Successfully Controlling Impedance in Mass-Termination Cable Breakouts
Test and integration applications often use large numbers of differential signals. While coaxial cables provide excellent signal integrity for high speed signaling, they are expensive and cumbersome to handle when dealing with large numbers of signals. Therefore many applications use mass-terminated connectors, such as VHDCI-68, DB37, DB25, and even RJ45, to transport high signal counts. This presents challenges when it is necessary to break the mass-terminated signal bundle out into separate cables for connection to test equipment, signal buffers, switches, patch panels, etc.
PRL offers several PCB assemblies for adapting to and from various mass-terminated connectors to the SMA connectors typically used on PRL fanout buffers, logic level translators, line drivers, and switches. Some assemblies are offered as standalone products, such as our PRL-RJ45-SMA adapter, and some are integrated into system level products, such as the PRL-4517 Level Translator.
All these PRL adapter assemblies are variations on a theme—routing pairs of pins on a mass-terminated connector to pairs of SMAs, via matched-length, controlled-impedance, 50 Ohm traces. The result is a solution that provides excellent continuation of the differential transmission lines required for high-speed LVDS, ECL, and RS422 signaling in demanding test and integration requirements in aerospace, defense, and automated test environments.
The following setup will show how closely impedance can be controlled when breaking out mass-terminated cables to SMAs.
PRL examined several of our adapter assemblies via basic reflectometry to illustrate how our products solve this class of problem.
We start with a differential, 50 Ohm signal source, the PRL-177A-100. This product produces a square-wave clock signal, with complementary TTL outputs. The 100 MHz base frequency can be divided by 1, 2, 4, or 8 on the primary output pair, or by 2, 4, 8, or 16 on the secondary output pair. We selected the PRL-177A-100 for its fast rise time (750 ps, typical) and good pulse symmetry. All complementary outputs have equivalent 50 Ohm output impedance, suitable for driving either single-ended 50 Ohm cables or 100 Ohm differential cables into 100 Ohm floating loads or into high-impedance loads over long cables.
To examine the signals “in flight” we connected the outputs of the PRL-177A-100 to the “Through” ports of a PRL-860D-SMA, Dual-Channel Inline Signal Monitor. This unique product is a high bandwidth, asymmetric, differential pickoff tee. The Through ports carry 95% of the original signals through a pair of 50 Ohm transmission lines, while the Monitor ports extract 5% of each signal via a 450 Ohm resistor, for routing to a 50 Ohm oscilloscope. A Weinschel 980-2 phase shifter was inserted along one leg to provide fine de-skewing of the input signal.
The Through ports were connected to various cables, boards, and loads, while the Monitor ports were connected to a two-channel sampling scope. A Math channel was defined as Ch1 – Ch2, to illustrate what a differential receiver will see when presented with these waveforms.
This configuration creates a low-cost reflectometer, suitable for characterizing transmission line effects into the low GHz range.
The completed setup is shown above. During the test the system was constructed, step-by-step, with oscilloscope captures taken at each step to examine the effect of each added component.
The Through outputs from the PRL-860D-SMA are connected, via a pair of 1’ 50 Ohm coaxial cables, to opposite ends of a PRL-SR-100, Series 100 Ohm Resistor module. The 100 Ohm floating load presents the equivalent of two 50 Ohm terminations to a “virtual ground” as the opposing pulses meet in the center of the module.
The Monitor signals on the oscilloscope capture show a well-matched 50 Ohm system, with fast rise- and fall-times and a clean pulse response on the leading edge. (The small spikes after the trailing edge of the pulse are caused by a small pulse-width mismatch between the differential outputs from the PRL-177A-100, but they cancel in the subtracted Math waveform. This effect will be discussed in a separate application note. For purposes of this application note we will be concerned primarily with the leading edges of the pulses and how they behave under various loading conditions.)
Differential TDR driving two 1’, 50 Ohm coax cables, terminated into a 100 Ohm floating load (PRL-SR-100).
The complementary outputs from the PRL-860D-SMA are now cabled into the differential inputs of a PRL-454LV, 1:4 LVDS Fanout Buffer. This input has a 100 Ohm floating termination in parallel with a high-impedance comparator input. The oscilloscope shows a very similar response to Step 1, with only a minor ripple caused by the slightly imperfect load at the input of the PRL-454LV.
Differential TDR driving two 1’, 50 Ohm coax cables, terminated into the 100 Ohm floating load of the PRL-454LV.
Now we remove the load. The pulse reflects from the open circuit at the end of the cables and back through the Monitor ports, where it adds to the original pulse and doubles the voltage presented to the scope. The section of waveform between the vertical cursors shows the transit through the pair of 50 Ohm transmission lines, before the pulse reflects back from the open circuit. In the scope capture below, the transmission line is ~3 ns in duration, representing the double propagation delay through the pair of 1’ coax cables, at approximately 1.5 ns/ft.
50 Ohm transmission lines through 1’ of cable, then into open circuit.
We connect the cables to a PRL-RJ45-SMA Adapter module. The matched-length trace pair presents a continuation of the 50 Ohm transmission lines, adding approximately 1.2 ns to the “good” part of the waveform, before the signal hits the open circuit at the far end of the adapter assembly.
1’ cables, followed by PRL-RJ45-SMA adapter assembly.
We connect an off-the-shelf Cat6a patch cable to the output of the PRL-RJ45-SMA. The nominal characteristic impedance of each twisted pair in Cat6a cabling is 100 Ohms ±15%. As shown in the scope capture, this ~100 Ohm twisted pair looks very much like a continuation of the 50 Ohm coax pair. There is a slight increase in pulse amplitude, which indicates that the pair in this cable has an impedance of slightly > 100 Ohms, but still well within the Cat6a specification. The Math waveform shows that the effect at the comparator input of a differential receiver would be trivial. This section of the trace is approximately 8.2 ns in duration, which corresponds well to the double propagation delay through the 3’ length of the connected cable.
PRL-RJ45-SMA assembly, followed by 3’ Cat6a cable.
We connect a second PRL-RJ45-SMA module to the end of the Cat6a patch cable. The matched length, impedance controlled pair of 50 Ohm traces adds another ~1.2 ns to the “good” part of the waveform.
Cat6a cable connected to a second PRL-RJ45-SMA assembly.
We connect a final pair of 1’, 50 Ohm coax cables to the SMA outputs of the PRL-RJ45-SMA module. This extends the 50 Ohm transmission lines by a further 3 ns.
Second PRL-RJ45-SMA connected to a pair of 1’ 50 Ohm coax cables.
Finally, we terminate the differential pair into the 100 Ohm floating load input of a PRL-454LV, 1:4 LVDS Fanout Buffer. Now the matched impedance system is complete, with a pair of 50 Ohm drivers driving a pair of 50 Ohm transmission lines (or a 100 Ohm twisted pair) into a floating 100 Ohm load. With a properly terminated signal, there is no reflected energy, and the signals return to their original levels.
The waveforms show only minor ripples at each connection point, and in general the signal quality is excellent .The Math channel on the oscilloscope shows a clean pulse all the way through, with no distortion, ringing, or overshoot that might cause spurious triggering.
The waveforms looks very similar to those in Step 1, where the driver was driving a pair of perfect 50 Ohm lines into a perfect 100 Ohm floating load. This illustrates that the mass-termination-to-SMA assemblies and the intermediate cable bundle do not present any material discontinuity to the transmission lines, and that they are suitable for carrying high-speed differential signals with excellent fidelity.
Terminated into the 100 Ohm floating load at the input of the PRL-454LV.
The reflectometry reveals interesting characteristics of the transmission path, but what really matters is what gets delivered to the end of the cable. We changed the location of the PRL-860D-SMA to probe the transmission line at the far end of the adapter assembly, as the pulses enter the PRL-454LV differential receiver:
At 100 MHz clock rate, the received pulses are clean, despite traveling through two PRL-RJ45-SMA adapter assemblies and a 3’ Cat 6a cable.
100 MHz Clock Pulses After Two PRL-RJ45-SMA Adapter Assemblies and 3' Cat6a Cable.
Inexpensive, off-the-shelf network cabling also allows us to test this configuration over much longer cables.
Using a Category 6a 10G Shielded Solid Patch Cable from Stonewall Cable (SC-914S-E61), we ran a 50 MHz clock signal over 100’ with good results.
100 MHz Clock Pulses After Two PRL-RJ45-SMA Adapter Assemblies and 100' Cat6a Cable.
We repeated this experiment with a different adapter assembly, PRL part number 56004475. This PCB adapts selected pins from VHDCI-68 connector to 12 pairs of SMAs. As with the PRL-RJ45-SMA adapter, all PCB traces are matched-length and controlled-impedance, to maintain 50 Ohm transmission line pairs.
As with Step 4, the coax cables and adapter assembly show a good 50 Ohm transmission line pair up until the open circuit at the unconnected VHDCI connector, with the PCB presenting approximately 3 ns of double-propagation delay following the 50 Ohm cables:
Cabled into a VHDCI-SMA adapter assembly.
We connected a 1m VHDCI cable assembly (Molex P/N 79918-0080) to the output of the adapter assembly. The impedance specification on this cable is 125 Ohms. As shown in scope capture below, the impedance mismatch distorts the pulse response more than it was in Step 5. This is apparent in the Math channel, which shows an appreciable “step” where the signal exits the PRL adapter assembly and enters the cable assembly.
This “step” through the 1m cable has a duration of ~9.6 ns, showing the pulse hitting the open circuit at the end of the cable and reflecting back.
Connected to a 1m VHDCI cable assembly.
We connected a second VHDCI adapter assembly, which returns the system to 50 Ohms, for a duration of 3.0 ns.
Connected to a 1m VHDCI cable assembly.
We connected a final pair of 50 Ohm coax cables.
Connected to a 1m VHDCI cable assembly.
And finally, we terminate into the floating 100 Ohm load of the PRL-454LV.
Connected to a 1m VHDCI cable assembly.
At 100 MHz clock rate, the received pulses are clean, despite traveling through two VHDCI adapter assemblies and a 1m VHDCI cable with a slightly mismatched impedance.
100 MHz Clock Pulses After Two VHDCI Adapter Assemblies and 1m VHDCI Cable.
The above screen captures demonstrate that commonly-used mass-terminated cable assemblies and connectors can carry differential signaling at high data rates and/or over long distances with acceptable signal quality.
Even with minor impedance mismatches caused by cabling with differing specifications, the effects at a differential receiver with proper floating termination are immaterial at clock rates of 100 MHz and beyond.
PRL adapter assemblies can adapt many common mass-terminated standards, such as DB37, VHDCI, RJ45, etc., to SMA connectors and cabling, for easy connection to standard test equipment, including PRL fanout buffers, level translators, switches, etc. The controlled-impedance 50 Ohm trace pairs are matched with the 100 Ohm twisted pairs in many commercial cable assemblies, and much of effect of the discontinuity in the mass-terminated connectors (which are typically not impedance controlled) disappears at the differential receiver.
The PRL-RJ45-SMA is available as an off-the-shelf product, and other adapter types are either available as special-order products or can be designed to your specifications, with affordable NRE charges and fast lead times. Please contact email@example.com for more details.
Future articles will address issues such as rise/fall time, signal attenuation, skew and duty-cycle mismatching, and AC- vs. DC-coupling.