Recently some questions have surfaced on the real need and the
benefits (or lack thereof) of feeding a SDR program with quadrature
(I/Q) signals. Let me summarize here my thoughts on the subject.
- Case 1) You have a classical radio, with a last IF of, let's say, 15
Khz (or a similar frequency up to about one half the sampling
frequency of your sound card). The radio has a crystal filter which
has already taken care of the unwanted sideband. E.g., your radio has
a first IF at 8.3 MHz (like some Japanese rigs have), then the second
IF is at 455 kHz, and you have added a further conversion down to 15
kHz (like many do for DRM reception).
If this is the case, you can feed the same signal to both the left and
right channels of the sound card. The 15 kHz signal is further brought
to baseband inside the software, using a quadrature software NCO with
a complex mixer followed by a complex passband filter centered at 0
Hz, and the I/Q components are generated inside the program, then used
for the demodulation process, which in any case needs them. You can
fully use the filtering and the demodulation capabilities of the
software, but you will lose the use of a tunable band equal to the
sampling frequency, you will be limited to just a half of it.
- Case 2) You are implementing in hardware a Direct Conversion
receiver, with a hardware Local Oscillator at a frequency which we
will call LO. Then, if you want to be able to separate the wanted from
the unwanted sideband, you *must* do the conversion to baseband (or
near baseband, as the final fine tuning is done in software) using a
quadrature mixer, fed with an analytic LO signal, i.e. with I/Q
components. Ideally, also the RF should be separated in I and Q
components, but the advantage of doing this is just the elimination of
a couple of low pass filters, so normally it is preferred to feed the
same RF signal to the complex mixer, and have just the LO in
quadrature form. A couple of low pass filters is much more easy to
implement than a quadrature hybrid that keeps the phase angle at 90
degrees over a rather large span of RF frequencies. Also the LO should
keep the phase angle between the components at 90 degrees,
irrespective of the frequency. So, solutions like the RC network used
in Softrock V5 can work as the LO signal is generated by a quartz
crystal, so at a fixed frequency. I would have preferred to see the
possibility to tweak the values of the RC network, to achieve a near
perfect phase angle, but small differences can be taken care of by the
software. If you use instead a DDS, I am afraid the RC network can
work only on a limited range of frequencies, in this case a Johnson
counter fed with a frequency 4 times that wanted, seems to be the best
solution.
Returning at the topic of this message, if you are in Case 2) the need
for I/Q conversion is dictated by the need of suppressing the unwanted
sideband, otherwise unattainable in a DC receiver. An added plus is
the fact that now the tunable range goes from LO - fs/2 to LO + fs/2
where fs is the sampling frequency, the double than in Case 1).
This is my view on the subject. If I haven't been clear enough, just ask.
73 Alberto I2PHD
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/soft_radio/
<*> To unsubscribe from this group, send an email to:
[EMAIL PROTECTED]
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
