350W linear regulator looks very poor?

[cupoftea]

Hi,
Today I tested a 350W offline linear regulator (Mains transformer/rectifier/capacitor/regulator).
It did appear a little “drifty” and the current limit seemed to limit at a different (higher) current level when the unit was hotter.

It was for 69.5V and 5A output.

I had no BOM for it, just a dodgy schematic which had no component values, also the schem had “crossing” traces with no way of knowing if the track cross points were joined or not. (no “blobs” at the track junctions to show connection). I re-drew the schematic as in the attached, but cant be sure about the exact trace connections shown.

I believe the attached is very poor. Surely its way too susceptible to temperature change issues with the BJTs and zeners? As such, will try to find a better way to do it.
Do you know why they did it like the attached? Is it a recognised configuration?

(sorry if attached schem is not super accurate due to the above mentioned “dodgy” schematic.)
I wasn’t permitted to reverse engineer it. Its all thru-hole, and should always stay like that.
The 15v and 80v are from separate secondaries on the mains transformer.

--- Updated Jan 28, 2022 ---

Woudl you agree that the attached is a far better way to do this...no trim pots needed at production time, no dependence on semiconductor temperature.......

Why didnt they do it liek this in the first place?

(LTspice simulation and PDF schem attached)

 

[cupoftea]

Hi,
Linear regulator 85V to 69.5V at 350W.
You are going to use parallel BJTs
Why would you ever use an NPN instead of a NPN_Darlington for this?

Eg why MJ15022 instead of the Darlington MJ11022


MJ15022 (NPN)

https://www.mouser.co.uk/datasheet/2/308/1/MJ15022_D-2316114.pdf

MJ11022 (NPN Darlington)

https://www.onsemi.com/pdf/datasheet/mj11021-d.pdf

 

[KlausST]

Hi,

Without looking at the datasheets I guess it´s a speed problem ... and less speed in a regulator means less stability, higher overshot, longer ringing time, maybe even oscillation.

You may need paralleling of the transitors because of the huge power dissipation.

Klaus

 

[cupoftea]

Without looking at the datasheets I guess it´s a speed problem ... and less speed in a regulator means less stability, higher overshot, longer ringing time, maybe even oscillation.

Thanks, yes, but the straight NPN needs another NPN to be added to it anyway, to make it a Darlington anyway, in order to allow it to pass the current....do you mean the "discreet Darlington", can be made slower, so that it has less chance of oscillation, but the "inbuilt" Darlingtons are always way to fast, and likely to go unstable?
....though in either case, couldnt the feednback loop compensation components simply be adjusted to take account of the increased Darlington gain?

(Thanks I agree they will need paralleling)

 

[KlausST]

Hi,

slower does not mean more stable.
Faster usually means more stable.

See it this way:
Drive a car ... and there is a 5 seconds delay from steering wheel to the front wheels.
Are you more precise?
The car is too much left so you try to move right, car does not react, you even more try to move right, car still does not react. you even more....... now car moves right ... and right ... and right... too much right .. and too much left..

Klaus

 

[cupoftea]

Thanks, yes, i explained wrong, really i meant to say that the compensation gain and phase could account for the high gain of the Darlington....and so it would be OK to use the "inbuilt Darlington"?....these Darlingtons are always used , i believe, in amplifiers or regulators, so it is OK?

 

[dick_freebird]

Darlingtons which do not expose the lower base will have very asymmetric switching, and if you saturate then turnoff time blows way out. A discrete lash-up would let you tune the balance between on and off delays.

--- Updated Jan 30, 2022 ---

Darlingtons which do not expose the lower base will have very asymmetric switching, and if you saturate then turnoff time blows way out. A discrete lash-up would let you tune the balance between on and off delays.

And to that point, letting the Darlington get hot will aggravate all that. So be sure you keep to the intended cooling setup and maybe put a thermocouple on the heat sink base.

 

[cupoftea]

Thanks, it sounds liek its best to make your own Darlington so that the lower base can be exposed.....so this means, eg, having a resistor from the lower base to its emitter?...

 

[dick_freebird]

That, maybe a Schottky if you like (and can find one with a proper Vf), squaring circuit or whatever you find gives full-envelope performance / consistency?

 

[cupoftea]

Thanks, in the attached schem (which is actually from the top post), why do they use Q3 & Q4 to drive Q6, which in turn drives Q7?......please ignore the part numbers there.........in the real circuit, Q7 is an MG15022G NPN.

In other words, why so many drive transistors?

I dont know what part nos are the other xtors.

BTW, actual product features nine Q7's in parallel.

MJ15022G datasheet

https://www.mouser.co.uk/datasheet/2/308/1/MJ15022_D-2316114.pdf

BTW, this is a real product, in use in industry

 

[dick_freebird]

To me this looks sort of like a "4-high Darlington" but imagine that
the final stage, is set up (by its resistor values) to act as a current booster
at high load, but let the second-from-bottom device get up on beta peak
earlier and hand off to the big rack a decade later?

At least each base has a discharge resistor, to keep things from getting
-too- soggy. As emitter followers (in that section) at least you don't
have to worry about saturation (unless R17 is too high for the combined
collector currents of Q3 & Q4, and then when railed high @ Q3.B you
might (but this would be, I expect, designed outside of spec range).

 

[crutschow]

You can save about 0.6V forward drop if you use a Sziklai Pair (sometimes called a complementary Darlington) in place of a Darlington, if you don't have an available base voltage source higher than the collector voltage.

Ltspice simulation below shows the ON voltage for both configurations with the same current.

 

[cupoftea]

Thanks
Thanks, do you agree that however POT1 is adjusted to give the current limit level....when Q7 gets hotter, that current limit level will rise?