NOKOMIS, FL, August 12, 2011
A head-to-head comparison of Agilent Technologies’ PXA Signal Analyzer and National Instruments’ new PXIe-5665 Vector Signal Analyzer was featured as part of the keynote presentation at NI’s annual NIWeek conference. During the demo, NI claimed that the adjacent channel power ratio measurements were similar at about -80 dBc but the NI product was as much as 15x faster. And, when the computations were performed on a Compact RIO FPGA, the speed differential increased to a factor of greater than 200: a truly amazing improvement.
Agilent experts refuted and clarified many of NI’s claims on Aug. 11 during a phone conference with editors at EE-Evaluation Engineering. On Aug. 15, NI issued a revised white paper describing the test setup used in the demo. Five main points emerged.
1. The NI and Agilent instruments saw identical input signals, both conditioned by a 248.6-MHz WCDMA filter. As NI stated in the white paper detailing the demo setup, the intention was not to measure the instrument ACPR, but rather to show that the best measurements from the two instruments were similar—in this case about -81 dBc.
2. According to Agilent, the PXA has a FAST ACPR mode that makes the measurement in 14 ms without averaging. NI stated that, because readings typically have to be transferred back to the host PC, the transfer time was included along with the actual measurement time. PXIe has much lower latency and a faster transfer rate than either LAN or GPIB. Taking all these factors into account, NI reported 420 ms for the Agilent PXA in IBW mode with 10 averages vs. 30 ms for the NI 5665. With averaging off and the PXA in FAST ACPR mode, the PXA takes 60 ms and the NI 5665 5 ms. ACPR measurements were similar for both instruments with and without averaging.
3. During the demo, NI calculated the PXA third-order intercept (TOI) to be +19 or 20 dBm. In fact, Agilent measured +36 dBm for the demo input signals. TOI is maximized by optimizing the signal levels reaching the mixer; and with 24-dB attenuation, the PXA really does have +36 dBm. In the demo, this test was performed with 0-dB attenuation for both instruments, resulting in +19 or 20 dBm for the PXA and +27 dBm for the NI 5665.
4. A number of harmonics were measured, and the NI 5665 was 1.7x faster: 404 ms vs. 720 ms. For both instruments, the list mode was used to measure the 2nd, 5th, 10th, and 14th harmonics of a 1-GHz signal. NI explained that, because each instrument changes band at 3.6 GHz, they wanted to include the time to change band. Agilent claimed that list mode could make the measurements in 233 ms. It is not clear if this time includes either the band change or data transfer time.
5. The cost of the PXA depends on what performance you actually need. For an apples-to-apples comparison, NI has added the cost of 50-MHz bandwidth ($15,435), an electronic attenuator ($3,093), a preamp ($1,900), and a microwave preselector bypass ($5,361) to the standard 40-MHz PXA list price ($61,956) for a total of $87,745 vs. the NI 5665 with PXIe chassis and MXI controller ($3,198) at $58,197. Agilent points out that the MXA and EXA models provide 25-MHz bandwidth but otherwise similar specifications start at about $30,000.
There are several important aspects to the situation. Today’s exceedingly complex instruments offer many ways to make an incorrect measurement. Each part of the test setup must be analyzed to ensure that the final number actually means what you think it does. Once that is ensured, you must be certain that the instrument has been set to its optimum mode to deliver the best performance.
When the music is rockin’ and you are with 3,300 fellow test system developers, a well-produced, high-tech demo looks pretty good. Although the speed of the Compact RIO FPGA cannot be disputed, almost all of the other claim/counter-claim acrimony and confusion could have been avoided by running a technically correct demo developed by an independent third party not affiliated with either company.