T10 Phy WG,
The transmitter common-mode voltage limit is specified in
two places, once in Table 61 as a broadband limit, and again in Figure 123 as a
per-frequency-band limit. Table 61 limits the overall amount to 30 mVrms
which translates to 84.9 mVpp if the AC signal is a pure sinusoid. Figure
123 imposes a limit in the band 100 MHz to 300 MHz (1 MHz measurement band) to
12.7 dBmV which translates to 4.3 mVrms and 12.2 mVpp if it is a pure sinusoid.
Furthermore, if the transmitter had energy at each band that
followed the limits in Figure 123 it would far exceed the limit in Table
61. A collection of discrete frequencies at the limit in Figure 123 would
violate the overall limit, such as (for example) spikes at 100 MHz, 200 MHz,
and so on up to 1400 MHz at the Fig 123 limit would violate the overall
limit. As another example, spikes at 100 MHz, 300 MHz, 500 MHz, and so on
up to 1900 MHz would violate the limit, as would a set of spikes at 750 MHz, 1500
MHz, 2250 MHz and 3000 MHz. Note that these spikes would violate
the “energy” aspect of the limit, where they are combined as a
sum-of-squares and then translated effectively into a sinusoid. The
actual combination of the spikes would depend on the relative phases; the
smallest I could find for the 750/1500/2250/3000 MHz case was approximately 40
mV, so that combination of frequencies could not all simultaneously be at the
Fig 123 levels.
Since these sinusoidal levels are quite small, I propose
that we remove Figure 123 and its limits entirely and retain only the limit in
Table 61. This absolute wide-band time-domain specification will be
enough to limit transmitted CM energy and still allow the silicon and system
designers enough leeway to specify their power supply noise and filtering
limits depending upon the particular frequencies in their systems. This
leeway should not jeopardize practical systems.
Mark Seidel
Principal Engineer
Intel Corporation