Hello all, I just powered on my allnetchina 8G Rock5B with the official Radxa 30W USB PD and Radxa Type C cable but I saw smoke and sizzling noise came from HDMI-1 near the Type C on boot phase. Immediately powered off after that happened. That HDMI is plugged into a cheap passive HDMI-VGA converter that worked before on my other devices. I have yet to check for physical damage, can anyone tell me what went wrong? Did I have defect Rock5B? or the HDMI-VGA fried something? Will update later once I check with real HDMI capable monitor and serial console.
Update: Returning the unit to allnetchina for replacement.
that sizzling sound usually hints towards a voltage(buck) converter not being able to sustain voltage. I’d assume that this isnt per se an HDMI Issue, especially when the adapter was completely passive as you said. Sounds more like a capacitor burned in the voltage converter or maybe a defect coil.
Check for visibly defective inductors and capacitors. I do not have my rock 5 b here(its at the office) and I actually havent looked too closely at the board, but from the pictures on the radxa wiki I would say the power conversion happens in the middle of the triangle between the actual soc, the e-key m.2 and the gpio header. theres a square chip surrounded by 3 inductors and 6 caps - check these for visible damage first. The actual smoke might have come from the mosfets north of the HDMI port - if the voltage regulator broke down, those might have quite literally gotten too much power and literally smoked down. Be aware that I am working off the lowres photos from the radxa wiki images here(and I know already I got the pd controller wrong, the ROCK 5 uses a Fairchild FUSB302 USB PD, which is a tiny 9pin BGA Chip, howeverthat just handles usb, according to the data sheet it doesnt do any switching of voltages etc).
Either way, if there is smoke, that means there is always some visible damage somewhere, and mosfets / power switching components are always a prime suspects to look out for.
I checked the board, this Chip (U2602 based on SMD PDF Rock5B v1.42) is the one that burned. Idk what this is but will check the schematics. I will try to contact Allnet for replacement.
This chip is near the USB Type C power and near the RAM.
if you happen to have a multimeter at hand I would (in addition) suggest you check the smd capacitors around that chip. Also note as per my understanding, this chip is only resonsible for negotiating power delivery modes with the power supply, but not actually switching it.(this is done via i²c by the looks of it. I would generally assume that there is some sort of protection in place to avoid a faulty power supply smoking up the entire sbc(likely a fuse). So as a workaround, you can try the following things: measure all the components that are around the fusb032, make sure all inductors are ok, check the vin pins of the fusb to make sure they do not short to gnd. Considering that the fusb is likely smoked, prime candidates for further death would be the capacitors C2612, C2613, C0402.
Finally, if you are of the daring kind, you can power the rock via(i think its those, PLEASE DOUBLECHECK) the GPIO pins 4+6 using 5V and see if it comes up. Alternatively use the GPIO Headers. DO NOT APPLY MORE THAN +5V on either. Then see if the board comes up. if it does, you might be saved by simply replacing the FUSB032 and any off the shelf smd components that you can find as damaged. IF it does not come back to life, there’s definitely further damage somewhere that you should investigate and fix. Btw the FUSB032 is a fairly off the shelf component itself, finding a more local replacement from one of the known suppliers like digikey or mouser should get you a cheap replacement part rather soon. considering that its actually not a bga (I only did a quick google for a datasheet on the chip which suggested a 9pin bga) this should be fairly easy to replace despite the small footprint of the chip.
A little postscript: you can ignore the proximity to the ram. The ram has to be placed closely to the SoC itself simply for signal reasons. Given this isnt a full scale desktop pc the memory controller is in the SoC itself, and needs a rather short path to the ram chips. they most certainly do not have any direct connection to the power inputs
Out of curiosity, were you using a barrel jack to USB-C adapter when this happened?
Others reported the same issue (FUSB302B burning) with broken jack-to-USB-C adapters that send the barrel jack’s voltage to the wrong pins (only GND and VBUS should be connected, but some adapters erroneously connect CC as well).
Currently, I do not have access to multimeter. For replacing the FUSB032, I do not have equipment for that as well. Allnet has already replied to me for the return of this unit.
Please don’t blindly put 12v on this board. Stuff like this will fry the board. The board should always be initialized at 5v, just as any of the acceptable power sources(dumb USB brick or pd Bd) would do. Anything higher WILL destroy stuff.
Out of curiosity @AXdaruser since you got the got the RMA approved already, could you try powering the board on using 5v on gpio 4/6? These kinds of failure are actually quite interesting as to how far the damage goes. The manufacturer will not report on this issue for good reason
There’s two Client Challenge in the power path. The 1st deals with the input voltage (4.5 – 26V) at the USB-C receptacle to transform it into sane 5.1V for USB3, HDMI, etc. and the 2nd then transforms this into 4V to feed this into the PMIC.
