Lessons learned on AMDC firmware for reading voltage on AMDS sensor cards

[elsevers]

Abstract

This show and tell summarizes the following two lessons learned on how to properly read voltage from AMDS sensor cards:

1. That it is important to prevent the unsigned 16 bit ADC digital value from being improperly interpretted as a signed value and then sign-extended to a 32 bit int.
2. The proper scaling factor for converting the digital value received from the AMDS into a voltage value.

Problematic legacy code

In our research group, code along the following lines has propagated through most of our research projects for reading the AMDS sensor card voltages:

    err = amds_get_data(AMDS_PORT, AMDS_CH_6, &out);
    *adc_volts_out = 0.00125 * (double) out;

    // Magic delta value to deal with odd analog design... :(
    // Specific to CurrentCard design, perhaps VoltageCard as well.
    //
    // ==> at 0A input, the CurrentCard ADC voltage input is mid-rail,
    //     about 2.4V, and -FS current is about 0.2V, +FS is ~4.8V.
    //
    //     The binary ADC codes for these are odd, they go from
    //     negative to positive weirdly. I haven't debugged if
    //     this is actually correct of not... So, this DELTA value
    //     shifts them around to get a nice linear output
    //     relative to the current input value.
    double DELTA = (double) 0xFFFF * 0.00125 / 2.0;
    if (*adc_volts_out > 0) {
        *adc_volts_out -= DELTA;
    } else {
        *adc_volts_out += DELTA;
    }

There are two problems here:

1. The scaling factor of 0.00125 does not match the ADC configuration (at least for the current card), which maps an input voltage of VRef to a digital value of 0xFFFF. See Fig. 47 on page 21 of the datasheet.
2. The "magic" code at the bottom actually has nothing to do with odd analog circuitry, but is rather overcoming a mistake of treating the unsigned 16 bit ADC digital value of the ADC as a signed 16 bit value and allowing that to get sign-extended to 32 bits.

Correct code

The problems in the above code are resolved in the following code:

    err = amds_get_data(AMDS_PORT, AMDS_CH_6, &out);
    *adc_volts_out = 4.5/65536 * (double) (out & 0x0000FFFF); 
    //Note: "4.5" corresponds to the ADC Vref of 4.5V on Rev C current sensor cards.
    // Rev B and earlier current sensor cards use 5V, and therefore for these boards this line should read:
    // *adc_volts_out = 5.0/65536 * (double) (out & 0x0000FFFF); 

Note that this code:

1. Instructs C to keep only the lower 16 bits of the int out
2. Scales the digital value by the full range of the ADC as $\frac{V_\text{REF}}{\rm 0xFFFF}$