[time-nuts] Near-perfect chip for Loran-C frequency receiver

[Bruce Griffiths]

Poul-Henning Kamp wrote:
> In message <14647AC2-BDFB-422A-9FCE-BE9D936DF3D9 at cox.net>, "Thomas A. Frank" wr
> ites:
>
>   
>> In as much as we know exactly the frequencies of interest, perhaps  
>> the analog front end could consist of multiple signal paths with very  
>> narrow band filters to separate out the signals of interest (a 60 kHz  
>> channel, a 100 kHz channel, plus whatever other channels are  
>> desired), then each channel gets an individual AGC amp, and the  
>> matched signals are summed back together for A-D?
>>     
>
> If you really want to do something like that, notch filters are
> a much better way to go.
>
> But as I keep trying to tell you guys: you don't need all that.
>
> Look at:
> 	http://phk.freebsd.dk/misc/L6/
>
>   
Nonsense, although it may work in favourable cases where the signal 
levels arent too disparate, in the more general case this isnt true.

> There are four GRI's: 7499, 6731, 9007 and 8000, each in their own row.
>
>   
Thats obviously fine for some Loran applications but you haven't 
demonstrated that its also true when attempting to simultaneously track 
another signal of interest that is relatively weak.
> The leftmost plot shows the entire FRI (2 times GRI) and is used
> to locate the individual signals (look for one green (master) and
> multiple blue (slave) peaks.
>
> The small plots show three stations in each chain, to the extent that
> I have been able to locate them by eyeball.  The X, Y designations may
> not match the official LORAN-C station names.
>
> The title on each of these plots show the relative signal strength,
> and as you can see the 7499M is 44 times stronger than the 9007Y.
>
>   
Thats only a 33 dB difference in signal level what happens when the 
signal strengths differ by 60, 80 dB?
> The data you see there is roughly 8 minutes worth of data at 
> 10 bits per sample, 1Msps, processed as fixedpoint in the
> time-critical path.
>
> The antenna is this one: http://phk.freebsd.dk/loran-c/Antenna/
> and it just has a simple Low-pass filter around 200 kHz.
>
> Unfortunately, the ST32M103 will have neither the RAM nor
> CPU power to track that many signals, but how many it can
> do remains to be seen.
>
>   
In which case a dedicated ST32M103 for Loran and another for 
MSF/WWVB/JJYDCF77 etc is appropriate together with optimised analog 
front ends, bandpass filters, notch filters and orientation of the 
individual antennae to null strong interfering signals should cope with 
most problematic sites. These refinements can be omitted at more 
favourable locations.


Bruce