Take ADC resolution into account

[ManuelMcLure]

Description

Take the ADC resolution into account in the AD595 and AD8495 temperature formulas.

Benefits

Makes the AD595 and AD8495 formulas work correctly on boards with 12 bit ADCs.

Related Issues

Related to #16993.

[AnHardt]

Does the 6.6 still happily apply with this change? There have been some questions about it. As I recall the original submission for the AD8495 was a little different, but to make it line up with the AD595 formula above it I changed the code so it just has the literal 6.6 value in the same spot as the 5.0. It was probably something like 660 instead of 6.6 * 100. I'll take a look at the original PR at some point.

#define TEMP_AD595(RAW) ((RAW) * 5.0 * 100.0 / float(HAL_ADC_RANGE) / (OVERSAMPLENR) * (TEMP_SENSOR_AD595_GAIN) + TEMP_SENSOR_AD595_OFFSET)
#define TEMP_AD8495(RAW) ((RAW) * 6.6 * 100.0 / float(HAL_ADC_RANGE) / (OVERSAMPLENR) * (TEMP_SENSOR_AD8495_GAIN) + TEMP_SENSOR_AD8495_OFFSET)

The best to read and likely most understandable version could be:
#define TEMP_AD8495(RAW) ((RAW) / float(OVERSAMPLENR) ... // the average ADC_value of the samples
... * 3.3 / float(HAL_ADC_RANGE) ... // Volt per ADC_LSB
... / 0.005 ... The direct number for the sensor gain from the datasheet - not its reciprocal value 200. If that's really speeding up the calculation the reciprocal is acceptable. That's the only number distinguishing AD595 from AD8495.
... * (TEMP_SENSOR_AD8495_GAIN) + TEMP_SENSOR_AD8495_OFFSET) // the user calibration.

Ideally the direct value for the ADC_REFERENCE_VOLTAGE 3.3 or 5.0 Volt should be taken from a define in the cores. Likely we have to define something like HAL_ADC_REFERENCE_VOLTAGE in the HALs because the cores, if they define something like that, don't name it uniformly. And likely it makes sense to make that over-writable by the configs.

Every value, except RAW, in this calculation is constant. Maybe we should help the compiler to see that by pre-calculating. Then the actual runtime calculation can be abbreviated to #define TEMP_AD8495(RAW) ((RAW) * a + b)

[thinkyhead]

Maybe we should help the compiler to see that by pre-calculating.

No need. The compiler can easily reduce the constant parts to a single multiply and add on its own.

I agree that if there's are reference values to be used —such as those at #14257— they would be better to define at the board level.