• Home
  • Help
  • Search
  • Login
  • Register
  Show Posts
Pages: [1]
1  Hardware and U-Boot firmware / Hardware / Re: SheevaPlug Power Supply Failure on: August 03, 2011, 06:09:20 PM
One of the biggest issues that I see is they have used a small heat sink on the power transistor, but its completely internal to the case and the electrical insulation they use will also add thermal insulation insuring that the heat is going to stay in the can... FAIL!

Ultimately, what I would probably do if this doesn't work is cut out the insulation and add the necessary hardware for electrical isolation and then bolt the transistor directly to the metal can. You want to get the heat out of the can and the best way to do that is turn the can into the heat sink.

In addition you really should have additional louvers in the power supply section of the case so that you get some convection, otherwise the plastic case is just another thermal barrier to the heat dissipation.

It leaves me scratching my head when I see they have such an innovative plug design where you can use a cord or the direct mains plug, but then they botch the power supply... in all fairness someone else probably botched the power supply and they just happened to OEM it.
2  Hardware and U-Boot firmware / Hardware / SheevaPlug Power Supply Failure on: August 01, 2011, 04:32:04 PM
I guess I'm a little late to the party. My SheevaPlug power supply ate itself over a year ago and I'm just now getting around to dealing with it.

  • I purchased a SheevaPlug in 2009.
  • I first brought it up 2009-11-15
  • I never had a chance to finish implementing the application software so it just sat, powered, in my development lab
  • In the first week of 2010 it began to reboot its self every 15 minutes.
  • Sometime later I noticed the LEDs pulsing and immediately knew that the electrolytic capacitors in the power supply had fried

I recently was prompted to restore it to life because the NSLU-2, which it was to replace, had been acting up and its necessary to have a backup strategy although I'm not very confident in the long-term viability of the SheevaPlug.

After opening the power supply, sure enough, the two output filter caps had the physical signs I'd expect blown/bulging pressure relief disks in the top of the cans.

Let me digress for a moment. Unlike most other passive electronic components, electrolytic capacitors are subject to degradation and one of the critical factors that plays a role in their decay is temperature. All too often people do not look beyond basic voltage and capacitance when replacing a capacitor and this is often a fatal mistake. In addition to capacitance and voltage specifications, electrolytic capacitors will also have an operating temperature specification as well as an endurance specification... (if yours doesn't you have no business using it). Failure to select the correct component for the target environment probably will result in the replacement failing sooner than the original, and sometimes even more catastrophically. Where good ventilation is present and high temperature isn't a real issue lower temperature components can be used safely. However, in an application like the SheevaPlug where the power supply is completely encapsulated in a metal can, and insulated both electrically and thermally, you have the makings of a real problem because of the tiny oven that has been created. Given that the SheevaPlug power supply design is a thermal nightmare, and we're seeing significant failures as a result, it is absolutely critical any repairs take this design consideration into account otherwise the repair is likely to be very short lived.

So, the two 1,000uf caps were rated at 105c which are a high temperature variety (typical general purpose variety are 85c). Trying to improve upon the rating proved next to impossible because even though there are higher temperature parts with greater endurance available they only come in the next can size which will not fit physically in the supply case. So, the only recourse was to select the best rating among the available 105c parts in an 8x15 case. I eventually settled on a Panasonic EEUFR1A102L http://industrial.panasonic.com/www-data/pdf/ABA0000/ABA0000CE132.pdf. This part is 105c rated with 8,000 hrs @ 105c endurance. They are readily available for $0.50.

After replacement I put the whole box back together with a thermocouple between the two cap cans. I've been monitoring for about a week now and the temperature has stabilized at 50c. Given the measured electrical and thermal operating values and the part specification calculated endurance is ~117 years. Just for contrast if you were to use a 1,000 hr @ 85c part calculated endurance would be 3.7 years... and that assumes that yours operates at the same temperature as mine which is unlikely with a higher mains voltage or greater current draw. Also, these are theoretical calculations. Actual results have a lot to do with the actual quality of the part and its likely that the original caps were not of particularly high quality which would explain why they lasted only ~9 months. I expect better results from the replacement I used, but time will tell.

Obviously, a good quality external power supply would be a solution, however, I think it kind of defeats the purpose. The SheevaPlug packaging was a big selling feature and it becomes a lot less compelling when you add an extra item to the package, hence my desire to try and salvage the unit without such modification.

Pages: [1]