Slightly evolved P3A simulation

[kalmara]

Hello, can someone tell me how good is this schematic. Are there any willing folks, that are good with the simulating programs to make me a virtual test.
Reason being - I have two pcb's almost assembled for a hearing test, but I'm not sure if the constant current source for the VAS is correctly calculated. And how low can the miller cap be reduced to improve the sonics, but without a risk of oscillation. Also - why did uncle Rod put a miller cap in the negative rail driver ? And can we reduce/remove that also?
This is the schematic.


p.s. When simulating /if someone ever does/ it would be good to put some test voltages in appropriate spots of the schematic for troubleshooting. Just in case.

[danfo098]

I did some quick simulations. If the voltage drop over the LED is 2.5 V then the VAS CCS quiescent current is 18.3 mA and the LTP CCS quiescent current is 3.39 mA.

When VR1 is set to 2k2 the output quiescent current is ~700 mA!
When VR1 is set to 0R1 the output quiescent current is ~24.5 A!

I'm attaching the LT Spice .asc schematic file and an image of it.
I might do some more simulations tomorrow to find better values for
R16, VR1 and R9 to lower the output quiescent current which is way to high now.

[kalmara]

Actually, I forgot to mention it in my circuit drawing program - the led is a diffused green one, advertised forward drop is 1.7V.
From that point on, I think both quiescent currents of the CCS's will be lower, and the output quiescent also.
I did the theory /circuit/ of this hybrid a while ago /year or so/, but still didn't get a chance to test it /in reality, heck even in virtuality/.
If I remember correctly with the diffused green led my calculations about the VAS were about 7mA or so, and about 1-2mA for the LTP.

Overall ... is it holding on? No strange things happening? The main point of these gymnastics was to add a little more linearity to the input and VAS stages and minimize the DC offset. And still remain the fair simplicity of the cult called P3A.

[MJL21193]

Hi K,
Maybe give the CC sources their own bias - I tried this before with poor results.
With the current mirror added you should consider emitter resistors in the LTP.
The vbe multiplier might need some work (as mentioned already).

[danfo098]

1.7 V drop over the LED gives the following from the sim:

LTP CCS: 1.88 mA
VAS CCS: 11.27 mA
OPS Iq @ VR1=0R1: 15.1 A
OPS Iq @ VR1=2k2: 628 mA

Still a bit high.

[jaycee]

Put a 1K resistor in the base of Q12, or give it its own reference, otherwise during clipping Q12 will pull the reference voltage down and affect the current source feeding the LTP. Also degenrate the LTP with 100 ohm resistors.

Personally I gave up when the CFP pair annoyed me too much.

[kalmara]

Today I was going trough the boxes that are on top of my wardrobe, and I found my test boards ... assembled!
My brain sucks BIG time! LOL I don't seem to recall when did I finish them, so anyhow...modifications are down to the bare minimum, needed not to fry the circuit. The linearity will be in the next revision of the pcb, beacause space is something I don't have, and everything is very tight. I'll post up an updated schematic later with your suggestions.

jaycee ... what's wrong with the CFP ? In my opinion it's the best thing - practically indestructible. I was running the original P3A with +/-42V rails and 4ohm load on a pretty small heatsink, when my fan stalled and I didn't even notice, for TWO hours was I pumping the **** out of that amp, and it didn't fail, without any thermal tracking for the bias. When I plugged it off, my skin could easily stick on the heatsink if I touched it.
p.s. And my output devices weren't the best things - 2SC3281/2SA1302, but from the turkish-chinise type ***** with a 1.5x1.5mm die, beacause toshiba stopped making them before a long time ago.

[kalmara]

My internet at home is dead right now, so I can't post any schematics. I have a question, thou - what will be the benefits /if there are any at all/ if we change the LTP with fets, like 2SK301 or so? Also I'm thinking of changing the CCS for the LTP with a dual-transistor type. /not sure how is the term in english/

And one last thing - my boards are assembled, and as I said - space is tight, my question is how can we correct the schematic to make it run as it is, and adjust like 60-100mA bias for the outputs, WITHOUT touching the trimmer - only R9/R16? The main goal is just to make them run, without frying anything, then I'll give these boards to a friend of mine, and we can start the evaluation of the circuit and board. When we're done with that, the final revision of the boards will be contribute to my ears.

p.s. What will happen, if I power the circuit without the LTP emitter resistors, bare in mind that Q1-Q2 and Q10-Q11 are very closely beta-matched. I chose them myself.

[Lumba Ogir]

kalmara,

we change the LTP with fets, like 2SK301 or so? Also I'm thinking of changing the CCS for the LTP with a dual-transistor type.

Very good idea. You should do that at once. Don´t use source resistors.

[Lumba Ogir]

kalmara;
sorry, clarification is needed .

2SK301 will be fine with source resistors.

The output CFP could be problematic. Look out for oscillation. I would use MOSFETs for Q7/Q8.