Re: [Discuss-gnuradio] Direct Conversion vs Superheterodyne

[Nick Waterman]

Crusty Curmudgeon wrote:
> Ok.. a direct conversion receiver is simple and cheap, but I think there are
> some flies in the ointment too:

I've recently become quite interested in the Tayloe 
Mixer/Detector/System. It's a remarkably simple circuit where you 
moreorless connect your antenna through a fast high-speed analogue bus 
switch which is clocked at N*f Hz, connecting it to N (usually 4) 
capacitors in sequence. 4's a particularly nice number as you end up 
with 4 nice quadrature signals 90 degrees apart - convenient for SDR 
work. Charge capacitors a,b,c,d,a,b,c,d, then use simple op-amp-ish 
circuits to give you a gain boost and make Q = a - c, I = b - d... and 
the rest of it feels like a digital circuit anyway!   :-)

> 1) It requires very good shielding of the local oscillator and great
> mixer-to-antenna isolation or else it will leak signals right on top of what
> you're trying to listen to, potentially jamming the channels. 

But in most true direct-conversion receivers, including the Tayloe 
detector, you're leaking signals at EXACTLY the frequency you're trying 
to listen to, producing (if anything) a simple DC offset. PC Soundcards 
are quite good (or arguably bad) at ignoring DC for you   :-)

That's an ADVANTAGE of direct conversion - any undesired mixer products 
appear as DC on your outputs, whereas in superhets they appear at IF+F 
and IF-F or whatever. If you have multiple IFs in your superhet,  you've 
got all kinds of accidental products produced by all kinds of products 
of multiple input carriers and IFs interacting in all kinds of odd ways. 
2 nearby carriers on adjacent channels can even produce all kinds of 
products with themselves. The superhet requires far more shielding and 
filtering because otherwise you end up with all kinds of messy products 
all over the place, whereas in the Direct Conversion receiver, you have 
ONE frequency to worry about, and if it leaks, it gives you DC.

> 2) It does not tolerate strong near channel signals well. That can lead to
> overload of the mixer(s) or follow-on baseband amplifiers and active filters.
> Very high level LO mixers can improve things a little, but that impacts the
> signal leakage problem too.

The Tayloe detector was originlly conceived as a tuneable narrow 
bandpass filter centered at exactly (clock freq)/4. In the same way that 
charging a capacitor through a resistor gives you a low pass filter, 
charging 4 capacitors in sequence acts like a bandpass filter at a 
quarter of the clock rate - basically a bandpass filter at the frequency 
you're trying to tune. Bonus! Tweak the circuit a bit if you want 
narrower/wider bandpass.

> 3) It is sensitive to any harmonically (sp?) related signals, so good bandpass
> filtering between the antenna and mixer is needed. Of course this make band
> hopping a problem. 

I've already said the Tayloe detector acts like quite a good bandpass 
filter. F/4 gives your perfect quadratures. F*2/4 effectively gets added 
to a&c and subtracted from b&d, so getting Q from a-c and I from b-d 
cancel this out. F*3/4 sort of cancels itself out by appearing 120 
degrees out of phase 3 times in a row (I'm probably not describing this 
too well)

> 3) Some sort of automatic frequency control or phase locking is needed on the
> local oscillator to make sure it's on the same carrier frequency as the desired
> signal or else the beat note will dominate the analog to digital converter
> output, chewing up dynamic range.

Now that's reasonably true if tuning an AM broadcast, but in SSB there's 
no such carrier to get in the way, and in FM your carrier is moving 
around and that's exactly what you WANT to dominate your A2D and eat up 
your dynamic range   :-)

Even in the case of AM, CW, RTTY etc with a strong carrier or 
carrier-like signal, you can tune (say) 10kHz away from it, treating 
your "direct conversion" more like an IF to get within the frequency 
response of your A2D, then tune down 10kHz in the software (or do your 
CW/RTTY/whatever decoding centered around 10kHz instead of DC).

> 4) An agc system is also needed to help handle the extreme amplitude shifts
> multipath fluttering causes to the signal or else you'll need a rather expensive
> analog to digital converter.

Not a disadvantage of Direct Conversion at all - exactly the same 
applies to superhets!   :-p

In my opinion the coolest thing about the Tayloe system, and what makes 
it a natural partner to SDR - It feels like half-analogue, half-digital 
technology - you're using a component that's designed to quickly select 
1 of 4 busses in a computer, and you're moreorless using it to 
"demultiplex" your desired frequency into 4 quadrature signals separated 
by 90 degrees. Now you just need a nice stable VFO, doesn't even need to 
be a nice pure sine wave as it's only being used as a clock - a square 
wave is in fact ideal, as long as it's nice and stable... Dare I say it 
might even be generated by digital hardware, or even software if you can 
get the stability?... and then you probably want a bit of AGC, and even 
THAT might be better controlled by software feedback rather than being 
all analogue. It's almost the ideal bare minimum analogue stuff, letting 
you do the rest in the digital and software domains.

See Fig 10, 12, 14 in "SDR for the masses part 1" at 
http://www.arrl.org/tis/info/sdr.html - some of the rest of the article 
even explains the harmonic/alias frequency rejection better than I can.

--
"Nosey" Nick Waterman, Senior Sysadmin.  #include <stddisclaimer>
address@hidden  http://noseynick.net/
Those who live in glass houses...shouldn't.