R: [Voyage-linux] Negative temperature reading on Wrap

[Tommaso]

I have performed again the temperature measurements monitoring the LM77
output while the WRAP cools down. It definitely seems to be a
signed/unsigned problem since when the temperature drop below zero the LM77
output passes from 0 to 4095500 and then it keeps decreasing 500 by 500 for
each negative 0.5 °C. 
Has someone any idea how to patch this problem? I'm not an i2c guru ;)
Bye

Tommaso

-----Messaggio originale-----
Da: Michael Renzmann [mailto:mrenzmann at otaku42.de] 
Inviato: venerdì 9 dicembre 2005 10.50
A: Beat Meier
Cc: Tommaso Cavanna - Wave s.r.l.; voyage-linux at voyage.hk
Oggetto: Re: [Voyage-linux] Negative temperature reading on Wrap

Hi all.

On Thu, 2005-12-08 at 21:34 -0300, Beat Meier wrote:
> > The board correctly booted-up, and I can enter the ssh login in order 
> > to read the LM77 temperature, but when I read the temp1_input value I 
> > get: 4095000 value.
> >
> > Since I expected a negative value, what does it mean 4095000?

This sounds like a "signed vs. unsigned integer" issue. Rudolf Marek
brought up another possibility when I talked to him today and mentioned
this problem:

=== cut ===
<ruik> or it can mean
<ruik> that some bus transaction was 0xFFFF
<ruik> and hence then the formula went nuts...
<otaku42> ruik: if that was the case, is there any way to find out?
where could one see the result of such a bus transaction?
<otaku42> (yes, i'm a i2c/lm_sensors newbie :))
<ruik> I must snoop the bus with external device
<ruik> like you can use another PC for this
<ruik> and then tap the bus and listen for the bus transaction
<otaku42> ruik: read: it's not easy to be done, right?
<ruik> It is easy
<ruik> there is a program for that
<ruik> you just need to build simple adapter
<otaku42> ruik: you have the url at hands?
<ruik> 2 rsistrors some diodes...
<ruik> yes
<ruik> http://warmcat.com/milksop/milk.html
<ruik> http://warmcat.com/milksop/cheapi2c.html
=== cut ===

> I had the same problem when I read in an intervall of 1 second. There 
> seems to be a problem if you read "rapidly" the values
> or something else goes wrong.

Asking him about this issue, he recalled that this could be a
hardware-related problem (bad design, noise).

> Rudolf was busy (as I remembre) and so was I ...

He still seems to be quite busy. It took three or four days until I
could catch him in IRC, but that also was caused by the fact that I was
busy as well and wasn't online anymore when he was available :)

Probably the best way to discuss these issues is to take them on the
lm_sensors list.

> Just about 2 days I had a question from a person about this.

Yep, that was me :)

Bye, Mike