It would have taken us 2 weeks (more than anticipated) to put our finger on the problem… We would have never found the root cause of the problem. That would mean a good chance of a recall in the future!
Maor Schvartsburd, RF Engineer | BAE Systems Rokar, Israel – August 14, 2016
BAE Systems Rokar evaluated the EMxpert ERX+.
We finished the analysis of the EMI problem.
Please see the description of the problem and the process of elimination using the EMxpert ERX+.
- We encountered a problem after changing a firmware version in a well-established hardware.
- We tried to put our finger on the problem using a simple sniffer probe and a list of potential frequencies capable of interference by themselves or their harmonics, the list was extremely large and had no baseline of power for comparison. We estimated it will take ~ 6 days involving 3 people and a mechanical setup to find the source of the interference. Therefore we decided it is an excellent opportunity to prove the EMxpert ERX+.
- After a short setup time and explanation (4 hrs) we started broadband scanning of the old Firmware version, we then changed the firmware version and repeat the process.
- By using MS EXCEL, we narrowed the possibilities of EM levels to 4 frequencies with an increase of 6 dB or more than reference.
- After a more narrow scan, we started running a high resolution spatial scan at all 4 frequencies and found 2 locations where the EM power was maximal: the DDR RAM and the FPGA.
- After pinpointing the sources (interference was ~ 1.3 GHz and 1.6 GHz) we placed an RF absorber (a 1” sq sheet ) on the DDR RAM (our initial feeling), but there was no difference in system performance. We then moved the absorber to the FPGA, the difference was dramatic.
- After showing the data to the FPGA people they ran some tests and we feel we have the solution and understand the problems.
Problem: There are several PLL used in the FPGA for internal computation. That part was not changed in all versions, but in the last version it seems to have caused a big problem.
Solution: The FPGA synthesis program has a random seed when compiling a new FPGA code. Using the EMxpert ERX+, we proved that the bulk of the design causing the problem was moved to a different corner in the FPGA (a big FPGA), and GND coupling or decoupling capacitors’ design is not optimal. We then asked the FPGA people to remove the specific PLL causing the problem, and tested it. The system functions well. So, we need either to move the internal frequency to some better frequency, or force the FPGA to start from the correct corner (in our opinion).
The EMxpert ERX+ learning, tests and analysis took us 13 hrs in total and was mostly done by 1 person. After understanding the problem, we are positive that it would have taken us 2 weeks (more than anticipated) to put our finger on the problem with the first suggested method.
Moreover, as a part of the process we would ask the FPGA people for a different version, and since the compilation is random, there is no way of knowing if the problem would not have disappeared, and we would have never found the root cause of the problem. That would mean a good chance of a recall in the future, when a new firmware is sent to the clients and the system performance is not to spec.
EMSCAN is the world leading developer of FAST magnetic very-near-field measurement technologies and applications since 1989, providing real-time test solutions to antenna and PCB designers and verification engineers, without the need for a chamber.
The EMxpert, a family of compact EMC and EMI diagnostic tools, and the RFxpert, a family of antenna measurement tools, enable engineers to quickly optimize their designs. EMSCAN solutions dramatically increase designer productivity and substantially reduce time-to-market and project development costs. www.emscan.com