Saturday, May 12, 2012

High-Performance HF Transceiver Design

From EDN Online.
It's an excellent article that describes the State-Of-The-Current-Art in radio design.
I'd expand on the article a bit more, but I'm too tired to care about much tonight.......


  1. It's nice, but he's still a bit short of what certain folks I know are doing ;-).

    Take the Perseus SDR (bottom) architecture and replace the 14 bit ADC with a 16 bit converter with better than 100 dB SFDR and it gets better still. Really, all the magic is in the DSP and it does require a lot of throughput. Nothing a modern digital team can't do.

    Then ask yourself "Why should I bandpass filter the input? If the dynamic range of the converter can handle all the signals, nothing off frequency is going to bother it." So you can literally replace filters with current in amplifiers. You set the IP3 so that any expected signals won't cause an IM product above some level you figure you can tolerate. Then remember, you might hear harmonics of service at a lower frequency, and there's no way to avoid them (you probably hear them now). The filter won't remove the harmonic - it's on the channel you're tuned to. THe linearity requirement is you create less harmonic than the external stations provide.

    I've played for long periods with receivers that had virtually no filtering at all (except for the lowpass/anti-aliasing filter at the ADC). They work great.

  2. Yeah, the guys that do the coding make about 90% of the magic happen.
    My FlexRadio 5000A gets better and better with each software release, even though the hardware hasn't changed since I bought it seven years ago.
    When I lived in Torrance, I was a couple of miles away from KNX 1070, a 50kW clear-channel powerhouse. I wound up having to put a high-pass filter with a 3MHz cutoff in front of the radio to keep the crud out.
    KNX has a reputation of being able to be heard on your fillings in Torrance even though they take great pains to have a signal much cleaner than the FCC requires. It's "The Rusty Bolt" effect times about a million due to the area they're located in.
    Way back when I worked for Hughes Aircraft, I was involved in some of the early SDR research. The ADC's available at the time were pretty crude, along with the software and computers, and you absolutely HAD to have bandpass filters on the front-end or you'd get _nothing_ usable out.
    Some of the stuff out now in the Military and various 'spook' markets is just flat out amazing.

  3. Check out the LTC2207/2208 family. They still seem to be the peak of the performance curve in HF ADCs. I've run one at 100 MHz and really gotten a 105 dB SFDR. Linear Tech sells an eval kit that makes it into a wiz-bang spectrum analyzer - the best I had in the HF lab.

    One effect I noticed that isn't well-documented is that these converters generate harmonics of their input and it comes out of the input pins! We were failing a 60 dB harmonic rejection test and pulling our hair out because the analysis said it should have been over 100 dB down. I disconnected the RF path into the converter, and sure enough, it was impossible to find. Then we hooked up a transformer coupler onto the input and saw the harmonic coming back out of the ADC at -50 something dBc.

    The flex radios are cool, but their hardware architecture is "so last century". Even a low end OTS decimator chip like the TI Graychips or the Analog Devices 6624 can make a really excellent filter.

    The down side is that us RF engineers become almost non-existent in the radio companies. The typical program has an RF guy, a processor designer and 10 SW engineers.

  4. WELL......FlexRadio Systems just released their newest model.
    Read all about it here:


Keep it civil, please....