Er... Digital signals exist in the digital domain, by definition. There is nothing analogue about 0, 1, 0xdeadbeef etc.
Adding the time domain changes absolutely nothing of that.
Okaaaayyy... I'm a little busy packing a 454 for shipping to its new home, but of course I've built a Mod-5 using flip-flops. Normally need at least two transistors to make a discrete flip-flop, and then something for the reset, but I remember reading a Hackaday about a DTL circuit that uses a single transistor as a flip-flop, but as tested it was limited to under 1 MHz operation. Not sure if that's what you had in mind, or something more along the lines of "analog computer" design wherein "operations" are performed as a function of RC/LC,etc time constants and LF heterodyning.
The circuit bd139 showed indicated one way. I think (without verifying it) that the Tek 184 uses a different way. No flip flops involved, only "glugs" - to use the lovely and descriptive) term frequently found in high precision voltmeters.
I remembered seeing bd's circuit from a while back, but couldn't be arsed to dig it up.
In my defense, I'm a little preoccupied saying goodbye to an old friend that's been with me for decades, but I haven't treated very well of late.
The only difference between a PWM voltage and a serial comms signal is convention (what is the acceptable amplitude/duration) and the fact of sampling over a given time window. Both have to exist as analog signals first.
In that example, the 8-bit digital signal is transmitted by the modulation scheme you mention. The serial comms voltages are a first-order hold of the sampled digital signal.
There are, of course, many other ways of transmitting digital signals. A 64 bit digital signal could be transmitted on 64 separate frequencies e.g. COFDM, or as multiple amplitudes and phases e.g. 64QAM with added noise:
This concept is obviated by the fact that multi I/O pins on modern processors often treat both interchangeably; creating an alternate serial data channel by modulating a PWM out at given frequency and amplitude to send serial data, or sampling an ADC IN similarly to receive serial data.
That is indeed
one implementation technology that is around
at the moment.
There have been many over the years, e.g. one from near where I grew up:
Come on man; we're arguing over semantics & terminology here.
It is a deeper distinction than that, and philosophically quite interesting.
But it isn't (pure) TEA, and while it may deserve a separate thread, it isn't too relevant to a TEA thread. (Not that direct relevance has been too important in the past!).