On 07/29/2013 03:47 AM, Angelo Compagnucci wrote:
Hello list,
I'm implementing an LDD for the MCP3422 ADC from microchip. I'm heading
at kernel inclusion, so I'm asking here before doing it too much wrong.
I'm reading Documentation/hwmon/sysfs-interface, and I find that the
chip adhere poorly to that specification:
0) The chip has an 18bits ADC, but the unit of in0_input should be in
mV, I don't want to waste the 18bits precision. Right now I have the
unit in nV and it works quite well, but If I understand correctly, it's
not right for hwmon specs.
No it isn't.
1) The chip provides up to 4 differential input channels, but only one
at a time could be read after setting the corresponding channel in a
registry. How should I manage 4 different inputs that maps the same
device registry to be compliant?
You wouldn't. Each attribute should be associated with exactly one
input. The code would have to multiplex to provide respective readings.
2) Cause the device has 4 differential inputs, it could output negative
values. Are they permitted?
Sure, voltages can be negative.
3) The device has various configuration parameters (like varialbe
12/14/16/18bits precision, variable gain), could them be added as
external attributes?
I don't think so.
4) The update_interval depends on the precion (device is faster with
less precision), could it be updated at runtime if user changes precision?
Yes, but ...
Have you considered writing an iio driver instead ? That might be more
appropriate given the more likely use of this chip for generic ADC, not
for hardware monitoring. If it is used for hardware monitoring, the
iio/hwmon bridge driver could be used.
Thanks,
Guenter
_______________________________________________
lm-sensors mailing list
lm-sensors@xxxxxxxxxxxxxx
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
