Text preview for : 5989-3248EN.pdf part of HP 5989-3248EN HP Publikacje 5989-3248EN.pdf
Back to : 5989-3248EN.pdf | Home
Backplane Differential Channel
Microprobe Characterization in
Time and Frequency Domains
White Paper
Table of Contents
Differential Channels Will Proliferate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3
The Bottleneck of SMA's . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5
Advantages of Microprobing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6
Design for Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7
Physical Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8
Understanding 4-port Mixed Mode Analysis . . . . . . . . . . . . . . . . . . . . . . . . .10
Design Case Study: XAUI Backplane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11
Comparing Time Domain and Frequency Domain Data . . . . . . . . . . . . . . . .12
Coupling Pulls Down Differential Impedance . . . . . . . . . . . . . . . . . . . . . . . .13
Eye Diagram Simulation Using 4-port S Parameters . . . . . . . . . . . . . . . . . .15
Non-Ideal Differential Signaling (AKA Mode Conversions) . . . . . . . . . . . . .16
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18
Abstract
The chief difficulty with routine characterization of differential channel paths
is the current requirement for SMA connectors to interface with the test
equipment. With little extra space available on functional backplanes, and
the problem of line loading produced by the SMA stubs, it is not practical to
use SMA connectors. This means only specially designed test boards can be
routinely characterized. This paper introduces a new methodology for testing
the passive interconnects associated with a differential channel in a backplane
assembly, which can be used for functional, populated backplane assemblies
and shows how the pad layout can be optimized for routine probing without
impacting the functional system performance.
2
Differential Channels Will Proliferate
We are in the beginning of a revolution in input/output formats. All high-speed
interface specs are migrating toward the use of differential signals imple-
mented with a differential pair of transmission lines. Figure 1 is a partial
list of some of these high-speed serial link formats.
All Next Generation High Speed Serial Links will use Differential Signaling
Serial ATA 1.25 Gbps XZUI 3.125 Gbps
Hypertransport 1.6 Gbps PCI Express II 5.0 Gbps
AGP8x 2.1 Gbps OC-192 9.953 Gbps
Infiniband 2.5 Gbps 10 GbE 10 Gbps
PCI Express 2.5 Gbps OC-768 39.81 Gbps
Serial ATA II 2.5 Gbps
Figure 1. Partial list of many new high speed serial link formats
In addition to being differential signals, they are all in excess of 1 Gbps. This
translates to an analog bandwidth of at least 2 GHz. In the case of XAUI, this
is in excess of a 5 GHz analog bandwidth. In order to verify the performance
and the quality of models for use in system level simulation, measurements on
the differential channel properties must be at least twice the application
bandwidth, or up to a 10 GHz measurement bandwidth.
Characterizing a differential channel can be accomplished in two ways,
either by building an equivalent circuit model based on a topology of ideal
circuit elements that all simulators can understand, or with a physical layer
characterization based on performance metrics. The model extracted from
physical layer characterization is usually called a behavioral model, as it
uses the direct measurements as the model itself. A behavioral model is
often called a "black box" model because it cannot be determined what specific
structures are inside.
A topology based model can be used in any simulator and provides useful
design insight on what physical features influence which electrical features. It
requires more work to construct and the effort usually increases exponen-
tially with increasing bandwidth and complexity. Creating a model for a simple
structure such as a balanced transmission line (microstrip or stripline)
can be difficult at high speeds. Creating a topological model for a large
backplane with multiple channels is daunting even for a team of experts.
3
The alternative characterization approach is to use the direct measurements
as the characterization. This behavioral model can then be used directly in a
system level simulation. Careful manipulation of the direct measurements
can present the information in a format from which useful performance
information can be directly extracted.
In this latter approach, there are a few metrics that describe the performance
of the differential channel. They are: