[time-nuts] OT: AC voltage standard

[Bruce Griffiths]

Didier Juges wrote:
> The problem with a mercury relay is that the switching delay is significant and not well controlled, so the duty cycle of the resulting waveform is not well controlled, and so would be the RMS value.
>
> I believe CMOS analog switches would provide better control, and with series resistance that is easily below 10 ohm, that would give you negligible error when driving loads in the megohm, such as a voltmeter.
>
> Four switches in a full bridge configuration will give you a true AC square wave output, and if you know the DC voltage feeding the bridge (using your voltmeter calibrated with the Weston cell), you will have an accurate AC source that will not require further calibration, at least good enough for most home lab uses. 
>
> Now, for a sinewave, it's another matter, but Bruce's suggestion of a DAC powered from a precise DC source would work extremely well (limited by the DAC) and provide a low distortion sinewave, which is just as important as controlling the peak voltage. A simple microcontroller is all that's required to drive the DAC. Make sure you understand the delays involved with making software loops. Alternately, a counter driving a suitably programmed EPROM driving the DAC will take software out of the equation, but it sounds like the 70's all over again...
>
> Didier KO4BB
>   
One can always adapt the techniques used in NIST's quantum AC standard.
This device is basically a single sigma delta DAC that turns a very
accurate and stable voltage source on and off at 10GHz or so.
The modulator output is then low pass filtered.
To generate a stable AC voltage, at least for low frequencies, use CMOS
switches controlled by a sigma delta bitstream to switch the input to a
low pass filter between ground and the stable voltage.
AC couple the output to eliminate the dc offset.
Using a single bit DAC simplifies the calibration process over that when
using a multibit DAC.
NIST's AC standard is currently useful for generating frequencies up to
100kHz with 10MHz the projected useful limit for a 10Gb/s bit stream.
With say a 1Mb/s bit stream output frequencies up to 1KHz or so should
be feasible with high amplitude stability.

Bruce