Crosstalk (Noise) in Digital Systems
This
letter to a French speaker(?) in Canada refers to my 1967 paper; Ivor Catt,
IEEE Trans. Com. Vol. EC-16 No. 6, dec1967, pp743-63.
Je viens d’ajouter deux et deux at je suis
arrive a quatre.
Je vois que peut-etre il y a un peu du
Francais atour de vous; votre nom, et votre pays.
I number of factors come together. You have
French connections. You said you were involved in Electricity Board power
lines.
I will take you through the way I came upon
the idea of two velocities of propagation.
In
Motorola Phoenix Arizona in 1964 I investigated the problem of interconnecting
high speed (1 nsec) logic gates. To avoid interconnection delays, the system
had to be reduced in size, leading to a 13 layer printed circuit mother board.
Alternate layers had signal lines and voltage planes.
I
decided that one key problem was to determine the amount of crosstalk between
two parallel lines, either surface or buried. First, I found a flat topped
pulse getting onto the other line, which disproved the idea that dv/dt or di/dt
created the noise. However, there was a further puzzle; a big spike at the end
of the flat topped noise pulse.
My
mentor Ken C Johnson, whom I had left when I moved from Ferranti in Manchester
England, happened to run in to me in the corridor of Motorola Phoenix. I grabbed
him, and he told me there were two modes involved. He then shot back to
England. However, he got them wrong. If the active and passive lines are A and
P, and their mirrors (below the voltage plane, which acts as a mirror) are B
and Q, he said the modes were, firstly, between A and P on the one side to B
and Q on the other; and the other mode was between A and B on the one side, and
P and Q on the other. This second mode he gave incorrectly. However, he gave me
the important clue, to look for two modes. (See pp32, 33, figs. 39 and 40 in my
1995 book which you bought from me, for the correct two modes.)
In
due course I worked out the correct two modes, A and P to B and Q (Even Mode)
and A and Q to B and P (Odd Mode). I proved my theory with high speed pulses
and photography, see my 1967 paper.
Before
Ken turned up, I read Jarvis, who had brewed up the correct equations. However,
then, to “simplify” the problem, he removed certain “minor” terms. This was a
disaster, because it unbalanced the equations and prevented him from getting
the correct result. In fact, his resulting formula was ridiculous, with noise
increasing more and more until it could be larger than the incident signal. I
said either he was wrong, or we had an amplifier!
I
worked out my own equations, and solved them, see p30 of my 1995 book, now at www.ivorcatt.com/em.htm .
Jarvis
referenced Cotte, and after I had worked it all out (as I remember) I got two
articles by the Frenchman Cotte, dated 1947 and 1954. He had the equations, but
upside down, and he did not resolve them as I did (see p30 of my 1995 book). I
am writing all this from memory, because the Cotte papers are somewhere in the
roof here. I am certain he did not “solve” the equations, and that he could not
do so because, as with Jarvis, they were upside down.
Some
time later I discovered that in Microwave, the two modes were known. However,
working in digital electronics, knowledge in the Microwave camp was not
available to me. I reference Oliver as my 4th reference at the end
of my 1967 paper. I reference Cotte, who wrote in French, at my 5th
and 17th references. (I spoke French at that time, and do so now.)
As
I remember, Cotte worked on overhead power lines.
I
wrote this because (1) you have a French name and are in Canada; (2) Cotte, who
wrote in French, was involved in the problem of crosstalk early on; (3) my
memory says that, like you, he was involved in overhead power lines.
It
is likely that you are not concerned with crosstalk, but merely with transients
(switch on/off of lightening) in overhead power lines. However, since Cotte got
into crosstalk, and (as I remember) worked on overhead power lines, then
crosstalk in that situation may be of importance.
Ivor
Catt 25jan02