Help with repairing a Samsung HT-BD8200 Soundbar SMPS

[eigiman]

Dear EEV Friends,

I have many years' experience in SMPS and general electronics design and repair, and generally know my limits, but am a little stumped by this little bugger (see attachment).

Basically, there is a 5.6V standby circuit, which is used to produce a P_ON signal which switches everything else on.
Everything else does come on and is well within tolerance, except the +/-28V and -5V supplies generated by the MR4040. This will come on intermittently, most often not though.

All the standard culprits have been checked - electrolytic capacitance and ESR checked, diodes, optocouplers, resistances, dry joints. MR4040 has been replaced. Nothing. I've partially localised the issue (I think) - the measured Vcc on pin 4 of the MR4040 is at around 7V when it doesn't switch on, but is at +11V when it is functioning. I can't find any info on the Vcc requirements for the MR4040, though.

You can see that during the transition from standby to on, initially the Vcc is supplied from a rectified output on the primary of the ICE3B1565J via RB16, DB6 and RB10 (all checked), but I don't understand how this voltage can be too low if all the secondary voltages are correct.

I'm beginning to suspect the ICE3B1565J, a little out of desperation rather than logic...
Any suggestions or ideas?

This appears to be a common failure mode for this particular board based on internet searches (power LED flashes white/red, sometimes starts if allowed to "rest" for a few days), however I could not find a solution...

Cheers
Ilya

[PKTKS]

2 cents of rant .. first..

Every single day from last 2 decades...

someone somewhere design a "NEW!" supervisor chip
claiming it outperforms a "NEW"!!! class of IGBABCD ..MOSFET class..

A complete mess...

This weird combination  is just another one. From two
almost unknown and rarely found chipsets...
period.

Rant over fallback to classic trusty things:
- Those 2 zeners up ahead the optocoupler are part of the
 trans-conductance   feedback loop error amp.
- they should meet tight and very specific requirements to
 put the loop into the proper margins.
 - any deviation over time or incompatible weird thermal stress
 may render the loop - and the soft start and the stability to crap.

I would try messing those 2 with some tests and the optocouplers
to make sure the feeedback loop error amp is really stable.

Paul

[PKTKS]

and BTW... there are indeed a plethora of fake  TL431 out there
... which is the de facto main trans-conductance error amp
of that feedback loop..

If they have the mishap of having fakes TL431..
your loop is just as good or bad as that faked TL431.

90% trans-conductance amps based on TL431 have
more or less this faith.. as long as TL431 has become
so that widely used for that.

Paul

[eigiman]

Rant over fallback to classic trusty things:
- Those 2 zeners up ahead the optocoupler are part of the
 trans-conductance   feedback loop error amp.
- they should meet tight and very specific requirements to
 put the loop into the proper margins.
 - any deviation over time or incompatible weird thermal stress
 may render the loop - and the soft start and the stability to crap.

I would try messing those 2 with some tests and the optocouplers
to make sure the feeedback loop error amp is really stable.

Paul

Thanks very much for your response, Paul. So you reckon the problem is within the MR4040 loop, and not a startup Vcc issue? I'll replace the zeners ZD7 and ZD8 when I get a chance and see what happens. As for the TL431, I can't (don't want to) believe that a fake one would work for years and then suddenly become intermittent... I can swap the two TL431 chips (UB4 and UB3) around though.

TBC...
Ilya

[PKTKS]

Nono don't  just replace them...

As I said .. you already checked several points...
what is left are those misbehaves caused by bad loop compensation
or component out of specs.

Those zeners and a few other caps and resistors conform
the loop compensation network. The zener capacitance like it
or not is part of the feedback network

If replacing them you should very carefully check what
type of capacitance is driven the feedback loop otherwise
the feedback may and will cause instability much like leading
to not working output.

There must be also a capacitor in this loop which dominates
the phase compensation. Even with  proper model things
change not only over time but also by manufacturer.

Literally tons of faked TL431 exists. The specs vary a lot.
If replacing try matching exact model within tolerance.

otherwise you have another dead end.
with bad soft start or unstable output tripping OVP or OCP

Paul

[eigiman]

Paul, I understand your reasoning about the feedback loop and respect your experience. However, when I try to approach the fault logically, there is one thing that throws me.

If Vcc is 7V when the controller is off, and Vcc is 11V when it is working, and knowing that the current draw is very small (insignificant voltage drop across RB10), surely this is a clue?

Vcc is derived from 2 sources, and although I have no way of monitoring the value of Vcc during the transient stages from standby to on, would it not make sense to provide an independent source of 11V+ on the primary side (independent of P_ON), and to see whether oscillation starts reliably under those conditions?

Or perhaps just reducing the value of RB16, to 71R say?

Am I missing something?
Ilya

[PKTKS]

Reason being of my guessing of a bad FB network is that:
- I have not found the datasheet for MR4xxx series.
- As far as I can educated guess the "resonant" design
  actually is some sort of RCC (Ringing Choke Converter)
- The RCC is capable of "sense" "microloads" and adjust properly (ornot)
- RCCs are bad. not matter how much effort you put that third winding
which feeds the Vcc on MR4040 depends on the proper soft start of the
 chip - which includes the "LOAD CURRENT SENSING" (via TL431)

Put together the soft start should load enough the TL431 to convert
the sensed current into a proper voltage on pin 2 (FB) while that
is not stable the voltage on the VCC winding will float.

I don't bother much of VCC floating because it is being fed by that
third winding which depends on the proper RCC ringing

** IF ** and very if... my assumption educated guess of that "resonant"
chip is nothing more than an RCC designed in silicon...

and RCC designs depends quite a lot in the FB loop to start and keep stable
ringing the choke...

I will say again  I have not found the datasheet.  I am assuming this is
another attempt to do RCC into something  stable... I have seen others...

I would try to use a 4 chan scope or a LOGIC ANALYZER to see at least
4 wires into this network starting...

1) VCC if it is stable or just gone wild...
2) the DRAIN driver signal
3) the FB converted "sensed" load current (which drives the "resonant" thing)
   (e.g. if the sensed current is bad.. so is the ringing...)
4) the output

This design is not very much reliable if this is really an RCC chip

If someone can find the datasheet just drop it here for another
closer look into this design... which looks roughly like RCC...

Obviously I may just guessed wrong the chip...
I will try  a bit more to find that MR4xxx datasheet..
actually part of routine nowadays is finding these things...

Paul

[PKTKS]

Looks like this one...

https://datasheetspdf.com/datasheet/download-pdf.php?id=616515

And indeed it seems to be yet another RCC designed  in silicon...

Everything points in that topology as the only one very
specifically designed...

From that point... the analyzer is a must ..

Paul

[PKTKS]

Also from that datasheet...  I could locate the proper
items in the topology..

As for the resonant components are
- Capacitor CB2 2KV (ouch... a quite spike one)
-  Resistor RB2 over 3W should be hot...

That point should be also part of the Logic Analyzer View...

[PKTKS]

After a fast overlook of this new chip to me..

I would consider that fast damping diode (FR type) DB4
if is still capable of damping the ringing or if is just not
within specs...

by elimination
Paul

[shakalnokturn]

Without wanting to contradict the other Paul who is actually going to the trouble of thinking about the problem, I'd swap the ICE3B1565J.
I've been round in circles before on intermittent illogical SMPS faults that after waisting a lot of time were caused by one of the ICE family IC's.
Now as soon as I see one of these I do the basic checks then swap the PWM...

[eigiman]

Thanks very much. I had already ordered a new ICE3B1565J. So I'll be systematic and replace one component at a time.
I'll post what happens on here.
Ilya

[goaty]

So did you check the pin 7 of UB2 if this is correct voltage (<=24V) ?
Also you could measure across RB8 and RB16 for unreasonable currents.

[eigiman]

Hello and thanks, 14V on pin 7 of UB2 and insignificant current draw through RB8 and RB16...
Ilya

[eigiman]

Just spent 40 minutes on the train staring at the MR4040 application note and Samsung circuit again (attached). It's bizarre, they talk about Vcc and its different possible levels all the time, but not once do they mention an absolute value...
Also, the app note states that C112 (resonance frequency selection capacitor) should be 820-2200pF. However, the Samsung design uses only 100pF for the equivalent component (CB2). Is that not very odd?
Ilya

[eigiman]

Just came across another site (allaboutcircuits) where someone doesn't expect the MR chip to start oscillating until Vcc hits 14V: https://forum.allaboutcircuits.com/threads/mr5060.95515/

So I think I'm going to try and reduce the value of RB16 and see what happens...

[goaty]

So, if UB2 Pin7 is 14V, but UB1 Pin4 is not, then - if the schematic matches what you have - there must be a problem with RB16, DB6, RB10 _or_ ZD3, DB3, CE3.

Measure voltage drop across RB10, DB6, RB16. Where do the 14V-5.6V = 8.4V go ?

[eigiman]

But these aren't DC voltages before the rectifiers. I can't directly measure and compare the voltages across DB6, DB7, and RB8 and RB16 - or at least I don't know how in a meaningful way. In any case, assuming a similar current draw from both controller chips, wouldn't the value of RB16 (100R) compared to RB8 (36R) explain the voltage difference of 7V (14V - 7V, the 5V6 is only present on the secondary side as a regulated DC output).
Thanks, Ilya

[goaty]

Could be so that the 127Ohm into 47uF is too slow a time constant for the voltage to go up.
It basically charges the capacitor up until the zener clamps at 24V max.
Hm. Youre probably right with reducing the resistor. Strange design though.

[eigiman]

I'll have a play tomorrow and report back.

[PKTKS]

Without wanting to contradict the other Paul who is actually going to the trouble of thinking about the problem, I'd swap the ICE3B1565J.
(..)

Not at all .  Please do so...
So I can review and rethink my logic procedure.

This chip and such design of two mixed RCCs are totally new to me...

i am following...

Paul

[eigiman]

OK, it's fixed.

First, I tried the feedback components around the MR4040 - no change.

Then I replaced the ICE3B1565J - no change.

Then I reduced the value of RB16 until the MR4040 was switching on reliably - 47R. Reconnected the power supply. The flashing red/white was much slower and eventually the machine went into standby. I could then switch it on, but it wouldn't read discs nor open/close the tray.

I checked the supply voltage to the tray motor control IC - nothing. Traced it back to a very hot 7808 V regulator. Then checked all the wiring connected to the +8V line on the main board, disconnecting the flex cables one by one. The short was coming from the 40-way cable to the optical sled in the BD/CD drive...

Dismantled the drive and ... just where the first 180 degree bend is in the flex coming from the sled, on the outside the +8V and AGND lines had touched and come loose/broken from overheating.

I stripped back just enough of the flex to get decent copper, bridged the tears and strenthened everything with PVC tape perpendicular to the flex. It's not perfect, but the connections are there, without a short.

All working now, let's see how long it lasts. In any case, I can easily get a replacement flex.

So now for the fault analysis... The extra load on the +12V line from the SMPS (which feeds the 7808) was causing a change in the voltages on the primary side, so that the MR4040 wouldn't start up.

Thanks for all your help! We may have been barking up the wrong tree, but your encouragement was very important for me.

Another device which won't end up on the dump for another few years!
Ilya

[PKTKS]

Very nice... traced the culprit finally.

Those things I take notes for reference what supervisors
are really worthy more than 30 minutes of hassle or not.

A last question in my mind would be to disconnect all loads
eventually using a dummy load to see if this RCC would soft start...

RCCs are not reliable and over time then do fail to start...

Very good first contact with this chip.
Paul