Sifos Technologies PhyView PVA-3000, Reference Manual

Page: 11 / 134

PVA-3000 Reference Manual
December 2, 2019 Sifos Technologies Page
11
increased both the equipment and labor costs associated with testing.
A second dilemma associated with traditional transmitter testing came along with 100BaseTx. Most 100BaseTx
PHY’s were actually 10/100BaseT and used auto-negotiation with a link partner to activate a 100BaseT transmission.
When a transmitter pair is connected to a test instrument, there is no link partner and therefore, there is no 100BaseTx
transmitted signaling. This meant that the Ethernet port-under-test either needed some special test mode to activate a
100BaseT transmission, or it needed to start from a linked state with a 100BaseT link partner, then have the transmit
pair separately disconnected so it could get routed to the test instrument. IEEE 802.3 did not define any special test
mode registers so in many cases, especially when testing
finished product, the only option was the latter scheme.
1000BaseT made this even more difficult. When linked,
each pair simultaneously transmits baseband signaling in
both directions, using local hybrid cancellation on each
pair end to sort received signal from transmitted signal.
Additionally, PAM5 signaling is digitally filtered, or pre-
distorted, in order to shape it to the 100BaseT spectra.
So what shows up on each twisted pair is unintelligible.
IEEE 802.3 clause 40 addressed this be defining standard
test signals, standard registers to activate those test
signals, and standard methods for testing those test
signals (
see Figure 1.3 ). This assured that PHY
transmission measurements would be unavailable given
typical finished product.
1.2.5. Packet Testing: Pros and Cons
Given the complexity, cost, and expertise requirements of traditional PHY conformance testing, it is no surprise that
this type testing is done only in design verification, or at a very low sample rate basis, or in a test services setting such
as UNH-IOL. Higher volume testing required a much simpler method and in many cases, that method was packet flow
testing. In fact, many engineers and technicians equate packet flow testing with physical layer testing even though
packet flow testing is purely functional in nature and does not directly measure
any physical layer parameters.
Packet flow testing is attractive because it is easy and often low cost. Since Ethernet links universally discard defective
packets, the core requirement is a device that can transmit, receive and count Ethernet packets. Commercial packet
analyzers from manufacturers such as Spirent and Ixia have been widely deployed for years to help with Ethernet
interface and protocol testing. Lower cost, software-based solutions and even Ethernet switches with modified
firmware can serve the purpose of a packet generator / counter.
As a method for evaluating the physical layer, packet flow testing is in truth quite limited in utility and defect coverage
for several reasons:
Limitation #1 : It is inherently difficult to resolve packet loss to defects in a transmitter versus a receiver. If two
unknown ports are linked and packets drop in one or both directions, there is no information to suggest whether this is
a transmission problem or a receiver problem. On the other hand, if a known good port is connected to an unknown
port and packets are pinged (meaning the port–under-test has a full protocol stack including ICMP available), any
packet loss is still ambiguous as to whether the unknown port failed to receive them properly or corrupted the
transmission back. This limitation can be partially overcome by accepting Limitation #2.
Limitation #2 : The device under test must be a hub,
switch, router, or repeater with 2 or more ports, one of
which is a
known good port . Many of the twisted pair
Ethernet ports ever produced are in fact multi-port
devices that can bridge traffic between ports. This allows
known good (or instrument grade ) ports to be connected
to the known good port and to any other unknown port
so that the direction of packet loss on the unknown port
should point to a transmission versus receiver problem.
Limitation #3 : The degree of bit errors in a link may
provide hints to the severity of a problem but does not
shed any light on the nature of the problem. Figure 1.4
depicts the types of relationships found when comparing
Bit Error Rate (BER) measurements to various physical
impairments or defects. From this figure, it is apparent
that bit errors typically do not appear until the degree of
Figure 1.3 Specialized Transmitter Test Signals
Figure 1.4 BER versus Physical Impairment