JAS - Just Another Sziklai

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
OK,
As I have said in some other thread, here is a DIY project that I am working on.
It does not want to be nothing very special, but just something that we can play with and hopefully sounds well!
It does not want to be a production item so it might be necessary to tweak it to reach maximun performances (i.e. THD) and stability, but that is the juce of this game, for me.
The schematic is pretty standard and resembles many around: OS´s BX Mongrel, P3, Baby AKSA, etc.
The output is CFP (thanks, Mr Sziklai!) and can be arranged in two ways: a "pure" CFP with onlly one pair and a "modified" version suitable for two or more pairs. I see a lack of multiple pair CFPs around, so this is my proposal.
I have arranged a PCB that can support both the versions and lets space for some tweaking.
I am posting first the multiple pair version (-200) that is the more interesting, I believe.
The four output devices should be mounted on a separated Heat Sink, and their emitter resistor simply wired. (Nothing forbids you to make an add-on PCB)
Just think that ALL the four OP devices can be put directly on the SAME Heat sink WITHOUT insulators, provided that the heatsink itself is insulated from the chassis.

Driver devices AND VBE device MUST share another, small heat sink ( a small alluminium bar will suffice). Also the VAS needs a small heatsink, alone.
Hope that somebody is intersted. Comments, suggestions, questions are highly appreciated.
 

Attachments

  • JAS200_1b.pdf
    22.8 KB · Views: 911
Administrator
Joined 2007
Paid Member
It looks fine... hard to comment without building and listening.

Only thing that comes to mind is R10 and R27. Is there enough current drive for the drivers and outputs ? CCS maybe.

The IN5400's used as clamps are only 50 volt devices ! They will see twice that under full drive. Use IN5404's or 5408's
 
It looks fine... hard to comment without building and listening.

Only thing that comes to mind is R10 and R27. Is there enough current drive for the drivers and outputs ? CCS maybe.

The IN5400's used as clamps are only 50 volt devices ! They will see twice that under full drive. Use IN5404's or 5408's

True, for the diodes, I just used the wrong generic symbol. I meant to use a 3A -600V Fast device.
R11 - R12 should allow about 400mA that should be enough for the bases, considering a total OP hfe of 30 at least.
Don't see the problem for R27.
Thanks a lot for the suggestions.
 
MJE15034/35 are good gain.
So, I think it will work with the bootstrap, 2k7 and 3k9

I should use a normal RED LED for CCS.
It will follow the transistor Q4 well at different temperatures.
There is a reason many designers use RED LED.
But this is a rare occation with BLUE :)
 
Administrator
Joined 2007
Paid Member
It just looked a little on the low side for VAS current thats all... I know its bootstrapped... just thinking worst case, low impedance load, high output swing, high current at high frequency and so on. As I say it was just something that struck me.

See how it performs :)
 
MJE15034/35 are good gain.
So, I think it will work with the bootstrap, 2k7 and 3k9

I should use a normal RED LED for CCS.
It will follow the transistor Q4 well at different temperatures.
There is a reason many designers use RED LED.
But this is a rare occation with BLUE :)
In my opinion Blue LED has some advantage. I has a l bigger bias voltage that helps the transistors (in case you use not very high Vce models like BC546) and lets you use a bigger Re for the same current, that makes the CCS more stable. You could use a WHITE as well.(ab. 3.9 V). I believe the main reason that they are not so common is that they did NOT exist in the past. Blue and UV leds are GaN devices, a more recent technology than the GaAs used for Red , IR, Yellow and Green, thanks to the wonderful work of Mr Nakamura. Of course if you would have to use the same reference for the VAS CCS, the bias voltage might be too big, but it is not the case since we have a simple bootstrap here.
Thanks for the interesting comments.
 
In my opinion Blue LED has some advantage. I has a l bigger bias voltage that helps the transistors (in case you use not very high Vce models like BC546) and lets you use a bigger Re for the same current, that makes the CCS more stable. You could use a WHITE as well.(ab. 3.9 V). I believe the main reason that they are not so common is that they did NOT exist in the past. Blue and UV leds are GaN devices, a more recent technology than the GaAs used for Red , IR, Yellow and Green, thanks to the wonderful work of Mr Nakamura. Of course if you would have to use the same reference for the VAS CCS, the bias voltage might be too big, but it is not the case since we have a simple bootstrap here.
Thanks for the interesting comments.

2/3 reds = blue w/ better delta (temperature) characteristics. In my next hawksford cascode I will use 3 reds per rail to set my base voltages. SUPERIOR. Best red led is GaAs/1.65-1.8Vf.
PS - on a LIN amp cascode , I got 10V by chaining 6 SMD LED's together. Final strip was 3 X10mm w/ leads (like a small transistor).
OS
 
Last edited:
ATAUDIO,

You can, but it is not. You have to build the circuit to see if it`s stable. I doubt it.
I WILL buid, by sure, I have all the components ready and made the PCB layout. About stability, the single output version is pretty standard. On what grounds you base the instability assertion? Just because its is a CFP? Can you elaborate a little more?
I hope you have some intersting infos to give me. Thanks

P.S.
OS, you are probably right about 2/3 x red LEDs. Nice Idea indeed.
I am just trying to keep the active components count small, and I guess the blue LED is a good compromise.
Thanks
 
The CFP can have stability problems, due to the high amount of local feedback. However, when it works it does a good job. It needs much less quiescent current than a Darlington - see Doug Self's book (or my articles (Better Buffers) in Electronics World - unfortunately not available on the web).
I fully agree. I have readen the D. Self book and the conclusion is that CFP has defenetly LESS cross-over distorsion than standard EF. He gives scientific proof of that, not based on "souns better" or "Happiness"
I agree also that you CAN have stability problems but:
1) This is a DIY project, not a production item, so the hobbist can carefully tune it (i.e. compensation) to make it stable. Personally I believe that it is EASY with this SIMPLE circuit and that EF is BORING.

BTW, did anybody notice TMC applied to this Boostrapped VAS? I did not see many around and I believe it is worth the try. I have readen on the Bob Cordell´s thread about it that thre are no sound arguments against it on such topology.

2)Any good amplifier that looks for GOOD PERFORMANCE has to be carefully tested for stability, regardless of the OP topology.
Thanks for your comments.
 
Last edited:
I assume WuYit is a FET fan, and is concerned about carrier storage in BJT?

My guess is that most of any ugly crossover distortion from CFP would come from setting the quiescent current too high, so getting gm-doubling in the central region. Correct CFP quiescent current is counter-intuitively low, especially if your previous amp used Darlingtons. CFP is 5-20mA, Darlington is 100-200mA? Set a CFP too high and it will sound bad.
 
the "1st watt" principle of heavy output stage bias is quite popular with "name" amp designers here

few aside from Self seems to push "Optimum Bias" - which appears too subject to thermal variation for the advantage to be robust with dynamic program material into complex Z loads

high feedback has less problem with relatively smooth gm doubling compared to dynamic dead zone

playing with the http://www.diyaudio.com/forums/solid-state/188293-load-invariant-ltspice-ringing.html sim I see more improvement from higher bias, particularly of the input Q of the CFP
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.