Thursday, September 18, 2008

The FLEX-5000

I recently joined the ARRL, in response to their announcement that members would have access to their archive of QST all the way back to 1915. That was what it took to push me over. A few days later, I get a nice fat envelope from them that contained a welcome letter (which I nearly threw out without first detaching the membership card that was Intel-boogered to it), a solicitation to join the Diamond Club (seems mighty expensive for what little it gets me), and the most recent issue of QST.

So I've been browsing this issue, while recovering from the flu. I'm feeling well enough now to form coherent sentences, so I'm going to blog some reactions. First one is to an ad (QST appears to be about two-third ads), for the FLEX-5000 SDR being sold by FlexRadio Systems. I'm very interested in software defined radio, but frankly the $2799 pricetag for the FLEX-5000A is out of my league. And the $5299 tag for the FLEX-5000C seems way overpriced, given that it appears to just be the 5000A with an imbedded Wintel system (and a speaker), which I can't really see being worth $2500. It's my general impression that high-end ham gear is markedly overpriced, probably because the people who can afford to buy this stuff for a hobby are generally just drowning in spare cash and don't really care about the difference between $2000 and $4000 (or even $10,000). So while I'm interested in SDR, I'm not this interested. Even the USRP2 (at $1400) is less of an investment, although the USRP needs a lot of additional gear to make it into a full-fledged HF radio (mainly amplifiers, as the USRP RF daughterboards have relatively low power outputs).

2 comments:

  1. USRP and Flex5000 are pretty different types of SDR.

    Flex5000 could be considered the current pinnacle of the evolution of "narrowband" SDR, the sort of SDR which is most popular with ham radio.

    In narrowband SDR people use high resolution (often 24bit and with good enough SNR to make use of most of those bits) 'sound card' interfaces running at 48KHz - 192KHz, using either 1 channel (normal sampling) or 2 channels (complex sampling).

    The higher end these systems include multi-stage IFs with deep crystal filtering allowing for close spaced blocking dynamic range well over 110dB. This allows for the sort of "I can DX with a QRP station 20kHz away from my neighbor's 1.5kw station" performers that contesters care about but which no one else should.

    USRP is an example of hardware which is mostly useful for wide-band SDR. Wide-band SDR is of great interest to researchers working on all sorts of spread spectrum stuff in the SHF ISM bands, but so far of less interest in the ham community. This may be because wide-band has only one obvious application on HF and VHF: Concurrently listening to every frequency all at once.

    The deep filters and enormous blocking dynamic ranges of narrow-band SDR are not found in wide-band SDR, since hardware filters would obviously get in the way of running really wide modes.

    Though on the SHF+ ham bands there is plenty of bandwidth to make use of wide-band SDR.

    For USRP you'll want to provide your own filtering and likely low noise preamps, otherwise your selectivity and sensitivity will suck, but since you'll probably be using filters many MHz wide, the fact the the filtering is a bolt-on external to the USRP isn't much of a problem and at higher frequencies you'll really want the pre-amp at the antenna for really good performance.


    You didn't mention the USRP1 which is a more accessible path into SDR than the USRP2. ($700 vs $1400)

    USRP1:
    Four 64MHz ADCs(12bit) and 128MHz DACs(14bit) (used in pairs performing complex sampling, so it can capture a 64MHz bandpass; the fact that the DACs are faster allows for software oversampling which results in less need for TX filtering).

    The two pairs of ADCs and DACs means that you can have two different daughterboards, for example a 0-30mhz transceiver and a 1200-1300mhz transceiver.

    The FPGA on the USRP does decimation/upsampling to match the ADC/DAC sampling rate to the sample rate send/received from your computer.

    USRP1 is interfaced over USB2.0 which limits you to 32MBytes/sec in each direction which means that with the (normal) 16bit samples you're limited to an 8MHz chunk of spectrum, or 4MHz if you're using two daughter-boards concurrently in the same direction.

    You could use smaller 8bit samples and double the bandwidth, but usually the larger bandpasses are when you need more dynamic range.

    Theoretically some processing could be done on the FPGA ahead of the bottleneck (such as demodulation), but the USRP1 FPGA is pretty much out of room.

    USRP2 has a pair of 100MHz 14bit ADC and 400MHz 16 bit DACs, along with a gigabit ethernet interface back it up. It's able to pass a 25MHz band-pass down to the host using 16bit complex sampling. So it has a lot more bandwidth and dynamic range, but it only supports one daughterboard. (you can gang up multiple USRP or USRP2s to drive antenna arrays or many bands, if your wallet can support it :) ).

    USRP2 has some nicer clocking options: it supports a 10mhz refclock. If you want to use a GSP locked OXCO with USRP1 you'll need to generate a 64MHz signal and do a little soldering. (the clock on the USRP1 is not super accurate)

    ATV needs 6MHz, and I'd expect any data mode for ham radio you're likely to cook up yourself would work well enough in an 8MHz. My digital voice modes are far narrower than that. So for ham use the I think the only obvious application for bandwidth beyond a few MHz is the whole monitor everything at once technique.

    So for HF SDR, A single USRP2 could capture/concurrently transmit on the entire HF spectrum all at once. USRP1 could not. However, antenna and dynamic range issues means you probably wouldn't want to do that: You wouldn't want 1.5KW stations in 20m drowning out low power stations in 30m. Instead you' probably be better off with two USRP1 and four daughterboards, each working with a couple of bands. It would cost the same as USRP2 but you could grow into it incrementally.

    For VHF: The 2m allocation fits within the USRPs bandpass entirely. USRP2 isn't broadband enough to cover any adjacent bands, so you'd likely be better off with a USRP1 for VHF work. With a single USRP1 you could capture the 2m band twice in parallel, allowing you to perform software antenna phasing of a two antenna array. (However, there is currently no transceiver daughter board available that covers 144MHz, so if you want to transmit you'll need to build a transverter or wait for the new WBX0510)

    For 1.5m the story is the same as 2m.

    For 70cm USRP2 would let you work with the pretty much whole band at once where USRP would not.

    33cm is like 70cm, although interference from ISM gear in this band may create dynamic range problems that push you back into USRP1 level bandwidths.

    All the higher allocation (except 2300-2310) are so wide that neither the USRP or USRP2 comes close to capturing the whole things at once. 23cm is 60MHz wide, for example.

    Perhaps you're only interested in some subset of the band where USRP2 captures it and USRP1 does not or perhaps the increased dynamic range of USRP2 might be useful for you… But generally I think USRP1 is better suited to ham use. At $700 it's more affordable, around the price of a basic modern HF rig, and not too far beyond once you add the required accessories. It's unfortunate to not see more hams messing with this stuff.

    (Bias disclosure: I own two USRP1s, and am not likely to be able to justify spending on a USRP2 anytime soon)

    ReplyDelete
  2. FlexRadio Systems has just introduced the FLEX-3000 which is a mid range software defined radio with a 100w full HF - 6M transceiver with a built in ATU standard. This unit is starting at an introductory price of $1499 until November 30. This product will be shipping early January.

    Check the website for details

    ReplyDelete