Ultima Thule said:
Ok calm down now boys! 😎
How about my suggestion that seem to have passed unmarked(?) in favour of "less adult" behaviour here! 🙂
Cheers Michael
PS: Carlos, I have noticed that you have been quite upset for last 2 months or so, lot have happend and GM left us, I feel it's easily boiling under the surface of you and it blast too easily because of your inside pressure from past frustrating months, it's for me absolutely understanable... now, take a fiesta and tomorrow everything is ok, Pavel have made a great job and also discovered symasym have some issues with Toshiba transistor BUT working on it making symasym even more versatile for DIY:ers giving them a broader choice of output transistors, good designs should be able to work with many different compatible transistors making it appealing to even more DIY:ers me thinks! 🙂
Re: Re: Order PCB and construction units yesterday
I orderd MJL 3281A / MJL 1302A and i wrote to the order it must be ONS types. i'll wating.
Please see the order:
MikeB said:
I orderd MJL 3281A / MJL 1302A and i wrote to the order it must be ONS types. i'll wating.
Please see the order:
Hi Carlos,
This would not be the first amplifier that oscillated with replacement output transistors. Many commercial amps do that, even some that use STK power paks. I feel the design is maturing.
The designer of an amplifier will normally not test the design with different numbers of parts, just the target parts. I feel PMA is working very hard to make this a better amplifer with the variety of different parts out there. My hat is off to every one working on this. I still can not get the traces right drawing them by hand. The lines are too thick with my method. Still trying though.
-Chris
This would not be the first amplifier that oscillated with replacement output transistors. Many commercial amps do that, even some that use STK power paks. I feel the design is maturing.
The designer of an amplifier will normally not test the design with different numbers of parts, just the target parts. I feel PMA is working very hard to make this a better amplifer with the variety of different parts out there. My hat is off to every one working on this. I still can not get the traces right drawing them by hand. The lines are too thick with my method. Still trying though.
-Chris
I was thinking how to get out, as there is much work done. Finally I have chosen the Ignore List. Not very gladly, but better than shut the door. I am back for technical aid.
Dimitris,
I tried to increase base stoppers from 1.2R to 10R. Unfortunately, it is the wrong way. High DC offset, higher noise and same stability.
Dimitris,
I tried to increase base stoppers from 1.2R to 10R. Unfortunately, it is the wrong way. High DC offset, higher noise and same stability.
Hi,
although I've not followed this tread from the beginning, but after building tons of oscillating amps and trying to cure it, why not connect the 22p feedback cap to the output of the VAS stage? This trick helped me on a lot of amps.
dx.master,
who is jo-ton.de? I found a JO-TON Veranstaltungsservice but no PCB manufacturer. Have you an URL?
although I've not followed this tread from the beginning, but after building tons of oscillating amps and trying to cure it, why not connect the 22p feedback cap to the output of the VAS stage? This trick helped me on a lot of amps.
dx.master,
who is jo-ton.de? I found a JO-TON Veranstaltungsservice but no PCB manufacturer. Have you an URL?
Bocka,
you are on the same wave. 5 minutes ago I desoldered the feedback cap of 22 pF. What happened? Stable!! Of course! This cap is a small circuit design flaw. The feedback cap may be "problemloss" used only for circuits stable at 0 dB gain. It brings CL 0dB line at HF and can interesect OLG at the place of high order slope. And it is happening here.
So Michael, how to handle?
1) decrease the cap
2) put the cap in series with resistor
3) omitt the cap completely and maybe change another compensation?
you are on the same wave. 5 minutes ago I desoldered the feedback cap of 22 pF. What happened? Stable!! Of course! This cap is a small circuit design flaw. The feedback cap may be "problemloss" used only for circuits stable at 0 dB gain. It brings CL 0dB line at HF and can interesect OLG at the place of high order slope. And it is happening here.
So Michael, how to handle?
1) decrease the cap
2) put the cap in series with resistor
3) omitt the cap completely and maybe change another compensation?
Now the symasym is stable even for Iq = 1A, when runs completely in a class A (not recommended, I have large heatsinks).
Yes, as i told, this cap is problematic, but necessary for nice step response. Yesterday i lowered it to 10pf, it gave already degraded stepresponse, but still acceptable. I often had to lower it to stop oscillation (with previous designs). Pavel, have you checked squarewave without this cap ? Or stability with 10pf ?
Bocka, i already considered placing this cap to vas output, but has the disadvantage of moving the outputstage out of feedbackloop for higher freqs, resulting in massive amounts of high order harmonics. It's correct, with this cap it will never oscillate again, giving a max phasehift of 50° to feedbackpoint...
Pavel, i will investigate all 3 options, number 2 sounds promising !
I had suggested No1 already to Dimitris, No3 will not be easy, but the nicest...
I do not understand the increased DC-offset with increased basestoppers ?
Mike
Bocka, i already considered placing this cap to vas output, but has the disadvantage of moving the outputstage out of feedbackloop for higher freqs, resulting in massive amounts of high order harmonics. It's correct, with this cap it will never oscillate again, giving a max phasehift of 50° to feedbackpoint...
Pavel, i will investigate all 3 options, number 2 sounds promising !
I had suggested No1 already to Dimitris, No3 will not be easy, but the nicest...
I do not understand the increased DC-offset with increased basestoppers ?
Mike
Yup, I know. But will this also affect the sound quality? If I would have the time I'd build this amp, unfortunately I've too many projects to be realiszed...
Michael,
the step response is not ideal into dummy load (but only for several us). I believe in feedback cap in series with resistor. The resistor value allows to find less severe point than 0dB at OLG characteristics.
Pavel
correction: the step response looks quite fine. I forgot to use the "IGND" (above 10R resistor) when I wrote the sentence hereabove
the step response is not ideal into dummy load (but only for several us). I believe in feedback cap in series with resistor. The resistor value allows to find less severe point than 0dB at OLG characteristics.
Pavel
correction: the step response looks quite fine. I forgot to use the "IGND" (above 10R resistor) when I wrote the sentence hereabove
This amplifier`
I really don't see where the fuss is, I built two different versions
of this amplifier, the very first one, after a few hick ups *mistake
on the PCB and so on" it really funcioned well. Its quality was bit
not to my taste, but....
This latest version is much better and UNCONDITIONALLY stable.
Very pleased with the results and that's why I don't see where
the problems are....
I really don't see where the fuss is, I built two different versions
of this amplifier, the very first one, after a few hick ups *mistake
on the PCB and so on" it really funcioned well. Its quality was bit
not to my taste, but....
This latest version is much better and UNCONDITIONALLY stable.
Very pleased with the results and that's why I don't see where
the problems are....
jmateus,
the feedback cap can be used only under certain defined conditions. If you wanted to know more, please find Ron Mancini's papers about voltage feedback amp compensation, or review Analog Devices website.
For those lucky guys with 3281/1302 the problem did not appear (yet), for those unlucky with 5200/1941 it appeared.
I am sure that a very small change at this position will make the amp matured, device-independent and easy-for-diyers-to-built /under any conditions/ - and that's how the kit should look like.
the feedback cap can be used only under certain defined conditions. If you wanted to know more, please find Ron Mancini's papers about voltage feedback amp compensation, or review Analog Devices website.
For those lucky guys with 3281/1302 the problem did not appear (yet), for those unlucky with 5200/1941 it appeared.
I am sure that a very small change at this position will make the amp matured, device-independent and easy-for-diyers-to-built /under any conditions/ - and that's how the kit should look like.
After having the feedback cap removed, I prepared the worst conditions for the symasym, square waves from 40Hz to 16kHz at 90% of full output into dummy load, the only result is, that I burnt a portion of my dummy 😉.
The symasym has survived. No trace of problem. If everything goes like now, I will complete the 2nd channel and start the listening tests.
The symasym has survived. No trace of problem. If everything goes like now, I will complete the 2nd channel and start the listening tests.
Hi John ! It's been a long time... 
I did not know that you have tried the latest version ! The actual problem is to get symasym stable with "all" kind of output devices, in this special case the Toshibas driving symasym into oscillation.
If i remember correct, you used 20mhz Sankens ?
Pavel, i made some sims, and omitting this feedback cap looks like a very bad idea, it moves down the 180° freq to ~1mhz, which is inacceptable. Even pure openloop it has this point at ~6mhz.
This compensation has to do something with compensating some kind of reactive input at feedbackpoint.
I think you can't call this cap "design-flaw", just difficult... 😉
Mike
I did not know that you have tried the latest version ! The actual problem is to get symasym stable with "all" kind of output devices, in this special case the Toshibas driving symasym into oscillation.
If i remember correct, you used 20mhz Sankens ?
Pavel, i made some sims, and omitting this feedback cap looks like a very bad idea, it moves down the 180° freq to ~1mhz, which is inacceptable. Even pure openloop it has this point at ~6mhz.
This compensation has to do something with compensating some kind of reactive input at feedbackpoint.
I think you can't call this cap "design-flaw", just difficult... 😉
Mike
Michael,
do not believe simulation results in HF region for cases like this. The pure feedback cap must not be used for system with several poles in OLG amplitude characteristics, like symasym. It is only a game of luck if the OLG gain curve and compensated loop gain curve does not intersect at unstability point. You can see how it is dangerous right now. And the cure should be quite easy, resulting in no potential instability for the future.
If there was only one dominant pole (above 0dB) in the OLG, then the feedback cap would be no problem (this is not the cure, but the condition to use the feedback cap).
If you wanted, I could send the papers.
do not believe simulation results in HF region for cases like this. The pure feedback cap must not be used for system with several poles in OLG amplitude characteristics, like symasym. It is only a game of luck if the OLG gain curve and compensated loop gain curve does not intersect at unstability point. You can see how it is dangerous right now. And the cure should be quite easy, resulting in no potential instability for the future.
If there was only one dominant pole (above 0dB) in the OLG, then the feedback cap would be no problem (this is not the cure, but the condition to use the feedback cap).
If you wanted, I could send the papers.
feedback cap with Toshibas
Well, from my point of view, and according to my results, I can accept and recommend for Toshiba 2SC5200/2SA1943 the feedback capacitor to be 3.3pF, instead of original 22pF, that were evaluated by Michael for MJL3281/1302.
Well, from my point of view, and according to my results, I can accept and recommend for Toshiba 2SC5200/2SA1943 the feedback capacitor to be 3.3pF, instead of original 22pF, that were evaluated by Michael for MJL3281/1302.
can i use other output bjts like Sanken 2SA1695 / 2SC4468 ?
@MikeB and the World,
can i use these Sanken 2SA1695 / 2SC4468 bjts for the MJLs ?
Datasheets Sanken 2SA1695 / 2SC4468
@MikeB and the World,
can i use these Sanken 2SA1695 / 2SC4468 bjts for the MJLs ?
Datasheets Sanken 2SA1695 / 2SC4468
Generally speaking - the faster output devices, the more troubles with feedback cap type compensation, as OLG char moves to higher frequencies.
All the substitutes should be tested, but there is a good premise, that slower devices will not make problems (just a premise).
All the substitutes should be tested, but there is a good premise, that slower devices will not make problems (just a premise).
Pavel,
How about introducing phase lead in conjunction with lag comp?
Say 5pF from active VAS collector to feedback node........
I've used this approach with good results.
Cheers,
Hugh
How about introducing phase lead in conjunction with lag comp?
Say 5pF from active VAS collector to feedback node........
I've used this approach with good results.
Cheers,
Hugh