Aquadopp Profiler Signal To Noise

Hi There,

I'm afraid that after combing this forum and reading the manual I still do not understand the concept of a signal to noise ratio as it applies to current data collected using our 2mhz AquaDopp current profiler.

Elsewhere (for example [1]) it has been noted that the AquaDopp profilers do not record a 'noise ping' prior to taking the current profile.  If this is the case then what metric does Storm use to determine which bins to remove (using Processing > Settings > Low signal SNR threshold)?  Or is this just as arbitrary (meaningless?) as the what happens when you use ExploreP (a tool not designed for this device, as I understand) to perform similar filtering (also at [1])? 

Am I correct in thinking that our profiler does not record a value for 'noise' (I can't find anything that might be it in our output!)?  If so is an arbitrary cut-off the only option - or is there a recommended way to derive a reliable 'noise' metric from the data directly?

I don't have any particular use for the SNR here - I'm just trying to understand what Storm is doing with our data!

Regards,

Dan

[1]: http://www.nortekusa.com/en/knowledge-center/forum/software/644656765

Dear Dan,

Your observations are right, there are no 'noise ping' in our modern instruments.

What Storm does is to assume a noise floor as given below. 

To get a rule of thumb cut-off at 3dB you need to find the noise floor of your instrument and compensate for this if it deviates from the hard-coded values too much. An example are found below.

We realize this is a bit cumbersome, but SNR is not giving the full picture either. Needed SNR to obtain good current measurements have decreased during the refining of these instruments through the years. Different conditions may require different SNR to give good readings.

A good check of the validity of the data is to check that the vertical velocity is close to zero. Current profiles that show large current variations vertically or vertical structures should be flagged suspicious.  The ones that are most doubtful are those that are physically unrealistic (e.g. 1 m/s changes over a few meters).  The measurements in a profile that are far from the mean are best removed. 

Best regards
Jonas Røstad

Aquadopp 25 counts

Aquadopp Profiler 400kHz 33 counts
Aquadopp Profiler 600kHz 33 counts
Aquadopp Profiler 1Hz 25 counts
Aquadopp Profiler 2MHz 18 counts

AWAC 600kHz 33 counts
AWAC 1MHz 20 counts

Continental 190 kHZ 45 counts
Continental 470 kHZ 45 counts

Vector 45 counts

Example of SNR compensating in Storm/Surge: Your Aquadopp Profiler has a noise floor of 25 counts. Storm assumes 18 counts as noise floor. You will then need to add (25-18) counts*0,43dB/counts=3 dB. If you like to use the 3dB rule of thumb threshold, you actually need to input 6dB

Hi Jonas,

This explains a lot and answers my question perfectly.  Thank you for the prompt response.

Regards,

Dan