Snubber circuit in TTi PSU overheating

[g0rsq]

I have been given a nice TTI EX354 PSU which will make a nice addition to my test equipment if I can repair it.
It is a 300W SMPSU and effectively working, but the snubber resistors burn out after only a few seconds (30) even when no load applied. The snubber resistor is 680 Ohms 4 watt.

To test the PSU I run it through a 60 watt light bulb (in series), to limit the current, and the bulb is quite bright, so the circuit is drawing too much current also. The snubber resistors get very hot also!

I have tested all the circuit around the snubbers, and the switching mosfets, and everything seems normal.

I would like to eliminate the secondary components (they contain MagAmps) to ensure they are not loading the primary.

Is it safe to disconnect the secondary windings in this SMPSU? There is no feedback control from the secondary to the switching control.

Thanks

[Andy Watson]

If anything, I would expect un-loading the secondary would leave more energy for the snubbers to dissipate. Have you checked that C17 and C101 are blocking D.C. ? 

[T3sl4co1l]

C17, C101 maybe are cooked, and thus delivering much more snubber power?  (Or DC, as Andy suggests?)

Not really any obvious operational failure that should cause that, otherwise.  Slower switching speed or partially failed switches/diodes should cause greater dissipation in those components, and less if anything in the snubber.

If you need a replacement for C17/C101, I would suggest a 1n, >= 400V, C0G ceramic (disc or MLCC) or snubber-type polypropylene film capacitor.  Avoid X7R and other type II ceramics, and polyester and other cheaper film or other types that have higher losses.

Tim

[wraper]

Unlikely that anything other than faulty D8 or C17/C101 could cause this. But if they are ok, check D2-D5 too.
BTW, If you do not desolder D8 (or resistors in parallel), obviously you cannot check if it is open.

[TheBay]

TTI will send you via email more in depth schematics if you ask them nicely 

[wraper]

To test the PSU I run it through a 60 watt light bulb (in series), to limit the current, and the bulb is quite bright, so the circuit is drawing too much current also.

Glowing 60W bulb is completely expected even with normally working 280W PSU, especially considering that this is lab PSU made not about low load efficiency in the first place. If you don't want it to glow, you need >= 150W bulb at least.

[g0rsq]

Thanks for all the replies, very helpful.

I have checked C101, and it measures 1n (C17 is not fitted). D8 is also good. Both these parts tested out of circuit. I think I will re-test them, but use a 500V megger to see how they behave under stress!

I made an in circuit test of D2 to D5 after removing the FET's and got good results, again with low voltage testers, so will remove these and test them at higher voltages.

Probably a good idea to replace all these components though as a precaution.

OK about removing the secondaries, wont help me diagnosing the issue. I just want to eliminate the MagAmps as they could be causing a reverse action through the transformer as their fields break down, but really do not understand MagAmps and their application in this PSU. All the description says is they keep the voltage to the linear regulators to the output voltage plus 1Volt! To reduce heat dissipation in the linear stages.

Ive got a TEK portable oscilloscope (with isolated inputs), so will start probing the oscillator circuits tonight and check the frequency. Should be 80kHz. There is some circuitry that regulated the pulse width depending on what the HT voltage is, so will also check this is OK.

Thanks again

Peter

[wraper]

but really do not understand MagAmps and their application in this PSU. All the description says is they keep the voltage to the linear regulators to the output voltage plus 1Volt! To reduce heat dissipation in the linear stages.

Without them it would not be possible to regulate output voltage of the SMPS in a such a wide range reliably.

Did You check C14?  Guess it might affect snubber if open. Did you check voltage on C13 and if it is OK too?

[Rufus]

but really do not understand MagAmps and their application in this PSU.

I didn't either so read up a bit. At the end of a PWM cycle the MAGAMP is saturated in the forward direction and has little inductance looking like a short. When the transformer voltage goes -ve at the start of the next PWM cycle the reverse current in the MAGAMP is limited by the control circuitry and it demagnetises the MAGAMP by a controlled amount. When the transformer goes +ve again the MAGAMP has high inductance and looks like an open until that demagnetising has been undone and it saturates again.

So the MAGAMP is off for a proportion of the +ve half cycle controlled by how much current you feed it during the -ve half cycle. It is like a time delay switch for the +ve half cycles.

[g0rsq]

Rufus

Thanks for taking the time to explain the MagAmp. The only information i could find on the interweb was for a "true" amplifier using two separate cores.
In this circuit I can now picture it in my head as an AC current limiter, that is controlled by a voltage. The control circuit allows just enough current to flow through it, each cycle, to just maintain the input voltage to the regulator at 1 volt over output voltage.

Wraper

Re-checked all diodes and capacitors at a higher voltage, and all are OK.

In fact I have measured the voltage across the primary winding, and the snubber resistors, and to me they look OK. I attach the waveforms here. I have loaded the output to about 1A and everything is working fine, except the resistors getting very hot (They are mounted right next to the C13) and after about 10 seconds are hot enough to burn my fingers!

What I did notice when testing the outputs of the PSU was that the 5V output only runs at 4.2V, and the 3.2V output runs at about 2.5V. The voltage on FS2 is about 5.5V. I cant see how this could affect the snubber circuit. However experience has told me to always fix the faults you can see first, so I will move my attention to this next.

I am working away from home for next few days, so will try and get back to this at weekend.

Thanks again for suggestions and information.

Peter

[g0rsq]

Just to update on this fault.

Never really got to the bottom of the problem. All components tested good, but resistors still overheating!

I put a scope across the resistors (Isolated input scope), and took a RMS measurement across the resistors and got 120V RMS.

Doing the maths:-

P=V^2/R gives 20 Watts....No wonder they are over heating!

I worked backwards and worked out I can only have a maximum of 60Vrms to stay inside the resistor power ratings!

I bought some 8 watt 3k3 resistors to experiment with.

If I increased R120 resistance from 750R to 1k5, the RMS voltage across them went up to 170Vrms, giving once again 20 Watts dissipation!

If I decreased the resistance, to 500R the voltage dropped to 100Vrms, giving......... 20 watts power dissipation  !

I simulated the circuit in LTSpice, and found that if C101 was reduced in value, then the power dissipated in the resistors went down.

I settled on 560pF and replaced C101 with a 3KV 560pF capacitor.

Now the PSU runs fine, and the resistors, while getting hot, are running fine. The RMS voltage across the resistors is 45 Vrms, resulting in only 2.7 Watts in the resistors, (2x 2.4 Watt resistors).

Not sure if there are any implications from changing this capacitor value, only time will tell.

Thanks for the help

Peter

[Neilm]

Reducing the value of C101 will increase the voltage spike on the FET when it turns off. Make sure that this does exceed the voltage rating of the FET. I have seen some that were hit with small over voltage spikes in this sort of application - it survived for a surprisingly long time before it failed short circuit.

[g0rsq]

Thanks

Will check this out. I presume I am looking at the voltage across the Drain-Source?

Peter

[T3sl4co1l]

Yes.

[g0rsq]

No problems

Voltage measured was less than 80V P-P using a 100Mhz scope (no spikes observed).

IFR740 is rated to 400V DS.

Yes PSU is quite large for its rating, only single side PCB's used. Basically it is a box with 90% air by volume!!!

Anyway thanks for all the help on this one, I think I can box it up and start using it.

Peter

[Andy Watson]

That is a 2 switch forward converter, isn't it? Why even have a snubber, the switches are clamped to the input anyway?
....
Answered one of my own questions, the snubber is for EMI.

An interesting observation. Any back-emf should be clamped to the input supply and the snubber is there to reduce EMI. It's as if there is too much energy in the core.

In fact I have measured the voltage across the primary winding,

Is the scale on the waveform correct, and is the measurement really across the primary, or is it one side to ground?
If it's the latter, ok. However, if it's the former I would expect to see 2x340V.

And then there's this:

What I did notice when testing the outputs of the PSU was that the 5V output only runs at 4.2V, and the 3.2V output runs at about 2.5V. The voltage on FS2 is about 5.5V.

What if:
One of the power fets has failed short-circuit. The other good-pair would still make the switching but the flux in the transformer core would not decay before the next cycle - leading to possible saturation and lack of output on the secondary.

Edit: "Ground" - I mean the negative of the high voltage supply.

[Eugen]

Hi g0rsq,
I designed this PSU in 1994 .
The waveform across the transformer primary is not correct if working from a nominal 230V mains supply .
It appears to be working at maximum pulse width (50% duty) , whereas it should be about 30% duty at nominal mains voltage .

IC7 (555 timer) is triggered from a negative pulse on Q21 collector to start a new cycle .
C68 is then charged by R82, R90 to the threshold voltage on IC7/pin6 and the pulse is terminated .This simply reduces the operating pulse width with an increase of mains voltage .
If for some reason C68 does  not reach the threshold voltage, the pulse is terminated by Q19,IC5-F and D51 charging C68 to the threshold voltage .
It seems to me that C68 is not being charged by R82, R90 and D53 . Please check those components .

Regards,
Eugen

[g0rsq]

Thanks Eugen 

I did read about this "feed forward PWM" but as I was seeing a 50% duty cycle had assumed this was correct (not having any other information to hand).

I will certainly check this area as soon as possible (I can run it through a variac and see if the pulse width changes).

I am working in Ireland for the next week, so earliest I can get back to it is next weekend, though at the moment I have the PSU connected to a misbehaving Rubidium oscillator.

Thanks again

Peter

[g0rsq]

Sorry for the long delay getting back to this! Life never goes to plan!!!

The Feed Forward PWM does seem to be working fine, as the mains voltage in my house is only 220VAC!

I checked the values of R90, R82 and C68, and were all OK.

The function of IC7 is working fine as output on pin 3 starts at -ve pulse on trigger, and ends on threshold value pin 6.

Using a Variac I have altered the mains input voltage from 180VAC to 245VAC and the results are shown below.

At 240VAC the duty cycle is about 34%, but at 220VAC the duty cycle is nearer 40%.

Interestingly the duty cycle across the primary winding (TX2F) is different, at 45% at 240VAC and 50% at 220VAC and below

Not sure if this is normal or not, but I will make some checks around TX1A, B and C.

Peter

[Andy Watson]

What exactly are you measuring? Are those waveforms across the primary, or are they one-side to ground? Either way, they look very square - I would expect to see some exponential component in there as the transformer core recovers.

[g0rsq]

Sorry, should have added more information!

They are the output of the 555 timer on pin 3, basically monitoring the duty cycle of the PWM that drives the Q2 and Q3 and the gates of the power mosfets.

I think the results show the Feed Forward regulation control is working, but by the time the drive signal gets to the primary winding (TX2-F) it increases in duty cycle.

A few quick measurements show that the voltage across TX1-A (primary of the drive transformer) has already increased in duty cycle from 40% to 50%, so will check out this area as soon as I can.


Peter