OT: a digital signals primer, as a 23min video

[BigEd]

This video was rather good, I thought, introducing some basics on how digital signals work, and aiming to clear up some common misconceptions. It's somewhat targeted at digital audio, but the same things surely apply to digital radio, or indeed to digital scopes.

Quote:

Xiph.Org's second video on digital media explores multiple facets of digital audio signals and how they really behave in the real world.

Demonstrations of sampling, quantization, bit-depth, and dither explore digital audio behavior on real audio equipment using both modern digital analysis and vintage analog bench equipment, just in case we can't trust those newfangled digital gizmos. You can download the source code for each demo and try it all for yourself!

(The presentation style is refreshingly simple and calm, and while it starts slow, there's plenty of stuff in there.)

There's an associated wiki too:
https://wiki.xiph.org/Videos/Digital_Show_and_Tell

[GARTHWILSON]

You're talking my language! Thanks. One of the many things I've used my workbench computer for was to vary the audio sampling rate on the fly, in realtime, to see how many samples per second we needed to record air traffic controller audio with good intelligibility, and to 100% interactively evaluate the effects of immediate changes. Same for sampling depth in bits, with and without dynamic compression, and experiment with anti-aliasing filters. (A recent post elsewhere questioned the 65816's computing power to do, in Forth, anything of the sort, not to mention the '02. Well, I've been doing it.) In a semi-medical project about five years ago, I used 8-bit sampling to produce pure tones that were previously thought to need 24-bit and a DSP. Initial experiments were done on the workbench computer, showing the results to the higher-ups and knocking their socks off, and then implementing the method on a PIC16F74.

The square-wave part was applicable to something that gets discussed here much more often. He showed that the 1kHz square wave was made up of frequencies up to, and way beyond, 20kHz. Similarly, a 1MHz square wave, will have a lot of harmonic content at even 20MHz and beyond. If you take those away, it's no longer square. It doesn't need to be perfectly square for the digital work; but the circuit's response to those higher frequencies can get you in trouble, and running a 20MHz part at 1MHz doesn't necessarily keep you out of trouble.

[barrym95838]

I had to drop my Signals and Systems class back about 30 years ago because my college course units were just too much in that particular semester, and the professor seemed to have no qualms in assigning a heavy (at least to me) homework load. My Computer Engineering degree ended up not requiring it anyway, so I never took the class again. I do find the subject fascinating, however, and I like the way Monty presents the material.

Mike B.

[BigEd]

I thought it was a nice touch to use an analogue spectrum analyser to inspect the outputs, as well as an analogue scope. I think it's extremely useful to be able to think in both the time domain and the frequency domain, but if you think of spectrum analysers as necessarily digital (using FFT for example) then you're one step removed from thinking of analogue signals as having a spectrum.

[BigEd]

In a similar vein, here's a nice short video explanation of Fourier Analysis, or the relation between the time domain and the frequency domain:

But what is the Fourier Transform? A visual introduction.