How does one get the raw data scaled in milivolts?

[busypete]

I am using the HX711 on an experiment with a strain gauge, and do not need to convert or scale to weight units, I only need the output in +-20 milivolts (128 gain). However, I am struggling a bit around the concept of the conversion from the raw serial data from the ADC to the IEEE754 float mantissa. I cannot seem to match the data bits seen on the scope to the ones saved on the float. How can I extract or scale this data to represent the input in volts?

[RobTillaart]

Interesting question, I need to do some reading.

[RobTillaart]

However, I am struggling a bit around the concept of the conversion from the raw serial data from the ADC to the IEEE754 float mantissa.

The ADC returns a 24 bit value which is read in read()

  v.data[2] = _shiftIn();
  v.data[1] = _shiftIn();
  v.data[0] = _shiftIn();

these are the raw bits converted to a signed long (mapping with union) By design I use convert this long into a float (as most math of the lib is float)

  return 1.0 * v.value;

The bit layout of a float uses some tricks to encode a value, details see IEEE754 That is why it does not match the bits of the long value.

It would be possible to add a function to access the rawADC bits of the last measurement. Store the bits received in a private variable and wrap a function to fetch this value around it.

[RobTillaart]

I only need the output in +-20 milivolts (128 gain).

Process how I would try to get mV's (not tested)

Calibration

Put 0.0 mV on the A port, Call read() to get the zero point. (Expect the ADC returns zero +- some noise) Put 20.0 mV on the A port Call read to get the max value. Put -20.0 mV on the A port Call read to get the min value.

Interpolation

Then interpolate between those points to get your measurements in milliVolts. If needed more interpolation points can be needed (check HX711_MP)

Example code (not tested)

If 0.0 mV equals a read of 12345 And max equals 24365155

In code you need something like

float getmV()
{
  float zeroV = 12345.0
  float maxV = 24365155.0;
  float raw = HX.read();    <<<<<<<<<<<<<<<<<<< you might use different modi like average or median etc.
  return 20.0 * (raw - zeroV) / (maxV - zeroV) ;
}

Can you do the above test, I have no HW setup free to do so. Please post your results.

If it works well I can add it as part of the interface of the HX711 class.

[RobTillaart]

Idea!

In a bright(?) moment I had the following thought:

• Call tare() when 0.0mV is applied
• Call calibrate_scale(20000) when 20.000 mV is applied.

then the float get_units(uint8_t times = 1); should return voltages

[RobTillaart]

@busypete Did the above replies answer your question?

[RobTillaart]

@busypete As there is no answer to my question I assume the issue is solved. Feel free to reopen if this is not the case.

[RobTillaart]

Added a section in the readme.md of my local develop branch to keep this idea.