|Subject:||Re: [Discuss-gnuradio] ham/amateur getting started|
|Date:||Fri, 25 Dec 2015 19:09:32 +0100|
|User-agent:||Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.1.0|
On 12/25/2015 03:20 PM, Ralph A. Schmid, dk5ras wrote:
I do agree; but:
For example, if you have bad luck, the B2x0 might produce a third harmonic of the oscillator it generates to mix up your signal; e.g. if you tune to TX at 1GHz, you might see power emitted at 3GHz, which might even be only 15dB weaker than the tone at 1GHz!
Now, most antennas don't work well over full multiples of their center frequency.
Of course, typical antennas like the dipole (and hence, the monopole imagined as half a dipole mirrored by a ground plane) work well for frequencies that are multiples of their designated operating frequency in theory: If the current distribution is the typical , with an antenna which has roots of the current distribution at its physical edges , then for a different frequency with , there will be roots at the same , because of the periodicity of the cosine.
In practice, it really depends; the architecture of how you connect the coax to you dipole starts to play a role; impedance matching you might need to do to get most of your energy from cable to antenna (or in RX, the other way around) will have a limited bandwidth, even your cable might have a much stronger attenuation. Also, for more complex antenna designs, there might be much more than a one-dimensional dipole involved, so you might get a completely different radiation pattern and efficiency behaviour for those.
Problem is that many popular "ham bands" are relatively low frequency (compared to the 6 GHz upper limit of the B2x0), meaning that things like affordable cabling still work pretty perfectly over a large range; for example, RG-59 has maybe a few dB more attenuation at 300MHz compared to 100MHz per 100m cable length. For 7.2 GHz vs. 2.4GHz, things do look more drastically attenuated.
Another effect of "low" frequency is that antennas do tend to be simple dipoles, or monopoles, or for this aspect, relatively broadband types like Yagis. Hence, yes, don't overestimate your antenna as a filter, especially since the things you want filtered out are at frequencies that are a multiple of your desired band.
You should still look at how well your for example your PA, if you use one, handles multiples of your operational frequency; if they suppress it by a couple of dozen dB...
One thing I can't fully agree on:
Well; they are used for experimentation, but since you can directly run e.g. a 2G/3G base station on them, I'd argue they are not "experimental" (that sounds so much like they'd halt and catch fire as soon as you stop looking ;) ).First of all, the USRP radios are kind of experimental radios, using them for real ham radio operation on antennas will require filters and PAs. "Out of the box" it will only be some proof of concept when you create a ham radio application with it.
I do agree that, given the discussion regarding harmonics, a preselection filter might be a good idea if you know that there might be significant interference. Luckily, these can really be "easy" LC filters etc, because they really only have to have a transition width of about their passband edge's frequency. Things that people tend to spend a lot of money on are channel selection (tunable/baseband/IF) filters, and you can most of the time really get nice results without them; that's the beauty of digital signal processing with a potentially massively oversampled passband signal.
Take my block diagram of the B2x0 architecture as an example:
So, signal comes in on one of two antennas, goes through a band selection filter (not shown), which, as discussed, might need a little help if your external circuitry/ antenna isn't very selective itself, and then gets mixed down to baseband with a local oscillator generated by the B2x0 itself. Then, there's the adjustable analog baseband filter (or, two parallel filters for the "real-valued" I and Q baseband signals, if you want so). That filter is set so that the analog-to-digital-converter can do its job without seeing aliasing, but you can chose to set it even narrower.
After the ADC comes another stage of mixing, but this time with a digitally generated tone in digital logic, followed by a decimator (the "M"), which reduces the high physical sample rate after filtering the signal digitally to what you request as a sample rate. That sample rate would typically be only a bit higher than your channel bandwidth, but the ADC clock might be a large multiple of that rate; that results in oversampling followed by low pass filtering, which implies a significant SNR gain.
Assuming you're after some FM channel with let's say 50kHz total bandwidth. Your computer really won't bat an eye if you tell the USRP that you want 1MS/s, so you choose to do that, and set the ADC clock to 40MHz, because you feel like using 40x oversampling in the USRP (you don't even have to do that -- UHD choses a sensible ADC rate automagically as soon as you set the sampling rate, if you don't specify otherwise).
The digital filters in the B2x0 are pretty OK, so everything out of +- 500MHz of your center frequency is well surpressed. Because you want a super high quality filter, you decide to use gr_filter_design to design a (real) low pass filter with 1MHz sampling rate, 24kHz cutoff freq (->passband width = 48kHz) and a 26kHz stopband. gr_filter_design takes quite some time to design and analyze that filter: Its really overly sharp. Now, after it finishes calculating the coefficients and a few pretty pictures illustrating filter behaviour, you learn it's a 2727 tap filter. That's what I'd call an oversized filter, and it means that for every input sample, you'd have to apply 2727*2 multiplications + additions. You then use the filter coefficients in a GNU Radio FFT filter block, set to a decimation of 10, so that the output signal has a nominal sampling rate of 100kHz (=1MHz/10); you attach a graphical sink to see how your spectrum looks like, add a noise source as test signal and also observe the achievable sampling rate on your oldish laptop with damaged cooling, which tends to thermally throttle your CPU:
Ok; the signal looks like this:
Notice how this filter achieves 70dB attenuation, (though designed for even higher attenuation, due to numerical accuracy) over less than 2kHz. It costed you nothing in analog filters :)
And you can still get something like 1.1MS/s after the tenfold decimation, so you have still 90% processing power headroom.
So: If you have or feel able to build a preselection filter, you should be fine to RX signal directly off the antenna.
In TX, always remember to be a good neighbor; if you have a PA, you'll probably want to have something to suppress the harmonics that one produces.
|[Prev in Thread]||Current Thread||[Next in Thread]|