On 06/01/2015 03:06 PM, Chris Lesiak wrote:
[ ... ]
.
I think things could be improved in one of two ways:
1. Accept the precision that you get by doing the calculation millivolts.
Simplify things by using millivolts throughout, although pullup_uv
needs to remain becaus it is part of the interface.
2. Create an additional patch doing the calculations with microvolts
instead of millivolts.
What are your thoughts?
You explain above that doing the calculations in mV looses precision.
Given that, what would be the point of the first proposal, especially
since the uV are already available (and may still be needed in
ntc_adc_iio_read if conversion through iio is not available) ?
Thanks,
Guenter
I do think that option 2 is the better choice. If you'd like, I'll create a patch.
Your call.
But if 1 were implemented, ntc_adc_iio_read would give a value in millivolts
even when iio_convert_raw_to_processed failed. Something like:
ret = iio_convert_raw_to_processed(channel, raw, &mv, 1);
if (ret < 0) {
/* Assume 12 bit ADC with vref at pullup_uv*/
mv = (pdata->pullup_uv * (s64)raw / 1000) >> 12;
}
But what would be the point ? I don't see how that would improve anything.
Since iio internally supports uV, at least optionally, you would then
always loose accuracy. So we might as well leave everything as-is.
Thanks,
Guenter
_______________________________________________
lm-sensors mailing list
lm-sensors@xxxxxxxxxxxxxx
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
