Flashing Senoko PVT board (Page 2 of 3)

OK, I'm back at my own place and ready to work on this again.
I get the following with four (real, ish) cells wired up, the TX pins on U302 shorted appropriately, and in general things wired up "properly" so the chip should be happy:-

ch> stats
Manufacturer:       Texas Inst.
Part name:          bq20z95
Firmware version:   0x5009
Charge FET:         off
Discharge FET:      off
State:              normal discharge
Time until full:    65535 minutes
Time until empty:   65535 minutes
Chemistry:          LION
Serial number:      0x0001
Charge:             69%
Max capacity:       2866 mAh
Design capacity:    4400 mAh
Temperature:        25.24455 C
Voltage:            15482 mV
Current:            -4 mA
Average current:    -1 mA
Target voltage:     0 mV
Target current:     0 mA
Number of cells:    4 cells
Cell 1 voltage:     3819 mV
Cell 2 voltage:     3875 mV
Cell 3 voltage:     3993 mV
Cell 4 voltage:     3796 mV
Charge status:      0x0
    Charging allowed?   yes
    Can suspend?        no
    Can precharge?      no
    Can maintenance?    no
    Temperature limit?  no
    Temperature limit?  no
    Can fastcharge?     no
    Pulse charging?     no
    Pulse disable CHG?  no
    Cell balancing?     no
    Precharge timeout?  no
    Fastcharge timeout? no
    Overcharge OV?      no
    Overcharge OC?      no
    Overcharge?         no
    Battery empty?      no
Alarms:
    TERMINATE CHARGE ALARM
    TERMINATE DISCHARGE ALARM
Charge state:
    Battery initialized
    Battery discharging/relaxing
No errors detected
No safety alerts
No safety status messages

No amount of "gg cells 3", "gg capacity 3 5000" will make it write to the chip well enough to update the output of the stats command. Also, setting the charger directly doesn't seem to work very well either:-

ch> chg
Charger information:
        Charge thread:    running
        Manufacturer ID:  0x0040
        Device ID:        0x0007
        Current:          0 mA
        Voltage:          0 mV
        Input:            3840 mA
ch> chg set 1000 16000 4000
Setting charger: 1000mA @ 16000mV (input: 4000mA)... Ok
ch> chg
Charger information:
        Charge thread:    running
        Manufacturer ID:  0x0040
        Device ID:        0x0007
        Current:          896 mA
        Voltage:          16000 mV
        Input:            3840 mA

---I wait 10 seconds here---

ch> chg
Charger information:
        Charge thread:    running
        Manufacturer ID:  0x0040
        Device ID:        0x0007
        Current:          0 mA
        Voltage:          0 mV
        Input:            3840 mA

At the end of the process the stats output is the same as at the start. The only thing I see of note is the stuff under "Alarms" in the gg status. I can't seem to find a way to reset those, even though the battery status is otherwise fine. gg pfreset and reboot doesn't help either. Forcing the charger output on the gg on (with gg chg +) doesn't seem to affect anything either.

Any ideas?

Well, now it's got a 4 in it consistently (0x80000040, I think)...

The I2C pins are pulled up to 3V3, so I think the connections are just fine. Not sure about the connections to the chip, but they good to me, as do the ARM. Given that it WAS soldered in backwards for a while, I wouldn't be surprised if it had died. Regardless, I've got a new one on the way. I'll try sticking it on (the right way!) and see if that works. If not, back to debugging.

I don't think there's any hardware problem with the board from your end, mclien. The gas gauge is notorious for doing this; the TI forum for them is full of developers hit with the same issue of read-only dataflash, and they get the same unsatisfactory answers about it from TI. Reflashing seems to be the best way to unlock the chip, which is what I'm working on.

I think the bug is triggered by some subtle difference in initialisation; possibly the order in which power, battery and senoko first execution is applied.

xobs, do you know when you'll get a chance to dump your board?

Thanks! I'll have a look at them this evening.

Haha, that could have gone very, very badly.

My chip's awake!

...and unplugging the AC power made everything go off. Whoops!

Either the forced reboot or the "gg auto 1" I ran afterwards seems to have fixed it, and now the power button works, and battery power works, and I can hot-plug the AC power with no apparent ill effect. Woohoo!

Now I just have to get this code to the point where someone else can (kinda) safely run it.

Excellent work!  Looking forward to the code, I guess very shortly I will no longer have an excuse to not finish my case.

It's not the senoko firmware that needs to be reflashed, but the data in the gas gauge IC. Flashing a known-working dump from xobs' gas gauge fixes the bad state that causes the gauge to be stuck in read-only mode.

I've finished putting verification code into the flashing tool; I don't think there's anything else I can do to make it safer to run than it is now. So if someone with a locked chip wants to give it a go?

https://github.com/pelrun/novena-gg-tools

Looking good so far.  I had to try the flash command quite a few times to get it to work, but eventually it did.

gg commands appear to be working.

About to try a battery....

xobs: It's just flashing the data; your firmware dump was identical to mine. (thankfully the firmware isn't storing extra data into the instruction flash, which was a possibility I needed to rule out.)

GregRob: Did it fail at the beginning of the flash process or during the verify step? I'm hoping it's just that the gauge wasn't communicating at all (i.e. like when a battery isn't connected) instead of getting partway through the flash then dying.

There's not too much actually on the i2c lines between the stm32 and the charger IC, at most a resistor? So the I2C lines are probably ok (but check continuity anyway). Is power getting to the charger IC?