Tricks of the Trade: Signal Integrity Devices

October 28, 2019 Jason J. Ellison

As with any job, Signal Integrity engineers use niche tools. SI tools are often thought of as S-parameter viewers, 3D EM solvers, or measurement equipment. These tools have one thing in common: analysis. What about the devices that SI engineers deploy to make their lives easier? In this article, I’ll discuss three devices to save time and money: multiport test cables, RF switches, and the old RC filter.

Multi-port RF Cable Assemblies

Today’s high-speed serial transmission standards guide us on how to develop 56Gb/s channels. To adequately evaluate these channels, we need to measure channel performance up to at least 28 GHz. At that frequency, we use 2.92mm coaxial connectors. The vertical compression mount style connector, which has been the most popular version for over a decade, is expensive, and the compression pin of these connectors moves into the connector body upon application. When the connector is removed and applied to another board, you have a 50-50 chance of it working. Typical SI test fixtures have 32 of these connectors, and we SI engineers use a lot of test fixtures! The costs add up quickly and could be crippling. 

Fortunately, advances in compression mount technology have made this a non-issue, because there are reusable cable assemblies that act as vertical surface mount coaxial connectors. The initial investment of these assemblies is high; however over the course of a year, these assemblies can save your company hundreds of thousands of dollars over coaxial connectors. These cable assemblies have between 8 and 32 connections which means you can use them with your 8 to 32 port VNA and apply only one connector per test. Also, the signal density is much higher than what is achievable using coaxial connectors, and that enables lower loss designs that can be de-embedded out to high frequencies. 

I have two examples of such products that use compression mount technology. First is Samtec’s BullsEye connector and second is Ardent Concepts Terminator.

Multiple rendered images of Samtec BullsEye BE70 connector systems capable of 70GHz of bandwidth. Image courtesy of Samtec, Inc.

Samtec’s BullsEye BE70 connector system capable of 70GHz of bandwidth. Image courtesy of Samtec, Inc.

RF-broadband Switches

RF switches can be used to speed up testing, save space on a PCB, or save money by using fewer coaxial connectors. There are two major types of broadband switches.

  • Semiconductor

  • Electromechanical

Electromechanical switches come in a large number of bandwidths and port counts. The bandwidths range from 1 GHz to 67 GHz, and the switch can be as large as a single-pole-12-throw, or SP12T. The poles are the number of common nodes between ports, and the throws are how many positions the switch can be connected to. That means that by connecting the poles of four SP12T switches to each port of a four-port VNA, you can make 48 measurements without changing the connections! That is very powerful, but I do have some cautionary notes before you go off and buy a set of four of these. 

1. Not all switches are created equal.

There are two tiers of these switches

  • Standard
  • High Repeatability

Standard switches are great when phase is not a concern. They work wonders for most SerDes, antennas, or two-port devices. High repeatability switches are needed for multi-mode measurements. The phase does not change between throws which results in consistent measurements and reliable mode-conversion results. 

2. Some assembly required.

RF switches are basically cans, and they are not great just laying on the lab bench. In practice, they are best used when they are mounted to a chassis or a rack. Designing the mounting device is relatively easy, but it is something else you’ll need to do if you want to use these. When I started with these, I made rough rig for a proof-of-concept. So don’t be afraid to just get in and make something!

Various lab devices in the background with partially loaded SP12T Radiall switches mounted to a plexiglass frame in the foreground.

Partially loaded SP12T Radiall switches mounted to a plexiglass frame.

Here are two RF switch websites to take a look at!

www.dowkey.com

www.radiall.com

The semiconductor switch is small and easy to use. The general control circuit is basically the same as the electromechanical, but the semiconductor switch form-factor mounts straight onto a PCB. That could be a positive or a negative. If your DUT (device under test) is on the same PCB as the switch, then these enable you to eliminate many connectors from the design with one or two switches. Since the circuit is already on the PCB, you don’t need to design additional mounting hardware. Another advantage is that these switches are extremely phase stable. As a result, there are no worries about mode-conversion issues with these! They also have faster settling times and are very inexpensive when compared to electromechanical switches. On the other hand if you need to connect from the PCB where the switch resides somewhere else with a coax cable, the result is more connectors, cost, and loss. 

Semiconductor switches thrive when used for optical device testing, SerDes testing, and high-volume RF devices testing. The first two scenarios are great because you can place a repeater after the switch to get great repeatable signal integrity into your device under test. In the third scenario, you can set up arrays of devices that can be tested as they are placed into the fixtures. 

You might be thinking that there is no place for electromechanical switches when you can get phase stable, inexpensive switches that don’t need extra mounting hardware. The two drawbacks of semiconductor switches that make electromechanical switches relevant are insertion loss and power limits. Semiconductor switches are usually a few dB of loss per channel, and the PCB trace that connects to the switch adds additional loss. The additional loss limits the effective bandwidth for post processing operations like de-embedding. Also, you cannot use these for any high current applications, or they will burn out. 

Here are two links to semiconductor switch products!

www.analog.com

www.psemi.com

RC Circuits for Equalization

When you think equalization, you probably think of SerDes with built-in auto tuning schemes, but what happens when you are running a relatively simple channel and you don’t have access to these high-end SerDes? Well, you can always use an RC filter as an equalizer. 

Schematic diagram of a parallel RC filter

Parallel RC filter

These work very well all the way up to 10 Gb/s. While you’re not going to make a 112G system work with these, it provides a cost effective solution for legacy designs or PCIe control signals. The impedance for this element is

Z = - j * R / (ω * R * C - j

Where, R is the resistor’s resistance and C is the capacitor’s capacitance. Plotting the magnitude of the impedance, you can clearly see this is a low pass filter. 

Figure showing an inverse relation between impedance and frequency of an RC filter with R = 100 Ohms and C = 10e-12 Farads

Impedance of an RC filter: R = 100 Ohms, C = 10e-12 Farads.

To utilize this filter, start with targeting a data rate. For this article, I’ll use 10 Gb/s. Next, find the loss at the Nyquist frequency. In this case, the Nyquist frequency is 5 GHz. I created an example transmission line with 13 dB of loss at 5 GHz. Its plotted insertion loss is shown below.

Figure showing insertion loss in an almost linear relation in dB scale where the magnitude of a signal decreases as its frequency increases.

Example transmission line with 13 dB of loss at 5 GHz.

To calculate the approximate resistor value, use the following equation.

R = 2 * Z₀ / (10 ^ (loss / 20)) - 2 * Z₀

I have found that this equation over-equalizes a little bit. So for this example, I use -10 dB for loss and 50 ohms for system impedance, and the resulting resistor is about 216 ohms. Next, calculate the capacitor value using the following equation.

C = 1 / (f_Nyquist * R

Now, let’s look at what this does to the eye pattern at 10 Gb/s.

Before:

After:

Wow, what a difference!! Now, there are more advanced schemes for creating these component values, but as you can see, this works quite well. There are solutions like this that are integrated into one circuit, but why pay more when you can do this?

Talk to an Altium expert today to learn more or continue reading about signal integrity tools for PCB designers in Altium Designer®.

About the Author

Jason J. Ellison


Jason J Ellison received his Masters of Science in Electrical Engineering from Penn State University in December 2017.

He is employed as a signal integrity engineer and develops high-speed interconnects, lab automation technology, and calibration technology. His interests are signal integrity, power integrity and embedded system design. He also writes technical publications for journals such as “The Signal Integrity Journal”.

Mr. Ellison is an active IEEE member and a DesignCon technical program committee member.

More Content by Jason J. Ellison
Previous Article
Hardware Design for Edge Computing vs. Cloud Computing
Hardware Design for Edge Computing vs. Cloud Computing

Explore the differences in hardware design for edge computing vs. cloud computing in this article.

Next Article
Say Hello to the First Space Tow Truck
Say Hello to the First Space Tow Truck

Northrop Grumman's Mission Extension Vehicle-1 (MEV-1), whose mission is to essentially act as a space tow ...