I noticed a number of posts on this forum reporting bricked Compute Sticks even though the factory power supply was used. I followed the recommendation to press the power button for 3 seconds and my latest bricked computer came back to live! The two previous ones are still dead, presumably damaged by those voltage spikes. So, it looks like this product has some serious flaws related to power supply... disappointing. Hopefully, it keeps working going forward...
I am with Intel Customer Support and I just would like to let you know that we haven’t tested or validated any battery powering and therefore our contribution to this scenario is very limited.
I encourage other community members to share their experience.
Funnily enough I have been testing 5V power supply over USB lately. So here's a few items I can think of.
1) A power capacitor across the output should help alleviate the spike, and any sags perhaps caused by spikes in demand.
2) How much are your little 'buck' converters worth? Are they any good? If you've got a $200 PC, spend a bit more than $2 on the power supply which follow.
a) How gracefully does the buck handle changes in power demand? If it suddenly increases, e.g. you plug a USB3 device in, will there be a droop whilst it catches up? In which case you're likely to suffer a crash or even another 'brick' experience.
b) what's the ripple like?
c) Do they handle overcurrent demands gracefully?
d) Do they maintain voltage output up to their rating? I have quite a few cheap devices that don't measure up to their 5V 2A rating or anywhere near it. The voltage droops far before (crash/brick)
I have some 12V to 5V 4A DC-DC converters that are worth ~ say $20 each. They're industrial ones, I believe now out of production. But there should be a plethora of similar ones available. However I have tested 5 of them and found:
a) They are at 5.00V +/-0.05V with only one at 4.95V
b) They all maintain voltage up to the 4A easily
c) They have an "overhead" of 20% i.e. they will run 5V up to 5A no problem (I test for say a minute) whatsoever. Testing done on an electronic load tester.
d) No heatsink required. They only get mildy warm.
e) They are larger than a simple "buck" a circuit board 105x50mm.
f) Can't remember but I think input is tolerant up to 18V
g) I think they are at least 80% efficient
h) If you wish to look them up they are model PP25-12-5 using a module from Densei Lambda, supplier in Australia was Amtex, now under heliosps (see this link for some specs)
3) I haven't tested the Intel Power supply, but I will now- if I can take of the power from the plug effectively without damaging it.
4) USB cable (and connector) quality is critical. And this actually is what my testing was all about- to be able to charge 5V 2A devices at 5V 2A. Cheap aftermarket cables will exhibit considerable resistance increasing with current and this leads to voltage drop (crash/brick). The brand name cables from USB high current charging devices, e.g. Asus, Samsung, Apple all are much better as they have thicker wire for the power wires. For example, at ~1A I couldn't get better than 4.6V through a cheap cable, and 4V at 2A. The Asus cable was ~4.9V at 2A. The Microsoft cable will have to support the 3A so you're probably better off cannibalising one of those power supplies for the cable to power your device.
5) 5V vs 5.16V and voltage drop. Was the 5.16V open circuit? And see point 4 above, if under load the 5.16V may be to allow some voltage drop in the cable under heavy load.
6) Also what is your instrument calibration? To be honest ours aren't directly calibrated however I have cross-checked voltages on several Fluke multimeters against the load tester and they were all in agreement within +/-0.02V at 5V DC.