Sigma Delta ADC for AC signals ?

Alex.Stanfield

Has anyone encountered a clever way to do sigma delta ADC on AC signals ?

I plan to measure mains AC current with a current transformer (50Hz in my country) and sigma delta. I originally planned to have an absolute value converter (aka: full wave rectifier) since the signal will have negative values. But then maybe someone already figured it out in a simpler way.

Thanks for your input.

Alex

Mike Green

One way or another, you have to shift the signal voltage to match what the ADC is capable of. In the case of the ADC using the Propeller's I/O pins, the voltage has to be positive. You could use an op-amp operating off positive and negative supply voltages to buffer and shift the voltage to be measured so it's above zero. That would be better than a full-wave rectifier, particularly close to zero.

Here's a tutorial on the concept.

jmg

SIgma Delta are charge balancing, and collects a rolling sum from that balancing.
To measure AC, you would just need to zero-cross detect the AC, and alternate the +/- totals that each half cycle gives.
ie synchronise the result-reading to the half cycles.

Exact phase precision of the Zero cross is not critical, as the residual will be fixed as a scale factor.
Some noise filtering on the Zero cross would be a good idea, along with a Schmitt

In theory, you could run a faster differencing check on the ADC, and derive a phase-lock from that, but that's a lot more SW work.
eg check at ~ 10KHz, and find the Min and Max to derive phase-info for when to sample the slower ADC total.

jmg

Mike Green wrote: »

One way or another, you have to shift the signal voltage to match what the ADC is capable of. In the case of the ADC using the Propeller's I/O pins, the voltage has to be positive. You could use an op-amp operating off positive and negative supply voltages to buffer and shift the voltage to be measured so it's above zero. That would be better than a full-wave rectifier, particularly close to zero.

If the current transformer has reasonable output (some volts) then it can be AC coupled, (with phase sync'd reading as described above) and the Transformer 'gnd' leg can be lifted to close to the ADC threshold, to reduce the Power-On-charge-time-effect of the coupling cap.

jmg

Thinking some more about this

jmg wrote: »

In theory, you could run a faster differencing check on the ADC, and derive a phase-lock from that, but that's a lot more SW work.
eg check at ~ 10KHz, and find the Min and Max to derive phase-info for when to sample the slower ADC total.

Using the Current Transformer in Differential mode, with two ADCs, I think there is a simpler polling option.
In SDM, Zero scale gives a 50% Adder rate so ( PHSA ~ (CNT SHR 1)
-ve IP peaks will see PHSA gain on (CNT SHR 1), and +ve IP peaks will see PHSA fall behind (CNT SHR 1)
you could just poll the (PHSA - PHSB) difference for peaks.

This self-sync's off the AC-modulated nature of the charge sum, and avoids any hard timing at all. Would work on 60Hz & 50Hz.

Alex.Stanfield

Mike Green wrote: »

One way or another, you have to shift the signal voltage to match what the ADC is capable of. In the case of the ADC using the Propeller's I/O pins, the voltage has to be positive. You could use an op-amp operating off positive and negative supply voltages to buffer and shift the voltage to be measured so it's above zero. That would be better than a full-wave rectifier, particularly close to zero.

Here's a tutorial on the concept.

Yes, maybe I'll do the absolute value conversion anyway since just shifting the ground level will impose more work on the SW side. Also by doing an absolute value you double the resolution.

I found this circuit which is simpler and allegedly has better linearity near zero and can also be used with a single power supply. Will give it a try.



The complete article is here http://www.analog.com/library/analogDialogue/archives/44-04/absolute.html
BTW: It's funny it's on the same site as your reference..

Alex

Alex.Stanfield

@jmg

Thanks for you input, as you say it could be done with differential simultaneous ADCs, however the complexity added on the SW side will impose more hours working on the measuring instead of working on the "judging" of the signal which is what I want after all, so it seems wiser to just drop in the absolute value circuit and focus on the interpretation.

Thanks to all

Alex