Please critique this design concept

[PartialDischarge]

I'm finishing a design to sample the voltage at the mains input, send it via a fiber optic and reconstruct it in a scaled down version.
It is made up of 2 modules, I've posted a simplified diagram of what I intend to do.

The first module converts the 230v AC mains into 5V DC. I don't want/need isolation so I'll use a linear regulator (described as "black box") to do this. Then I create a 2.5V virtual ground as a reference to the scaled down voltage. I use a resistive voltage divider, again no isolation, to feed and op-amp, filter and a 10 bit  ADC.

Data is transmitted via fiber optic to a 2nd module that reconstructs the signal, removes the DC component with a capacitor and feeds the signal into a power amplifier driven by symmetrical rails. This 2nd module is of course powered from a different place.

Forget the linear regulator and the resistive voltage divider, ie don't suggest a switching PS, as I'd prefer those to be like that.

What problems do you foresee?
Common mode at the input?
I'm worried about the 2.5V virtual ground and how it would affect the reconstruction, as I need high precision. More specifically maybe I should include another 2.5V reference in the second module and subtract this component with analog circuitry. In this latter case however any mismatch in voltage between the references will create a DC component.

 

[PartialDischarge]

Any reason not using AMC1200/1301?

Yes, both systems would be meters apart from each other, so I'd prefer to use optical Tx/Rx, instead of using cables for the signals + the need to supply the isolated side of the AMC1200. Although it would be a solution that maintains isolation for this and others reasons I'd prefer not to use isolated op-amps

[orolo]

Excuse me if I'm wrong, but if you use full bridge rectification, won't your 0V DC ground be floating with respecto to neutral and live? Half-wave rectification would be perhaps preferable in this case.

[TheDane]

Measuring mains voltage isn't easy, and it's a harsh environment. (Surges, voltage spikes, etc.)

I would recommend an optical isolator/optocoupler like HCPL-0500. https://www.broadcom.com/products/optocouplers/industrial-plastic/digital-optocouplers/1mbd/hcpl-0500
It is possible to get a scaled voltage output, relative to input voltage.

If you have newer worked with an optocoupler, do read the app note:
AN 9512 Linear Applications Optocouplers - https://docs.broadcom.com/docs/5954-8430E
(it's on the documentations page - https://www.broadcom.com/products/optocouplers/industrial-plastic/digital-optocouplers/1mbd/hcpl-0500#documentation)

More specs would be usefull, like what is your need ~ sampling rates/frequencies. (eg, are you just measuring min/max/rms voltage, or are you trying to get a noise figure, voltage drop-outs. Also be aware that you might want to do a 0-crossing detection circuit, so you know when the AC line shifts polarity, otherwise your measurement could be quite noisy depending on how you sample and process it)

There are also new RF capable processors on the marked, capable of doing live mains interface. But since you state you want cable, I won't go into that 

[Kleinstein]

The power supply with full wave rectifier will not work properly. The more obvious solution would be one with a capacitor as a dropper and than half way rectification. With a dual half way rectification one could directly get +-2,5 V - so not more DC offset needed. Otherwise the concept could work.

[PartialDischarge]

Orolo and Kleinstein, you're right, I can't measure the full wave of the ac using a rectifier like that, half wave then.

[David Hess]

What bandwidth do you need?  Why not implement a voltage to frequency converter to drive the optical link one way?  Or self time the ADC to do the same thing.

[PartialDischarge]

What bandwidth do you need?  Why not implement a voltage to frequency converter to drive the optical link one way?  Or self time the ADC to do the same thing.

I need a few Khz. The CLK is drawn there conceptually, if the adc were to be self timed then I'd need another uP since CLK should not be continuous but a high speed burst on the sampling points.
I'll think about the VF converter anyways.

[David Hess]

What bandwidth do you need?  Why not implement a voltage to frequency converter to drive the optical link one way?  Or self time the ADC to do the same thing.

I need a few Khz. The CLK is drawn there conceptually, if the adc were to be self timed then I'd need another uP since CLK should not be continuous but a high speed burst on the sampling points.
I'll think about the VF converter anyways.

With that low of a bandwidth, I think a voltage to frequency converter would be ideal for this.  The input voltage range can be offset to handle negative values and the output frequency never approaches zero.

The alternative I would consider is self timing an ADC and sending the data either asynchronously which is easy to decode or with an inexpensive and tiny microcontroller on each end and some encoding which includes the clock.  In either case there is no need for a separate clock.