At a time personal electronics have become ubiquitous in our daily lives, high-speed digital standards like PCIe, SATA, and HDMI assume ever-increasing levels of importance. The faster the signal, the more sensitive the system is to impedance mismatch at the interface between the transmission line -- whether that's a cable or an interconnect -- and the transmitter/receiver. The mismatch generates primary and secondary reflections from the transmitter and receiver, which can degrade signal quality (see Figure 1).
Figure 1: As this eye diagram shows, signal quality for system without source impedance matching (left) is significantly worse than for the same system with impedance matching (right).
(Courtesy of Agilent Technologies)
At first blush there would seem to be an obvious fix for that -- match the impedance. Before you can match it, you have to measure it, though, typically using time domain reflectometry (TDR). The problem is that the impedance of a device while powered (hot) differs significantly from its level in the off state. In response, many high-speed digital standards call for impedance analysis while the device under test (DUT) is hot.
Hot TDR basically involves launching a step voltage into the system and monitoring the reflections generated by impedance mismatches. In the past, that's been performed using oscilloscopes, but we've seen increasing interest in the use of vector network analyzers to perform the task in the frequency domain.
With the ever-increasing buzz in the test community regarding hot TDR, we took the opportunity to chat with Craig Kirkpatrick, senior applications engineer at Agilent Technologies and co-presenter at a half-day TDR short course held at DesignCon 2013.
Kristin Lewotsky: Oscilloscope-based TDR measurements have traditionally focused on passive devices -- cables and interconnects, and boards not powered up. Why is that?
Craig Kirkpatrick: What it comes down to is you’re putting out a pretty small-signal step voltage and the reflections from that can be pretty small, just a couple of millivolts. If you turn on most circuit boards, especially anything with a digital board, it creates a lot of noise which can mask the small signal of interest.
K.L.: I've been hearing a lot of buzz about using vector network analyzers (VNAs) for hot TDR measurements. Most labs and production facilities already have oscilloscopes. Can't you get TDR results that are almost as good with traditional methods?
C.K.: With the VNA, we can actually have the DUT transmitter powered up and transmitting serial data streams, so you can see what the source impedance of the transmitter is as it is operating.
K.L.: In theory, you can do hot TDR with an oscilloscope though, can't you? What does the VNA approach bring to the table?
C.K.: The VNA is a signal source, but it’s also a narrowband signal receiver. You can step across any frequency and have the receiver look at the exact same frequency. The benefit of the narrowband receiver is it filters out the unwanted signals from the hot DUT. There are some games you can play to attempt to do hot TDR with an oscilloscope, but anyone who’s done it would just say it’s not pretty. You don't have much in the way of results and [the process] lacks repeatability.
Back when the serial ATA standard came out, we used oscilloscope-based TDR to measure source impedance. It could take about an hour to get some data that you could believe, but the times I did it, I had no confidence in the measurement. What breeds measurement confidence? Measurement repeatability, the fact the you can take a measurement and hand the whole pile of equipment to someone else and say, "You measure it," and get identical results. There was too much operator judgment and skill, and too much tweaking involved to get it to work right. Doing hot TDR with a VNA takes a couple of minutes, it’s very repeatable, and it doesn’t depend on operator skill or judgment.
K.L.: Certainly when it comes to R&D, most signal integrity labs have traditional oscilloscope TDRs. Is there enough of a value proposition to justify getting a separate instrument for TDR?
C.K.: Besides hot TDR we just talked about, there is another measurement that can benefit greatly from the VNA approach to time domain measurements. That is the measurement of stressed eye diagrams. For example, in the early days, when HDMI was being engineered, the people promoting the standard worked with test companies to develop a technology for testing the HDMI cable. HDMI cables transmit a digital signal, so they thought they needed to test it with digital transmitters, and it wound up taking more than half a million dollars of equipment (see Figure 2). That was purely the cost of test, and that got passed along to the consumer. If you were an early adopter of HDMI, you could pay $100 for a cable. But just because a cable is intended for operation with a digital signal doesn’t mean you have to measure it that way.
Figure 2: The traditional solution for an HDMI cable test (left) requires more instrumentation, introducing costs and complexity. The vector-network analyzer approach (right) is now the preferred approach for the task.
K.L.: You mentioned measurement repeatability earlier. Do the results from TDR measurements taken by VNA correlate with those taken by oscilloscope?
Figure 3: VNA captures return loss (left) and impedance (right) with good correlation to oscilloscope measurements, but in a fraction of the time.
(Courtesy of Agilent Technologies)
K.L.: But an oscilloscope is a pretty flexible tool that can be used for other tasks.
C.K.: An oscilloscope is a versatile time domain tool, but a VNA is a versatile frequency domain tool for tasks beyond just TDR such as insertion-loss measurements. The dynamic range of the VNA is also superior to the oscilloscope TDR. Quite commonly, signal integrity labs possess an oscilloscope TDR, but it limits the different types of devices that you can actually address. Adding a VNA to your signal integrity lab will give you more versatility.