Detecting small signals with large noise

The current problem of finding the "black boxes" from the Air France
flight that crashed would seem to be another small signal/large noise
problem, maybe one that signal processing could help.

My first thought is that they should use exponential backoff
on the "ping" device to conserve battery power.  That is, ping
slower as time goes on giving searchers more time, though with
reduced ping frequency.  

Now, assuming that the design is meant to be found in the ocean,
are there design features that would make it easier?  
Maybe signal modulation methods that would allow for digital
processing allowing for detection at longer distances with
more sophisticated signal processing in the detectors?

I would expect the actual underwater (as in not surface) noise
sources to be relatively small.  If appropriate cancelation
techniques could be used to reduce surface noise in the detection
system, one might be able to detect the ping a long way away.

-- glen

glen herrmannsfeldt wrote:
> The current problem of finding the "black boxes" from the Air France
> flight that crashed would seem to be another small signal/large noise
> problem, maybe one that signal processing could help.
>
> My first thought is that they should use exponential backoff
> on the "ping" device to conserve battery power.  That is, ping
> slower as time goes on giving searchers more time, though with
> reduced ping frequency.
>
> Now, assuming that the design is meant to be found in the ocean,
> are there design features that would make it easier?
> Maybe signal modulation methods that would allow for digital
> processing allowing for detection at longer distances with
> more sophisticated signal processing in the detectors?
>
> I would expect the actual underwater (as in not surface) noise
> sources to be relatively small.  If appropriate cancelation
> techniques could be used to reduce surface noise in the detection
> system, one might be able to detect the ping a long way away.
>
> -- glen

I think you can forget about cancellation of random noise from a long 
distance / dispersed source.

I think you can forget about modulation as a likely method .. too much 
multipath, etc.  This ain't radar!

Actual underwater noise sources are multiple and not necessarily related to 
the surface.

So, the thinking has to be about how to improve SNR as best you can.....

Attenuation increases with frequency.  Noise decreases with frequency 
(partly due to the attenuation).  Directivity for a given array size 
improves with frequency.  So, there is an optimum frequency depending on the 
situation.

Fred 

Fred Marshall wrote:
> glen herrmannsfeldt wrote:
>> The current problem of finding the "black boxes" from the Air France
>> flight that crashed would seem to be another small signal/large noise
>> problem, maybe one that signal processing could help.
>>
>> My first thought is that they should use exponential backoff
>> on the "ping" device to conserve battery power.  That is, ping
>> slower as time goes on giving searchers more time, though with
>> reduced ping frequency.
>>
>> Now, assuming that the design is meant to be found in the ocean,
>> are there design features that would make it easier?
>> Maybe signal modulation methods that would allow for digital
>> processing allowing for detection at longer distances with
>> more sophisticated signal processing in the detectors?
>>
>> I would expect the actual underwater (as in not surface) noise
>> sources to be relatively small.  If appropriate cancelation
>> techniques could be used to reduce surface noise in the detection
>> system, one might be able to detect the ping a long way away.
>>
>> -- glen
>
> I think you can forget about cancellation of random noise from a long
> distance / dispersed source.
>
> I think you can forget about modulation as a likely method .. too much
> multipath, etc.  This ain't radar!
>
> Actual underwater noise sources are multiple and not necessarily
> related to the surface.
>
> So, the thinking has to be about how to improve SNR as best you
> can.....
> Attenuation increases with frequency.  Noise decreases with frequency
> (partly due to the attenuation).  Directivity for a given array size
> improves with frequency.  So, there is an optimum frequency depending
> on the situation.
>
> Fred 

Fred Marshall wrote:
> glen herrmannsfeldt wrote:
>> The current problem of finding the "black boxes" from the Air France
>> flight that crashed would seem to be another small signal/large noise
>> problem, maybe one that signal processing could help.
>>
>> My first thought is that they should use exponential backoff
>> on the "ping" device to conserve battery power.  That is, ping
>> slower as time goes on giving searchers more time, though with
>> reduced ping frequency.
>>
>> Now, assuming that the design is meant to be found in the ocean,
>> are there design features that would make it easier?
>> Maybe signal modulation methods that would allow for digital
>> processing allowing for detection at longer distances with
>> more sophisticated signal processing in the detectors?
>>
>> I would expect the actual underwater (as in not surface) noise
>> sources to be relatively small.  If appropriate cancelation
>> techniques could be used to reduce surface noise in the detection
>> system, one might be able to detect the ping a long way away.
>>
>> -- glen
>
> I think you can forget about cancellation of random noise from a long
> distance / dispersed source.
>
> I think you can forget about modulation as a likely method .. too much
> multipath, etc.  This ain't radar!
>
> Actual underwater noise sources are multiple and not necessarily
> related to the surface.
>
> So, the thinking has to be about how to improve SNR as best you
> can.....
> Attenuation increases with frequency.  Noise decreases with frequency
> (partly due to the attenuation).  Directivity for a given array size
> improves with frequency.  So, there is an optimum frequency depending
> on the situation.
>
> Fred

But, you might do line enhancement vs. broadband noise if you have a good 
frequency reference.  That's the same as bandpass filtering.  It can't be 
better than that.

Fred 

On 9 Jun, 20:52, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:
> The current problem of finding the "black boxes" from the Air France
> flight that crashed would seem to be another small signal/large noise
> problem, maybe one that signal processing could help.
>
> My first thought is that they should use exponential backoff
> on the "ping" device to conserve battery power. &#4294967295;That is, ping
> slower as time goes on giving searchers more time, though with
> reduced ping frequency. &#4294967295;
>
> Now, assuming that the design is meant to be found in the ocean,
> are there design features that would make it easier? &#4294967295;
> Maybe signal modulation methods that would allow for digital
> processing allowing for detection at longer distances with
> more sophisticated signal processing in the detectors?
>
> I would expect the actual underwater (as in not surface) noise
> sources to be relatively small. &#4294967295;If appropriate cancelation
> techniques could be used to reduce surface noise in the detection
> system, one might be able to detect the ping a long way away.
>
> -- glen

The physics, not the DSP, is the limiting factor. The
transmitters lie in or near the mid-atlantic under-sea
mountain range, meaning that the might be bathymetric
shadowing effects at the outset. The devices lie in very
deep water, where the sound speed profile is generally
upwards refracting, which generally tends to narrow the
detection range.

Since a deep-sea sub apparently is en route to the area,
the best search strategy would be to use aircraft to
deploy sonobuoys in the general area and hopefully
get a crude position fix. Some reports suggest a
nuclear sub is in the area. If so, this could be used
to get better position estimates for the transmitters.

Fred Marshall <fmarshallx@remove_the_x.acm.org> wrote:
(big snip)

> But, you might do line enhancement vs. broadband noise if you have a good 
> frequency reference.  That's the same as bandpass filtering.  It can't be 
> better than that.

Well, for the existing case that is likely true.

But if one considers the problem in advance, are there modulation
methods that would allow for better detection at a given source
power level and background noise level?

-- glen

According to the following, the underwater beacon is a pinger at
37.5Khz:

http://www.marinebuzz.com/2009/06/08/black-box-locating-flight-recorder-of-air-france-flight-447-in-atlantic-ocean/

I don't know how many companies make them, but Benthos was mentioned
in the above:

http://www.benthos.com/undersea-underwater-aviation-locator-pinger.asp

If it survived the crash and subsequent depth, then it should operate
for 30 days.  And, as mentioned in the first link, it�s probably in
very deep water, which would refract any sound waves strongly upward.
And it may be bouncing the signal off a mud bottom or underwater
mountain or something.

Terrible tragedy - I hope they find the recorders so at least they can
get a better idea of what happened.

Kevin

On Jun 9, 11:47�pm, kevinjmc...@netscape.net wrote:
> According to the following, the underwater beacon is a pinger at
> 37.5Khz:
>
> http://www.marinebuzz.com/2009/06/08/black-box-locating-flight-record...
>
> I don't know how many companies make them, but Benthos was mentioned
> in the above:
>
> http://www.benthos.com/undersea-underwater-aviation-locator-pinger.asp
>
> If it survived the crash and subsequent depth, then it should operate
> for 30 days. �And, as mentioned in the first link, it�s probably in
> very deep water, which would refract any sound waves strongly upward.
> And it may be bouncing the signal off a mud bottom or underwater
> mountain or something.
>
> Terrible tragedy - I hope they find the recorders so at least they can
> get a better idea of what happened.
>
> Kevin

One way to increase the SNR is to integrate the signal over multiple
one second frames. I would assume the searchers are doing that.

John

John <sampson164@gmail.com> wrote:
< On Jun 9, 11:47?pm, kevinjmc...@netscape.net wrote:
<> According to the following, the underwater beacon is 
<> a pinger at 37.5Khz:
(snip)
 
< One way to increase the SNR is to integrate the signal over multiple
< one second frames. I would assume the searchers are doing that.

As I wrote previously, if you plan ahead it might be possible
to make a signal that is easier to detect in a noisy environment.

Integrating over time is one way, but it helps if you know
the relative phase.  If the source used a crystal reference,
it might be that the phase delay between pings would be
(to the crystal tolerance) known.   

In any case, it is possible that with enough signal processing
a very small signal could be extracted that would otherwise
not be possible to detect.  

-- glen

On Jun 11, 4:55&#4294967295;pm, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:
> John <sampson...@gmail.com> wrote:
>
> < On Jun 9, 11:47?pm, kevinjmc...@netscape.net wrote:
> <> According to the following, the underwater beacon is
> <> a pinger at 37.5Khz:
> (snip)
>
> < One way to increase the SNR is to integrate the signal over multiple
> < one second frames. I would assume the searchers are doing that.
>
> As I wrote previously, if you plan ahead it might be possible
> to make a signal that is easier to detect in a noisy environment.
>
> Integrating over time is one way, but it helps if you know
> the relative phase. &#4294967295;If the source used a crystal reference,
> it might be that the phase delay between pings would be
> (to the crystal tolerance) known. &#4294967295;
>
> In any case, it is possible that with enough signal processing
> a very small signal could be extracted that would otherwise
> not be possible to detect. &#4294967295;
>
> -- glen

Battery life is another concern; you could do better if the beacon
could send a PN sequence continuously for 30 days. There is only so
much you can do with 9 ms/sec of energy.

John