Delay-Line Memory Demo Board

In post #1, 'p-lab' wrote:

" This time I have built demonstration board that implements a Delay Line Memory, entirely out of TTL ICs. "

" This type of sequential volatile memory has been used on computers such as the UNIVAC-1 and the Olivetti Programma 101 and others. "

Uncle Bernie disagrees:

The UNIVAC-1 and Olivetti Programma 101 had   a c o u s t i c   delay lines which are a completely different animal than TTL based shift registers. Still, for people totally unfamiliar with shift registers, your demo board may help them to get some insight in bit serial memories. And no, these TTL shift registers are not  'volatile'.

The memory of the UNIVAC-1 was based on liquid mercury tanks, see the 8th picture with the funny device looking like a mechanical hedgehog on the UNIVAC-1 wikipedia page:

https://en.wikipedia.org/wiki/UNIVAC_I

.... while the Olivetti Programma used a spring steel wire wound into a spiral, see here:

https://ub.fnwi.uva.nl/computermuseum/p101dl.html

The closest "modern" equivalent would be one of the 1960's or 1970's music reverb units based on long spiral springs with acoustic transducers at the ends. In the 1970s, the PMOS "bucket brigade" ICs which could shift analog quantities through them started to replace these mechanical contraptions. Nowadays everything is digitized anyways, so finding these components is difficult.

With acoustic transducers for ultrasound, you could actually turn any "echo" chamber into a memory device for a computer.

I once experimented with turning a spool of coax wire into a serial delay line memory, but it turned out that this has limits caused by dispersion of the impulse trains sent in. This is not exotic at all - such a delay line is in most analog oscilloscopes (to make visible the signal before the trigger event happens) and wirebound communications channels (such as Ethernet and its faster offspring) can be treated as delay lines, and if you amplify and restore the impulses at one end of such a channel and send it back through another channel, you could use the copper cable infrastructure of your local telecom company as a serial storage device. You could also build a powerful modulated laser and bounce its beam back from the moon - and use that as as delay line for a memory. Oh, and last but not least, any 'analog' PAL color TV has an acoustic delay line based on a glass block. Which is able to store one analog color difference signal for one TV line. This could also be turned into a computer memory. You could use fast, high dynamic range DACs and ADCs to convert a multi bit digital word to an analog signal and back to digital.

The possibilities are endless. But believe me, 'real' digital memory ICs are much, much cheaper nowadays, so turning things into acoustic memories is all but a futile exercise.

- Uncle Bernie

In post #3, 'p-lab' wrote:

" And no, P101 memory was magnetostrictive, not just  a c o u s t i c ... "

Uncle Bernie responds:

The actual delay line effect used in the P101 is acoustic: it's propagation of torsional stress waves in the spring steel wire. The only "magnetostrictive" effect is at the begin and end of the delay line, in the things called 'transducers' which convert electrical pulses to these torsional stresses in the wire (begin) and the arriving stresses back to electrical pulses (end).

Propagation of mechanical stress waves in any medium qualify as  an   a c o u s t i c  effect in any case.

Just as a side note, why did they use such a complicated effect like torsional stress waves ? I think the answer lies with the sensitivity to environmental effects of any other type of acoustic wave. For instance, the typing on the keyboard or running the printer of the Olivetti Programma 101 makes a lot of mechanical and acoustic shock waves and these will find their way to the delay line spiral, despite it is suspended by dampers. However, the "magnetostrictive" transducers only react to torsional domains and ignore all other kinds of mechnical stresses. This clever trick makes the delay line memory much more robust against said interferers. As a counterexample, if you (gently) tap on the housing of a spiral spring based reverb unit, you can"hear" this tapping in the audio.  Which is bad. These are not robust against such interferers and would not work in the Olivetti Programma 101.

Oh,  and I think the use of complicated words like "magnetostrictive" only confuses most people and puts them off. Nobody wants to be doused in  technobabble. If you want to demonstrate and explain technical things to nontechnical people, keep wording as simple as possible.

As for you video, yesterday I had no opportunity to watch it, due to lousy internet speed, but today my internet access was better, and I watched your video, it was fun to watch, and very well made, and certainly helpful to educate people about serial memories (I recognized this educational motive as mentioned in post #2).

As for the "volatile", I must admit I used the wrong word, what I meant with "volatile" in this context was "need for refresh of the data" which TTL does not need, but MOS dynamic logic (and the analog bucket brigades) and all other forms of delay line memories with loss effects in the medium need such a refresh (or restoration) of the data. But I don't know the english word for "need for refresh of the data", and "volatile" seemed to be fitting. Despite in the stricter sense, TTL is indeed volatile when the power is turned off. Same thing with all these delay line memories. Magnetic core memory is an exception, it will keep its contents when power is turned off, although the circuitry must be designed such that collapsing and upcoming power supplies do not cause spurious write pulses within the core memory matrix.

Enough nit-picking and indulging into details. I'd recommend readers just to watch the video. It explains the principle of serial memory very nicely.

But here is a final thought based on human psychology and how the human brain works: by using TTL shift registers you have moved the "magical mystery" focus which mind-boggles the layperson away from a mechanical device (the acoustic delay line) to the even "more magical mystery" of solid state electronics, little black plastic boxes with many electrical pins coming out. But what goes on inside ? How come that this works at all ? --- I've been pondering for years over the yet unsolved problem how to explain electronic components to laypersons, but found no good solution yet. Which is a pity, because there are only very few IC designers in the world who can work on the transistor / physical layout level, and most are old (like me) and soon will die off. Most of the folks who call themselves "IC designers" actually only are "coders" who write Verilog or VHDL. Sure, that's also a very much needed profession. But what the design kit gives them are ready made standard cells and automatic tools to tansform their RTL to a circuit using these standard cells. Which were made by the few IC designers in the world who can work on the transistor / physical layout level (oh yes there are CAD tools which make these cells automatically, but this method has limits and human brainwork is still needed). So who (of the young people) will grow up into being able to sustain and further develop this technology ? When I look at all the billions of people living their lives with their noses stuck into a "Smartphone" who are completely and utterly oblivious to the technology which makes these possible, and which are totally disinterested into pursuing a STEM career which sustains and develops this technology, I get worried. So any attempt to lure people into our field of expertise must be welcome, and if your video catches a few "fish" then it was a good job. 

- Uncle Bernie