Measurements on the captured samples, with every cycle in the statistics: asked by 0xdeadbeef

What was asked

the MHO98 seems to either use the display data or at least decimated data instead of the raw data for automatic measurements which leads to aliasing errors in the measurements.

Time measurements (frequency, period, widths) come from the captured samples, never from decimated screen data, at any time base.

the MHO98 always performs only a single measurement per trigger even if the data displayed on the screen (and in memory) would enable thousands of measurement.

The statistics take every cycle in the measured range, not one value per trigger (up to 1000 per trigger).

The important point here is most probably to use adjacent measured data points instead of performing the measurements on decimated data.

Edges timed on adjacent samples at the full sample rate.

What NovaOS does now

Every cycle in the statistics

Period, frequency, widths, duty cycles, rise and fall times and slew rates put one value per cycle into the statistics, up to 1000 a trigger. On a DHO924S, a 975 kHz square from its own generator, one SINGLE counts 4 at 500 ns/div (the four periods on the screen), 102 at 500 us/div and 816 at 20 ms/div. Voltage items still count one per trigger. :NOVAos:MEASure:STATistic:CYCLes OFF, or Each cycle Off on the Measure sheet's Settings, counts one per trigger as the stock does.

Measured on the captured samples at any time base

The edges are timed on adjacent samples at the full sample rate, to a fraction of a sample, never on the screen's decimated points. The mean reads 975.000 kHz at 500 ns, 500 us and 20 ms/div, where the MHO98 read 970 and 847 kHz. A SINGLE always measures its own record.

Shown in the video above.

The statistics on the screen, to six figures

Analyse, Measure, Results shows each item's current, mean, maximum, minimum, deviation and count, the measurement strip's own items too, with times and frequencies to six figures: 975.000 kHz, 975.543 kHz maximum and 974.417 kHz minimum over 922 cycles.

Shown in the video above.

Where it stands

✓ DoneTimes from the captured samplesWhen the screen holds fewer points than the record, NovaOS reads up to 131,072 adjacent samples at the full sample rate from the middle of the screen and times every edge in them to a fraction of a sample. The screen trace (a min/max envelope of the record) gives only the voltage extremes. A 975 kHz square at 4 GSa/s reads 975 kHz to 10 ppm in the model, also where each screen point spans several periods.

✓ DoneEvery cycle in the statisticsPeriod, frequency, widths, duty cycles, rise and fall times and slew rates put one value per cycle into the statistics (up to 1000 a trigger), so a screen of four periods counts four, not one. :NOVAos:MEASure:STATistic:CYCLes OFF, or Each cycle Off on the Measure sheet's Settings, counts one per trigger as the stock does.

✓ DoneOnly the visible samples, and no double countingAt fast time bases the measured stretch is the screen's own samples (at 500 ns/div, the four or five periods you see), never the samples around it. The deep record is read again every 250 ms, and the triggers in between show its value without counting it a second time.

✓ DoneChecked on our scopesYour test on our MHO934 (G1 into CH2, 500 Mpts at 2 GSa/s): one trigger counts 4 at 500 ns/div and 63 at 500 us/div and at 20 ms/div, where it counted 1; the mean reads 974.93, 975.00 and 974.99 kHz. Recorded on our DHO924S (its generator into CH4, 50 Mpts): 4, 102 and 816 cycles per SINGLE, 975.000 kHz on the scope's own Results page, min and max showing the generator's own jitter.

✓ DoneUp to 25,000 cycles a trigger, from the whole screenAs on the SDS5000X (we read its firmware 1.1.5.3: the first AIM cycles from the left, 1000 by default): the most cycles a trigger adds is settable, 10 to 25,000 (Each cycle on the Measure sheet's Settings, :NOVAos:MEASure:STATistic:CYCLes:LIMit). On a deep record NovaOS also reads six stretches spread over the whole screen beside the middle one, so the cycles counted come from all of it, not from its middle.