At 12:24 AM 8/25/2006, Ian Wade wrote: >All, > >I have just been reading a review on the SDR-1000 by Peter Hart, G3SJX, >in the June 2006 issue of RadCom. > >He is generally upbeat, but he has a concern about image rejection. He >is talking about v1.6.0 of the software, so maybe things have improved >since then. He says: >
>~<snip> >Similarly on transmit there is also an image. This can be separately >nulled but a second receiver or preferably a spectrum analyser is >needed. >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > >I don't have any SDR hardware (yet), so I'm not able to see these >defects. However, it does raise some questions in my mind: > >1. Is it really a matter of having to automatically/manually null the >image every time you change band or move through a large frequency >increment? It's actually a bit more complicated than that... But, overall, you're right.. -> The SDR1000 isn't like a microwave radio or cellphone with an sub-octave bandwidth analog image reject mixer, where you can tweak it once and have it work everywhere. Think of it more like two separate receivers, fed with Local Oscillators that are (approximately) 90 degrees out of phase, with each receiver having slightly different IF characteristics. You need to calibrate both the LOs and the IFs for each, and both have a multi-octave (if not multi-decade) bandwidth. Also, to get 80 dB image rejection, you need a pretty impressive phase/amplitude match: approximately arctan(0.0001) (0.005 degrees). 20 dB only takes 6 degrees, 40 dB about 0.6 degree, so to get those last 10 dB is definitely down in the gnat's eyelash territory The I and Q Local Oscillator signals run through different low pass filters (between DDS output and QSD/QSE), so their relative phases and amplitudes change (slightly) as the frequency changes over a multiple octave range (just because of component tolerances, if nothing else). This effect is fixed, and could be calibrated over frequency once. For any given "DDS tuning frequency" there's a few calibration parameters that could automatically be applied. Based on my measurements, I think that the Tx and Rx sides are pretty consistent here (with a fixed offset between Tx and Rx), so one lookup table (LO cal parameter vs LO frequency) would do. {There's a temperature dependence too, but for amateur applications at room temp, you can ignore it} Then, the I/Q audio signals run through separate channels, so there's a "audio frequency" dependent variation between I and Q. This is fairly fixed, independent of frequency, but is also not particularly well represented by a single phase/amplitude calibration value (as currently used in PowerSDR), especially when used with wideband audio interfaces (e.g. 96 kHz sampling, etc.). The audio interface manufacturers do a fairly good job keeping phase and amplitude matched beween channels in the middle ranges (say, 100 to 10 kHz), but not so wonderful farther out: their primary criteria is making sure it "sounds right" and phase/amplitude problems in the lower end of the range would result in "stereo imaging" artifacts {Phase difference between L and R being one of the big cues for how you tell what direction a sound is coming from}. I don't want to give the impression that all is lost and hopeless, because the matching between channels is quite good, just because they are identical components from the same lots assembled together and operating in the same environment. The simple single point calibration that's currently used works pretty well. It's just not perfect. >2. Do you really need a second receiver/spectrum analyser to make sure >you're not transmitting an unacceptably high-level image? Do you have to >keep checking every time you change transmit frequency? Not as a practical matter. Recall that (in the U.S. at least) the spurious emission requirement is only 40 dB. From on-air observations of a variety of signals from a variety of transmitters, there's an awful lot of ham signals that do NOT meet that requirement, mostly because of overdriving or misadjusted equipment. Fortunately, most of the signals that *I* receive are only about 20-30 dB over the noise floor, so I can't see those spurious emissions. {Cynically, I would comment that the current obssession with low receiver LO phase noise is superfluous, because the transmitted signals have terrible "far out" noise skirts that completely mask a quiet receiver}. If a more sophisticated calibration/compensation process were used in PowerSDR, the LO side would be calibrated once (and could be checked periodically) using the receive side only, and that would probably hold for the transmit chain as well. The audio part is different between Tx and Rx, so independent correction is needed, and there's a slight complication here, because part of the chain is inside the audio interface and part is in the SDR1000. You can calibrate the audio interface part by looping the audio out to the audio in, but there's no way in the SDR1000 to loop the audio chains "inside" the SDR1000. You could calibrate this once, and be done with it, because it probably wouldn't change much over the life of the radio (perhaps it could be done at the factory during checkout, and the calibration data supplied as a data file?) There IS an impulse generator in the later revs of the SDR1000 (not in mine, as it happens) which can be used to calibrate the receive chain, although the current version of PowerSDR does not support its use. I don't know the details of where the impulse is injected, nor what the spectral properties are, so I can't comment on how well it would work (esp if you wanted 80 dB kinds of IR). There have been a couple examples of the I/Q data from the impulse generator published which look encouraging. From a theoretical standpoint, this can work quite well. I can also say from some laboratory measurements on a set of 4 SDR1000s using fairly grotty on-mobo sound interfaces that it is possible to get image cancellation over a limited bandwidth (a few kHz, at least) that is better than can be measured with state of the art instruments. >3. Is there any plan to have a frequency-based look-up table for optimum >rejection, both on receive and transmit? (Similar, I guess, to automatic >ATU tuning). I've heard it talked about for several years (ever since the new rev of the board with the impulse generator was released), however, since the existing scheme is "good enough", it's probably not real high on the priority list. While conceptually simple to do, there are a fair number of practical details to integrate it into dttsp. I'm sure that the dttsp developers have some partially implemented versions of it lying around, but, as with all software, the last 20% of the implementation is 80% of the work. > James Lux, P.E. Spacecraft Radio Frequency Subsystems Group Flight Communications Systems Section Jet Propulsion Laboratory, Mail Stop 161-213 4800 Oak Grove Drive Pasadena CA 91109 tel: (818)354-2075 fax: (818)393-6875 _______________________________________________ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz Archive Link: http://www.mail-archive.com/flexradio%40flex-radio.biz/ FlexRadio Homepage: http://www.flex-radio.com