Undergrad How was electronics propagation delay measured historically?

[Muhammad Usman]

TL;DR

How electronics propagation delay was measured historically ?

Hi,

I know in todays world we have more sophisticated equipment to measure the propagation delay of electronics equipment, but for my own curiosity i want to know how historically the propagation delay was measured for electronics device such as transistor.

 

[Muhammad Usman]

TL;DR Summary: How electronics propagation delay was measured historically ?

I know in todays world we have more sophisticated equipment to measure the propagation delay of electronics equipment, but for my own curiosity i want to know how historically the propagation delay was measured for electronics device such as transistor.

or in more simple words how pico seconds propagation delay is measured or experimentally verified or experimentally discovered in early days of electronics

 

[Baluncore]

or in more simple words how pico seconds propagation delay is measured or experimentally verified or experimentally discovered in early days of electronics

The propagation delay of an inverter, could be measured by making a phase-shift oscillator from a ring of three inverters, the frequency of oscillation is related to the speed of the propagation delay through the inverters.

Microseconds, and then nanoseconds, were measured using the length of transmission lines. I do not know of individual picoseconds being measured historically, although it can now be done using non-linear transmission lines, NLTLs, and sampling circuits.
https://www.mwrf.com/technologies/c...2972/nltls-push-sampler-products-past-100-ghz

In one nanosecond, light can travel 300 mm, one foot. In the 1970s, time domain reflectometry, TDR, using sampling oscilloscopes, could resolve about 10 mm which represents 30 ps.

 

[sophiecentaur]

It was only recently (in RF history) that time domain measurements were possible and pulses have the disadvantage that they are only there once per repetition and huge bandwidths are involved. Phase was much more convenient to measure, earlier on. There can be a problem with knowing how many whole cycles were involved but using different 'probe' frequencies would give a 'slope' for working out the gross timing.

There's a whole history of how we measured things in the past, according to the technology available. You'd be daft not to use Time Domain for most applications, these days.

 

[tech99]

I suppose the earliest delay measurements were made in the late 19th Century on the trans Atlantic telegraph cable, using the ink printer called the syphon recorder, used as a simple mechanical oscillograph.. This led to the understanding of transmission lines (Heaviside) and the invention of inductive loading. Measurements of sound delays were made around WW1 using good mechanical oscilloscopes, such as the Enthoven String type. When HF radio was discovered in the 1920s, the delays and multipath effects of ionospheric propagation were studied using analog fax machines. By 1930, practical cathode ray tunes became available and simple oscilloscopes were possible. However, due to the cost and difficulty, the TV pioneers often worked without an oscilloscope, instead observing effects on the TV picture. By WW2 and the introduction of practicable radar, CRT displays for observing pulses and time delays became widespread, but only working up to a few MHz. With the introduction of digital techniques in the 1950s the interest in short delays increased and modern oscilloscopes became widespread, gradually getting faster and becoming digital themselves towards 2000.

 

[phyzguy]

The propagation delay of an inverter, could be measured by making a phase-shift oscillator from a ring of three inverters, the frequency of oscillation is related to the speed of the propagation delay through the inverters.

It doesn't have to be just three inverters. Any odd number will do, and the more inverters, the easier it is to measure the delay. When I used this technique to measure propagation delay in integrated circuits, we would typically use on the order of 100 inverters. This is called a ring oscillator. If there are an odd number of inverters, the ring oscillator will free oscillate, and the period of oscillation is just the propagation delay of one inverter times 2 times the number of inverters. So even if the propagation delay is 1 ns, the frequency of the ring oscillator for say 101 inverters in the chain is 202ns. This is about 5 MHz, which is easy to measure.

 

[sophiecentaur]

So even if the propagation delay is 1 ns, the frequency of the ring oscillator for say 101 inverters in the chain is 202ns.

1ns in 202ns is tight spec for resolution unless the rest of the circuit is pretty steady. But there many ways to skin a cat and we all have designed or used them.

You have to take your hat off to the early workers who couldn't rely on any of the components being what the label said. Everything needed to be checked again and again.

 

[Baluncore]

It doesn't have to be just three inverters. Any odd number will do, and the more inverters, the easier it is to measure the delay.

You appear to be ignoring the harmonic possibility of there being more than one pulse travelling in the inverter ring at the one time.

 

[sophiecentaur]

You appear to be ignoring the harmonic possibility of there being more than one pulse travelling in the inverter ring at the one time.

You can resolve multiple values by a range of frequencies.

 

[phyzguy]

You appear to be ignoring the harmonic possibility of there being more than one pulse travelling in the inverter ring at the one time.

That may be true in theory, but in practice we routinely used ring oscillators with large numbers of stages as production monitors to monitor the propagation delay. They virtually always oscillated at the fundamental frequency. It's possible to design a power-up mode to force it to have only one pulse in the chain, but we typically didn't find that to be necessary.

 

[Baluncore]

That may be true in theory, but in practice we routinely used ring oscillators with large numbers of stages as production monitors to monitor the propagation delay.

That may be in IC times, but when vacuum tube and transistor gate circuits were first being developed, the cost of three devices was very significant. The idea that you would need to build more than three to estimate or verify the propagation delay of a circuit could not be entertained.

The measurement of microwave frequencies has always been ahead of the speed of switching logic gates. You do not need accuracy when estimating propagation delay, so two or three digit accuracy will do.

Lecher lines met that requirement, and could measure wavelengths down to a couple of centimetres, (67 ps; 15 GHz), before the invention of transistors. That is how the wavelengths of magnetrons, for radar, were measured during WWII.
https://en.wikipedia.org/wiki/Lecher_line

 

[Nik_2213]

The 'inverter stages in series' thing...
Back when transistors were germanium and had inch-long legs, I had a bunch of S-Decs, almost the first solderless breadboards to use internal rather than external clips...
Flip-flops were fun, a two-stage non-complementary RC multi-vibrator just about fit on one S-Dec.
Three stages did a 'flap-flap-flop'. Four, despite heroic differences in RC values, always gave two flip-flop pairs in counter-point / push-pull. Five gave a 'four' plus a 'runt pulse'. Six, three pairs. Beyond that, I ran out of S-Decs, OC-71s and mini-bulb driving OC-72s...