Re: [lm-sensors] How to compute a voltage that may be positive or negative

[Jean Delvare <jdelvare@suse.de>]

Hi Christopher,

On Fri, 22 Aug 2014 18:31:22 -0400, Christopher Cordahi wrote:
> I'm wondering if this is a valid use case.  I need to perform both a
> multiplication and a subtraction to compute an original voltage.
> 
> lm sensors 3.2.0 is working fine on Linux 2.6.39.4 reading an ADS1115 ADC
> chip that measures a voltage from 0 V to about 2 V.
> 
> # uname -srmo
> Linux 2.6.39.4 armv5tejl GNU/Linux
> # sensors --version
> sensors version 3.2.0 with libsensors version 3.2.0
> # sensors --bus-list ads1115-i2c-2-4a
> bus "i2c-2" "i2c-1-mux (chan_id 0)"
> bus "i2c-1" "DaVinci I2C adapter"
> # sensors --no-adapter ads1115-i2c-2-4a
> ads1115-i2c-2-4a
> in4:         +1.14 V
> in5:         +1.14 V
> in6:         +1.14 V
> 
> My problem is that I'd like for the sensor utility to convert the ADC
> voltages to the original voltage which are in the range from about -10 V to
> about +10 V.
> 
> If I modify the sensor configuration as follows, it appears to work
> correctly.
> 
> root@centaur-3_0009:~# tail -2 /etc/sensors3.conf
> chip "ads1115-*"
>     compute in4 @*10-10 , (@+10)*10
> root@centaur-3_0009:~# sensors --no-adapter ads1115-i2c-2-4a
> ads1115-i2c-2-4a
> in4:         +1.41 V
> in5:         +1.14 V
> in6:         +1.14 V
> 
> Unfortunately when I adjust the values to their more accurate values, it no
> longer converts and reports the original ADC value instead.

I tried but could not reproduce the bug here. I also looked at the
source code and there is no arbitrary limit to the number of digits.

> root@centaur-3_0009:~# tail -2 /etc/sensors3.conf
> chip "ads1115-*"
>      compute in4 @*9.72340425532-9.95823902214 ,
> (@+9958.23902214)/9.72340425532

Note that you have a wrong factor in the second half here, even though
that's not related to your problem.

> root@centaur-3_0009:~# sensors --no-adapter ads1115-i2c-2-4a
> ads1115-i2c-2-4a
> in4:         +1.14 V
> in5:         +1.14 V
> in6:         +1.14 V

Well, 1.142 * 9.72340425532 - 9.95823902214 is 1.145. So I think the
value is properly scaled, just it happens that for your specific value,
the scaled value is almost the same as the original value, and that
confused you.

The limited resolution of "sensors" admittedly does not help here, try
"sensors -u" to see the 3rd digit.

-- 
Jean Delvare
http://jdelvare.nerim.net/wishlist.html

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors