Comments on MiCOM Labs test of xMax

On June 1, 2007, xG Technologies released a report by MiCOM Labs of their xMax technology. The report itself was dated 16 October 2006.

The test compared xG's 900 MHz transceivers against an unspecified commercial 900 MHz digital radio modem while carrying digital video from an MPEG-4 encoder. The most relevant part of the test included a short radio link and measured the peak RF power required to "close" each link, i.e., to allow the receiver to function.

There are several problems with this test, the main one being that the commercial digital radio used for comparison was not specified. There are many digital radios on the market. Some are good and others are bad. We aren't even told the comparison radio's data rate; we only know that it's greater than that required to carry a MPEG-4 video stream in real time. If the commercial radio operates at a significantly higher peak data rate than the xG transceiver, it would have required considerably more peak RF power than an equally performing radio operating at the average rate of the MPEG-4 encoder.

A more meaningful measurement would compare the average RF powers of the two radios while operating at the same average data rate. Even then we'd have to ensure that the commercial radio did not "pad" unused bits in a way that would inflate the required RF power.

MiCOM's use of a short RF path in the tests was baffling. This merely complicated the tests and introduced additional factors that had to be controlled. End-to-end cable paths with sufficient attenuation would have been simpler and more stable.

Where's the Eb/N0 Measurement?

But there's a more fundamental problem with the test. Operating power measurements are useful when comparing complete systems, but they are not meaningful measures of a modulation and coding scheme -- which is where xG claims to have innovated. A higher RF power requirement by one receiver could just as easily reflect a noisier preamplifier or mixer as an inferior modulation and coding scheme. To measure only the performance of the modulation and coding, the effects of receiver noise must be eliminated.

This can be done with an extra piece of test equipment, a RF noise generator. The idea is to add enough noise to the input of the receiver under test to "swamp", or render unimportant, the noise produced internally in the receiver. Then the signal-to-noise ratio (SNR) seen at the demodulator will be the same as at the antenna terminals where it can be controlled.

This test should take the user data rate into account so that the required Eb/N0 can be computed. Eb/N0, the ratio of the energy per bit to the noise power spectral density, is a fundamental figure of merit for every digital modulation and coding scheme, and it is the most meaningful way to compare one scheme to another. There are several ways to perform the measurement, and equipment exists to do it automatically. It can also be measured manually with a spectrum analyzer, a wideband RF noise generator, an RF attenuator, a hybrid combiner and a hybrid splitter. The manual procedure is as follows:

  • Connect the signal source and the noise generator to the input ports of the hybrid combiner. Use the RF attenuator to vary the signal level into the combiner if the signal generator output cannot be varied. Connect the output of the combiner to the input of the splitter. Connect one port from the splitter to the receiver under test, and connect the other port to the spectrum analyzer.
  • Turn off the noise generator and turn on the test signal generator. Make the resolution bandwidth (RBW) of the spectrum analyzer wider than the test data signal. Adjust the signal level using the RF attenuator so that it is strong, but not so strong that it overdrives the receiver. Note this level.
  • Turn off the signal source and turn on the noise generator. Set the resolution bandwidth of the spectrum analyzer to the system bit rate. For example, if the data rate is 5 Mb/s, set the RBW to 5 MHz. Note: do not set the RBW to the signal bandwidth or to the modem symbol rate. The bit rate is whatever the modem user sees.
  • Adjusting the analyzer's video filter as necessary, monitor the noise level on the spectrum analyzer while adjusting it to reflect the Eb/N0 to be tested. For example, if the signal amplitude is -30 dBm with the RBW capturing the entire signal, then setting the noise level to -40 dBm in a RBW set to the user data rate will establish a Eb/N0 of +10 dB.
  • Turn on the signal source (so both it and the noise generator are on) and see if the demodulator is working. Measure the bit or packet error rate as appropriate.
  • Repeat the process for other values of Eb/N0 until the packet or bit error rate vs Eb/N0 "waterfall" curve can be adequately characterized.

    Summary and Conclusions

    MiCOM's report doesn't really say anything about the xMax equipment except that it works if the received signal is sufficiently strong. It was compared to an unspecified commercial transceiver whose performance is not characterized, and the most important measurement of any modulation and coding scheme, Eb/N0, was not made.

    Phil Karn, 4 June 2007