The tradeoff between crosstalk and loss
High Speed Serial standards such as 100G and 400G Ethernet and the OIF-CEI 28G and 56G standard-like "information agreements" try to allot designers as much freedom as possible in how they satisfy performance requirements without sacrificing interoperability.
Newer specifications require more than one stressed receiver tolerance test: a noise tolerance test that focuses on how robust the receiver is to ISI (intersymbol interference), random noise and jitter, and sinusoidal jitter that tests clock recovery and equalizer performance and at least one separate interference tolerance test that probes receiver performance with crosstalk. In the latter, crosstalk is balanced against channel loss in kind of a cool if arcane way.
To explain how it works, we need to sort out ICN (integrated crosstalk noise). It doesn't matter whether the signal is PAM4 or NRZ/PAM2. Consider a multi-channel system, like 4 x 25 Gbit/s for 100 GbE or 8 x 50 Gbit/s for 400 GbE. Multi-channel systems have crosstalk aggressors for each pair of signal carriers. In Ethernet terminology, aggressors are called "disturbers." The rms near-end and far-end crosstalk (NEXT and FEXT) for multi-channel systems are parameterized by these unwieldy expressions:
The sum is carried out over all pair-wise combinations, Δf is the frequency step over which the crosstalk is summed, W is a weighting function that models the receiver’s frequency response, and MDNEXT and MDFEXT are the equivalent crosstalk losses in decibels. ICN is then given by:
You need the differential S-parameters of the whole system—including scattering between separate channels—to calculate or simulate ICN.
Let's go back to design flexibility.
One approach at 28.5 Gbits/s requires levels of ICN that depend on the channel insertion loss measured at the signal’s fundamental harmonic frequency. Figure 1 shows the range of compliance and the tradeoff between insertion loss and crosstalk.
Figure 1. Balancing crosstalk, ICN, against insertion loss.
Emerging specifications such as 400G require that receivers pass two separate interference/crosstalk tests: one with a high loss channel and low crosstalk and one with low loss and high crosstalk. The channel loss requirements are specified by differential frequency response masks, like those in Figure 2. The high and low crosstalk conditions are specified implicitly by requiring COM (channel operating margin) < 3dB for both tests.
Figure 2. Masks for the frequency response of low and high loss channels (Courtesy of Tektronix Instruments).
COM is essentially the ratio of the signal amplitude to the combination of all signal impairments, including crosstalk and channel loss—like a generalized signal-to-noise ratio. COM is calculated from the combination of S-parameters, models of the transmitted and received signals, random and deterministic jitter, and voltage noise, and the effects of equalization schemes at both the transmitter and receiver.
Both the low loss test 1 and high loss test 2 require COM to be less than but close to 3 Db—the performance spec for the channel is COM >3 dB, so this counts as maximum, even extra stress. With the same COM value for both tests, the increased channel loss of Test 1 implies a decrease in crosstalk and, for Test 2, vice versa.
By specifying a minimum compliant COM, standards provide flexibility in how designers budget the combination of ISI, random noise and jitter, and crosstalk. By combining that spec with two channel loss performance criteria, the flexibility is reduced but interoperability is enhanced.
Want to learn more? Attend DesignCon 2016, the premier conference for chip, board, and systems design engineers. Taking place January 19-21, 2016, at the Santa Clara Convention Center, DesignCon will feature technical paper sessions, tutorials, industry panels, product demos, and exhibits. Register here.
DesignCon and EDN are owned by UBM Canon.