> I have done this, and i have read back the shadowed register values. > > The max6626 data sheet says: > Pointer Register - D7 to D2: Will read all zeros, cannot be written. > > Because the D2..D7 cannot be written, the leaving pinter register bits > are 0..1 > which shows the register shadow effect. Please understand that there are two different phenomenons on the LM75: 1* Registers cycle over 8-addr boudaries. 2* addresses 0x04-0x07 are not mapped to any register. If I read you correctly, the MAX6626 is somewhat different in that it cycles over 4-addr boudaries. This would be enough to reliably differenciate between both chips. Note that this is really a detection/identification trick. Both chips only really have 4 registers at addresses 0x00-0x03. I would really like to see the output of i2cdump. Care to post it here? > O.k. I aggree, but the problem I currently see is, > that I have to distinguish between the LM75 and MAX6626 to select > the appropriate conversion and detection functions. > Because both chips use the same addresses I think it is not possible > to do this in automated way. What do you think? According to what I said above it doesn't sound impossible (unless I missed your point). That said I see no reason to differenciate. On the LM75 the extra bits (10-12) will always read 0 which is compatible with the MAX6626. Considering that every (LM75-compatible) chip has 12-bit resolution shouldn't hurt. Note that in 2.6 all temperatures are reported to user-space in millidegree C anyway. If you are able to differenciate between both chips, then you can introduce a new kind for the max6626 and handle the 3 extra resolution bit for it only. However, a new kind means that you have to add support to libsensors, sensors, and virtually any application relying on libsensors. I seriously doubt it's worth the pain, which is why I'd stick to the lm75 name for everyone and extend the reolution to 12 bits for everyone as well. Thanks, -- Jean Delvare http://khali.linux-fr.org/
