Dispersion in Coax

n to you.

very long coax cable and look at the output of the cable with an oscillosc ope to see that the high frequencies are traveling faster than the low freq uencies. Anyone in Tapper land do an experiment like this or a similar var iation?

But I wondered if there was some other transmission line

g all the time. 'Scopes were EG&G specials with rise times in the 10s of p s and frequency responses out past 40 GHz. Transmission lines were long (s everal 100 feet) runs of 3/4" and 1" cables (e.g. RG-211, RG-219) with mass ive equalization for pulse signals. There was always a precursor before th e main signal rise.

A long time ago (1974) I worked on a prototype digital television studio distribution system at BBC Research Labs. The video was sampled at 8 bits/primary colour at 3 x colour subcarrier, about

13.3MHz. Each bit was transmitted on a separate twisted pair of a telephone cable along with the 13.3MHz sampling clock on its own pair. Multi-pair telephone cable was chosen because of the huge quantities already installed in studios.

The distance between repeaters was I think 50m. The system was never reliable because with some picture content there were "sparkles". The worst picture was a horizontal grey scale going smoothly from black at one side of the screen to white at the other. Half way across the screen there would be a vertical line of occasional errors. No amount of tweaking of clock delays would fix the problem for all picture types. The cause, which was easy to see on a 'scope was that low frequency square waves (most significant bit of a grey scale, so one cycle per horizontal scan) arrived with a different delay from high frequency square waves (least significant bit). The delay variation was enough to ensure that the available timing window for latching the parallel data bits was sometimes violated. The project was abandoned and distribution was done in a different way.

Sending a high frequency clock (of 10 or 20 MHz) alongside a variable-frequency square wave divided down from it, both having their edges synchronised at the sending end differentially along a length of telephone wire would be enough to demonstrate dispersion on an oscilloscope.

John

possibly dielectric absorbtion.

--
  When I tried casting out nines I made a hash of it.

There's mode distortion in twisted pair as there is in microstrip, because some of the field is confined to the dielectric and some is in air. The mode shape depends on the ratio of the conductor separation to the wavelength. It shouldn't amount to much at 13 MHz, but might build up over 50 metres.

It might also have been crosstalk, i.e. signal coupling back and forth between lines. IIRC phone cable didn't use different twist pitches the way Ethernet cable does. A pair of coupled lines with the same pitch will eventually couple 100% of the signal into the adjacent line, and then 100% back again, with the amplitude following a cosine. That certainly could matter over 50 metres.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
https://hobbs-eo.com

I am fairly certain that crosstalk was not a factor - I did a lot of measurements under controlled conditions and I would have seen crosstalk. The dielectric constant of PVC insulation (which was used on telephone wire in that era) does vary significantly with frequency over the relevant range of 15.6kHz to 13MHz. The exact characteristics seem to be strongly dependent on the plasticiser formulation and processing parameters.

John

Running multiple NRZ streams with a separate clock stream is asking for trouble. The propagation delay of each channel needs to be very well controlled.

A better solution is using some self clocking code (such as Manchester) independently on each data channel. There may be some variations between channels, but the variations are easily handled.

A self clocking waveform will increase the highest frequencies present (and hence losses), but on the other hand, the lowest frequency is also up in the megahertz range, reducing dispersion issues as well as simplifying transformer coupling. A similar case was 1/2 inch 9 channel computer magnetic tape. To densities up to 800 BPI used NRZ data, with constraint alignment problems when reading tapes written on different machines. Newer systems with 1600 BPI and up, used independent self clocking waveforms without problems with different machines, with much higher densities.

That distance sounds very short. The first video link from London to the channel cost in the late 1930's used telephone wires with repeter/equalizers every kilometer or so. The TV-standard was system A with 405 lines and nominally 3 MHz video baseband.

For 13.5 MHz sample rate, the highest NRZ frequency for ..0101010.. sequence is Fs/2 square wave. To receive something looking even closely like a square wave, some third harmonic at 3Fs/2 or 20 MHz is needed.

I totally agree. There were coaxial delay lines that had to be trimmed to get the clock synchronisation right.

Yes

The distance requirement for the prototype was governed by the locations of two labs that needed to be connected. The designer thought that this would be a simple project for me (a trainee) to implement and test. It was all hand-wired schottky TTL. Parallel data plus clock was already being used for communication between devices over short distances in the labs without any problems, and initially this had seemed like a trivial extension of that approach. It turned out to be a very educational project.

John