Power supply voltage drop...any suggestions?

[Los Frijoles]

For about two weeks or so, I had my raspberry pi on continuously running xbmc with two network connections (wireless (usb dongle plugged in directly to the pi) and wired). After a while I noticed that I could no longer ssh into the raspberry pi and although when I plugged in the usb keyboard directly it would respond, attempting to do anything involving network would cause it to stop responding on all fronts for a while. It wouldn't respond to pings or anything like that either afterwards.

I suspected the power supply was going bad after I noticed that plugging in my wireless dongle would cause the raspberry pi to reset when previously it did not. I then measured the voltage across the input capacitor with my crappy multimeter and noticed that before the device started it was 5.1V and would hover between 4.7V and 4.8V when active. I think the high current draw from xbmc's idle state (100% cpu...compared to 40-60% when playing an MP4 over the network ) may have caused it to slowly degrade since I believe it was more stable before and didn't drop as much.

So, I am in the market for a power supply that will be more reliable, yet still on the not incredibly expensive side. Does anyone have any suggestions? There are a lot of power supplies out there, so I figured I may as well ask around to see what people here are using.

[btidey]

Before replacing the power supply make sure that it is not voltage drop elsewhere. The USB cables can drop a significant amount if they are not thick wire and also if they are long. If that is your problem changing the power supply will have little effect.

Unreliability due to low power can have a very small margin so if you think it has got worse, it may be just a very small change.

[mahjongg]

The F3 polyfuse is also a candidate, especially if it seems to be slowly degrading, which would be a strange thing for a cable to do.

Leave the PI alone for a couple of days, and see if afterwards it is behaving better.

If the polyfuse has had a "bang", it could have an increased resistance and be on the verge of "blowing", it would explain the "slow degrading".

[Los Frijoles]

Hm.. I don't think it was the cable. It is one of the thickest cables I have. However, it is a little lengthy (1m or so). I can try using my kindle's cable (~30cm) and see if that is any better.

As for the polyfuse, the resistance across it is nearly 0, but I know that could mean nothing unless I measured the voltage drop during operation. Is there somewhere that lists the part number of that fuse so I can have a replacement ready to solder in if it does fail?

[mahjongg]

The fuse that is (most often) used is a Raychem Tyco MINISMDC075F-2


It can be bought from Farnell here, also a nice source for data on it.
http://nl.farnell.com/te-connectivity-r ... dp/1175840

a typical resistance for it would be 0.2 Ohm, (it can vary a lot though, which explains why peoples mileage with it varies, also its value depends on what people have done with it, and even how hot is has been during soldering!) but for an average sample it means that when a practical current of 500mA runs through it, it would generate 0.1 V of drop, if the micro-USB cable had no loss, and the PSU would output exactly 5.0V at 500mA load current, (both fictitious, I know) then the PI would receive 5.0 - 0.1 = 4.9V, and at the polyfuse trigger threshold of 0.75A (750mA) it would drop 0.15 V.

So if you measure across the polyfuse and it reads, 0.35V you should consider that there is something wrong with it!

Its easier to measure the voltage between both end of the polyfuse, than it is to measure its resistance (obviously while the PI is off), as to measure the resistance the multimeter sends a very small but constant current through it, ten measures the voltage, and obviously its easier to measure the voltage while 500mA is flowing through it than when just 0.5 mA is flowing through it (as the voltage is a thousand times higher).

Some other people are saying that you should measure the voltage (to GND) at one end of the polyfuse then measure the voltage at the other end and subtract the two different values, seems an unnecessary complicated way to do so to me, as there really is no reason not to measure the drop directly.

[btidey]

I replace mine with a component with a higher holding current 1.5A MC36229 which one can get quite cheaply from eBay suppliers. Yes, there is a little more risk in fault current flows but normally one is worried about large trip currents rather than sustained intermediate values. The 1.5A version I use has about 80 milliOhm resistance after soldering which helps keep the on-board voltage up. It also allows a bit more current to flow through to the GPIO and USB assuming the power supply has the current capability.

You can also get pretty good measurements of its resistance using the lowest ohm ranges on a decent multimeter by putting both leads together on one side of F3 and then moving one to the other. Subtracting the two readings takes out the multimeter lead resistance (typically 0.2 Ohm) and gives you the actual F3 value.