Voltage reference for MCP3202 ADC

Don M

I am thinking of using a voltage reference (3.0 VDC or 3.3 VDC) for the 12 bit MCP3202 for stability. Looking at TI/National LM4132 series.

My question is does the voltage reference (value) make any difference?

I will be using both channels of the ADC- one measuring 0 - 40 VDC and the other 0 - 5 VDC using voltage dividers on both.

Is the resolution the same regardless of what my reference is since I'm using dividers anyway?

40 / 4095 (9.8 mV / bit) for channel 1 and 5 / 4095 (1.2 mV / bit) for channel 2.

Mark_T

You seem to want high accuracy...

You need to worry more about the accuracy of your resistors in the dividers and the input resistance of the ADC - if not extremely
large it will affect the divide ratio and you'll have to consider a buffer stage with a JFET or CMOS opamp to drive the DAC.

This sort of thing is called working with an error budget - you have to add every source on inaccuracy together to see the worst
case total error and see what spec each part has to be. For 1% accuracy you'll probably need 0.1% resistors, a 0.5% reference,
and careful attention to grounding.

The relative performance of the ADC at 3.3V or 2.048V Vref won't be the largest problem I suspect.

frank freedman

The input impedance of the mcp32xx seems fairly low at about 1k. Probably want to buffer the ADC inputs so they don't load the sampled system. By buffering the sampled valued via an op amp, you could then scale the inputs cover the full range of a single reference voltage for multiple ADC input ranges. Since the input will then generate the full range of 0-4096, you need only tell the software what the volts/bit should be to convert the binsry count to actual value.

FF

Duane C. Johnson

Technically the low source impedance requirement of 1KΩ is only for the fastest sampling rates of about 50K/S.
If, say, you only need 5K/S rates the source impedance would only need to be 10KΩ.

There is a trick though.
The source impedance of the voltage divider doesn't have to be only done with resistors.

The Capacitors form the primary high speed Voltage Divider with the Resistors providing the lower speed DC Voltage Divider. But the DC current draw is significantly lower.

This technique is not all for free. The impedance ratios of both the capacitors and resistors must match.
Further, the capacitors must be high quality and accurate over the temperature range. Read this as ceramic won't work.
Op amp buffers are no a cake walk either at 50K/S to 12bit accuracy.

Ultimately, if you can afford the current load caused by 1KΩ resister dividers that's the way to go.

Duane J

kwinn

Don, it really comes down to a trade off between the accuracy you need, the power supply voltages available, the choice of precision resistors available, and the kind of math (integer or floating point) you want to do.

If the ADC is to be powered from the 3.3V supply then you are limited to a reference no higher than 3.3V. If it runs off 5V you could also use a 4.096V reference.

If you want to keep it simple (no floating point math) I would suggest a 3V reference.
A 10K/702 ohm resistor divider for the 40V range, and a 904 ohm/1K for the 5V range would result in simple shifts, adds, conversion to decimal, and placing the decimal point in the correct spot to output the voltage. Two 22 turn 1K pots could be used in place of the 702 and 904 ohm resistors.

Of course with floating point math the resistor choice is much simpler.