[time-nuts] Difference in antennas

[Gerhard Hoffmann]

Am 22.11.19 um 23:29 schrieb Hal Murray:
> jra at febo.com said:
>> I like the idea of inserting attenuation until the SNR or Cn values start to
>> go down.  That may be the most practical solution.
> Inserting attenuation is a good trick for the tool box.  It is also used to
> measure error rates on fiber links.
>
> With a reasonable fiber setup, the error rate is so low that it is hard to
> measure.  At low error rates, there is a simple relation between signal/noise
> and error rate.  So insert enough attenuation until you can easily measure the
> error rate, then compute what it would be without the attenuation.
>
Unfortunately, in real life it's not so simple. There are more 
contributors than just SNR. In fact, when we were building 10 GBPS XFP 
fiber optic tranceivers, we got no bit errors at all. Just in one test, 
there was 1 failing bit per day or so. It turned out that the polynomial 
used to generate the data for that test had unusual long and unluckily 
grouped runlengths of the same polarity.

Someone had decided that 10 nF coupling capacitors on the differential 
CML lines were large enough for 10 GBPS and they did fit better onto the 
microstrips than the usual 100nF ones.

That did cost us a lot of time and money.

regards, Gerhard