SNR calculation

Hi,

I have a question regarding the calculation of the SNR in ADVs. I know that it is calculated as:

SNR=Factor*(Signal amplitude-noise)

I understand that there is some sort of logarithmic amplifier of the amplitude counts, and that the output amplitude and noise is in fact a natural logarithm transformation of the actual levels. This is the reason why you actually substract the values instead of making the ratio between them.  Then, in order to express it as decibels, you have to multiply by log10(n)/ln(n), which happens to be 0.4343. This is about the value I have found in the literature as well as in this forum for Vectors. However, the Vector manual in p. 33, as well as the Nortek technical note 3, say that this factor ranges between 0.4 and 0.47 in different velocimeters. Why is this factor changing so much?

thanks in advance,

Òscar Guadayol

Dear Òscar

I am sorry it took some time to answer this but your question about a possible relation between the natural log and the factor 0.43 threw us a little since  I don't think anybody at Nortek had thought about this before.  We have looked at the circuit but we still do not know if there is a relationship between the scaling factor and some sort of natural response function in the receiver.  However, what we do know is:

a) The response of the receiver is not fully linear so the scaling factor depends on where you are on the amplification curve.   I think to do some extent this is compensated by the variability in the amplitude measurements (if you look at the raw data) so after averaging for some time, using a constant scaling factor isn't necessarily so bad.

b) The scaling factor varies a little from system to system and system

In conclusion, I don't think the general recommendation are bad. If you want more precise data, you almost have to measure the response of the individual system.  This can be achieved by connecting an extra transducer to a frequency generator, attaching the extra transducer to the instrument and then measuring the response for different input levels (controlled by the frequency generator).

Best regards, Atle Lohrmann

Thank you very much Atle,

òscar

Previously Atle Lohrmann wrote:

Dear Òscar

I am sorry it took some time to answer this but your question about a possible relation between the natural log and the factor 0.43 threw us a little since  I don't think anybody at Nortek had thought about this before.  We have looked at the circuit but we still do not know if there is a relationship between the scaling factor and some sort of natural response function in the receiver.  However, what we do know is:

a) The response of the receiver is not fully linear so the scaling factor depends on where you are on the amplification curve.   I think to do some extent this is compensated by the variability in the amplitude measurements (if you look at the raw data) so after averaging for some time, using a constant scaling factor isn't necessarily so bad.

b) The scaling factor varies a little from system to system and system

In conclusion, I don't think the general recommendation are bad. If you want more precise data, you almost have to measure the response of the individual system.  This can be achieved by connecting an extra transducer to a frequency generator, attaching the extra transducer to the instrument and then measuring the response for different input levels (controlled by the frequency generator).

Best regards, Atle Lohrmann