Stepping motor torque ripple

Not to change the subject (never!) I've been experimenting with the jitter implications of PWM, namely how precise in time can a PWM system be? Better than the PWM period? How much better?

It's an interesting perversion of the Shannon/Nyquist sampling theorem. Maybe I'll post some sims.

PWM is waveform generation; you can synchronize a feature of that waveform with good precision, and the PWM period only determines the (at steady-state) repetition interval, not the features within that interval. Useful examples of PWM usually alter the features according to some criterion other than timing, i.e. output voltage under load.

What specific time precision is under consideration? Old IBM game controllers got pinged at a crystal-clock-synchronized interval, and gave 8-bit resolution.

Can you elaborate on that?

The PWM of a microcontroller is a combination of the clock jitter and the jitter of the PWM counter/compare capture unit, plus noise on the supply I guess

The abstract problem is to send a DC-coupled logic level over an AC-coupled telecom-type link. Imagine the link has AGC or logic repeaters so we can't send different amplitudes to code things.

One could lowpass filter the logic input, sample and digitize, send the sample values in serial packets, deserialize and DAC and filter and compare at the receiver. Pure Shannon Sampling Theorem with some ADC quantization. Slow and painful.

FSK would work but would be very slow and maybe jitterey. OOK would be really ugly over a telecom type link. PWM might work.

I'm thinking about PWM at a "carrier" pulse clock rate of, say, 1 GHz.

30% duty cycle is a 0, 70% is a 1. The logic level to be sent is not synchronized to the PWM clock.

It will certainly work, but what's the jitter like? Can it be less than the 1 ns PWM pulse period? I'm thinking it can.

It's just a thought experiment, with a bit of Spicing, right now.

tirsdag den 1. februar 2022 kl. 01.40.26 UTC+1 skrev John Larkin:

will your duty cycle survive the AC coupling on the link? isn't such things normally DC balanced with bit stuffing or similar?

Yes, telecom links assume balanced data, NRZ or 8b10b or something. That makes it hard to send DC!

I've tested the proposed link and it can propagate duty cycles, numbers like 30% and 70%; maybe 35 and 65 to be conservative.

Just now I can't think of a better modulation scheme, at least one that's practical to actually build.

The edges of a PWM pulse can be positioned with arbitrarily small time resolution, so sub-clock timing is possible.

You might think about the use of quadrature coding to send a 10kbit/sec data stream over a telephone line.

Those modems were quite popular when telephone pairs were our only way of getting our home computer hooked up to the internet.

On a sunny day (Mon, 31 Jan 2022 16:40:14 -0800) it happened John Larkin <jlarkin@highland_atwork_technology.com> wrote in snipped-for-privacy@4ax.com:

In video (that has a DC conponent reference) we used FM modulation for example to record to tape. Used it over phone line too for slow scan TV. That was before all went digital.

Does the link preserve polarity - after an arbitrary number of inversions 30% and 70% duty cycle may be indistinguisable. How about 50% as one state and 30/70 as the other?

piglet

Seismometers are too low frequency for direct tape recording, so they're always frequency-converted for field tape-recorder stations. The tape drives have a multiplicity of tracks, so one of 'em gets a timing reference, the others record data. It's done without a bias frequency, which saves precious battery power.