Hi Jocko.
I think that you mostly speak about the ouput CFP or Sziklai pair. For the moment, i am not working on the output stage.
I must say that my first schematic was not clear on this point.
But i am very interested by all your advises !
I am trying input stage and VAS, in Microcap, in bench, and soon i will listen...
I would like to test a few thoughts, i have in mind.
http://www.diyaudio.com/forums/showthread.php?postid=422778#post422778
An idea of the design i am working on...
I think that you mostly speak about the ouput CFP or Sziklai pair. For the moment, i am not working on the output stage.
I must say that my first schematic was not clear on this point.
But i am very interested by all your advises !
I am trying input stage and VAS, in Microcap, in bench, and soon i will listen...
I would like to test a few thoughts, i have in mind.
http://www.diyaudio.com/forums/showthread.php?postid=422778#post422778
An idea of the design i am working on...
An externally hosted image should be here but it was not working when we last tested it.
"What is the purpose of the "CFP" in the input stage????? Probably the last place that I would use one."
So I would have thought, but I now find both D. Self and Ben Duncan refer to them in differential amps at the input. Where you normally have a differential pair you use to CF pairs (more likely called Sziklai pairs in this context) to provide local fedback.
I moded an existing LT Spice circuit to insert these to see what happens. "THD+N" (mostly the "N", I think) is slightly reduced. We're talking a change of ~.0001%! But in real life the transistors are neither ideal or perfectly matched to the real benefit may be greater. No mention is made of potential foroscillation. In the Duncan diagrams it is indicated that all four should be thermally coupled which sounds like an entertaining endeavor for people who like to fiddle with small objects! I would want to be thoroghly convinced I had a problem before applying this fix.
So I would have thought, but I now find both D. Self and Ben Duncan refer to them in differential amps at the input. Where you normally have a differential pair you use to CF pairs (more likely called Sziklai pairs in this context) to provide local fedback.
I moded an existing LT Spice circuit to insert these to see what happens. "THD+N" (mostly the "N", I think) is slightly reduced. We're talking a change of ~.0001%! But in real life the transistors are neither ideal or perfectly matched to the real benefit may be greater. No mention is made of potential foroscillation. In the Duncan diagrams it is indicated that all four should be thermally coupled which sounds like an entertaining endeavor for people who like to fiddle with small objects! I would want to be thoroghly convinced I had a problem before applying this fix.
Re: Tricky means......
Sure OSCILLATIONS problems !!!! But i progress, slowly but i progress.
O.K. for the input stage YEAAAAHHHH !!!!!
Just a question:
What is an audio power amplifier else, than a modulated power supply ? (Voltage source from 0Hz to 100khz or more).
Well I speak about CFP's in class A not in AB or even worse B, not only due to crossover THD, but in class A you have less changes in the transistor characteristics, and i think it's very important for our purpose. Today I think in a CFP output stage, but maybe i will change my mind, I need tests. The diamond buffer looks very interesting too...
For the moment :
JFET + BJT / CFP = marvelous FFT plot (fine spectra), nice looking square wave, no input capacitor, no CCS ! What moreover ! .... LISTEN maybe ! But I have to solve a few problems with VAS, before ...
Jocko Homo said:
Sure OSCILLATIONS problems !!!! But i progress, slowly but i progress.
O.K. for the input stage YEAAAAHHHH !!!!!
Just a question:
What is an audio power amplifier else, than a modulated power supply ? (Voltage source from 0Hz to 100khz or more).
Well I speak about CFP's in class A not in AB or even worse B, not only due to crossover THD, but in class A you have less changes in the transistor characteristics, and i think it's very important for our purpose. Today I think in a CFP output stage, but maybe i will change my mind, I need tests. The diamond buffer looks very interesting too...
For the moment :
JFET + BJT / CFP = marvelous FFT plot (fine spectra), nice looking square wave, no input capacitor, no CCS ! What moreover ! .... LISTEN maybe ! But I have to solve a few problems with VAS, before ...
Well, after a few small changes in the design… Lots of problems with small H.F. oscillations, about 4/5Mhz, so Miller cap adjust, base resistors, cascode configuration, and very, very important, good ground, and decoupling power supply rails. (I must say, that my stabilized power supply is not very good).
Now the prototype works fine, haem… in test bench !
I just need to buy... batteries… In old listening tests, for my “active” crossover, I notice that my stabilized power supply sounds very bad ! And this time, I want to get rid of it.
Below, some FFT graphs. I can’t show oscilloscope traces of square wave signal, I have nothing for capture or photograph the screen, but there is quasi no ringing. Band-pass is 2 MHz and the gain is 20 dB. At 0dB input, VAS output is about 20V peak to peak.
T.H.D. graphs.
Input -20dB
Input -10dB
Input 0dB
I.M.D. graphs.
Input -20dB
Input -10dB
Input 0dB
Now the prototype works fine, haem… in test bench !
I just need to buy... batteries… In old listening tests, for my “active” crossover, I notice that my stabilized power supply sounds very bad ! And this time, I want to get rid of it.
Below, some FFT graphs. I can’t show oscilloscope traces of square wave signal, I have nothing for capture or photograph the screen, but there is quasi no ringing. Band-pass is 2 MHz and the gain is 20 dB. At 0dB input, VAS output is about 20V peak to peak.
T.H.D. graphs.
Input -20dB
Input -10dB
Input 0dB
I.M.D. graphs.
Input -20dB
Input -10dB
Input 0dB
Re: DON'T SAY THAT!!!!!
Gedday Jocko,
How is sunny Texas,
WRT CFP in PS, I totally agree they have a good balance
between BW, settling and OP Z, and I used them for years.
Funny though, I gave a few customers a "short out" option
for the CFP tranny, ie; in an N channel CFP follower, short the P
channel E - B and turn it into a plain vanilla single tranny EF.
They all preferred the plain vanilla version.
This was in DAC's where very wide BW would be important,
nevertheless it was an interesting experiment. I was actually
biased toward the CFP, probably placaebo because I knew
it had lower OP Z.
I would also like to note you can do a three stage CFP
and get even more "OLG" for lower OP Z and even more
stability problems
Cheers,
Terry from sunny Sydney.
Jocko Homo said:
Gedday Jocko,
How is sunny Texas,
WRT CFP in PS, I totally agree they have a good balance
between BW, settling and OP Z, and I used them for years.
Funny though, I gave a few customers a "short out" option
for the CFP tranny, ie; in an N channel CFP follower, short the P
channel E - B and turn it into a plain vanilla single tranny EF.
They all preferred the plain vanilla version.
This was in DAC's where very wide BW would be important,
nevertheless it was an interesting experiment. I was actually
biased toward the CFP, probably placaebo because I knew
it had lower OP Z.
I would also like to note you can do a three stage CFP
and get even more "OLG" for lower OP Z and even more
stability problems
Cheers,
Terry from sunny Sydney.
Hi Justcallmedad,
I have successfully used CFP (with cascodes) in amps without oscillation problems in both input stage and VAS stage. I was "inspired" by Gerard Perrot patent # 5 635 874 and peufreu web site
http://peufeu.free.fr/audio/memory/
about Memory distortion and also Halcro patent #5 892 398.
I have found that the use of so-called low memory circuit arrangement (in replacement of single transistor usually used) provide a more define and detailed music (for the same circuit topology).
In the future I am considering extending this principle also to the output stage like one schematics ( using CFP) I found on the web at http://www.ne.jp/asahi/evo/amp/J49K134/zu3.pdf
Keep us posted on your results.
Fab
I have successfully used CFP (with cascodes) in amps without oscillation problems in both input stage and VAS stage. I was "inspired" by Gerard Perrot patent # 5 635 874 and peufreu web site
http://peufeu.free.fr/audio/memory/
about Memory distortion and also Halcro patent #5 892 398.
I have found that the use of so-called low memory circuit arrangement (in replacement of single transistor usually used) provide a more define and detailed music (for the same circuit topology).
In the future I am considering extending this principle also to the output stage like one schematics ( using CFP) I found on the web at http://www.ne.jp/asahi/evo/amp/J49K134/zu3.pdf
Keep us posted on your results.
Fab
Hi fab.
It's exactly the way i am working on, I don't use CFP in the classical way "inverted darlington", that's sure.
I am trying configurations that improve BJT's caracteristics for audio use.
Gerard Perrot was a friend of mine, his other name was Hephaistos, he was the founder of "Lavardin Technologies",
he died 5 years ago.
I was beginning audio design hobby since a few months, when i met him in a writers meeting for "La revue de L'Audiophile",
a french Hi-Fi revue.
I begin amplifier design, with him, his research results, and all that I learned from him.
Since i learned a lot more from D. Self, B. Duncan, W. Jung, M. Leach, Jon Marsh, D. Reich, Charles Hansen, John Curl, Hugh Dean, Nelson Pass, Erno Borbely, for the most "popular", and all the people here, like you, less commonly known, but quite as qualified.
Now I am trying my own thoughts, I am doing some "little" research, and very important: a lot of listening tests.
I have not yet, sufficient electronic background to explain all, or understand all, but sometimes and paticularly in our field,
this fact is rather interesting to leave the current paths.
Somme common electronic rules are not always good for audio use : CCS, current mirrors, feedback... etc...
Mathematically they are often right, but their implementation for audio seems to bring drawbacks...
So, we need you, we need you to better understand and explain the facts that we have note.
Thank you to be here !
Well, with regard to my project, headphone listening is really very, very good, tubelike but with best highs, but i am not very familiar with this form of listening, it has just help me to tune the circuit, when my test bench don't see the difference between two optionsresistor vs ccs with bjt or fet), or when, between two drawbacks, wich one choose ?
Following the good results i reach, I have decided to go further and implement the output stage, i must yet buy batteries,
big batteries... heatsinks etc...
And... listen in place of my usual components : 300B (without feedback) for medium/high TAD horn loaded, Kaneda amplifier for bass with VOT.
Question : in a listening way wich drivers sound best :Toshiba ?
ON-semi MJ 15032/33 ? If someone has tested them...
I have a lot of 2SC3281/2SA1302 so i think in 2SC4793/2SA1837
high gain, high bandwith... I will use them at 150 mA even more.
It's exactly the way i am working on, I don't use CFP in the classical way "inverted darlington", that's sure.
I am trying configurations that improve BJT's caracteristics for audio use.
Gerard Perrot was a friend of mine, his other name was Hephaistos, he was the founder of "Lavardin Technologies",
he died 5 years ago.
I was beginning audio design hobby since a few months, when i met him in a writers meeting for "La revue de L'Audiophile",
a french Hi-Fi revue.
I begin amplifier design, with him, his research results, and all that I learned from him.
Since i learned a lot more from D. Self, B. Duncan, W. Jung, M. Leach, Jon Marsh, D. Reich, Charles Hansen, John Curl, Hugh Dean, Nelson Pass, Erno Borbely, for the most "popular", and all the people here, like you, less commonly known, but quite as qualified.
Now I am trying my own thoughts, I am doing some "little" research, and very important: a lot of listening tests.
I have not yet, sufficient electronic background to explain all, or understand all, but sometimes and paticularly in our field,
this fact is rather interesting to leave the current paths.
Somme common electronic rules are not always good for audio use : CCS, current mirrors, feedback... etc...
Mathematically they are often right, but their implementation for audio seems to bring drawbacks...
So, we need you, we need you to better understand and explain the facts that we have note.
Thank you to be here !
Well, with regard to my project, headphone listening is really very, very good, tubelike but with best highs, but i am not very familiar with this form of listening, it has just help me to tune the circuit, when my test bench don't see the difference between two options
resistor vs ccs with bjt or fet), or when, between two drawbacks, wich one choose ? Following the good results i reach, I have decided to go further and implement the output stage, i must yet buy batteries,
big batteries... heatsinks etc...
And... listen in place of my usual components : 300B (without feedback) for medium/high TAD horn loaded, Kaneda amplifier for bass with VOT.
Question : in a listening way wich drivers sound best :Toshiba ?
ON-semi MJ 15032/33 ? If someone has tested them...
I have a lot of 2SC3281/2SA1302 so i think in 2SC4793/2SA1837
high gain, high bandwith... I will use them at 150 mA even more.
fab said:
Justcallmedad,
Having a deeper look at it, the connection points wer not clear on the printout I have made of the above circuit and finally it is not a CFP output stage but a normal inverting output stage with an inverting driver stage. Sorry if I mislead you.
Regarding the driver transistors I can not help you but I am also looking to replace my 2N5415/2N3440 (to-39) with other types with to-220 package for which heatsinks are more available.
If you plan to have 150ma for Q5 and Q105 (if you are not talking about a next driver stage) then you will need a big heatsink and you will probably also need a darlington in lieu of Q3 and Q103 if you do not want to load too much the input stage.
Salut
Fab
Hi fab.
For the drivers, i talked about the next stage, never mind, I will test both types Toshiba/On-Semi.
Concerning the pair 2N5416/2N3440, they are quite good devices, but like you said low power handling. Maybe the best thing, is to cascode them with more recent drivers, or even with mosfet 2SJ76/2SK216, it will enhance their linearity and the power handling will be supported by these devices. If no cascode, as you need high Hfe, I tried in others designs, the pair 2SA1306/2SC3298 with very good results.
Without an idea of your schematic, i can't help you more precisely. Something like Zenquito ? it's the pair usually used for it.
Salut.
For the drivers, i talked about the next stage, never mind, I will test both types Toshiba/On-Semi.
Concerning the pair 2N5416/2N3440, they are quite good devices, but like you said low power handling. Maybe the best thing, is to cascode them with more recent drivers, or even with mosfet 2SJ76/2SK216, it will enhance their linearity and the power handling will be supported by these devices. If no cascode, as you need high Hfe, I tried in others designs, the pair 2SA1306/2SC3298 with very good results.
Without an idea of your schematic, i can't help you more precisely. Something like Zenquito ? it's the pair usually used for it.
Salut.
thanh said:
Hi Than
What input stage circuit are you using (bjt, jjet, differential, symetric complementary diff., characteristics of transistors,...) ?
Anyway, the CFP have already been used in several amps design without problems and tested at frequencies a lot higher than 20 KHz. In an amp design, there may be other things than CFP that creates oscillations you know. For the VAS, the CFP can be more difficult to use but for input stage it is not a big challenge I believe.
I do not want to imply that simulation are not good but try to really build one CFP and see.
By the way, it is CFP and not CFB for clarification purpose.
Fab
Justcallmedad said:
Hi Justcallmedad
In fact I used small power bjt in the VAS except for the cascode transistor where I use the 2N5416/2N3440 for dissipated power consideration. What I am looking for is a BJT pair with TO-220 package with similar (or better): Hfe, VCE max, BW. And max power of about 10W. I mean like a close "direct" replacement. The schematics is not important. Are the 2SA1306/2SC3298 BJT? If so, I would need to verify if the 2SA1306/2SC3298 are easily available in Canada.
Also, if you use a next driver stage at 150ma, I suppose that you have power BJT for output devices and not mosfets?
Thanks
Fab
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.