Re: [Discuss-gnuradio] Direct Conversion vs Superheterodyne

[David Bengtson]

On Mon, 13 Jan 2003 11:19:16 -0800, you wrote:

>On Mon, Jan 13, 2003 at 01:33:34PM -0500, Tanner Lovelace wrote:
>> > Why does everyone insist on using an if?  I seem to recall from my radio 
>> > classes (admittedly quite a few years ago) that direct-to-baseband IQ 
>> > conversion keeps all the modulation information present in the signal.  
>> 
>> Well, admittedly my radio knowledge isn't as much as it should be, but
>> from what I understand, the main reason for superhet designs was that
>> it was easier to process signals at a fixed frequency than over a 
>> range of frequencies.  While this may be true in analog, it's entirely
>> possible that direct conversion would be better for an SDR.  In
>> fact, I was just looking at direct conversion circuits last night
>> in the ARRL Handbook and wondering this very thing.  Perhaps someone
>> with better knowledge could speak about this?
>
>The theory you're talking about here is called undersampling.  You 
>still need to tune and mix your signal to an IF (eg: 455kHz), but if
>your signal is only 10kHz wide, you can sample your 455kHz IF at at least
>20kHz and aliasing will effectively do the final mix-down to base band.
>Aliasing is usually a Bad Thing(tm) that makes a 11kHz tone sound like
>a 9kHz tone when sampled at 20kHz.  In this case, it works in our favor.
>

undersampling and Direct Conversion are slightly different approaches.

 In a Direct Conversion receiver, there is a Mixer stage that mixes
the desired signal down to an IF frequency of zero. In other words,
you would had a mixer stage at, say 10.7 MHz, and a local oscillator
injection signal at 10.7 MHz. These two signal's mix, with one mixing
product at DC, and the other at 2*10.7 MHz, or 21.4 MHz. This can be
sampled out quite easily by a simple low pass filter. 

In an undersampled receiver, the desired signal, at an IF frequency
(Use 10.7 MHz) is sampled by a A/D operating at a clock rate of 500
kHz. You then sample signal's from DC to 250 kHz, in the typical
fashion. Signals from 250 kHz to 500 kHz are also sampled by the A/D,
as are signals at every frequency presented to the input of the A/D. 
In this approach, you must filter out the unwanted frequency bands,
since any signal present will alias into the passband, and then appear
as an in-band jammer. Notice that this also applies to DC offset's
that appear in circuits. 

Another kicker with undersampled receivers is that the baseband SNR is
dependent on the sampling jitter of the A/D, and you can get some
pretty hairy specs for clock jitter. 

Analog Devices has a decent app note by Brad Brannon that discusses
Digital Radio Fundamentals on their web site, Titled Radio 101. I
don't have a URL at hand, though. 


EDN Magazine Reference

http://www.e-insite.net/ednmag/index.asp?layout=article&articleid=CA46249


I hope this helps


Dave