With the announcement of HDMI Rev 2.1 and its new bandwidth requirement of 48Gbps, it is vital to able to measure equipment performance. But even for a lab with the right equipment, the intricacies of the yet-to-be-released spec make testing a huge undertaking.
Here at DPL Labs we always try to keep an eye out for new state-of-the-art test equipment, and after months of research, studies and test trials we were able to work closely with Tektronix, one of the leading test equipment companies, to develop a new set of procedures to provide us with the proper equipment that could handle a 48Gbps envelope.
Every time a test trial begins, it becomes a team effort by the test equipment company and the testing agency. You never know when issues will arise with cable assembly limitations, fixture operations, and of course the myriad software typically used in today’s test equipment.
During this particular trial period working with the folks at Tektronix, it took almost an entire day to get the equipment up and running to our expectations, hindered by anomalies with many of the support products used to test HDMI devices. And working in the 48Gpbs world, a whole new set of factors enters into each test process.
In this case, there’s an initialization process, which takes about 20 minutes to set up. Then the software goes on its merry way doing its routine. But each time the routine was executed, each test performed with a DUT (device under test) failed.
Time and time again we could not get through the process to generate the 48Gbps Eye Patterns we were looking for. Finally, we had to dive into the bowels of the system and pull out all the data that was being used to execute these waveforms.
In one case a series of S Parameters is used for part of these measurements and when examined very closely we found a small glitch positioned approximately 11ps before the test pulse began its routine.
This small glitch was only about 480mv high, but because it was so close to the pulse start, it caused the system to interpret the waveform as something else — forcing it to lose sync and destroy the measurement. Now the question was, where did it come from? Using many of the same processes custom integrators employ in the field, we began a process of debugging by using various DUTs, cable assemblies and connectors.
At the end of the day we finally determined the culprit. It was the HDMI fixture commonly used when testing cables and such. But there was more. After substituting in a replacement fixture the same glitch appeared. No matter how many new fixtures we used, this nasty glitch would not go away.
If you look close at the waveform presented below, you can see where the spike occurs. What it is showing us was a reflection anomaly being produced by the HDMI connector itself. Since the same connector was used on each test fixture, they all produced the same error.
So we contacted a fellow RF engineer that specialized in connector issues and found that his firm built instrument-grade HDMI connectors. Not cheap, but if it worked we would be home free.
Well, that solved it.
We replaced the connectors on the fixtures and the problem went away and we finally were able to establish a perfect 12Gbps Eye Pattern on the oscilloscope.
So the next question comes around — will we see more tiny glitches like this as we move deeper and deeper into high-speed data rates? Can small anomalies like this cause issues in the field as these new standards roll out?
We’ll keep you posted.