At Milliwave Silicon Solutions, we are witnessing first-hand the fast pace development of a multitude of new millimeter-wave (mmWave) solutions addressing three main market pulls: ACC (77GHz), 802.11ad and 802.11ay at 60GHz, and of course 5G (NR) mmWave. These solutions, due to the nature of mmWave propagation, rely heavily on the quality of their beamforming capability at the algorithm, RF, and antenna level.
Traditional anechoic chamber setups for microwave radiation pattern measurement, even when adjusted specifically for mmWave, have some critical shortfalls:
1) Form factor
There are two obvious conflicting forces for defining the correct size of radiation pattern test fixtures.
a) Large chambers are desirable because measurements should be done in the far field. For mmWave, far field may not be as far as one might think. For example, a 6cm array at 28GHz would be in the far field at about 60cm distance and standard chambers are often oversized.
b) Small chambers are desirable because they help to resolve two issues: cable length and the number of chambers that can fit in a given lab space. The cable loss is high at mmWave, and shorter cables give the best signal integrity. The number of chambers in the lab directly impacts the time to market. In mmWave, unlike for most RF design and due to the importance of beamforming performance, multiple teams of engineers from SW to baseband to RF to antenna will compete for access to the chamber to adjust their designs, and some of the tests may take hundreds of hours to complete.
Having one big chamber is not practical and can slow the engineering progress. Instead, our approach has been to have mini-chambers which fit on a single lab bench and still leave space for engineers to place their laptop and equipment. After years of experimentation, we narrowed it down to 2’ x2’ x 4’ as the most versatile chamber size for mmWave. On top of it we created a solution which can be extended by adding 2’ sections such that in certain cases a 2’ x 2’ x (6’ or 8’) can be created starting from the same original 4’ fixture.
Ultimately the best mmWave lab configuration contains a number of benches with chambers on them.
2) Metal frame
For so many years of radio development, RF test chambers were made as a big metal box lined with anechoic absorbers. Some are made as a copper mesh Faraday cage. In our experience, those do not translate to the mmWave measurement requirement. In a mmWave environment, metal is not a friend when developing high-gain phased arrays. When we started over-the-air (OTA) measurements over 10 years ago, we slowly eliminated metal from the radiation pattern test chamber. Instead we used wood, PVC, and PLA for the frame and nylon screws for the entire assembly. Now in our chambers we have less than 1% metal (the motors) and everything else is either wood or plastic. Our philosophy came from the observation that we are not as susceptible to noise from the outside as we are to multipath inside the box. To further eliminate reflections the inside on the chamber is padded with 4” thick, corrugated absorbers that are specially loaded at mmWave. They provide about 50dB attenuation across the mmWave bands. This is enough to confine the signal inside and avoid interference from another chamber nearby.
Cost here takes into account a number of factors.
a) Our solution’s price tag is only a tiny fraction of what other test equipment makers charge for an equivalent application. This is important because everybody has a given budget to spend on test equipment. What we have learned the hard way is that for mmWave development, to some degree, the more chambers you have the faster the beamforming solution will come true. One chamber is never enough and if the entire budget has to be allocated to it, then the entire effort will be throttled down when engineers form a line trying to get to it. This happens over and over in our experience with new mmWave teams, and this is due to the microwave legacy when one chamber for a whole team is usually enough. That’s because the beam forming effort is often underestimated and it is overlooked that this effort will involve multiple teams who traditionally did not need chamber access or only sporadically. In our experience planning on three chambers is a reasonable assumption for a team of 15 to 20 engineers.
b) Cost is also the fact that lab space can be expensive, back to form factor, and the fact that aluminum and copper are so much more expensive than wood and PVC in our solution.
c) Finally the last aspect impacting cost is the cost of maintenance, the cost of downtime, and the cost of customization. Our solution is made 100% from standard parts. The only parts we buy for our solution are available off the shelf online. All other parts that are specific to our solution are made by us. We can replace any part of our design in a few days and get you back on-track. On top of this the entire solution is self-maintainable and anybody who can assemble an IKEA shelf can entirely take apart and rebuild our complete fixture in a few hours with our included instructions. Obviously the robustness and reliability of our solution is the first line of defense, but in our opinion there is no repair service, 24/7 hotline or maintenance plan that can beat the ability to quickly and easily fix things on one’s own and if necessary, get a replacement part shipped overnight.
4) Cable length and wiring
Another factor completely overlooked that we are seeing on daily basis, whether RF engineers build or buy radiation pattern chambers for mmWave, is the importance of cable routing capability and wiring flexibility. After years of tangling wires ourselves, we finally came up with a unique solution to get any wire, coax, or ribbon from the DUT to the test or control instrument without a passthrough connector and without tangling the wires when the DUT rotates. Additionally, our solution reduces the cable length to the minimal needed, 50cm to 1m is enough in most setups. This has many implications in the ease of use of the system, lower cost of cables and connectors, and improved signal integrity. A bad approach to wiring can ruin a chamber solution which looked great on paper. The motor control is achieved using a daisy-chained, single 3-wire bus (12V, G, D) going to a power supply and a USB dongle. That’s all there is to it and that’s enough.
Through the years we have spent working on mmWave designs, we know that plans change and needs evolve. It is often a limiting factor when solutions are either too custom for a particular application or just a standard build with no flexibility at all. What differentiates our solution is that it is modular at every level.
a) Flexible with tools vendors:
Our solution is not tied to any specific instrument vendor. As long as they support GPIB that’s all we need.
b) Software/OS agnostic:
The motor control is done through a USB to serial dongle with open source high level APIs for Python, MATLAB, Labview, C, etc. The motors are extremely popular in small robotics and are very well documented online and heavily crowd-supported with open-source GPL2 license. We include the entire Python sample code for it, but if one wanted to integrate within a larger framework or use specific postprocessing tools and application, it could be done. Ultimately, the fixture dumps out a .csv data file with coordinates and matching captured value. From there, many types of rendering (3D, heatmap, etc.) and post-processing can be done for to suit one’s needs. We have our own favorites, but we let the user decide for himself.
c) Flexible framing
The chamber frame itself is made of a set of parts which can all be assembled to one another and reorganized indefinitely. The door position and axis can be changed to fit the lab usage or accommodate the user. New sections can be purchased to extend the length or the height. Feet extensions can also be purchase to create a shelf for the instruments below the chamber deck and save bench space.
d) Accommodate DUT form factor
Our solution comes standard with a generic platform where the DUT is bolted onto. This platform can accommodate small and large DUTs up to 200mm x 160mm x 160mm. The platform has 20 mounting holes and 4 cable routing options which can satisfy most designs whether dipoles, patches or mixed antenna arrays. As the rest of the design, if one wanted a particular DUT mounting which cannot be handled by the standard platform, we can design and ship custom made platforms quickly.
6) Angle resolution
The resolution angle of a radiation pattern test chamber is often seen as an important criteria. What we have seen is that a 1-degree resolution for (-90 +90) x (-180 + 180) rotation in (elevation) x (azimuth) gives a full data set of 64800 data points in a single plot and this is more than enough. Our motors could be programmed to go down to 0.08 deg in elevation and 0.008 degree in azimuth, but this would clearly be an overkill, and also a huge waste of time. Three plot resolution options (quick scan at 10deg, full plot at 5deg, fine plot at 1deg) gives the right amount of balance between accuracy and time. Lastly, our solution comes with a built-in laser cross-hair pointer which is used to check and calibrate the nominal (0,0) center point visually.
Traditional microwave type test chambers do not address the needs of mmWave radiation pattern testing very well; they have a lot of shortfalls and are expensive and bulky. Instead, we always chose to build our own, as there was nothing else out there that met our needs, but this required a lot of engineering time. We had many iterations along the years, reworking them each generation, with new parts and vendors. Now our mmWave partners are asking us to help them do these test fixtures for them. At Milliwave Silicon Solutions, we believe we can really help the exploding mmWave marketplace by finally coming up with a simple affordable solution that addresses the needs for a vast majority of cases, by incorporating all the know-how we accumulated over the years doing it for ourselves and our partners.
Our Solution is called it MilliBox™ and it will be unveiled in January 2018.
For more information contact me at Jeanmarc.firstname.lastname@example.org