[RFI] ethernet EMI revisited - questions for the pros

Martin Ewing martin.s.ewing at gmail.com
Mon Nov 17 15:54:14 EST 2008

In my experience, the RFI from Ethernet is mostly *not* broad spectrum.
(This is distinct from any other hash generated by router processors, etc.)
It appears as a set of carrier tones every 62 KHz or so (at varying
strengths).  I generally hear a cluster of tones corresponding (I think) to
the clock oscillators in each Ethernet device, which are offset a little.

I believe we hear these tones at submultiples of the "100 MHz" clock (which
may actually be a multiple of 100 MHz).  The submultiples come from the
fixed frame lengths (every so many clock periods) that divide up the (idle
mode) Ether channel.  (I haven't looked at the 100bT format spec, but it
must be something like this.)

When you have actual data transmission (which is only a small percentage of
the time for me), I would expect to see broadband noise also, but the
spectra power density would be so low that it would be hard to detect. Or so
I'd guess.

A 1000bT signal would obviously have a much higher basic clock rate, but it
would be framed in a similar way.  If everything scaled up proportionately,
the carriers spaced by "62 KHz"  would go up to 620 KHz spacing, which would
mean less trouble for us at HF.  However, real life is probably not so

IIRC, a 10baseT connection emits little or no carrier when there is no data
transfer, while 100bT always sends carrier, making more potential trouble
for us HF types.

Still and all, I have to work hard to hear any of these Ethernet signals
through my normal antenna systems.  I had to make a simple wire probe to
hear them clearly.

It would be great if someone could produce "authoritative" spectral power
measurements for these 3 Ethernet systems.

73 Martin AA6E

On Sun, Nov 16, 2008 at 9:27 PM, Roger (K8RI) <k8ri at rogerhalstead.com>wrote:

> aa8ia at aa8ia.org wrote:
> >> From: "Martin Ewing" <martin.s.ewing at gmail.com>
> >> Subject: Re: [RFI] ethernet EMI revisited - questions for the pros
> >> This is an interesting discussion.  I would add a few points/questions:
> >>
> >> -Does anyone know of published "typical" spectra of 1000baseT vs 100 vs
> 10?
> >> I agree that 10bT is much quieter on HF than 100bT, but it is possible
> that
> >> 1000bT pushes the spectrum up enough to be quieter on HF. I did some
> work on
> >>
> I doubt that is what is happening.  You have the basic clock frequencies
> in the chips generating or regenerating the basic data frame. Typical
> systems that run Gigabit also run 10 and 100 base-T. Whether the clock
> actually changes I don't remember.  The data stream as you have surmised
> generates a noise that is pretty much random. That noise not only
> depends on the basic speed but character string. Typically the highest
> frequency would be generated by a stream of alternating 1's and 0's.
> Unfortunately life is not simple in the network world and much depends
> on the chip set and the implementation.
Martin Ewing, AA6E
Branford, CT

More information about the RFI mailing list