[SOLVED] How to detect micro volt changes in a voltage?

How to use a PIC ADC to detect micro volt changes in a voltage?

Using a PIC16F88 chip I can get a resolution of 5V/1024 = 5mV. Even using a 16 bit ADC will only give me say .08mV resolution.

So. How to get detect a relatively fast 10 uV change in the input voltage using the 16F88 chip?

2 ideas I had. What are their pros and cons?

1. Measure the input voltage V_in using the 10 bit ADC, then create two reference voltages V_in +- 1mV and apply them to the V_ref+ and V_ref- pins and then monitor the ADC value of V_in between these two reference values. That would give me 2uV resolution.

Will the PIC ADC work OK if V_ref+ - V_ref- = 2mV?

I thought that PWM would be too noisy to generate the reference voltages, but considered charging a low loss capacitor (controlled by the PIC chip) until its voltage reached V_in + 1mV, using that for the V_ref+ input (via a unity gain op amp buffer) and an op amp voltage subtractor to get the V_in - 1mV for the V_ref- pin.

The idea being that the V_in change would happen much quicker than the V_cap voltage decay which could be monitored and topped it up as required.

2. The other method was to charge a low loss capacitor until V_cap is close to V_in and then use a high gain op amp amplifier to monitor the voltage difference between V_cap and V_in for fast changes in V_in compared to the relatively slow changes in V_cap.

Again, as V_cap and V_in slowly drift apart, monitoring could pause and the capacitor could be charged or discharged accordingly.

Any better ways to detect micro volt changes in a voltage?

Thanks.

I feel solution 1 is not practical, the ADC won't behave with such a low Vref.
A different approach to solution 2 is just AC-couple the signal to a high gain amplifier, and possibly add another AC coupling at its output. Choose the time constant longer than the expected duration of the voltage change.

Some additional analog processing before going digital is appropriate. Why stand on your head digitally-speaking, when a simple analog solution is at hand?

One nice way to AC couple is to use a R + C filter on the signal, and present its output along with the signal to a differential amplifier. The latter could be a follower with a difference amplifier, or it could be an instrumentation amplifier. Such an approach insures that you'll not go beyond the common-mode range of an amplifier with large step inputs, as can easily happen with the standard simple C + R ac-coupling approach.

hello,


:-xPost #1 this soluce is impossible ! delta +VRef and -Vref must be > 1,5Volt
:smile:You can try MCP 3424 or 3421 ADC delta sigma converter up to 18 bits
i did a lot of test on it .. i get 2µV / point with 17 bits of resolution ( reference =2,048V)

see code C18 and results on **broken link removed**
You easily can adapt the code to your MCU

Thank you all for your responses. 18 bit resolution ADC is not an option with my programmer. Sigh.

What I am trying to do is monitor a voltage input (that may fluctuate slowly anywhere between 0.5V and 4.5V) for relatively abrupt changes of perhaps 10uV over a period of 0.2s.

Win_Hill, I can see what you are saying. If I have a 1 Hz high pass filter and the voltage goes from 2V to 2.00001 V in 0.2s, what will the output of the high pass filter look like? Would it be a pulse 0.4s long?

Likewise, if the voltage dropped from 2V to 1.99999 V would the output show a voltage pulse in the -ve direction?

I have only considered filter circuits to be useful for AC signals rather than a single "half" pulse (as in a full pulse would return to 2.0 V)