Keywords

aeronautical telemetry, equalization, estimation

Abstract

This is the Phase 1 report for Preamble Assisted Equalization for Aeronautical Telemetry (PAQ).

Executive Summary:

To fully leverage the benefits of a periodically inserted preamble, the preamble is used to estimate the state of the channel. To this end the following algorithms have been developed, tested in simulation, and ported to the GPU-based real-time system:

1. The preamble detector, which scans the received samples searching for the presence of the preamble. The location of the preamble in the received samples is required to use the received samples to estimate the frequency offset, channel impulse response, and noise variance. The preamble detector works in the presence of an uncompensated frequency offset and unknown channel.

2. The frequency offset estimator, whose result is used to compensate for a large frequency offset in the RF carrier. The frequency offset estimator operates in the presence of an unknown channel.

3. The channel impulse response estimator, whose result is used by the zero-forcing and minimum mean-squared error equalizers to compute the equalizer filter coefficients. The channel impulse response estimate is also used by the CMA+AMA equalizer for initialization.

4. The noise variance estimator, whose result is used to estimate the signal-to-noise ratio parameter in computing the optimum minimum mean-squared error equalizer filter coefficients.

The performance of three equalizers has been evaluated in simulation using eleven test channels derived from channel sounding experiments at Edwards AFB, CA. The equalizers are the zero-forcing (ZF) equalizer, the minimum mean-squared error (MMSE) equalizer, and the combined constant modulus algorithm, alphabet matched algorithm (CMA+AMA) equalizer. The bit error rate performances of all three over the eleven test channels has been performed. A longitudinal comparison shows that the MMSE and CMA+AMA have almost equivalent bit error rate performance (and the performance of both is superior to that of the ZF equalizer). Other equalization options are also discussed (see Section 9).

Most of the hardware needed to implement the real-time demonstration system has been acquired. These items include the following:

1. Modified L/S-band and C-band transmitters. The modification was the periodic insertion of the iNET preamble and ASM fields.

2. A modified telemetry receiver that outputs inphase and quadrature samples at 2 samples/bit.

3. Two NVidia GPUs for performing the computationally complex equalization algorithms. These cards reside in two rack-mounted host computers, that have also been purchased.

4. An 8-channel bit error rate tester.

5. Portable, shock-proof, racks for housing the equipment, along with the hardware, tools, connectors, cables, fans, etc. necessary to support the experiments.

Because the emphasis of Phase 1 was primarily algorithmic, the report focuses on the description, analysis, and performance of the algorithms that form the equalizers. The Phase 2 report will provide a more detailed description of the hardware, C/C++ code, and the software architecture used to implement the equalization algorithms.