I just worked on a PSU for an Apple computer. It differs from ATX supplies by having a 25V 100W output instead of the normal +5V standby. The inductor was so badly overheated the wire had turned black! It still worked though, I just thought they painted it that way to radiate heat more effectively... Until suddenly there was a short from the +5V winding to the winding used for -12V. The resistance between +12V section and +5V section was about 10-20 ohms.
Pictures:
http://elektronikforumet.com/forum/viewtopic.php?t=25106#332639
Yes: That black stuff on the floor is (was) the wire enamel! The caps have already been replaced.
I have since removed the extra wire loop on the inductor, it was just there to make sure there was enough wire for an extra turn if I had miscounted the turns on the original one.
This power supply uses a single ended one transistor forward converter with current mode control. The low duty cycle of about 30% is used to cut down the voltage stress on the MOSFET as the bus is already 400V. I think they used a 900V part. The current sense resistor measurment ringing is probably due to using wrong impedance coax, nothing printed on it and I couldn't find anything marked 50 ohms. Much better than the probe though! The signal drowned in ringing when measured with the probe. Good enough to see current ripple and possible saturation in case the core was something else from the beginning. Impossible to know the colour code of the overheated core
Further up there are pictures of the dummy load I built
Black heatsinks are suboptimally attached with only one screw but this was good enough for the power I dissipated. I cheated and let car headlights take most of the 12V output power. The IRFP250 controlling them was only desaturated during startup when the lamps were cold.
Questions:
Have anyone seen this happen before?
Does anyone have a better method for measuring the voltage over the current sense resistor?
Pictures:
http://elektronikforumet.com/forum/viewtopic.php?t=25106#332639
Yes: That black stuff on the floor is (was) the wire enamel! The caps have already been replaced.
I have since removed the extra wire loop on the inductor, it was just there to make sure there was enough wire for an extra turn if I had miscounted the turns on the original one.
This power supply uses a single ended one transistor forward converter with current mode control. The low duty cycle of about 30% is used to cut down the voltage stress on the MOSFET as the bus is already 400V. I think they used a 900V part. The current sense resistor measurment ringing is probably due to using wrong impedance coax, nothing printed on it and I couldn't find anything marked 50 ohms. Much better than the probe though! The signal drowned in ringing when measured with the probe. Good enough to see current ripple and possible saturation in case the core was something else from the beginning. Impossible to know the colour code of the overheated core
Further up there are pictures of the dummy load I built
Black heatsinks are suboptimally attached with only one screw but this was good enough for the power I dissipated. I cheated and let car headlights take most of the 12V output power. The IRFP250 controlling them was only desaturated during startup when the lamps were cold.Questions:
Have anyone seen this happen before?
Does anyone have a better method for measuring the voltage over the current sense resistor?
Hopefully this is a very unusal failure. It is possible to see the inductor through the grille on the inside just where the cables come out. The caps had bulging tops too due to the heat from the bad inductor. If it's not burnt or has bulging caps I wouldn't worry.
The only possibility I can think of is that the inductor itself is bad, no other failure could cause this with the computer still working.
The only possibility I can think of is that the inductor itself is bad, no other failure could cause this with the computer still working.
Past a certain stress level, some types of iron powder cores may progressively suffer from unexpected aging and increased power losses. The process goes faster and faster as more heat is produced until something fails.
See this link:
http://www.micrometals.com/thermalaging_index.html
See this link:
http://www.micrometals.com/thermalaging_index.html
Interesting! Thanks for the link Eva. The "new" inductor runs at approx 70 deg C at full load, maybe 90 then when mounted in computer. Flux excursion is ~70mT pk-pk at 70kHz, the core is T-106 size yellow colored from another PC PSU.
Does this seem OK?
What's your experience with measuring over the current sense resistor in this kind of PSU? Would a ferrite core on the probe lead suffice? I didn't have one big enough to try it out. Measuring with the probe made it impossible to see the current waveform due to 1MHz or so ringing.
Does this seem OK?
What's your experience with measuring over the current sense resistor in this kind of PSU? Would a ferrite core on the probe lead suffice? I didn't have one big enough to try it out. Measuring with the probe made it impossible to see the current waveform due to 1MHz or so ringing.
This is probably too hot, it's likely to fail again after a few months.
Concerning the sense resistor, you should get a clean signal if everything is right in the PSU and in your probe. I assume you are using an isolation transformer. A common-mode choke would help at frequencies way higher than 1Mhz, like 30Mhz and above.
Concerning the sense resistor, you should get a clean signal if everything is right in the PSU and in your probe. I assume you are using an isolation transformer. A common-mode choke would help at frequencies way higher than 1Mhz, like 30Mhz and above.
How hot do you think is safe then for this material? For the materials they discussed in the Micrometals link it seemed like even 100 deg C would be safe. But this material (#26?) isn't very good is it?
I do understand that the core hasn't got very good thermal conductivity but most of the loss should be from the windings, estimated at about 6W for the DC part. From core loss graphs of #26 material core loss would be ~2W if the core is fresh.
Yes, I used back-to-back transformers for isolation. You can see it in the pictures. The mains outlets are not grounded but this couldn't matter, right? What if the psu is very noisy? AC mains comes in through a 3cm long ferrite bead-like core and then there are three (!) cascaded filter stages after that if I remember correctly. Then comes the PFC. Layout is pretty strange with an upper level board with PFC diode and coil. PFC transistor and filter caps are on the lower level and they are connected together with wires. The wires do not follow each other down, they are spread out.
I do understand that the core hasn't got very good thermal conductivity but most of the loss should be from the windings, estimated at about 6W for the DC part. From core loss graphs of #26 material core loss would be ~2W if the core is fresh.
Yes, I used back-to-back transformers for isolation. You can see it in the pictures. The mains outlets are not grounded but this couldn't matter, right? What if the psu is very noisy? AC mains comes in through a 3cm long ferrite bead-like core and then there are three (!) cascaded filter stages after that if I remember correctly. Then comes the PFC. Layout is pretty strange with an upper level board with PFC diode and coil. PFC transistor and filter caps are on the lower level and they are connected together with wires. The wires do not follow each other down, they are spread out.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.