Unable to flash firmware due to auto reset timing issues on windows & mac
Closed this issue · 37 comments
Two users of v1.3 hardware are unable to flash their devices from Windows or Mac computers. Both units were flashed multiple times on my Linux computer before shipping. Replacement units have the same problem. Removing C2 improves likelihood of getting successful reset for flashing. I am investigating component changes.
More info here
I have flashed v1.3 hardware (NUC+Daughterboard) successfully on Windows and macOS, not tried it on linux but I suppose I could. Not had any issues once I'd fixed any "hand soldered motes of dust" issues.
If you want to send me a non working one to test, I'm game. If it's only two users it sounds like a them issue, not a you issue, I can try it on various bits of hardware and cabling and see if I can get a fail.
I have flashed v1.3 hardware (NUC+Daughterboard) successfully on Windows and macOS, not tried it on linux but I suppose I could. Not had any issues once I'd fixed any "hand soldered motes of dust" issues.
If you want to send me a non working one to test, I'm game. If it's only two users it sounds like a them issue, not a you issue, I can try it on various bits of hardware and cabling and see if I can get a fail.
One user has received 2 v1.3 units from me that both flash ok on my end and they cannot flash either one. They also have a v1.0 which flashes fine with the same cable/computer. As mentioned, removing the capacitor allows it to sometimes flash meaning we might be able to fix this in hardware. The new wrover-e is supposed to have a silicon fix related to this issue and there may be too much capacitance. I shipped another board yesterday with 0.1uF cap and 10k pull up resistor on EN and will wait to see the results.
Better to make it work now for everyone than after "one billion served" :D
I gained access to a Windows 10 laptop today that exhibited the same problem flashing. After some digging through the CP2102N datasheet I found that v1.3 does not have VIO connected to VDD and it should be. This is likely causing the CP2102-QFN24 to be unstable during reset. I'm not sure why it affects certain computers and not others. This issue is not present in v1.0 FujiNet since it used the QFN28 chip which does not have VIO. A Fix for v1.3 boards is to make a solder bridge on pins 5 and 6 of U2. I have added an image of the bodge to make it work on v1.3 boards in the wild now:
Reopening until the fix is confirmed by another user with the problem. It worked for me but another user still cannot flash after applying the solder bridge.
How about thinking around the issue? Why not implement OTA or SD based firmware updates then there's no need for the bugfix
Does the person who still has the issue also have a problem with a 1.0 board?
Does the person who still has the issue also have a problem with a 1.0 board?
Nope. The v1.0 works absolutely fine.
After some more troubleshooting we've got the v1.3 bodge working except on two different Windows 7 home premium machines. They were able to flash using ubuntu and windows 8.1 and the bodged unit.
Pin 4 & 5 or 5 & 6??
Pin 4 & 5 or 5 & 6??
Typo fixed. It's pins 5 & 6
Confirmed that this fixes the issue for at least 3 users and in my testing. Closing
Reopening.. again.. I missed an important note about the need for a resistor divider on VBUS. While the VIO bodge seems to fix the issue for most users there are likely situations where flashing will not be possible. One user can still not flash with Windows 7 but it now works on their other systems. I will be sending them another bodged unit for testing.
page 8: https://www.silabs.com/documents/public/data-sheets/cp2102n-datasheet.pdf
Of interesting note is that the v1.0 with QFN28 seems to be more reliable with this issue than the QFN24 used on v1.3. The v1.0 board works on the Win7 machine mentioned above.
Image of the fix done to add the resistor divider on the v1.3 board:
https://github.com/FujiNetWIFI/fujinet-hardware/blob/master/FN32ROV-1.3-Q24/CP210x_RESET-BUG_FIX_2.jpg
Well, #$%^&*( me.
…
Yeah, it's a chore. It might be alright to just do the solder blob fix. That has fixed all but one of the problem units so far.
Does the divider fix the issue? is this final and has the schematic been updated? Here's my schematic change and updated pcb
Image of the fix done to add the resistor divider on the v1.3 board:
https://github.com/FujiNetWIFI/fujinet-hardware/blob/master/FN32ROV-1.3-Q24/CP210x_RESET-BUG_FIX_2.jpg
Holy hell!?! That is a lot more than a simple bodge. Ugh, I knew I should not have gotten so many of these boards.
Does the divider fix the issue? is this final and has the schematic been updated? Here's my schematic change and updated pcb
The windows 7 user has not received the replacement yet so I cannot confirm.
This is my updated schematic. I've added and changed the capacitors and added the resistor divider for the CP2102. Also, have added pull up resistor (R17) for SIO_DATAIN as suggested by BigBen.
I was able to get a Windows 7 computer that could not flash any newer FujiNet I plugged into it and this process worked every time:
- Plug FujiNet into computer with MicroUSB cable
- Run FujiNet Flasher
- Select COM port of FujiNet device
- Click "Flash FujiNet Firmware" and immediately press and HOLD button A
- Continue holding button A until you see "Erasing flash (this may take a while)", then release button A
- Wait for it to finish flashing and you're done
It's worth noting that v1.0 was able to flash fine on this same computer. Any version I tried v1.2 and up required the use of the above procedure. v1.2 added the open collector buffers for the SIO bus but I'm not sure how this relates to the issue and why it's only affecting some computers/operating systems. Also of note, I made two v1.4 pcbs, one with the qfn24 and another with the qfn28 variant of the CP2102 which made no difference; they both exhibit the same problem.
I'm considering this issue closed as the above workaround should cover anyone with problems. I will leave it open until I finalize the v1.4 files and release them.
This is amazing! This is with or without the hardware fix?
Without the hardware fix it worked for me, YMMV. If you are shipping 1.3 hardware you should apply the hardware fix. Hardware fix is updated in the repo
Do we know when the fix will be finalized?
I'm waiting for some users to receive units for testing.
The hardware fix has been confirmed working by the windows 7 user.
Works fine on my Macbook Pro as well.
Worked for me with 1.3 and MAcOSX Mojave (though it is written that flasher is for Catalina). I have kept rightmost button C (reset). Button A does nothing. Did not work with 1.0.
However, this cannot be closed until this "problem" turns into "feature" and is mentioned in oficial upgrade procedure. I have spent 3 evenings on this trying various configurations and three operating systems, asking people and nobody knows etc.
Worked for me with 1.3 and MAcOSX Mojave (though it is written that flasher is for Catalina). I have kept rightmost button C (reset). Button A does nothing. Did not work with 1.0.
However, this cannot be closed until this "problem" turns into "feature" and is mentioned in oficial upgrade procedure. I have spent 3 evenings on this trying various configurations and three operating systems, asking people and nobody knows etc.
And you made the changes suggested in the fixes? added and change the caps as well as the bodge wire?
Hello! I have seen no mods for Fujinet 1.0, maybe I have missed something... I have seen mods for 1.3, QFN24 CP2102, but not for 1.0.
Try replacing R3 & R4 with 10K resistors and report back here if it works or not.
Oh, I am sorry, I thought you were talking about the rev 1.3 that has known flashing problems. I haven't had any with 1.0. I am sorry that I was confused.
Hello, Unfortunately, after resistors change, still does not upgrade
Have you tried flashing multiple USB cables and with a different computer?
Before you plug the FujiNet into computer, hold button A. After plugging in, all 3 LEDs should be dimly lit. Then try flashing procedure.