Rock5 does not work on most PD power supplies

#193

Totally agree as adding a secondary PD would not of been very ‘clever’.
To be honest as SBC increase in power and capability a fixed 12v becomes ever more important than a mobile or toy charger.

@DooMMasteR
There are no SMbus/I²C issues on arm as a hardware SMbus essentially doesn’t exist as its a software wrapper around I²C that merely needs to be implemented as some has when we run the sensors cli command.

tcpm_source_psy_4_0022-i2c-4-22
Adapter: rk3x-i2c
in0:           5.00 V  (min =  +5.00 V, max =  +5.00 V)
curr1:         1.50 A  (max =  +1.50 A)

Is the neogiated level not actual and when not negotiated with a dumb psu its stays at uboot initial level .
I guess awk '{print $1/172.5}' </sys/bus/iio/devices/iio:device0/in_voltage6_raw will also get added to sensors as work progresses.
x86 Bios & Arm uBoot are just different to how things are implemented and even if they are, where much is just purely a software wrapper to emulate difference or missing function.

#194

But I2C4 is mapped to the FUSB302BMPX and the M2_NGFF_KEY_E slot, so once a card uses the pins and possibly the same address, how is it going to be handled?
I have the issue that my RockPi5 won’t boot at all with my wifi card added, but another card works fine and the Intel one with masked pins works too. So I guess it is some issue related to the I2C pins.

The Intel WiFi card has the pins connected to GND which is probably what kills the I2C bus.

#195

Reading the previous workaround, I wonder if a usb-c PD trigger board would do the trick?

I’ll test to see if it changes anything as I’ve got some PD boards laying around, (but still requires adapting DC output +/- terminals to usb-c connector).

this would be the DIY/ very hacky method though
At the end, I’m glad I didn’t order more than one board and only the 8GB version…

#196

Yeah its just that chicken & egg of early revision and likely it will take time to evaluate effect, what cards and if software or hardware revision is needed.

I guess currently the answer is don’t use a intel card and use one from the shop or one like the ones in the shop or others the community says works.

#197

I guess there is a reason why the Nanopi has a dedicated STM32G030F6P6 that basically does nothing aside from setting up power.

1 Like
#198

“Tape” tricks not working always.
Just got ALLNET PD 65W from ALLNET Taobao Store. With some tries, R5B can boot with 20V3.25A. (but now 100%, sometimes fails and stuck.)

tcpm_source_psy_4_0022-i2c-4-22
Adapter: rk3x-i2c
in0:          20.00 V  (min = +20.00 V, max = +20.00 V)
curr1:         3.25 A  (max =  +3.25 A)
Smoke after direct 19v
#199

Ordered the Radxa psu now. I prey this will work.

I think Radxa should team up with the vendors to sell the Radxa psu at 50% discount for those who have ordered the 5B.

With shipping cost is usd20. Too high for a low end psu.

1 Like
#200

这个问题,预计什么时候才能修复呢

#201

Good luck. Their PSU didnt work for me.

#202

They have better engineers over there at friendlyarm, clearly.

#203

It seems more like a “lazy” solution, which is also less flexible, but it clearly works a lot better of the start, yes. But I guess they might fix the issues over time.
The decision to “share” the SMbus/I²C however is more an issue worth a revision of the board…

#204

Doh!

May I know your mounted accessories?
I have a Radxa A8 wifi module and Samsung nvme disk only.

#206

Success with an Apple 20W charger, booting from NVMe via SPI. I (pre-)ordered the Radxa 30W from AmeriDroid, that should, fingers crossed, put paid to this issue.

#207

Just to be clear:
Do you got nvme + wifi bt module + emmc + micro sd + mouse n keyboard + hdmi screen.
All hands on deck and with rApple charger it works fine?

#208

Same for me, since I used NVMe, instead of SD, my previous non working PD charger are working now.
I’m booting with wifi/bt module + mouse and keyboard wireless dongles + hdmi screen.

So definitely looking like a timing issue, and workaround provided in wiki, to use a faster booting device seems to be the quick way to fix it for now.

#209

Indeed a real pain dealing with these PD negotiation issues…

The u-boot commit you’ve mentioned is not present in the stable-5.10-rock5 branch.

To test it I’ve cherry picked it to realize that it doesn’t change anything at least with my DELL 65w with PD modes (5V-3A; 9V-3.0A;15V-3.0A;20V-3.25A) as the board still ends up in an infinite loop with both recent Debian and Ubuntu images using an old class 4 16GB micro SD.

I’ve followed instructions here to compile U-boot https://wiki.radxa.com/Rock5/dev/Debian, so unless I’m doing something wrong this commit is not cutting it.

The latest android 12 image flashed on my NVME works fine with my Dell USB Power supply, which confirms the theory that with slow storage the PD negotiation is not timely executed.

Using a Samsung 15w max charger and booting from TF I’m able to go a little further but I suspect it’s drawing too much current during the boot process and it ends up rebooting anyway.

As for the SPI flash process I’ve discovered that the only way to make it work reliably is to use a USB 2 port, it works nicely for me including flashing android from the windows RKdevtool app.

Anyway PD negotiation would need to happen early on with U-boot rather than in the kernel otherwise anyone with less than stellar performance storage will experience this pain and honestly unless someone can confirm successfully booting debian with a class 10 mirco-sd card I don’t want to waste my time anymore.

Status of PD support during u-boot
#210

I have some powersupplies that work with the updated uboot and some that don’t, but before almost none did work.

#211

I got 65W Allnet power supply, with which I was unable to boot 5b both from SD and nvme. The board booted from SD with a Raspberry power brick and a dumb 12V 3A power supply almost always, but only rarely from nvme, more frequently it got stuck. I tried different linux and SPI images. Frustrated, I revisited the Allnet power supply and lo and behold - the board booted (to debian) with the latest SPI development and debian images, negotiating 20.3 V. Since then I rebooted it from nvme with wifi on multiple times without a single fail, making it the most successful of power supplies I tried.

#212

Could you please point me to the “latest SPI development image” ?

Thanks!

1 Like
#213

Err, I mistyped - latest debug not developmental image, from here :
https://dl.radxa.com/rock5/sw/images/loader/rock-5b/debug/rock-5b-spi-image-g3caf61a44c2-debug.img