The Biggest Challenges in Measuring Crosstalk

During DesignCon 2018, we interviewed many PCB experts. Chris Scholz is one of them. He works as a Product Manager for vector network analyzers at Rohde & Schwarz North America and told us all about the biggest challenges in measuring crosstalk.

0:07 Tell us about your role at Rohde & Schwarz.

A vector network analyzer measures what is called S-parameters, scattering parameters. So in very broad terms, it is the transmission and reflection of electromagnetic signals through a PCB or through a transmission line.

0:25 Is vector analyzing relevant when dealing with high-speed signal channels?

Yes, absolutely. And that is when VNAs, Vector Network Analyzers, come into play. Other tools start having issues when you go to very high speeds, very high frequencies. VNAs do have a defined measurement accuracy at high frequencies. So, those are the tools of choice as opposed to, for example, TDR tools that are more popular for lower frequencies. And the nice thing about TDR is that they are very intuitive, you see a time domain trace and that is really nice, when you kind of know, when you are probing for example, for PCB. So TDRs are very powerful, very intuitive, and really easy to use tools.

1:07 What are the limitations of TDR and where does the VNA become better?

The advantages of TDRs is they are very intuitive, easy to use. The problem is when you go to higher frequencies, it is very hard to say where the transition is, a few gigahertz with TDRs, so the measurement uncertainty of TDRs becomes very high. What we have seen from some of our customers is that they measure a certain structure of the TDR, and it passes the measurement criteria. The next day they do the same measurement, same probe, same setup, and it fails.

With the VNA, the measurement certain is defined. You can have a very sophisticated reference plane calibration, and the repeatability is excellent of VNAs. That is something that TDRs cannot do. Plus TDRs become very pricey at high frequencies. So you have got a not very accurate instrument at high frequencies, and a very high cost. VNAs are the tool of choice when you go beyond a few gigahertz.

2:09 Are VNAs more expensive than TDRs?

VNAs used to be super expensive. They have come down in price a lot. They are very competitive. Now if you are talking about a few kilohertz, VNAs tend to be more pricey, but at gigahertz frequencies, given the measurement accuracy, given the false positive, false negatives, the overall cost of VNAs tends to be very beneficial if you consider everything.

2:37 What top frequency of VNA will you need for channel measurement?

So the question is, what top frequency of VNA do you need for, let’s say 40 gigabits per second channel measurements? That’s a difficult question to answer. I mean, the higher frequency the better, of course, but of course the cost gets higher. Usually, people tend to think three to five times the fundamental frequency is what they need. That is for VNAs too. Now for our VNAs, there is an enhanced resolution feature, where you can extrapolate measurements to higher frequencies than the hardware of the instruments can measure. It is a fairly sophisticated algorithm that is using minimal energy techniques. It works pretty well in most cases, and that is a way of using an existing VNA that does not have the frequency for high frequencies. And that is a workaround. This is not something I would say works in all cases, but if you do not have the possibility, you can use the enhanced resolution to get to higher frequencies.

DesignCon Rohde & Schwarz
“For our VNAs, there is an enhanced resolution feature, where you can extrapolate measurements to higher frequencies than the hardware of the instruments can measure.”

3:50 What are jitter and crosstalk? How do you distinguish both of them?

This can be a difficult thing to do. The key is in the tools that you are using. As a signal integrity engineer, you have mostly four tools: bit error rate test set, oscilloscope, TDR, and VNA.

4:07 What was the first test?

Bit error rate test set. A BERT, a BERT. It gives you an overall performance, it is very difficult to drill apart what is jitter and what is crosstalk. Oscilloscopes, they are sophisticated algorithms, where you maybe can identify some of the crosstalk, maybe not, I guess that is still under discussion. TDR we talked about this already, starts getting difficult for higher frequencies, mostly because of measurement accuracy and reproducibility. And with the VNA, it is actually very easy to identify crosstalk because you measure from your aggressor to your victim, and that is where they do crosstalk. So you measure the channel directly, and you identify the crosstalk very easily.

The problem is that the crosstalk tends to… Well, if you have a very high level of crosstalk, you are out of luck anyhow. If you have a relatively low level of crosstalk, you start getting into measurement errors or noise issues with the TDR. TDRs do not have a very high dynamic range. So that tends to be a problem with TDRs. So if you have a lot of crosstalk, you probably know it already, you can measure it with a TDR. If you have a low level of crosstalk, the TDR might just not be the right tool, and you really do have to use a VNA for these measurements.

5:32 What are the biggest challenges in measuring crosstalk?

So the thing I am usually hearing from the people who are new to VNAs, especially people who use very old VNAs that are twenty, thirty years old and still work, you need to be an expert to actually set up the VNA, and most people when they have it set up, they do not touch it anymore, because you are probably not going to get back to the same setup. That is a big problem. So it is a lot of usability problems that the old VNAs had. And with modern VNAs, this is not really a problem anymore. We are trying to make the VNAs easy to use. So with a little bit of training, pretty much anybody can use VNAs. It is a totally different ballpark than it used to be just ten, twenty years ago.

And probing is a really good topic because if you probe PCBs you never really know if your probe made good contact or not. And just looking at the frequency domain trace, that can be hard to distinguish if you are just a little bit off with the probe tips or not. So what modern VNAs have is that you can display a TDR trace and virtually transform at the same time as you display a frequency domain plot. And you can really see if the probes make contact or not.

6:59 Can you make that up?

Yes, you look at the impedance over distance and you see exactly where the probe’s contact, you see a little bit of a ripple in the trace, or you see if they are open and they’ll make contact.

7:12 Are there any guidelines you can suggest to a designer who wants to become better at probing?

Well, you need to make sure that your probe pads are part of your design, and you need to be able to de-embed the probes from your transmission line. De-embedding can be a very challenging thing, especially… And there are different levels of de-embedding that are out there. There is simply chopping off part of the traces, is a simple way that sometimes works. We do support much more sophisticated ways that give you causal results. So you can, depending what you need to do, how accurate you need to be, what kind of frequency you are looking at, you need to dip deep into the toolkit that is available to you to get reliable results.

8:03 Is there a certain part of the transmission line that you have to separate out?

The typical use case is that you have a chip on a PCB and you want to de-embed the PCB. You just want to characterize the chip itself. That is the tip of the embedding setup. So you start talking about issues and you have a reference trace somewhere on the PCB or a separate coupon. Now, given the manufacturing tolerances of a PCB, that can be challenging because your coupon is not always the same as the traces. You know, the more accurate, of course, the less variation you have in the PCB the better, but of course, the cost becomes an issue. So this is kind of something that signal integrity engineers are really concerned about.

8:54 Are jitter measurements possible with VNA?

Yes, VNAs do support jitter measurements. The VNAs measure the channel per se, so the response of the characteristics of the traces on the PCB. And then they use the channel measurements to simulate eye-diagrams or jitter, and so it is a simulated bit pattern and a receiver that is done in software. And that gives the designers a quick and easy way to estimate the channel capacity of a specificity sign before they actually populate a circuit board.

9:33 What brings you to DesignCon?

I have been going to DesignCon for ten years I guess. As a testing measurement company, I think we have a pretty good handle on trends in the industry. DesignCon is a good way of verifying things and kind of understanding what is real and what is kind of fud and what is going to bubble up, what will be a real trend in the industry in the coming future, so it is great to connect to the experts in the field and the users.

10:08 Any advice on how to maximize your experience at DesignCon?

Well, first of all, be early and find a good parking spot!

And then, it is just a great community, the signal integrity community is a great community everybody is really trying to help each other, connect with people. It is just a very diverse, very open community. It is great to hang out in the coffee shop down here and chat with some people and understand what is going on and what they are up to.

You might also like:

Z-Zero’s Stack-Up Tool for Digital Hardware Engineers to Collaborate With Manufacturers

Related Articles