TDR Interconnect analysis – applications, measurement caveats, tips and techniques
The Time domain Reflectometer (TDR) has come a long way since the early days when it was used to locate faults in cables. If your designs involve signals with rise times shorter than one nanosecond, transmission line properties of the interconnects are important. TDR is a versatile and intuitive tool to provide a window into the performance of your interconnects to quickly and routinely answer the three important questions: does my interconnect meet specifications, will it work in my application, and where do I look to improve its performance?
The TDR is not just a simple radar station for transmission lines, sending pulses down the line and looking at the reflections from impedance discontinuities. It is also an instrument that can directly provide first order topology models and S parameter behavioral models. We will look at five most important, from the most common to the more advanced, applications of 1-port TDR.
[1] Measuring characteristic impedance and uniformity of a transmission line
For an ideal, lossless transmission line, there are only two parameters that fully characterize the interconnect: its characteristic impedance and its time delay. This is the easiest and most common application for the TDR. The TDR sends a calibrated step edge of roughly 200 mV into the device under test (DUT). Any changes in the instantaneous impedance the edge encounters along its path will cause some of this signal to reflect back, depending on the change in impedance it sees. The constant incident voltage of 200 mV, plus any reflected voltage, is what is displayed on the screen of the TDR.
Figure 1 shows the TDR response from a microstrip transmission line and a reference open. The DUT is a two-section microstrip transmission line with characteristic impedance of 50 Ohms and 40 Ohms, the far end is open. The blue trace shows the TDR response when the cable is not connected to the DUT and defines the beginning of the cable. The yellow TDR trace shows the small reflected voltage from the SMA launch, followed by the 50 Ohm section, the small drop in voltage from the lower impedance 40 Ohm section and the open end of the trace.
From the TDR response, we can obtain the instantaneous impedance of each segment using trace markers or converting the vertical voltage scale into an impedance scale. Using this method, we can evaluate the impedance uniformity of the line. One caveat is we are assuming all the measured voltage coming back to the TDR is due to reflections from impedance discontinuities. This is a good assumption when there are only small impedance discontinuities up to the location of the marker.
In Figure 2, we have the measured TDR response of a nominally uniform transmission line, on an expanded vertical scale of 2 Ohms/div. The impedance at the center of the screen is set at 50 Ohms. The large peak at the beginning of the line is the inductive discontinuity of the SMA launch on this high resolution scale looks huge. On this scale, the uniform transmission line does not look to be so uniform. Is this variation real or some sort of artifact?
There are two important artifacts that might give rise to this behavior. Firstly, there could be rise time degradation in the incident signal. It may not be perfectly flat, like an ideal Gaussian step edge. After all, the reflected signal displayed on the TDR is really the reflection of the incident signal. If the incident signal has a long tail, we will see this long tail in the TDR response and may mistakenly interpret this as an impedance profile variation. One way around this problem is to use the calibrated response feature of the Keysight DCA 86100D TDR Sampling Oscilloscope, which we are doing in this case.
Another source of artifact is due to the lossy nature of the line. There could either be distributed series resistance in the trace or distributed shunt conductance in the trace. The series resistance will cause the reflected voltage to increase as we move down the line, while the shunt conductance will cause the reflected TDR response to decrease as we move down the line, as in this case.
One way to evaluate whether an impedance profile is really showing a variation in the instantaneous impedance of the transmission line or an artifact, is to measure the TDR response of the line from both ends. If it is real, we should see the slope of the response change, depending on which end of the line we launch from. If it is one of the two artifacts, the response will look the same on the screen, independent of which end we launch from, as shown in Figure 3.
[2] Measuring time delay of a transmission line
The time delay of a transmission line from one end to the other can be measured directly from the TDR screen using markers. Figure 4 shows TDR responses for an open cable and when the DUT is connected. To increase accuracy, the time from the mid-point of the two open responses are used. The time interval from the beginning of the reflection from the open end of the cable to the reflection from the open far end of the DUT is the total round-trip time. The time delay is half of this value.
To ensure measurement integrity from assembly artifacts such as the launch connector, test line can be designed in to aid in the characterization of the circuit board and each layer. For instance, reference pads with a known separation can be added at two locations on the transmission line, see Figure 4. These small imperfections or pads can be easily detected with the TDR when displayed on the 2 Ohms/div scale.
[3] Accurate measurement of signal speed in a transmission line
By using the end to end method to measure the time delay, we can get an accurate measure of the signal speed traveling down the transmission line, independent of the nature of the launch. This is done by dividing the physical distance between the two reference pads by the acquired time delay. Figure 5 shows the two negative dips from the reference pads with a known separation distance. The time difference between these two negative dips is the round-trip time between the pads.
[4] Extracting bulk dielectric constant of the laminate
The signal speed of a transmission line, v is directly related to the dielectric constant Dk, the signal sees. For a stripline transmission line, the effective dielectric constant can be extracted using the simple relationship shown below:
Dk = (0.3/v)2 – where 0.3 is the speed of light in m/ns.
However, in a microstrip, some of the electric field lines are in the bulk laminate and some are in the air. The signal sees a composite of these two materials, which creates an effective dielectric constant, Dkeff. It is this value which affects the signal speed and can be extracted from the measured speed of the signal.
[5] Building a model of a discontinuity or interconnect
Structures such as test pads, component leads, corners and gaps in the return path will create discontinuities. Discontinuities can be categorized as capacitive, inductive and resistive. These structures are nonuniform and may require a 3D field solver to calculate. Sometimes, the quickest way to evaluate their impedance is to build a structure and measure it.
From the measured response, we can empirically evaluate the impact on the signal if we match the TDR’s rise time to the rise time of the application. We can directly measure off the TDR screen the amount of reflected voltage noise we might see in the system. Alternatively, we can use the TDR to extract a simple, first order model for the structure and use this model in a system level simulation to evaluate the impact of the discontinuity.
For example, we can observe from the TDR that corners or 90 degree bends have a response that of a lumped capacitor. Using TDR measurement, we can obtain the capacitance value for the lumped capacitor model of a corner and use this model in a system simulation to evaluate whether a corner might pose a potential problem, or can be ignored.
For the same impedance trace, the amount of capacitance in a corner will scale with the width of the line. A good rule of thumb to remember is that the capacitance of a corner is about 1 fF per mil of line width for a 50 Ohm line. Therefore, 60 mil and 5 mil wide lines have a corner capacitance of roughly 60 fF and 5 fF, respectively.
Finally, if we need more accuracy or a higher bandwidth model than what we can get directly from the screen, we can take the measured data from the TDR and bring it into a modeling and simulation tool, such as SPICE or ADS to fit a more accurate model.
Reference:
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
