High-frequency suspended sediment concentration from ADV

Hi there,

I am trying to calibrate a Vectrino to find suspended concentration of kaolin in a lab environment. That works reasonably well for the range 10-1000 mg/l. The calibration curve is exponentially, i.e. log10(SSC)=a*SNR+b. That means that any noise on the SNR is strongly amplified in the ssc series, resulting in larger uncertainty on a single value at high frequency (e.g. 25 Hz). When a series of ssc produced in that way can be low-pass filtered with a window of e.g. 1 second I find that the uncertainty drops to reasonable levels, as also shown in a paper by Salehi et al (2011, CSR). However, if I want to use the high-frequency ssc signal I have to find a way to correct the SNR signal to remove background noise. Are you aware of any techniques or publications discussing that matter?

thanks in advance

Boud

Hi P.J.,

I'm taking this up again. Thanks for your reply back in december.

I find linear relations between log(ssc) and SNR and between log(ssc) and AMP. So, no linear relationship between AMP and SSC, but  that's ok according to Rayleigh scattering. Taking into account sediment attenuation as a linear function of ssc I can even use some of the higher range where the linear relationship is moving into saturation.

But, I have some questions on the AMP and SNR outputs. SNR (dB) is internally computed as 20log(I/N), with both I and N in sound pressure levels? And AMP (dB) would be then equal to 10log(I/Io) with Io outgoing signal? And the noise levels in dB are something like 20log(N/Io)?

thanks

Boud

Previously P.J. Rusello wrote:

I'm not aware of any publications specifically dealing with noise in the amplitude and SNR data, but I think conceptually it's pretty similar to the velocity signal. There might be a different structure. Maybe something useful in this paper: 

Zedel, L. (1999). A coherent Doppler profiler for high-resolution particle velocimetry in the ocean: Laboratory measurements of turbulence and particle flux. Journal of Atmospheric and Oceanic ….

In the Vectrino header file the noise levels in each receiver are specified in the Velocity Header section. You could use this value and simple look at the amplitude signal instead of SNR for a noise correction. However this will just wind up being a reentering of the signal since this is an average value of the noise. You can also look at the amplitude spectrum and probably develop a reasonable estimate of the noise variance from it.

If you switch to the amplitude data you should find a linear relationship between it and SSC.

P.J.

Amplitude and noise are reported is digital counts, no units. You can convert these to a dB level by multiplying by a factor of 0.45 (see the sediments technical note here: http://www.nortek-as.com/lib/technical-notes/seditments).

I don't know if that will really help you though since it's all then relative measures. You could express any of these in dB since this is just a ratio of two values. Let me check another reference as it might have more, I'll get back to you on this.

P.J.

Check this reference: http://www.nortek-as.com/lib/bibliography/poindexter-et-al-adv-streaming I think the sound amplitude was measured independently in Table 3. It's not something we normally look at since we're not necessarily interested in calibrated backscatter and I believe it depends on the supply voltage as well.

P.J.