SMPS feedback and compensation

[iWei]

As far as I know, a TL431 is used as an adjustable zener in most SMPS for feedback, except that it being an opamp means that it has to be compensated like one, meaning that one would have to know the characteristics of all components involved and do the maths on it.

https://www.onsemi.com/pub/Collateral/TND381-D.PDF

Since Im modifying a power supply, many values are unknown making compensation a crapshoot, though Ive noticed someone simply cutting out all that and replace it with a single zener diode

http://danyk.cz/iz350_en.html

The main advantages of using a TL431 is its temp stability and repeatability, but if I dont care about slow voltage drifts (in fact 10-20% is fine) and just need stable(not oscillating) bulk power eg for an audio amp (obviously not audiophile things), would a simple zener diode/string of series zener diodes result in a equally stable power supply as a correctly compensated TL431 network?

[PKTKS]

This issue can spawn several pages here...

The 431 is a compromise solution .. and is not a simple one.

It is NOT a simple zener.. it is a transconductance error amp
and while some folks swear by god that even TYPE 1 compensators
are fine usually you will find TYPE 3 compensated error amps w/431

There is NOTHING NEW using such devices..

In fact in your particular case (AUDIO AMP stabilized power)
an ancient device harder to find these days would be much more effective..

Cause can operate safely with proper AUDIO AMP voltages
and it is intended to audio and RF circuits..

https://pdf1.alldatasheet.com/datasheet-pdf/view/6777/NEC/UPC574/+457JuICzZTPENzY+/datasheet.pdf

It is OLD and so far I have never seen a good replacement for it..

Nevertheless any SMPS will require a TYPE 2 or TYPE 3 compensator
around any 431..

Paul

[magic]

I modded one SMPS for variable output and didn't find it to oscillate so far. I just left the compensation circuit as-is.
It's a series RC network from CATHODE to REF and in the event of any trouble I would experiment with increasing C or decreasing R because I have no clue how to do it any smarter

Make sure you understand all the problems that may occur on the primary side when the secondary output voltage is changed:
https://www.eevblog.com/forum/projects/modding-switchmode-psus-for-different-output-voltage-and-why-it-sucks/

BTW, I'm not sure that using a simple zener for regulation would guarantee stable operation in every such circuit. There is still a regulation loop (zener, opto, pwm, output voltage) and there are phase delays in various places which could accumulate to 180° at some frequency. But I don't know, I'm no expert on SMPS.

[PKTKS]

(..)
It's a series RC network from CATHODE to REF and in the event of any trouble I would experiment with increasing C or decreasing R because I have no clue how to do it any smarter
(..)

no just a zener will not regulate your load in full range...

To play with that without burning a lot of switching drivers.. (BJT or MOSFETS)
you need SPICE...  you need to run several simuls with load and type variations..

SPICE is handy cause you can program the whole thing in spice decks..
once you got it it may be reused.

Required models:

Code: [Select]

.subckt 4N25 1 2 3 4 5
R1 N003 2 2
D1 1 N003 LD
G1 3 5 N003 2 .876m
C1 1 2 18p
Q1 3 5 4 [4] NP
.model LD D(Is=1e-20 Cjo=18p)
.model NP NPN(Bf=610 Vaf=140 Ikf=15m Rc=1 Cjc=19p Cje=7p Cjs=7p C2=1e-15)
.ends 4N25

.SUBCKT TL431 7 6 11
.MODEL DCLAMP D (IS=13.5N RS=25M N=1.59 CJO=45P VJ=.75 M=.302 TT=50.4N BV=34V IBV=1MA)
*V1 1 6 2.495 ; used for fixed reference, replaced with B1 Limiter
B1 1 6 V=V(7,6) > 2.495 ? 2.495 : V(7,6)
R1 6 2 15.6
C1 2 6 .5U
R2 2 3 100
C2 3 4 .08U
R3 4 6 10
G2 6 8 3 6 1.73
D1 5 8 DCLAMP
D2 7 8 DCLAMP
V4 5 6 2
G1 6 2 1 11 0.11
.ENDS

Done that couple times lately but many times over in academia... 
Paul

[PKTKS]

oops  just noticed that I had wrongly c&p another opto model..

I kept them in the same  SPICE .lib  .. so fixed the wrong one..

And bonus adding others .. as well.. 

Code: [Select]

.subckt 4N25 1 2 3 4 5
R1 N003 2 2
D1 1 N003 LD
G1 3 5 N003 2 .876m
C1 1 2 18p
Q1 3 5 4 [4] NP
.model LD D(Is=1e-20 Cjo=18p)
.model NP NPN(Bf=610 Vaf=140 Ikf=15m Rc=1 Cjc=19p Cje=7p Cjs=7p C2=1e-15)
.ends 4N25

.subckt MOC205 1 2 3 4 5
R1 N003 2 2
D1 1 N003 LD
G1 3 5 N003 2 {Igain}
C1 1 2 18p
Q1 3 5 4 [4] NP
.model LD D(Is=1e-20 Cjo=18p)
.model NP NPN(Bf=610 Vaf=140 Ikf=15m Rc=1 Cjc=19p Cje=7p Cjs=7p C2=1e-15)
.ends MOC205


.subckt PC817 1 2 3 4
R1 N003 2 2
D1 1 N003 LD
G1 3 N004 N003 2 {Igain}
C1 1 2 18p
Q1 3 N004 4 [4] NP
.model LD D(Is=1e-20 Cjo=18p)
.model NP NPN(Bf=1200 Vaf=140 Ikf=100m Rc=1 Cjc=19p Cje=7p Cjs=7p C2=3e-15)
.ends PC817

They are all easy findable and working fine under SPICE .. w/431 models

Paul