Programmable DC Electronic Load - First Look

I have made a little progress...

Received these parts which more or less match the 2nd schematic drawn above:

• Bourns PWR220T-35-1R00FBourns PWR220T-35-1R00F TO-220 style 1% 1 ohm resistors
• IRLZ34NPBFIRLZ34NPBF Infineon N-Ch MOSFET
• MCP4725A0T-E/CHMCP4725A0T-E/CH Microchip 12 Bit DAC
• OPA192IDBVROPA192IDBVR Texas Instruments Op Amp
• 10C035410C0354 Bourns 2 Ch Encoder

I also recently purchased an Adafruit Feather M4 Express which has the ATSAMD51J19 microcontroller with 12 bit ADCs and 12 BIT DACs so I did some experimentation with that.  The DAC was programmed in the Arduino IDE to output various voltages and then measured with via ADC (2 pins away) and a digital multimeter.  The first ADC reading was thrown out and then the next 5 averaged. The following plot shows the deviation of the programmed DAC output and ADC readings from the meter.

The DAC output in blue is very linear and could be easily corrected to a large extent in software.  There is a lot of scatter in the orange ADC readings.  One ADC reading at around 1800 mV is in error by approximately 35 mV which is outside what the datasheet gives as maximum error - not sure what caused that.  I seem to remember the MSP432 giving much better results and had a look at the TI datasheet.  It reports error differently but does seem to be both more accurate and precise.  It might be interesting to do a real world comparison to the Atmel chip.  It is a shame there isn't a MSP432 variant with a DAC.

The following plot shows the percent error for the ADC and DAC on the Feather M4 Express.  The title is a bit misleading - the X axis is the meter reading in mV and the Y axis is the percent error for the DAC in blue and the ADC in orange.

Percent error in the DAC would improve if the software correction described above was implemented.

In summary, DAC output seems good on the Atmel SAMD boards so far.  The ADC is probably only good for 20 mV or so.  This may be good enough for reading the input voltage of the DUT but not good enough for reading current.  For that, it would be better to just report the DAC setting.

The Bourns 1 ohm resistors were measured with the milliohm meter and are within 0.1% of 1 ohm.

Hello Frank,

Your ADC seems to be very noisy.

Could you try two experiments:

1) 1000 samples with maybe 50mV input (the 50mV must be low noise - ideally source from a battery and divider with a big low esr cap across it.

2) 1000 samples close to full scale

The first measurement will tell you about the ADC noise, the second will tell you if the ADC reference voltage is noisy

From your results so far it looks as if you have both problems.

How has the ADC been set up. (in terms of acqusition time, sampling speed averaging etc) ?

If you don't need to sample very fast you should be able to get a decent noise performance out of this chip by oversampling.

MK

The screenshot above was taken when the Feather was powered from a USB slot on an Ikea power strip.  If powered from my laptop (which supplied power during all the measurements above) we get the following which is cleaner than the Ikea supply:

Powered by battery over Feather LiPo battery port, same settings as above

Crummy Ikea USB Source:  175 mV Pk-Pk

Laptop USB Source:  69 mV Pk-Pk

LiPo Battery Power: 38 mV Pk-Pk

I still don't know if this is the cause though....

I need to think about this one.

The amazing thing is that the noise is way lower with the input at 3V than when at 100mV.

If the ADC noise were constant that would give you either the same std dev at both or possibly higher at 3v due to reference noise.

The other difference is that you are driving the input from the batteries (nice low source impedance) or the divider (2k2 source impedance.)

Well worth repeating the low voltage experiment but with a 10uF low esr cap// 100nF ceramic across the input as well.

Unfortunately the Arduino code hides most of the ADC set up.

Trying a low noise power supply and the ADC internal reference might be interesting as well.

MK

Does the ADC reference readings agains an internal reference source?  What about the MKR1000?  It's not possible from the code to see how the analog reads are being taken, and this may be a bit of a tangent/red herring, but I found that using the Arduino internal reference source improved accuracy significantly.  Is it possible for you to look at the library code and tell?  If the analogRead is the standard Arduino library code then it may benefit from something like this:

/*
 * Obtain the actual Vcc value of the 4Duino from an internal reference source.
 * The nominal 5V Vin can fluctuate a lot and affect analog read accuracy.
 */
long readVCC() {
  long result;
  // Read 1.1V reference against AVcc
  ADMUX = _BV(REFS0) | _BV(MUX4) | _BV(MUX3) | _BV(MUX2) | _BV(MUX1);
  delay(2); // Wait for Vref to settle
  ADCSRA |= _BV(ADSC); // Convert
  while (bit_is_set(ADCSRA, ADSC));
  result = ADCL;
  result |= ADCH << 8;
  result = 1126400L / result; // Calculate Vcc (in mV); 1126400 = 1.1*1024*1000
  return result;
}

/*
  Return an averaged reading from the Analog pin
*/
float ReadAnalog(int pin) {
  uint8_t i;
  float average;
  float samples[NUMSAMPLES];

  for (i = 0; i < NUMSAMPLES; i++) {
    int value = analogRead(pin);
    float supply = readVCC() / 1000.0; // obtain vcc at time of read
    float valueCorrected = supply / 5 * value; // improve the accuracy of the analog reading
    samples[i] = valueCorrected;
    delay(10); // spread readings out otherwise they are likely to be the same value or very close
  }
  average = 0;
  for (i = 0; i < NUMSAMPLES; i++) {
    average += samples[i];
  }
  return average / NUMSAMPLES;
}

The useful bit is the readVCC() function and lines 27, 28 and 29.  The 'valueCorrected' variable is determined from a 5V range.  I don't think this will eliminate noise (probably just read that more accurately!) but may firm up readings.  I've not tried this on a MKR1000 by the way.

It's an interesting project to follow along.

Hi Andrew,

I think you are right that the internal reference may give better results and have that on my list of things to explore.  I used the Adafruit SAMD51 board as opposed to MKR1000 in these latest experiments because it has a 12 bit DAC and I want 1 mA precision.  I ran the functions listed in the Arduino reference for Arduino SAMD boards, e.g. MKR1000, through the compiler for the Adafruit Feather Express SAMD51 and it appears to support the following for analog references to the ADC:

•   analogReference(AR_DEFAULT);  // This is what I am currently using and is shared with VDD
•   // analogReference(AR_INTERNAL);   // Not supported on Feather M4 Express
•   analogReference(AR_INTERNAL1V0);   // Internal 1.0 V reference
•   analogReference(AR_INTERNAL1V65);  // Internal 1.65 V reference
•   analogReference(AR_INTERNAL2V23);  // Internal 2.23 V reference
•   analogReference(AR_EXTERNAL);      // External reference applied to AREF pin

I may have to revert to direct register writes to the microcontroller to get the control necessary which is what I normally do with the TI MSP430 and MSP432 microcontrollers.  I have been using the Arduino IDE because I am not really familiar with Atmel SAMD microcontrollers and wanted to keep things simple.  That is a good idea until it isn't :-).

Edit:  see new comment to Michael above

Feeling like a dull boy... New trials were set up using the microcontroller's internal reference source and capacitance on the input voltage at 66 mV separately and together with slight advantage to adding capacitance.  Then the original setup at 66 mV with no capacitance and the analog reference set to the "default" to try and replicate yesterday's results was set up and almost identical good results were obtained.  In other words, I cannot replicate yesterday's noisy results.

Default analog reference, Capacitance added

Raw Avg:  90.0

Raw Median: 90

Std Dev: 0.9

Internal 2V23 analog reference, no capacitance

Raw Avg: 90

Raw Median: 90.0

Std Dev: 1.3

Internal 2V23 analog reference, with capacitance

Raw Avg: 90.1

Raw Median: 90

Std Dev: 1.0

Repeat original test, Default analog reference, no capacitance

Raw Avg:  90.3

Median: 90

Std Dev: 1.3 (yesterday was 11.5)

Looking back at the photo taken above the differences in the setup are limited to moving the resistor divider to fit the capacitors but something in the original tests gave noisy results at low input voltages that no longer seems to be there.   I am going to move off of the breadboard, solder something up, and repeat the experiments with care over the weekend.  The DAC also needs to be tested for noise.

Feeling like a dull boy... New trials were set up using the microcontroller's internal reference source and capacitance on the input voltage at 66 mV separately and together with slight advantage to adding capacitance.  Then the original setup at 66 mV with no capacitance and the analog reference set to the "default" to try and replicate yesterday's results was set up and almost identical good results were obtained.  In other words, I cannot replicate yesterday's noisy results.

Default analog reference, Capacitance added

Raw Avg:  90.0

Raw Median: 90

Std Dev: 0.9

Internal 2V23 analog reference, no capacitance

Raw Avg: 90

Raw Median: 90.0

Std Dev: 1.3

Internal 2V23 analog reference, with capacitance

Raw Avg: 90.1

Raw Median: 90

Std Dev: 1.0

Repeat original test, Default analog reference, no capacitance

Raw Avg:  90.3

Median: 90

Std Dev: 1.3 (yesterday was 11.5)

Looking back at the photo taken above the differences in the setup are limited to moving the resistor divider to fit the capacitors but something in the original tests gave noisy results at low input voltages that no longer seems to be there.   I am going to move off of the breadboard, solder something up, and repeat the experiments with care over the weekend.  The DAC also needs to be tested for noise.

Well, the good news is that the problem is fixable. I've not used pluggable breadboard for a very long time -  so I'm all for soldering.

Good luck.

MK