Signal speeds and devices continue to get faster and operating voltages continue to fall, placing ever increasing demands on signal and power integrity. If you haven’t already been measuring PCB traces and planes, you will certainly have to in the future. Fortunately, time domain reflectometry (TDR) instruments are very simple and fast to use. By following a few simple guidelines, your measurements will be high fidelity and easy to read.
#1: Use high quality, low loss cables and connectors
Inexpensive cables are often acceptable for making basic measurements, and we all like to save a few bucks. But this is not the best place to cut corners. You should only use only high-quality, low-loss cables or special TDR probes if you want to achieve accurate TDR measurements.
The setup in Figure 2 illustrates the impact of poor quality connectors. In this image, the TDR is deskewed with no cables attached to set the reference plane at the TDR instrument front panel. A high-quality semi-rigid coax is connected to the front panel while a thru connector is used to attach a second high-quality cable. Finally, the second cable is attached to a PCB 50Ω test trace with a low-quality PCB edge connector.
The high-quality connectors maintain a very tolerable impedance variation while the lo- quality edge connector does not.
The second coax cable is then disconnected from the PCB test trace and connected to another port of the TDR instrument as shown in Figure 3. The TDR instrument has a very high-quality connector and 50Ω terminator, confirming that the poor measurement in Figure 2 is due to the poor-quality edge connector on the PCB. This is still a single port TDR measurement. The second port is only being used for its high-quality connector and terminator.
In the same way that low-quality connectors can result in very noisy measurements, low-quality cables can also result in very noisy and erratic results. The measurement in Figure 4 shows the performance of a low-quality 50Ω coaxial cable.
Low-quality cables and connectors result in very noisy plots, making it difficult to see the measurement of the PCB traces and planes we are interested in.
The message is clear that while inexpensive cables are often acceptable for making basic scope measurements, high-quality, low-loss cables are essential for TDR measurements.
#2: Sometimes the best cable is no cable at all
An additional issue with poor quality cables is that the cable itself slows the rise time of the TDR step reducing the resolution of the setting. High-quality low-loss cables will have the least impact. The cables should be kept as short as possible as well.
One final issue to note, the interconnecting cable is included in the measurement unless a complete Short-Open-Load calibration is performed on the instrument and cable setup prior to making the measurement.
It is possible to have a high-quality cable that also has relatively high losses. In order to illustrate this point, a SMA mounted 1Ω resistor is connected to the TDR instrument using two different high-quality cables. One cable is lower loss than the other. The measurements show the low-loss cable to be 100mΩ less than the other cable, despite both being very high quality cables and similar, though not exactly the same length. One measurement reports 1.2Ω while the other measurement reports 1.3Ω. Neither measurement is correct due to the cable loss being added to the measurement.
The same 1Ω mounted resistor is connected to the TDR instrument without any cables, just an SMA thru adapter to mount the board to the front panel of the TDR instrument as shown in Figure 6. The measurement is now very close to 1Ω and with a much smaller error due to the greatly reduced connection losses.
The measurement setup in Figure 7 shows a series of connections all without the use of cables. The upper red trace in this photo is the measurement of the DSA8300 sampling probe with an SMA adapter installed. The middle trace shows the result of all of the connections with the exception of the 50Ω terminator at the end of the PCB test trace. The yellow trace shows the complete measurement all the way through to the 50Ω terminator. Note that the connection to the PCB using a low-quality connector is also very noisy in this measurement as it was in the measurement of Figure 2.
#3: Properly torque connectors
Loose connectors can provide incorrect results. The image in Figure 8 shows measurements with a “finger tight” cable and a torqued cable.
Not all manufacturers recommend the same torque, and the range for SMA connectors is very wide, from approximately 3in-Lbs to 10in-Lbs. Stainless connectors can generally be torqued more than brass. Check with the manufacturer of your TDR instrument and invest in a torque wrench.
#4: Perform a complete calibration when measuring low values and always measure a known resistance first
While quality, low-loss cables are strongly recommended, most instruments allow full Open-Short-Load calibration. This removes most of the cable and connection losses.
With careful calibration, most TDR instruments can measure resistances below 10mΩ. The image in Figure 9 shows the result of measuring a 10mΩ SMA mounted resistor confirming the measurement accuracy. As we saw previously, the results can have significant errors if the measurement is not well calibrated and/or low quality cables and connectors are used.
#5: It doesn’t always need to be on high
It is true that the resolution of the TDR measurement is related to the rise time of the TDR step. This does not mean that we need to always use the highest rise time setting. Most instruments allow the rise time to be adjusted. Reducing the rise time reduces many of the reflections that result from pre-step discontinuities and post-step ringing making the signal we are interested in easier to see.
The measurements shown in Figure 10 illustrate the effect of the step rise time. The 50Ω trace on the demo board from Figure 1 is connected to the instrument using only an SMA thru barrel adapter. The upper trace is measured with a 22.3ps rise time, while the middle trace is measured with a 150ps rise time. For comparison, a high-quality coax cable is attached to the instrument to show the further improvement resulting from the high-quality connector.
A reasonable guideline for setting an appropriate rise time is to set it twice as fast as the expected signal edges in circuit. For example, ultra high speed CMOS logic gates have a typical rise time of approximately 350ps and so the rise time can be set to 175ps. One note of caution is the coaxial cables connected to the instrument will slow down the rise time of the step. This is another reason to use only high-quality cables and to keep the cables as short as possible.
#6: Place PCB markers for accurate timing measurements
This tip is included in the Agilent reference at the end of this document, but it is a very good suggestion and it is worth repeating here. If the goal is to accurately calculate the dielectric constant of the board, it is best to put small artifacts, such as a small square, crossbar, or round pad precisely spaced along a test trace. The “blip” resulting from each artifact can be precisely located in the measurement and accurate time measurements can be made since the precise spacing will be known.
Keep in mind that in the single-port TDR measurement, the signal travels the length of the transmission line twice: once from the instrument to the end of the line and once from the end of the line back to the instrument.
Using the demo board test trace as an example, the trace is 5in. long and we can see in Figure 10 that the time measured is 1.63ns. The velocity is, therefore, 5in. twice in 1.63ns.
If the velocity of the signal is calculated from the distance between the “blips” in inches per nanosecond the effective dielectric constant, εr can be found as:
In this example, no such artifacts are included in the test trace, and this makes it difficult to determine exactly where to record the time displacement.
An improvement in the measurement can be made by adding an additional connection at the far end of the test trace. The measurement is shown in Figure 11. Using a high-quality coax cable and connector at the far end of the test trace results in an aberration due to the far end connector allowing a more precise time measurement, in this case 1.6137ns.
Using the equation above, this results in a reduced effective dielectric constant of 3.625.
Since we do not know the exact location of the aberration within the test trace connectors, there is still some uncertainty. Placing artifacts a precise distance apart removes this uncertainty.
