[time-nuts] PN sequence generation using GPS

[jimlux]

On 2/26/11 4:58 AM, scmcgrath at gmail.com wrote:
> Hi Chuck
>
> I'd see you on my waterfall display.  A flexradio is a wonderful thing.  The Icom 7x00'es would also see the energy and display it.
>

Maybe, maybe not.
Depends on what other activity there is in the band and what the noise 
properties are.  In the land mobile VHF bands, there's a lot more 
ongoing activity, and your hops blend in with everything else. On HF, 
since you have noise bursts from thunderstorms, etc.

Your flex only sees, at most, 192 kHz of BW.  If you're hopping over a 4 
MHz band, only 5% of the hops will be in your detection bandwidth. If 
the band were quiet, those might be noticeable.  Of course, you'd also 
have to be looking when the hopper is transmitting.

then, you have to look at the resolution of your FFTs.. if you're 
running 512 or 1024 sample buffers, you do a transform every 5-10 
milliseconds, so a 10 Hz hopper would be fairly easy to see, depending 
on what the modulation looks like.

And this is why nobody who does this "for real" uses 10 Hz hop rates and 
4 MHz hop bands.  Run at 1000 hops/sec or 10,000 hops a sec and the 
"detection game" becomes quite a bit more challenging.  Back in 
"retrofit land mobile radio" days, 10 hops/sec was easy, and quite a few 
commercial hoppers (e.g. the Racal Jaguar V) hopped in the 100 hop/sec 
range, because they had to wait for the PLL to settle.  Today, with a 
DDS, phase continuous hopping at 10,000 hops/sec would be trivial.

And with GPS to give you a good sync, a decent crypto algorithm to turn 
time of day into sequence seed, it would be easy to build something that 
is quite literally undetectable.

Let's for sake of argument, assume we're sending digitized voice at 16 
kbps.  The strategy will be to use FSK, with one bit per hop, and the 
two FSK tones being chosen randomly for each hop.

16khop/sec is 62.5 microsecond hops.  Taking your flex example...Let's 
say 1000 bins in the FFT at 192kHz sample rate, so your bins are 192 Hz 
wide, and you do a transform every 10 milliseconds. There are 160 hops 
during that 10 ms.   Only 20% of the hops are in the receiver bandwidth, 
so 8 of the bins will see some power, and it will be randomly 
distributed across your band.  (it looks like snow on analog TV).  Don't 
forget that the signal's only in a given bin for 62.5 microseconds out 
of the millisecond, so it's down by 12 dB.. If the signal were 30dB 
above the noise floor, unhopped, then you'd see it 18 dB above the 
noise.  You can't average transforms to smooth out the noise, of course, 
so that's above the grass which is probably 10dB high.

If the transmission pulses on and off (like a typical QSO), you might 
notice the snow coming and going.

But this all presumes that you're actually looking for it.  If you were, 
say, on 144.200 trying to make contacts, and the ham boneheadedly didn't 
exclude that from their hop list.. I think we calculated 800 channels 
before, so at 16 khop/sec, you hit a given channel, on average, at 100 
times/second.

That would probably be hard to distinguish from just increased 
background noise.  Yes, if you had a directional antenna, and you went 
hunting for this source of noise, eventually you'd find it.

Such is the quandary for a spread spectrum designer looking for low 
probability of detection.  The game is to spread yourself wide enough 
that a total power detector (radiometer) won't see you, because your 
increment in total power over thermal noise is small enough that it's 
not statistically detectable over the normal variations in noise power.

Standoff range is a problem.. if you're trying to communicate with 
someone 1000km away, and the guy looking for you is 1km away, that's a 
problem.