DMM question?

That is single-period measurement. It has resolution problems for medium input frequencies, short period but not high enough frequency to cut over to classic gated counting.

The next step would be to time stamp all the edges and count how many edges were seen in 2 seconds. Take the delta-T of the first and last edge, divide by N, and invert to get frequency.

They need boosters to stand on, behind the lecturn. Like some short movie stars.

The waveform tends to be flat-topped from rectifiers. Less lately with more PFC supplies.

It's a reciprocal counter which measures the number of fast clocks in one input period, or maybe a universal counter, which measures M fast clock pulses in N input periods.

The frequency is then

f_in = f_clk * N/M.

With a fast enough clock, you can get any resolution you like.

Easy.

Cheers

Phil Hobbs

Improves display readability for varying or jittery sources.

Reading 88 on the last two digits wouldn't be much fun.

RL

Thanks, Bill.

=================================

The fraction of supply *current* due to rectifier loads is tiny compared to resistive loads. Even microwave ovens have a good waveform.

The AC wave does not improve at times of low use of electronic devices. Power generators ( 3 phase alternators) do not output pure sine waves, the top tends be flattened.

..... Phil

Subject to noise on the line, and in the zero-crossing detector.

Better accuracy (noise immunity) could be had by curve-fitting a sine to the input, and timing the zero-crossings of that fitted curve. There's probably a proper DSP term and theory for this but I don't know it.

Question: how much noise reduction can this gain?

CH

FFT is a good sinewave curve fit.

A lot.

What about using a PLL to generate the period window?

Yes, but far more computationally expensive than necessary. Five digits requires 2^17 points, and gives you all the harmonic data you don't need.

What's the (computationally) cheapest way to implement an accurate digital PLL?

CH

Some easy ways are PLL multiplication, and digital PLL (both of these only work on pure single-frequency input). Or, a digital period measure, of zero-cross events, with microsecond timing of multiple zero crossings. Period-to-frequency requires some smarts, though maybe only integer multiply and a bit of trial-and-error with successively smaller steps.

Well, yeah, but the easy way is to co-opt a music-box-like springy-tines resonator, with the tines of the fork set for resonances in the 2x40 to 2x 70 Hz range. Then with an electromagnet excited by the line voltage placed so it pulls on the tines (at double the AC power frequency) the user just glances at the assembly to see what the blurriest one reads on a nearby frequency scale.

Just one divide. Multiply and divide if averaging multiple cycles. Like PH said.

SNR is always a limit. Shannon's theorem says that the channel capacity of a noisy link is

C = BW * log_2(1+SNR),

So you can get 12 correct digits in 1 second (as my old EIP 578 microwave counter does routinely) as long as the SNR in 0.5 Hz is at least

SNR >= 10**12 (120 dB).

Curve-fitting works OK at high SNR, but falls apart at lower SNR because the least-squares approach overemphasizes outlying points, leading to a threshold effect.

A really narrowband PLL is a much better approach, because it relies on orthogonality, which is able to reject out-of-band contributions.

A PLL can win by the bandwidth ratio.

Cheers

Phil Hobbs

On a sunny day (Mon, 8 Mar 2021 17:08:49 -0800 (PST)) it happened whit3rd snipped-for-privacy@gmail.com wrote in snipped-for-privacy@googlegroups.com:

Ha, yes the rsonance meter, we had in in the lab.

Worked with a physicist who had a great deal of trouble believing it was possible to measure frequency like that. The reaction was like saying something didn't conserve energy, he never really believed it.

In a sense, he's right. The general signal (white noise, for instance) can't be analyzed by its zero-crossings alone, but most of the schemes noted here would do something, and give a frequency answer anyhow.

Spehro Pefhany wrote: =================

Met a CSIRO* electronics engineer at a gathering once and the topic of frequency measurement came up. I told him said I owned a 7 digit counter with pre-scaler to read signals up to 1GHz. Used it mainly for radio mic transmitters in the range 165MHz to 850MHz. Readings took about one second to appear.

He scoffed at the method and told me that is his CSIRO lab, they would count a frequency for several hours to get an accurate reading.

..... Phil

• formatting link