I/Q (de)modulation in hardware or in software

[mweymarn]

Dear signal processing experts,

I am planning the development of an aeronautical test transceiver. One of the signals to be supported is the future standard of LDACS1. It operates in FDD mode, in the aeronautical L-Band (around 1 GHz). The modulation is OFDM, with 50 sub-carriers and a symbol duration of 120 microseconds. The occupied bandwidth is 500 kHz.

With the bandwidth being fairly narrow, I am debating the modulation and demodulation approach. The conventional analog zero-IF architecture with dual converters and  a quadrature (de)modulator seems overkill. We would also need to deal with DC offsets and I/Q imbalances.

The alternative would be to use a single ADC or DAC respectively and do the (de)modulation in software (FPGA, Xilinx Zynq). In theory that should work. But are there any practical downsides to this approach?

Many thanks for your inputs.

Martin

[coppice]

What's conventional about an analog zero-IF architecture with dual converters architecture in 2017? People mostly want to use a single converter, to avoid matching issues between multiple converter channels, these days.

[emeb]

Sounds like a fun project.

I'm a big fan of single-converter IF sampling systems. You can use inexpensive RF downconverters to get your L-band signal down into a sub 10MHz IF and then sample with a low-cost ADC. In digits you tune down to DC and simultaneously convert to I/Q. Filter, decimate, AGC and demodulate. Easy.

[scatha]

I haven't read up on that particular system, but most OFDM waveforms tend to avoid using the centre carrier(s) meaning that a DC offset is no big deal in direct conversion systems. If it is a problem, just offset the carrier then mix the signal to 0 IF in software/FPGA. Likewise, I/Q balance in such a narrow-band system should be no biggy -  just characterize it and apply the appropriate affine transform in software/FPGA.

The alternative is to use a 70 MHz IF (*very* common satcom and aeronautical IF = lots of RF hardware available) then directly sample. Assuming the output signal is band-pass and the ADC has sufficient bandwidth you can exploit sub-Nyquist sampling - certainly sub-2MSa/s should be possible.

[cncjerry]

I would look at what ADI is offering.  They have so many I/Q modulators/demodulators today you could probably do it with one chip.  I know I'm exaggerating, but every time I look there is some new level of integration offered. 

[mweymarn]

Thanks everyone for your input. Yes, the DC subcarrier is indeed not used. I forgot to mention that they use variable modulation (QPSK, 16QAM and also 64QAM).

I have also been thinking along the lines of converting to 70, filtering and then going for sub-Nyquist sampling. Viable for 64QAM?

ADI has some neat stuff. It is tempting to just slap an AD9364 in there and be done with it. My concern is that this is to be a long-term design, and we're in trouble if these chips go obsolete. I'm hoping that ADCs/DACs will live longer, or at least will be easier to replace compared to one of these highly integrated radio chips.

[cliftonburton]

Sounds like a fun project.

[NiHaoMike]

What about a RTLSDR for the receive side?