Re: [Discuss-gnuradio] Real-only direct-conversion

[Moeller]

On 27.02.2011 17:28, Marcus D. Leech wrote:
> I was on a call the other night with someone who asserted that you
> didn't need an I & Q representation
>   for  a direct-conversion receiver, and that I and Q could be
> synthesized later from a real-mode-only
>   baseband signal.
...

> So, my feeling is that you *absolutely need* the I and Q "form" in order
> to disambiguate +/-
>   frequencies when dealing with direct-conversion baseband signals.
> Who's right?

As long as you receive the complete signal bandwidth, you can create the I/Q 
form later.
Of course you need the double sample rate, if there's only the real "baseband"
representation. I call it baseband, but you can also call it IF with the lowest
possible IF frequency. Strictly speaking it's not a "direct-conversion" 
receiver,
since there is a fixed IF in the middle of the baseband spectrum.
The "data rate" is the same for both, one has double sample rate but only
half the sample size (real vs. complex numbers).

Complex baseband (I/Q) reconstruction:
- Hilbert transform eleminates the (symmetric) negative frequencies
- Frequency shifting the IF frequency to zero by multiplying a complex 
exp(-j*2*pi*f_IF*t)

This is standard in digital down converters (DDC).
The TVRX-Board is working this way. According to the schematic,
only the bipolar A channel is connected to the tuner chip, a real input.
Other Dboards use both A/B inputs for separate I/Q channels.

I think both variants have their advantages and disadvantages.