Hi Sebastian,
Did the modified driver solved your problem? Becous if it does, I
would send this to a friend of mine who can commit to tinyos-main, and
hopefully this will be part of tinyos 2.1.2.
Andris
On Fri, Dec 9, 2011 at 6:58 PM, András Bíró wrote:
> Hi Sebastian,
>
> On Fri, Dec 9, 2011 at 2:30 PM, Sebastian Dölker
> wrote:
>> Hi Andris,
>>
>> Yesterday I continued with my i2c problems and discovered the following:
>>
>> The debug values:
>> 20:03:44 "cmd" 0x00 0x00
>> 20:03:44 "ir read done.."
>> 20:03:44 "init" 255 3
>> 20:03:44 "cmd" 0x04 0x01
>> 20:03:44 "cmd" 0x25 0x01
>> 20:03:44 "int" 0x08 0x01
>> 20:03:44 "cmd" 0x45 0xb5
>> 20:03:44 "int" 0x40 0xb5
>> 20:03:44 "cmd" 0x45 0xb5
>> 20:03:44 "int" 0x00 0x6a <- bus error
>> 20:03:44 "cmd" 0x05 0x6a
>> 20:03:45 "*start*"
>> 20:03:46 "*start*"
>> 20:03:46 "bb read done.."
>> 20:03:46 "***start***"
>> 20:03:46 "cmd" 0x04 0x6a
>> 20:03:46 "*error*" <- error means EBUSY is returned..
>> 20:03:47 "***start***"
>> 20:03:47 "cmd" 0x14 0x6a
>> 20:03:47 "*error*"
>> 20:03:47 "bb read done.."
>> 20:03:48 "***start***"
>> 20:03:48 "cmd" 0x14 0x6a
>> 20:03:48 "*error*"
>>
>
> Thanks, it's much more easier to fix with this help.
>
>> So the i2c-status is 0x00.
>> According to the atmel128 datasheet page 224 this means bus error.
>> To recover one should set TWSTO flag and clear TWINT („by writing a logic
>> one to it“?!)
>
> Yes, the TWINT flags seems a bit illogical, but that's how you can
> clear it. This translates to send a stop condition (TWSTO), and clear
> the interrupt flag.
>
>> I didn’t find this procedure in the device driver sources. Is the bus error
>> state handled in the driver?
>
> You're right, it doesn't handle it. Actually I don't like the
> interrupt handler (I2C.commandComplete) at all, since it doesn't care
> with the TWSR, it persumes that with its own internal state variable.
> However, when I fixed some of the bugs in this, I wanted to change as
> little as possible, becouse it's an old component, and some users may
> count on this not quite good solution.
>
>> I tried to add this procedure but I wasn’t successful..
>> Must this be added in the driver?
>> Have you an idea where?
>
> I attached the modified dirver, but I just tested it with a working
> i2c sensor. It should do the recovery procedure, disable the i2c
> interrupt, and return with FAIL.
>
>>
>> One more, in the attachment I added a picture of the sequence while the
>> error occurs.
>> Channel 1 is i2c-clock, channel 2 i2c-data. At the beginning the i2c sensors
>> are read out successfully.
>> Then a sensor pulls clock low for ca. 2.3 seconds. Probably because of that
>> the bus error occurs.
>
> The atmega1281 datasheet mention that bus error only occurs when an
> illegal start/stop condition happens. It would be more intresting to
> see the last part of the failed communication in detail, to check for
> this. Although if that's the case, we couldn't do much more, a slave
> device should never send start/stop, so this is a bug in your sensor.
> Handling bus error interrupt should fix it, but it's just a
> workaround.
>
> Andris
>
>>
>> Thanks in advance!
>> Best regards
>> Sebastian
___
Tinyos-help mailing list
Tinyos-help@millennium.berkeley.edu
https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
