Touchscreen Firmware reprogramming so slow as to appear hung

[wad]

On one XO-4 C1 laptop only (so far), the touchscreen firmware auto-upgrade process is INCREDIBLY slow. It is so slow that it only draws one dot on the screen per minute. On the serial console, it is more like one dot every five seconds, although sometimes there are bursts of two or three dots printed.

One time, it was slow for the first few writes, then ran at full speed for the rest of the programming.

If the slow programming is allowed to complete, the MSP430 has been reprogrammed correctly.

Other C1 laptops tested so far don't show this problem. This is a 1.2 GHz SKU.

This was tested with both q7b05, q7b07 and q7b07ja, with both 0.6 and 0.5 neonode software preprogrammed into the MSP430.

Manually running flash-nn, $flash-bsl, or update-nn-flash works fine, at full speed. The touchscreen tests seem to run fine on this laptop.

I have seen update-nn-flash return a prompt imediately, with no error messages, but typing flash-nn rom:\nn.hex right afterwards worked fine.

The timing relative to boot is critical. If I delay the boot process by dropping into CForth then typing ofw, the programming is just as slow. If I delay the boot process by dropping into CForth then typing ofw while holding down the rotate key, then type resume after a few seconds, the reprogramming runs at full speed but always reboots the laptop immediately after completing (before returning an ok prompt, by jumping to a reset vector --- the EC is not involved in the reset).

Looking at the async. serial BSL signals between the SoC and the MSP430, the waveforms are nice and clean and running at the right voltages.

[wad]

I'll try to hang onto this motherboard for further testing, otherwise we will have to wait until it is shipped to Cambridge.

[wad]

The failing laptop has serial number SHC24600A2B, and is labelled "PRE BUILD" with "new speakers" written on the label.

[Quozl]

Suggested:

• manual programming immediately after contacting the MSP430,
• holding reset after device close,
• check behaviour of data ready signal in relation to reset on working and non-working laptops,
• check signals between MSP430 and IR PCB.

[Quozl]

Please try q7b07ji.rom, which contains an additional 1.2 ms delay after receive specified by TI that was absent from our code, and also an unrelated fix to the progress bar.

[Quozl]

q7b07ji fixed it, says Richard.

[Quozl]

svn 3443.

[Quozl]

Is in Q7B08.