistors (0.1 - 0.22 Ohm) serially with OPs' emitters;
3) reduce the resistor value at the drivers' collectors to some 100 Ohm (to allow more power to the drivers);
4) re-adjust the bias and see how it behaves without the caps at all.
Depending on the result:
- add base stoppers (2.2 Ohm) to the OPs;
- introduce the cap (47-100pF) between B-C of the driver.
Something like that.
Cheers,
Valery
1) get rid of one of the drivers (source both OPs' bases from one driver);
2) add resistors (0.1 - 0.22 Ohm) serially with OPs' emitters;
3) reduce the resistor value at the drivers' collectors to some 100 Ohm (to allow more power to the drivers);
4) re-adjust the bias and see how it behaves without the caps at all.
Depending on the result:
- add base stoppers (2.2 Ohm) to the OPs;
- introduce the cap (47-100pF) between B-C of the driver.
Something like that.
Cheers,
Valery
This will be long ( as usual )
At the first place i will have to thank all of you for your input ...still i will have to argue with many ...
@valery
Please remove from any of my threads the word Audiolab 8000 i see it as contamination... you are wrong 8000 was a horrible machine that obviously never failed since the power supply was too small fro such an output stage but the worst was the sonic performance due to the overall design .As about the thermals that was the worst part of the 8000 and the machine could maintain specs of bias only when the room temperature was about 20-22 degrees average ...Increase this to 28 and call me back ...
@Gareth
We have been there before in our private conversations ...sonic performance cannot be simulated ...Take uncle Charly AKA Destroyer X amplifier and simulate it ..Then built it with an average pcb and power supply ...Then built a version that will have for example 2X10.000 psu caps on board and very close to the output transistors I bet you that the one that features the psu caps on board will sound far better Can you simulate that ? Obviously not ...
Bryston 2B can be as good as any Pioneer or Sansui or any consumer amplifier on paper the basic differences between them are very few but still make the 2B hit the sky
**Small caps only 2X4700 per ch but located on board where these are supposed to be very next to the output transistors
**Dual mono /modular design
**Excellent choice of semis and proper use of them
That was enough to create a legend
@ valery ( latest post )
You are actually instructing me how to convert the APD 15 to an Audiolab 8000 ...Sorry but no thanks that wasn't my intention
@nigel
You might be correct ... Since it is impossible to run a background check to all of you I will have to place here mine so we know where are we standing from .
I repair amplifiers i do this all my life I am a freelancer so i have been repairing from the cheapest AKAI till Perreaux that will cost 5-6000 USD or UBA of 15000 USD
For every amplifier we repair we take feedback from the costumers measuring and we also make listening tests to every new amplifier that joins the list
Last year only we repaired in total 1000 audio devices
What i am trying to say here is that beyond living from this we record real life results from listeners ...
That will mean that for 5 years now we give P3A amplifier to listen to costumers while their amplifiers are under repair ...Each and every one of them was able to tell the difference .
So if you like Sziklai just make a P3A and all will be more than fine !!!
But if you look for more power with similar sonic performance the APD 15 was notorious for the kick coming from the output stage and the beefy power supply but the sonic performance was bad due to huge miller capacitors Now can we fix that ?
To farther support more my theory look at most Elektor /giesberts designs that suffered from the exact same issues like way too many stages current mirrors /sinks /sources in diamond input stages that oscillated probably for the same reasons and since the amp was hfe/beta/gain depending then the only solution was to add capacitors here and there to prevent oscillation .
yet again electronic circuits cannot be treated with medicine You will have to cure the problem where it starts ...
Thanks for your time
Kind regards
Sakis
At the first place i will have to thank all of you for your input ...still i will have to argue with many ...
@valery
Please remove from any of my threads the word Audiolab 8000 i see it as contamination... you are wrong 8000 was a horrible machine that obviously never failed since the power supply was too small fro such an output stage but the worst was the sonic performance due to the overall design .As about the thermals that was the worst part of the 8000 and the machine could maintain specs of bias only when the room temperature was about 20-22 degrees average ...Increase this to 28 and call me back ...
@Gareth
We have been there before in our private conversations ...sonic performance cannot be simulated ...Take uncle Charly AKA Destroyer X amplifier and simulate it ..Then built it with an average pcb and power supply ...Then built a version that will have for example 2X10.000 psu caps on board and very close to the output transistors I bet you that the one that features the psu caps on board will sound far better Can you simulate that ? Obviously not ...
Bryston 2B can be as good as any Pioneer or Sansui or any consumer amplifier on paper the basic differences between them are very few but still make the 2B hit the sky
**Small caps only 2X4700 per ch but located on board where these are supposed to be very next to the output transistors
**Dual mono /modular design
**Excellent choice of semis and proper use of them
That was enough to create a legend
@ valery ( latest post )
You are actually instructing me how to convert the APD 15 to an Audiolab 8000 ...Sorry but no thanks that wasn't my intention
@nigel
You might be correct ... Since it is impossible to run a background check to all of you I will have to place here mine so we know where are we standing from .
I repair amplifiers i do this all my life I am a freelancer so i have been repairing from the cheapest AKAI till Perreaux that will cost 5-6000 USD or UBA of 15000 USD
For every amplifier we repair we take feedback from the costumers measuring and we also make listening tests to every new amplifier that joins the list
Last year only we repaired in total 1000 audio devices
What i am trying to say here is that beyond living from this we record real life results from listeners ...
That will mean that for 5 years now we give P3A amplifier to listen to costumers while their amplifiers are under repair ...Each and every one of them was able to tell the difference .
So if you like Sziklai just make a P3A and all will be more than fine !!!
But if you look for more power with similar sonic performance the APD 15 was notorious for the kick coming from the output stage and the beefy power supply but the sonic performance was bad due to huge miller capacitors Now can we fix that ?
To farther support more my theory look at most Elektor /giesberts designs that suffered from the exact same issues like way too many stages current mirrors /sinks /sources in diamond input stages that oscillated probably for the same reasons and since the amp was hfe/beta/gain depending then the only solution was to add capacitors here and there to prevent oscillation .
yet again electronic circuits cannot be treated with medicine You will have to cure the problem where it starts ...
Thanks for your time
Kind regards
Sakis
There should be little difference with "off-board" caps if inductance and resistance are kept low by using twisted wires of reasonable gauge.
Anyway, I think this amp uses kind of a 2-pole compensation. VAS gain is shelved, then the caps at the CFPs take over.
It's a bit of an odd concept. I'd think it's somewhat lacking in current gain. Input LTP tail current and VAS current would not seem to be unusually high. Also, TIP32C in VAS? Yikes.
Anyway, I think this amp uses kind of a 2-pole compensation. VAS gain is shelved, then the caps at the CFPs take over.
It's a bit of an odd concept. I'd think it's somewhat lacking in current gain. Input LTP tail current and VAS current would not seem to be unusually high. Also, TIP32C in VAS? Yikes.
Sakis, no problem, let's keep the brand you don't like aside. But your question grabbed my interest
I did some modelling. As we can see now, the main issue of the original APD15 desidn - transistors. Both TIP31/32 and TIP35/36 are awful for the purpose. Slow, huge input capacitance transistors, provoking oscillation at rather low frequencies, that has to be suppressed with large lag compensation capacitors. Not a good concept by definition.
See bode, see the phase response. THD is incorrect just because of heavy oscillation.
I also attach the model with transistors changed to some more decent ones. You can see the difference. Also, TMC compensation would improve the distortion at the higher end of the frequency response.
P3A is a proven design, but it's possible to spoil its benefits, as we can see. Sure, the models are not perfect, live performance will be different, etc. But they give an indication, are they?
Press the soft-button at the bottom-left corner of the image to see it full size.
Cheers,
Valery
I did some modelling. As we can see now, the main issue of the original APD15 desidn - transistors. Both TIP31/32 and TIP35/36 are awful for the purpose. Slow, huge input capacitance transistors, provoking oscillation at rather low frequencies, that has to be suppressed with large lag compensation capacitors. Not a good concept by definition.
See bode, see the phase response. THD is incorrect just because of heavy oscillation.
I also attach the model with transistors changed to some more decent ones. You can see the difference. Also, TMC compensation would improve the distortion at the higher end of the frequency response.
P3A is a proven design, but it's possible to spoil its benefits, as we can see. Sure, the models are not perfect, live performance will be different, etc. But they give an indication, are they?
Press the soft-button at the bottom-left corner of the image to see it full size.
Cheers,
Valery
Attachments
Sakis, no problem, let's keep the brand you don't like aside. But your question grabbed my interest
I did some modelling. As we can see now, the main issue of the original APD15 desidn - transistors. Both TIP31/32 and TIP35/36 are awful for the purpose. Slow, huge input capacitance transistors, provoking oscillation at rather low frequencies, that has to be suppressed with large lag compensation capacitors. Not a good concept by definition.
See bode, see the phase response. THD is incorrect just because of heavy oscillation.
I also attach the model with transistors changed to some more decent ones. You can see the difference. Also, TMC compensation would improve the distortion at the higher end of the frequency response.
P3A is a proven design, but it's possible to spoil its benefits, as we can see. Sure, the models are not perfect, live performance will be different, etc. But they give an indication, are they?
Press the soft-button at the bottom-left corner of the image to see it full size.
Cheers,
Valery
Lets suppose that with better transistors the all thing works better .
here is a question though ?
Does your software support a feature to add tolerance to any of the components ? even better can you mess up with the hfe of the drivers in the model to see how the results will come up ? I can imagine a 20%+or- just to see how its going to be
That will be an interesting thing to have answered .
Thanks for your time Valery
Regards
Sakis
In my experience with Sziklai pairs and stability the first pre driver transistor needs to be fast with high ft and low cob. The tip 31/32 should have a bandwidth higher than the slave output transistor, for example a ksc2682 at around 100-150 MHz followed by a 30mhz Mjl1301. Sziklais are tedious and definately need well chosen parts to avoid the capacitor band aids.
Colin
Colin
East,
Looking at the schematic the 100pf miller cap should be more than enough for compensation, in fact it may be possible to get away with less which improves the sound with some due care and attention. I should add that it gets more complex due to the sziklai outputs having their own feedback loop within a global feedback loop, and i had found what may work fine on its own turns into an oscillator within the global loop. Even if it doesnt oscillate outright in a static condition, a dynamic condition may trigger it especially if it is close to the edge once the global feedback is applied. As long as the outputs are the slower device with some decent internal capacitance it will aid stability, ideally this pole should be higher than the vas dominant pole. Im a huge fan of Sziklais in power supplies where larger currents are needed +25ma.
Colin
Looking at the schematic the 100pf miller cap should be more than enough for compensation, in fact it may be possible to get away with less which improves the sound with some due care and attention. I should add that it gets more complex due to the sziklai outputs having their own feedback loop within a global feedback loop, and i had found what may work fine on its own turns into an oscillator within the global loop. Even if it doesnt oscillate outright in a static condition, a dynamic condition may trigger it especially if it is close to the edge once the global feedback is applied. As long as the outputs are the slower device with some decent internal capacitance it will aid stability, ideally this pole should be higher than the vas dominant pole. Im a huge fan of Sziklais in power supplies where larger currents are needed +25ma.
Colin
Member
Joined 2009
Paid Member
But you do agree that simulations can be handy for the initial engineering, before we build and listen ?
Hi Colin,
I must admit I was skeptical about using a CFP in the output. I've used them in an LTP, as drivers and will try one in a VAS in a JLH design. If it wasn't for all the good things that I've read about the P3a I might never have tried a CFP output !
Last edited:
I think there are some are some assumptions about 2 pair CFP output stage amps that are a bit too generalised. LJM's L12-2 design works absolutely fine, is rock stable in 4 kit builds I have seen and listened to in the last year. Look at the LAPTs (those I saw were genuine) and other arrangements that normally lead to instability. Yep, this CFP Sziklai does sound great.
I think there are some are some assumptions about 2 pair CFP output stage amps that are a bit too generalised. LJM's L12-2 design works absolutely fine, is rock stable in 4 kit builds I have seen and listened to in the last year. Look at the LAPTs (those I saw were genuine) and other arrangements that normally lead to instability. Yep, this CFP Sziklai does sound great.
Ian, what are the OP devices used here?
Cheers,
Valery
Hi Valery
They are Sanken 2SC2837 & 2SA1186 which are no doubt the same die as some larger MT200 case version.
L12 2 Assembled Stereo Audio Power Amplifier Board 2pcs DIY KIT 2 0 Channel AB | eBay
http://www.diyaudio.com/forums/solid-state/196089-l12-2-cfp-output-amp-120w-2-8r.html
They are Sanken 2SC2837 & 2SA1186 which are no doubt the same die as some larger MT200 case version.
L12 2 Assembled Stereo Audio Power Amplifier Board 2pcs DIY KIT 2 0 Channel AB | eBay
http://www.diyaudio.com/forums/solid-state/196089-l12-2-cfp-output-amp-120w-2-8r.html
Thank you Ian. Those ones are the good ones. Though, if you mean by the larger MT200 ones - C2922 and A1216, those are twice more powerful (17A, 200W maximum).
Anyway, they are fast, stable, reliable.
Prices on e-bay make me a bit suspicious 36 USD for 2 assembled boards... are they giving them away? no fakes? What do you think?
I would buy just for testing purposes.
Cheers,
Valery
Anyway, they are fast, stable, reliable.
Prices on e-bay make me a bit suspicious
36 USD for 2 assembled boards... are they giving them away? no fakes? What do you think?I would buy just for testing purposes.
Cheers,
Valery
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.