HDMI Cable Modeling and Analysis with TDR Data

White Paper
HDMI Cable Modeling and Analysis with TDR Data
Abstract: This paper describes HDMI cable modeling process based on TDR/T data. Frequency domain
transmission parameters can either measured directly using TDT or can be approximated from TDR data using a
measurement-based model. With the TDR approach, after good correlation is achieved between measured and
HSPICE circuit simulation data for both time and frequency domains in open-ended structure, the circuit model
then can be terminated with 50Ohm terminations to approximate S-parameters. The transmission characteristics
of the HDMI cable assembly can be effectively obtained from the circuit model and then analyzed with an eye
diagram. A measurement-based model for a short cable can be used to predict behavior of a longer cable.
I.
Introduction
The High-Definition Multimedia Interface (HDMI) is an emerging consumer electronics standard that offers the first
industry-supported, all-digital audio/video, one-cable interface. The HDMI interface allows data rates as fast as
5Gbit/s though a single connector instead of several cables as used in the past. High data rates used in the HDMI
cables require careful design and analysis techniques to ensure that the product passes required compliance tests.
The time-domain reflection and transmission (TDR/T) measurement of the HDMI cable allows to locate and model
discontinuities caused by the geometrical features of a connector and by the frequency-dependent losses of the
cable itself. Topological models can be consequently built for each part of the HDMI cable assembly, verified with
the measurement data, and then used to predict time- and frequency-domain response for the longer HDMI cable.
The topological modeling methodology allows an accurate approximation of the electrical behavior of the device
under test (DUT) before the longer prototype is manufactured.
In this paper a 3-meter long cable is used to build a topological model from TDR data, then the model is scaled up
and electrical performance is predicted for a 10-meter-long prototype. The HSPICE simulation, linked with IConnect
modeling software, reveals excellent correlation between the prediction and the actual TDR/T measurement for a
10-meter long HDMI cable assembly for both in time- and frequency-domains. The model, generated from TDR
measurements, allows obtaining an eye diagram, which is then compared with an eye diagram generated from TDT
measurements for the fabricated HDMI cable prototype. Such eye diagram can be generated directly from TDT
measurements, but TDR-only approach works best when only one side of the DUT is accessible. The result of an
eye mask test from model prediction agrees with the test performed for the fabricated prototype. The presented
technique allows the designer to quickly accomplish interconnect modeling and analysis tasks, resulting in faster
design time and lowering design costs.
II. TDR Measurements
The TDR measurement instrument is a very wide bandwidth equivalent sampling oscilloscope with an internal step
generator. The TDR sends a step stimulus to the DUT, and based on reflections from the DUT, the designer can
deduce a lot of information about the DUT’s properties such as location of failures, DUT impedance and time delay,
and to generate an eye diagram for the system [2]. An engineer can also use Time Domain Transmission (TDT)
measurement to measure crosstalk or to characterize lossy transmission line parameters, such as rise time
degradation, insertion loss, and skin effect and dielectric loss. Frequency dependent behavior of the system can be
calculated from the time domain (TD) measurements using IConnect software that employ a so-called Time Domain
Network Analysis Technique (TDNA) [3].
The TDR measurements are visual and intuitive to the digital designers due to the transient nature of this technique.
As the incident step propagates through the discontinuities in the DUT, it causes the reflections which indicate the
exact locations of discontinuities and their sizes. The fast TDR rise time provided by TDS8200 oscilloscope from
Tektronix [1] ensures that a wide range of frequencies is captured during this broadband measurement. The
generalized diagram of TDR/T measurement setup is shown in Figure 1. Any of these measurements can be
performed in a differential or single-ended fashion. The differential, common, or mixed mode measurements require
at least 2 synchronized sources and a 4-port measurement setup, as shown in Figure 2.
Cable: Z0, td
Reflection
Reflection
VTDR(t)
DUT
Port 2
Rsource
Transmission
Port 1
V(t)
TDR Front Panel
TDR/T Block Diagram
VTDT(t)
Figure 1. General Time Domain Reflection and Transmission (TDR and TDT) block diagram. A similar diagram can be drawn
for reverse measurements (from port 2 to port 1).
Rsource
TDR Oscilloscope Front Panel
Cable: Z0, td
V
V
ZDUT, tDUT
Ztermination
Vincident
Vreflected
Cable: Z0, td
ZDUT, tDUT
Rsource
Ztermination
Figure 2. Block diagram for measurements of coupled interconnects. A differential, common mode or mixed mode
measurement would require a 4-port measurement setup.
The TDR/T response of the device under test (DUT) allows not only to observe different discontinuities and to
characterize the HDMI interconnects, but also enables an engineer to quickly create topological models. The
topological models capture distributed nature of the high-speed interconnects and allow determining a precise
impact of each individual discontinuity on the overall performance of the DUT. IConnect modeling software utilizes
the TDR/T modeling techniques to generate and analyze topological models of various interconnects including the
HDMI cable assemblies.
III. Differential Impedance Modeling of the HDMI Cable
The HDMI standard uses so-called transition minimized differential signaling (TMDS) technology which provides
differential signals with nominal amplitude transitions of 500mV. If just the differential signaling is considered then
the interconnect can be reduced to a two-port structure, and the model can be built using just a differential TDR
voltage waveforms. Moreover, if the HDMI test fixtures are not available, a differential probe connected to the
desired channels at the reference plane can be used. In this modeling example a differential probe P80318, shown
in Figure 3, is used to obtain the differential TDR response of a 2.5 meter long HDMI cable; if the second probe is
available the differential TDT can be acquired as well, and insertion loss and eye diagram of the cable assembly
can be obtained directly from the measurement, without resorting to the modeling process. The combination of two
models, a connector model and a lossy cable model, is then used to predict both S-parameters and an eye diagram
of the interconnect.
2
Figure 3. Tektronix differential probe P80318. The probe has 18Ghz bandwidth and variable pitch.
The modeling process starts from modeling losses for the HDMI cable. IConnect software uses two approaches to
extract the losses, “matched” and “open.” In a “matched” approach both TDR and TDT data are used to generate a
lossy line model. The “open” approach can be handy in cases when it not possible to acquire a transmission
waveform because it allows using a reflection data with the other port kept open to generate an accurate model.
The losses in this approach can be extracted and optimized based on the information from the signal’s rise time
degradation and on the slope of the TDR voltage in the DUT’s region.
The TDR data for open-ended configuration, shown in Figure 4, is acquired using a TDS8200 instrument and
loaded into IConnect’s lossy line modeling tool. After the extraction is activated, IConnect extracts the RLGC losses
for the DUT. Figure 5 shows the results for such extraction in both time- and frequency domains shown on the left
and on the right of the figure respectively. Note that the correlation is excellent in terms of the rise time degradation
and modeling of the TDR’s slope, however, the connector’s reflections are not modeled in this case. This is also
observable in the frequency domain correlation shown on the right of the Figure 5; the depth of the modeled
resonances is smaller than the depth of the measured ones. This behavior is accurately captured when the
connector’s reflections are modeled using a single line modeler tool of IConnect.
Figure 4. TDR measurement setup for open-ended configuration.
3
Figure 5. Lossy line modeling of the 2.5-meter long HDMI cable. Blue is measured and green is a simulated waveforms. Left
figure shows time domain correlation and extracted RLGC loss parameters; right figure shows the frequency-domain correlation
for the same model.
The single line modeler tool utilizes true impedance profile to generate a model for each discontinuity in a
connector-cable transition of the HDMI cable assembly. To generate such model, only the TDR data is required.
Figure 6 shows both true impedance profile for the connector-cable area (on the left), and the model’s correlation
with measurements in terms of reflected voltages (on the right). The discontinuities of the connector-cable area
shown on the Figure 6 are modeled using sections of the ideal transmission lines; however, lumped element
topologies can be selected as well. After the model’s parameters are adjusted, the single line model can be
combined with the lossy line model to represent both losses and reflections and to predict S-parameters as well as
the eye diagram for the cable under test.
Red – Measurement
Green – Circuit Model
Figure 6. Single line modeling of the HDMI cable assembly in odd mode. Left figure shows true impedance profile and model’s
partitions; right figure shows correlation of the circuit model (green) and measurements (red).
4
A model for the connector and the lossy line model can be combined in the composite modeler window. The length
of the lossy line model can be scaled down using “scale parameters by” feature of the lossy line model; this feature
allows adjusting the length of the lossy model to fit connectors’ models. There is no need to create another model
for the connector-cable transition because these areas are identical for both sides of the cable. Hence, the model
can be reused by interchanging the ports direction in the sub-circuit’s netlist. HSPICE circuit simulation of the final
assembly model reveals excellent correlation in both time and frequency domains shown in Figure 7.
Figure 7. Time- and frequency domain correlation of the model assembly. Note that both losses and reflections are modeled
accurately.
Once an accurate model is created, it can be used to predict an eye diagram and S-parameters for the cable under
test. In order to extract a 2-port S-parameters and an eye diagram, both TDR and TDT responses of the HDMI
cable model are required. By the definition of S-parameters, the response is measured when all ports are
terminated with matched terminations [4]. This is simply done by changing the termination impedance in the
composite modeler and simulating the response. The HSPICE simulated TDR/T waveforms are then used to
compute S-parameters, which are shown on the left of Figure 8; where the return loss is shown in blue and the
insertion loss is shown in green colors. The eye diagram shown on the right of the Figure 8 is produced by using a
symmetric coupled lossy line modeler and loading the measured reference and simulated transmission waveforms.
Figure 8. Differential S-parameters and eye diagram predicted from the circuit model. Eye diagram is generated at 1.65Gbs data
rate and 200ps 20-80% rise time using absolute eye diagram mask for sink requirements.
5
IV.
Fully Coupled Modeling of the HDMI Cable
Although, a differential model can be efficiently used in the system’s simulations, a fully coupled model provides
more accurate representation of device’s performance. Signals that propagate differential lines can be decomposed
into even and odd mode components, therefore, if the model is capable of accurately capturing these two modes of
propagation, then any signaling can be accurately represented in circuit simulations. When coupled models are built
with IConnect they can be simulated with linked simulator and the results can be compared with measurements in
both modes of propagation.
A process of building coupled models is similar to the one described in the previous section. Separate models for
the reflections and losses need to be created first, and then they can be combined into one model assembly and
compared with the actual measurements data. The measurements should include both even and odd mode
responses and the models can be built based on TDR only, or based on TDR and TDT measurements. In the
example described in this section, the model is built assuming the minimum availability of the measurement
equipment that is differential TDR capability only.
The TDR data is acquired for both odd and even modes of propagation. The DUT connection is performed
according to the Figure 4, with the reference plane set as an open at the end of the fixture or probes. The odd mode
is excited by setting opposite polarity steps and acquired by using the difference between two channels. The even
mode is excited by using the same polarity signals and acquired by summing the TDR responses of channels one
and two.
A coupled lossy line model correlation is shown on the right of Figure 9, and the connector-cable area is shown on
the left of Figure 10. After accuracy of the models for both reflections and losses is verified, they can be combined
into one model assembly shown in Figure 10. The model assembly shows excellent correlation for both even and
odd modes of propagation in terms of RLGC losses and reflections.
Figure 9. Reflections and loss modeling of the HDMI cable assembly using coupled line modelers of IConnect. Left figure shows
the correlation of the HDMI connector model, and the right figure shows loss modeling.
6
Figure 10. Time domain correlation and fully-coupled model topology for HDMI cable assembly.
V.
Model-based Prediction of an Eye Diagram
The eye diagram test is another key measurement required by the HDMI signaling standard. The measurement of
the eye diagram captures the deterministic jitter in the interconnect, which is caused by losses and inter-symbol
interference. Since the transfer characteristics of a cable assembly contain all the information required to reconstruct this deterministic jitter, the eye diagram computed from the time domain transmission (TDT)
measurements using a TDR oscilloscope is as valid and accurate as the eye diagram obtained using a pattern
generator and a sampling oscilloscope.
Modeling tools of IConnect can be used to estimate an eye diagram of a long cable from the measurements of a
short one, thus enabling a designer to predict interconnect performance before even manufacturing it. This is done
by creating an accurate model for the short interconnect using a topological modeling approach described in the
previous sections and scaling up the lossy line model to represent the longer cable assembly. In this section we use
a 3-meter-long HDMI cable to predict an eye diagram of a 10-meter-long cable. Then the actual measurements of a
10 meter long cable are used to verify the prediction.
To build an eye diagram with IConnect, reference and transmission waveforms are required. The reference
waveform can be acquired from the open configuration with the DUT disconnected, while TDT waveform can be
obtained from simulation of the scaled model using the reference waveform as a source signal. The TDT response
is then saved and loaded into a new lossy line modeler tool, and the eye diagram is then generated according to the
compliance specifications.
Figure 11 shows a time domain correlation of the scaled model originally built using measurements of a 3-meter
long HDMI cable with the real measurements of a 10-meter-long cable assembly. Both TDR and TDT responses
show excellent correlations. Circuit model allows to generate a transmission waveform with can be used to
generate an eye diagram. Figure 12 shows comparison of the modeled and measured eye diagrams for a 10-meterlong cable. The scaled model provides a reasonable estimation of the actual 10-meter-long cable performance.
7
Figure 11. Time domain correlation of the scaled model with the actual measurements of the 10-meter-long cable. Green
waveforms represent the scaled model while the blue waveforms correspond to the time domain measurements.
Figure 12. Correlation of an eye diagram obtained from the scaled model with the eye diagram obtained from the actual
measurements of the 10-meter-long cable. Eye diagrams are generated at 1.65Gbs data rate and 200ps 20-80% rise time using
absolute eye diagram mask for sink requirements.
VI. Conclusion
In this paper we demonstrated that IConnect software can be efficiently used to perform an analysis of the HDMI
cable assemblies. Accurate two-port differential models can be quickly built from only TDR measurement data. The
designer can also build coupled models from even and odd mode TDR measurements, and then use those models
to predict both insertion and return losses of the cable assembly. Finally, the topological models obtained with
IConnect can be scaled up to accurately predict performance of the longer cables, which was demonstrated in the
last section of this application note.
8
Bibliography
[1] Tektronix, Inc., Beaverton, OR, 97077, USA.
[2] “Eye Diagram Measurements Using TDR Oscilloscope Transmission Data,” Application Note, Tektronix Inc.,
WebID: 3059.
[3] “S-parameters, Insertion, and Return Loss Measurements Using TDR Oscilloscope,” Application Note, Tektronix
Inc., WebID: 3058.
[4] Simon Ramo, John R. Whinnery, Thodore Van Duzer, “Fields and waves in communication electronics,” 3d ed.,
John Wiley &Sons, Inc., 1993.
Copyright © 2006, Tektronix, Inc. All rights reserved. Tektronix products are covered by U.S. and foreign patents, issued and pending.
TEKTRONIX and TEK are registered trademarks of Tektronix, Inc. All other trade names referenced are the service marks, trademarks or
registered trademarks of their respective companies. 07/06 MH/WOW 85W-19812-0
9