• • • TEST & MEASUREMENT • • •
Embracing AI: High-speed interface compliance testing and beyond
High-speed-cable testing can take several days and require substantial effort. The Rohde & Schwarz automated compliance test solution reduces this test time to mere minutes
he internal and external high-speed interfaces face rapidly increasing performance expectations. Modern chatbots such as ChatGPT from OpenAI, Microsoft’s Bing Chat or Google’s Bard are driving the trend. Since these chatbots rely on high-performance computing, hardware vendors introduced a modularized computer with three goals: increased data transfer and storage capacity, improved maintenance quality and lower operational costs. A crucial challenge is managing these huge data quantities on high-speed cables where they are transferred from module to module and server to server.
T
High-speed technologies: IEEE 802.3 and PCIe
External cables, especially direct attach copper cables (DAC), often follow high-speed ethernet computer network technologies, such as IEEE 802.3100Gigabit ethernet that can support rates of 100 Gbit/s. The inner data traces and cables use peripheral component interconnect express (PCIe) technology. PCIe 6.0 supports rates up to 64 Giga transfers per second (GT/s). Internal cables and data traces are used to connect the host processor to nearby peripherals and add-in cards, such as display cards or network interface cards (NIC).
Fig. 2: High-speed cable with 2 lanes and corresponding measurements: Through (THRU) and crosstalk from transmit neighbor lanes to victim lane (NEXT, FEXT)
Testing high-speed cables High-speed cables are constructed of lanes, which are basically copper wires organized as differential signal paths. IEEE 802.3 cable verification requires testing the transmission (THRU) of each lane and crosstalk for all neighboring lanes: near-end crosstalk (NEXT) and far-end crosstalk (FEXT). For more information, see [1]
.
When higher data rates are needed, the number of lanes in a cable also increases. Manual testing is repetitive, time consuming and error prone. The
main cause is the increased number of reconnections and measurements with a 4-port vector network analyzer (VNA). Testing can take up to several days for each cable, such as eight-lane QSPF-DD/OSFP-cables, where users must perform 256 manual measurements and cable reconnections.
A further challenge is post-processing data collected in line with the standard to determine whether a cable passes or fails. The ever increasing complexity and variety of the metrics in various standard specifications make for a heavy workload. IEEE 802.3 defines the channel operation margin (COM) and effective return loss (ERL) as metrics. Excursion metrics for integrated return loss (iRL) and NEXT/FEXT component-contributed integrated crosstalk noise (ccICN) have been introduced for PCIe.
Time efficient turn-key solution Rohde & Schwarz has developed an automated test solution for high-speed cables to relieve test engineers of time consuming and repetitive tasks. The solution combines the precision and performance of an R&S ZNB vector network analyzer with the R&S OSP open switch and control platform and the R&S ZNrun VNA software suite for faster, easier and more reliable measurements. The R&S ZNrun Cable Test Client together with
the R&S
Fig. 1: Rohde & Schwarz solution for automated high-speed cable testing with the R&S ZNB vector network analyzer and three R&S OSP320 open switch and control units controlled by R&S ZNrun software.
30 ELECTRICAL ENGINEERING • OCTOBER 2024
ZNrun-K4xx compliance test automation options make for a straightforward test procedure with three simple steps: set up, calibrate and measure. Each step is optimized to minimize calibration and measurement times.
electricalengineeringmagazine.co.uk
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50
