SMPS Repair attempt

[olewales]

Hello fellow EEVBloggers. I might need your advice.

Recently I've acquired Dlink DGS-3224tg 20port managed gigabit ethernet switch. It's quite old unit but it specs seem to quite impressive even today and because I got it basically for free (only had to pay something like $10 for shipping) I thought that even if it does not work it might be interesting to see if its any different inside than low end consumer gear (I found its MSRP in 2002 to be $3,499.99).

It turned out that it does not work BUT I narrowed problem down to PSU module. The logic board itself works after supplying its power rails (+12V, +5V, +3.3V) from ATX power supply (at least switching fabric does work, have not checked management side but I expect it to be fine). I had to play around with my setup for a while because as it turns out this thing is quite heavy on 3.3V rail and voltage drop over long cables is an issue.

Now I only have to diagnose and repair PSU module. The problem is I have very little knowledge about construction and operation of switch-mode power supplies and I don't really know where to start with it.

The PSU is Phihong PSF-151-213C It does power up but voltage on rails is oscillating. Because similar thing is happening on all three rails I am guessing that the problem lies somewhere before power is split (primary side of the transformer?) but that is as far as my thinking goes. I attach photos of the PSU and screenshots from scope measuring output of different voltages. I hope that someone will lead me closer to solution or at least tell me that its not worth the effort.

High res PCB photos:
http://imgur.com/a/Nqyau

[Christopher]

Bad output filter caps (ESR) can cause ripple like this.

Or no minimum load on the PSU, but you have the thing connected to the other board .

[SeanB]

Replace all those blue, brown and black Nippon Chemicon capacitors in the power supply, using a low ESR version from the same manufacturer or using Panasonic. Badcaps.net will supply them or buy from RS/Farnell/E14 in Europe. While you are there replacing them resolder all the transformer pins and all the pins of the big diodes, it should then work fine again. Do not buy mysterLee caps off fleabay as they will at best be fakes and at worse blow up.

[shotgun25]

I have a question relating to the use of an oscilloscope in the PSU - What sort of bandwidth range would you expect to see in an SMPS in equipment of this nature?

I doubt I would need the full 150MHz range of the scope I bought, but I'm curious because I'm currently selecting some suitable probes to work on issues like this and wondered if 150MHz+ probes are overkill for this type of work.

[olewales]

Bad output filter caps (ESR) can cause ripple like this.

Or no minimum load on the PSU, but you have the thing connected to the other board .

Previous measurements were actually without load 
Below are screenshots from scope with switch main board attached. Waveforms actually look even worse.

Replace all those blue, brown and black Nippon Chemicon capacitors in the power supply, using a low ESR version from the same manufacturer or using Panasonic. Badcaps.net will supply them or buy from RS/Farnell/E14 in Europe. While you are there replacing them resolder all the transformer pins and all the pins of the big diodes, it should then work fine again. Do not buy mysterLee caps off fleabay as they will at best be fakes and at worse blow up.

Got it: Replace caps. Avoid fakes.
How low exactly is lowESR? On Farnell I can filter them by specific value (or what series should I stick to?)
Also, is there a way to confirm this without ESR meter or is it pretty much standard symptom for bad caps?

[wraper]

I don't think that output capacitors are faulty. Fault mode is completely different from what you would expect in that case. Those pulses are of low frequency ~6 Hz, this is not ripple. Much more likely small electrolytic capacitor(s) around PWM controller on the primary side.

[Jay_Diddy_B]

I don't think that output capacitors are faulty. Fault mode is completely different from what you would expect in that case. Those pulses are of low frequency ~6 Hz, this is not ripple. Much more likely small electrolytic capacitor(s) around PWM controller on the primary side.

1+

Try and find the datasheet for the control IC. It is likely to be a 3842 or similar and look at the reference design schematic. There will be a high value resistor, typically 150K to charge a capacitor from the high voltage, and a winding on the transformer to keep the capacitor charged once the power supply has started. There could be something wrong with the diode or resistor that keeps the capacitor charged.



Even the power supply does not use the uc3842, there will be a similar circuit.

Regards,

Jay_Diddy_B

[olewales]

Try and find the datasheet for the control IC. It is likely to be a 3842 or similar and look at the reference design schematic. There will be a high value resistor, typically 150K to charge a capacitor from the high voltage, and a winding on the transformer to keep the capacitor charged once the power supply has started. There could be something wrong with the diode or resistor that keeps the capacitor charged.

I followed your suggestion but I couldn't figure out anything on my own. (I haven't properly analyzed datasheet yet though, so please forgive me if I am missing the obvious) I measured each pin of the IC. Maybe this will tell someone more than it tells me.
Control IC is ML4824 located on right angled riser board. I've checked for cold joints and burned components but they look ok to me.

Please ignore scraped soldermask on one of the photos. I did it to get some ground point for scope probe.

[Jay_Diddy_B]

Hi,

Here is a schematic from the Fairchild datasheet:

You have a triangular wave on the Vcc pin. This indicates that C30 is charging through R27. The IC turns ON, indicated by the switching waveforms, but C30 is not being charged through the path that I have highlighted.

You should find similar components on your board. check the diode and the resistor in this path.




Please excuse my triangular wave, I am not good at drawing with the mouse.



The upper limit and lower limit of the triangular comes from here on the datasheet:



Good luck !!

Jay_Diddy_B

[wraper]

Now I see that this PSU does have PFC too, This changes things a bit. To save time, just replace C13 and and if replacing C13 helps, replace smaller electrolytic cap too (Edit, probably not worth). However, what I think now, actually the big high voltage electrolytic capacitor is dead. From my experience, this is exactly how PSU containing PFC behaves in such situation. Edit: and waveform on V_FB pin (resistor divider  + 1nF capacitor low pass filter connected to the high voltage) perfectly supports this assumption. It suggests that there is no stable high voltage supply.

[wraper]

Hi,

Here is a schematic from the Fairchild datasheet:

You have a triangular wave on the Vcc pin. This indicates that C30 is charging through R27. The IC turns ON, indicated by the switching waveforms, but C30 is not being charged through the path that I have highlighted.

You should find similar components on your board. check the diode and the resistor in this path.

Actually, according to the waveform, seems that it gets charged a tiny bit. However as HV sags when mosfets get open, no significant charging occurs.

[olewales]

I've done some checking on caps you suggested. I don't have ESR meter so I desoldered them and used my UT61E as a capacitance meter

C10 (HV) - Rubycon MXC 180uF/450V nominal - measures 166uF. Bearing in mind that my meter is probably not very accurate I am guessing its ok. I can change it but I won't be able to source any brand name caps locally.

C13 (close to IC, bigger one) - Ltec 220uF/25V niminal - measures 210uF. My quick research shows that its chinese brand and it is not considered to be very good. Cap seem ok though. I may change it if I'll be able to buy one locally.

C14 (small cap close to transformer) - Ltec 10uF/50V - measures 3.5uF and seems to vary quite a bit (maybe because its warming up in my hand). I'll replace this one.

I've done some basic tracing with continuity tester and C14 seems to be the cap that Jay_Diddy_B marked in the schematic. (positive terminal has ~75ohm resistance to VCC). C13 has dead short connection from its positive terminal to VCC.

[Evil Lurker]

Up in the very first picture, is that a blown MOV or is it silicon on the front of it? Looks like a big crack.