Help me understand this complex flyback PSU

[F1GWR]

Central structure of the primary stage of this Artesyn NFS177-7630 (used in an HP 35670A analyzer) is reverse-engineered here : see enclosed files

Please help me understand in the daughter (= fille) board around the UC3842A how work these weird structures:
- the oscillator network with Q3 Q4 and what should I expect as a working frequency
- the voltage reference with VR1 associated with Q1 and Q2

As this PSU is completely dead (but Hi DC voltage is OK when supplied by mains (= secteur) or 12-28VDC to DC converter), my aim is to inject low cyclic pulses in the gate to force it to waken somewhat and consequently be able to locate the faulty component(s). Therefore I need to know the approximate oscillator frequency and the reference values.

Complex secondary of T1 is not depicted. The links from there are represented by T3 and the two opto-couplers. The only inspection finding is in the main board R16 that is discolored, measured 140 Ohms in circuit both ways and PCB yellowish.

Please note: Attached schematics below have been edited with error corrections since this first message has been issued. Therefore some wise remarks in the following thread may not appear consistent anymore!

[wasedadoc]

Where is Vcc generated?

[fzabkar]

Where is Vcc generated?

CR5 and Q1, AFAICT.

[wasedadoc]

Where is Vcc generated?

CR5 and Q1, AFAICT.

OK. I'd start by checking Vcc.

[shakalnokturn]

Sorry for not answering all those questions.
Having repaired plenty often without going into all the details of how they're designed, like others I'd start by checking on PWM Vcc.
A fair chance that your problem is R2 or R4/5 on "platine principale".

I wouldn't recommend simulating gate control, even at low duty cycle there's potential to cause unnecessary damage.
If you must, add a series light bulb between primary bulk capacitors and transformer.

Isn't T2's primary on platine principale meant to be in series with T1/MOSFET for primary current sensing?

R16 is in the snubber, they often run hot, not necessarily a defect although it may have drifted up in value.

[F1GWR]

Thanks to all contributors.
Sorry that nobody could help with the strange circuits. I would have learned some more...
The two suggestions (wasedadoc and shakalnokturn) to check Vcc were the right ones! Was so obvious that I'm a bit ashamed not having checked that... Was so confused by complexity: "the woods that hide the tree" & "the nose in the steering bar"...
Anyway investigations results are :
- R2 open circuit (but with absolutely no apparent damage) >>> replaced: OK
- < 1 V on the  CR6 - C10 junction
- controller U1 removed
- consequently retrieved 19.6 V on the CR6 - C10 supply = U1 shorted
And now replacement part order follows.
Will keep you informed...

[shakalnokturn]

Thanks for the update.
I wouldn't have expected U1 to be dead too...
Any chance that it got killed during probing?

In any case, with a shorted PWM I'd check components from output pin to MOSFET gate and MOSFET itself (although they only fail "open" in a spectacular way) your drawing of T2 winding puzzles me a little and I can't help thinking that a shorted MOSFET could fuse the current sense transformer.
Just speculating but I usually like to be sure of all possible causes for future smoke and sparks when replacing a PWM IC.

[F1GWR]

Thanks!
Don't think so: U1 had low voltage at Vcc from the beginning. MOSFET had already been tested good with ohmmeter. But as it is currently unsoldered during the probing of the circuit (to avoid further damage), I can replace it by a new one (in case of HV leaks) at the same time as the controller as it's only 2,42 €.
Same for Q4 and CR20 that test OK.

I didn't unmount T1 to check the connections underneath. You were right about the current sensing transformer T2 as the wire that passes through the sensor (PE-51688 - left picture on datasheet page 1) is attached directly to the transformer winding and therefore inserted between drain and primary winding of T1 most probably.

[shakalnokturn]

If it tests good you should be OK reusing the MOS. (Well it's certainly better than risking a new fake depending on your source...)
I've seen UC384x with supply pins shorted on a couple of occasions but it was always due to excessive Vcc because of bad capacitors.
On your supply I don't see any auxiliary winding feeding the PWM once it has kicked in so to me it doesn't fit that we'll to have both faults you have.
Still, I don't know the equipment's history and coincidences happen.

[F1GWR]

If it tests good you should be OK reusing the MOS. (Well it's certainly better than risking a new fake depending on your source...)

Still, I don't know the equipment's history and coincidences happen.

My source is Reichelt so it should be reliable: they sell Vishay branded FETs. I don't buy those (at best out of specs) chinese components on the bay or Ali-and-the-40-thieves. For the time being my Tektronix 370 is under maintenance (must rebuild the three Gray encoders) and therefore cannot test the original FET at 340 V.

Don't know history as I just bought this HP analyzer second (or more) hand recently, out of order.

[magic]

Are R4, R5 and Q1 alright? Those going open could be another reason for the Vcc rail being low under load.

The most important thing about MOSFET is that it doesn't have a short between gate and the other terminals. Such short can easily blow up the PWM chip with HV.

[F1GWR]

Are R4, R5 and Q1 alright? Those going open could be another reason for the Vcc rail being low under load.

The most important thing about MOSFET is that it doesn't have a short between gate and the other terminals. Such short can easily blow up the PWM chip with HV.

Yes all checked alright. And as written this morning at 9:48:47 CR6 cathode measured 19,6 V after I removed the controller U1 (before was less than 1 V).

[magic]

This power supply can deliver up to 1mA of current coming from R2 through Q1 base-emitter junction, even if Q1 has zero β or R4,R5 are open. So the real test is to see if it can still maintain similar VCC voltage while delivering 5mA load current.

Did you test for a short between VCC and GND pins of the suspected chip with a DMM?

Are you sure that the rail wasn't pulled down by the Q1,Q2 circuit? These appear to implement some sort of emergency shutdown, shorting the VCC rail to ground when activated by VR1.

[F1GWR]

This power supply can deliver up to 1mA of current coming from R2 through Q1 base-emitter junction, even if Q1 has zero β or R4,R5 are open. So the real test is to see if it can still maintain similar VCC voltage while delivering 5mA load current.

Did you test for a short between VCC and GND pins of the suspected chip with a DMM?

Are you sure that the rail wasn't pulled down by the Q1,Q2 circuit? These appear to implement some sort of emergency shutdown, shorting the VCC rail to ground when activated by VR1.

Weird: no short between Vcc and ground of U1!
A friend of mine explained that Q1-Q2 as a binistor. Found very little info about it.

Therefore I come back with my initial question (sorry to quote myself):
Please help me understand in the daughter (= fille) board around the UC3842A how work these weird structures:
- the oscillator network with Q3 Q4 and what should I expect as a working frequency
- the voltage reference with VR1 associated with Q1 and Q2

[wasedadoc]

I do not know what Q1, Q2 and VR1 are doing as there is no used output from that circuitry. Is your schematic complete and accurate?

[shakalnokturn]

Same here, a few too many things look too iffy on the schematic.
Assuming all the rest tied to the oscillator section is for secondary side OVP/ OCP / thermal shutdown I'd just go by R16 / C5 for the operating frequency calculation.

[F1GWR]

Had a check about my reverse engineered schematic. Made some minor corrections to the daughterboard schematic. New one is uploaded in my initial message with name: "Schéma_fille".

As for the T1-T2-CR1 section couldn't find any mismatch. But I agree there seems to be no output (apart from acting directly on power supply lanes).
Maybe anyone wants to check the above assertion with the enclosed PCB pics.

Motherboard changes in schematic will follow later (will modify transformer faulty connections and check in the triac section the weird: "both anodes linked to ground!")... (to be cont'd next post)

[F1GWR]

(cont'd from previous post)

...and PCB routing in Kicad (warning: components references in yellow are placed outside and consequently maybe superimposed over a neighbour!)

[F1GWR]

Base chopper (SchemaHacheur) schematic has been corrected due to inconsistencies and also hidden tracks on the PCB. New diagram has been dropped in the first message attachement of this thread in order to avoid confusion with previous wrong data. Sorry for these mistakes.
Triac circuit has been bebugged, but still does not seem correct...

[F1GWR]

Hi folks,
After a long time against this SMPS, it came to life again .

In summary:
1- main board 330K, 1/2 or 1 W start-up resistor R2 was open circuit; replaced by a bigger 2 W one
2- protection triac SCR2 was routed the wrong way (Artesyn swapped A1 and G  ); could not find a reverse TO-220 equivalent, so I had to lift pins 1 & 3 and solder wires to swap connections (fortunately the hack is now hidden )
3- some parts of the circuit are stlll a mystery...

Green LED was lit. All P1 output voltages checked quite accurate with no load (except maybe -18 V measured at -19 V, quite acceptable)

See below waveforms of the HiPot (Ch1) and U1 output (Ch2). 0 V reference = bottom. Trigger point: start of 2nd div. Supply voltage: 15 VDC.

HP 35670A test log now reveals DSP error 6 (X/P RAM address test). If valuable info I'll start a specific thread...

Thanks to all that involved in this repair. 
And hope that my reverse engineering schematics will help others in case of this PSU failure...