recall a classical MOSFET power supply

[digi01]

I have revised this classical mosfet AC power supply circuit again,and have making a module block now,Used for serving my series headamps.

this is the original sch scan:

[digi01]

this is my block psu,still in making...
Who can talk about the detail of this circuit? Or some suggestions?


-digi

[hjelm]

The E102 are they current diodes?

I have begun to wonder why the long tailed pair is not more often used in these kind of comparator scenarions. was actually thinking of doing something for a sine to square converter in a clock with a LTP, simulate at least.

[digi01]

This is a standard bunch mode power supply. I have done some substitution according to the device in the hand:
K213/J76-->>K214/J77,
E102-->>K246,
2SC1775/2SA872-->>B716/D756¡Á2.
it has good performance and work fine.

[CheffDeGaar]

The Kubota Reg, uh ?

Nice working reg, but it should deserve some improvements IMHO. Just have a look at this post by Jonathan Carr (and the following posts too) :

http://www.diyaudio.com/forums/showt...268#post135268

I'm currently willing to implement such regs in my future preamp, and I've made some simulations, trying to improve the output impedance and the line regulation.

For the moment, I ended up replacing the emitter's resistance of the diff pairs by CCS (just a source degerated JFet for example).
Improving the voltage reference filtering with a RC network won't certainly hurt, and adding a decoupling cap for the sensing resistor bridge helps a lot.

I've attached the schematic I've used so far, but I'm sure things can be further improved, though I don't know how...

In the schematic, R2/C3/C4 provide low pass filtering for the reference. The C2 cap showed a great influence on the reg's performance (PSRR), the higher being the better...
I've added the C5-C6 caps. I found them mandatory to further reduce the output impedance below 10mOhms.

Feel free to comment/improve/flame...

[hjelm]

Nice!

Just one question (for now) the c5 c6 they "short" the higher frequency voltage changes for the sensing thereby moving the input of the diff pair faster than without?

If so does it require some matching of the values?

[CheffDeGaar]

Yep, sort of...
AFAIK, they provide a lower impedance network, both for the sensing bridge and the whole output, to which it is connected. I didn't invent anything, it appears in many discrete regs, even the Walt Jung's one. Here is a plot of the output impedance, showing that (in theory), it improves the Zo by a 8 factor...