[RFI] Rogue beer fridge caught by Telstra 'robot'

[Roger (K8RI)]

On 6/4/2013 2:12 PM, David Robbins wrote:
> that sounds like a nice idea, but is much harder than it sounds.  first, the software would have to know how to sort out the noise from the desired signals.  it would also have to be able to scan wide bandwidths, maybe even outside the range of most hf receivers, just using 3khz wide audio is impossible in many cases... the grow light i found had a fundamental around 100khz with differing amplitude even and odd harmonics to above 4mhz where they started to overlap making an almost continuous noise, video systems could have fundamentals as low as 15khz, and harmonics can cover very wide bandwidths, doing a thorough analysis may require covering a large portion of the spectrum while ignoring many different legal signals.  plus since switching supplies are becoming plugin components there may be many devices using identical supplies and thus creating identical signatures.
>

It's both much simpler and much harder...

You only need to look for specific characteristics and generally these 
can be determined by looking at only a few spectrum segments and 
matching them against known offending signals..

As you are looking for specifics you don't normally have to sort any thing.

The difficult part is developing and cataloging the signals into the 
database. Once the database is developed, you only need the complicated 
software to identify new signals.  Like Voice recognition and finger 
prints you are matching key points so even if there are other signals 
present you only need match the key points. Of course you do need to be 
able to hear those points.  So if looking for the ID of an S4 signal 
under a 20 over 9 SSb signal you will need for them to take a breath.

This is one of the few legitimate reasons for "the cloud", A place that 
should be avoided unless you have no choice.

The difficult part is developing the database and recognizing the new 
sources.  Once developed it becomes one of matching key segments.

So some parts are very difficult while others can be relatively easy as 
algorithms already exist for the pattern matching. OTOH pattern matching 
often takes multiple samples, some times for extended periods gaining on 
and off cycles.  It may come down to statistical matching which does not 
normally result in quick answers.

>
> Jun 4, 2013 01:32:21 PM, w4byg at att.net wrote:
>
> Interesting article.
>
> As a result a thought: Wouldn't it be great if someone could develop
> software that when the output from your HF receiver is connected to your
> sound card, it could help identify the kind or type of noises received?
> It seems this would help decrease the time it would take to track down
> the source(s).
The software probably exists that could do this now, but if the 
algorithm is involved in pattern matching, the chances are the software 
is not available outside of it's proprietary use.


73

Roger (K8RI)


>
> Ray, W4BYG
>
>
>
> On 6/4/2013 7:50 AM, Steve Sacco NN4X wrote:
>> Fascinating RFI story from Australia.
>>
>> http://www.itnews.com.au/News/345494,rogue-beer-fridge-caught-by-telstra-robot.aspx
>>
>>
>>
>> 73,
>> Steve
>> NN4X
>> _______________________________________________
>> RFI mailing list
>> RFI at contesting.com
>> http://lists.contesting.com/mailman/listinfo/rfi
>>
>
> _______________________________________________
> RFI mailing list
> RFI at contesting.com
> http://lists.contesting.com/mailman/listinfo/rfi
> _______________________________________________
> RFI mailing list
> RFI at contesting.com
> http://lists.contesting.com/mailman/listinfo/rfi
>
>