S/N value

Hi!S/N stands for "Signal to Noise Ratio" and basically is a "quality" metric for your measurement. Your measurement will be more precise if the "signal" (the light from the star you want to do photometry on) stands out clearly against any noise. CSHeinz

Hubert,The SNR ratio also allows you to place an estimate on the precision of your measurement. As I just got done with three 16-hour days of observing at CHARA atop Mt. Wilson and have yet to return to Denver, I don't have any books so I'll have to do this from memory (and upon reading the text below, it's clear the lack of sleep is showing). Arne Henden's photometry book and the AIP4WIN manual/book will discuss some of this, but probably not everything I've included below.If I recall correctly, the basic equation is:precision = log(1 + 1/SNR)Where log is log base 10.So if you have a SNR of 100:log(1 + 1.0/100) = 0.0099 ~ 0.01Or with a SNR of 1000:log(1 + 1.0/1000) = 0.0099 ~ 0.001Where the numbers have been intentionally rounded up to provide an upper-bound on the uncertainty. Therefore a measurement of a magnitude, say 3.14159, with a SNR 100 is 3.14 +- 0.01 mag. This uncertainty provides others with a metric for the validity of your measurement. When I look through data, I always look at the uncertainties and if they aren't listed, I often ask for them.Now, let me type in a *BIG DISCLAIMER*: Really it isn't this simple. Often the error bars written in AAVSO publications is standard deviations and not a true uncertainty in the measurement. This requires much more math, time, and a detailed understanding of the rules of error propagation. Continuing on, you might have a high SNR on your target object, but a low SNR on the comparison object which bumps up your uncertainty, making the above estimation technique invalid. Or, you might have great SNR for the target and calibrator, but have a poor estimate for the uncertainty. I've written an example of the second case below, in which the comparison star has a large uncertainty. Granted, this example is contrived, but I've recently dealt with a similar problem with poor estimates for catalogue magnitudes in my J and H photometry work so it does happen. On to the example!Say you are doing differential photometry and have the following data:Target_IMag = 3.14159, standard deviation 0.005, SNR 1000Cal_IMag = 1.23456, standard deviation 0.005, SNR 1000Cal_CatMag = 2.7182 +- 0.125Where Cal is the Calibrator, IMag is the measured instrumental magnitude, and CatMag is the known catalogue magnitude with uncertainties from a trusted source (like the AAVSO calibration star database or SIMBAD). Ignoring extinction and all of that stuff for a minute, we can calculate the instrumental calibration factor from doing algebra on the differential photometry equation:Catalog_mag = Instrumental_mag + Cal_FactorWhich gives us a Cal_Factor of:Cal_Factor = 1.484 +- 0.125Even though the uncertainties add in quadrature (discussing what this means and why it is so will require a separate forum to be opened), I've ignored the uncertainty from the standard deviation and broken with good data analysis practice because those values are insignificant in comparison. Okay, back to the topic.Now that we have the calibration factor, we can estimate the true magnitude for our target star. Again applying the differential photometry equation, with Cal_Factor from above:Catalog_mag (from the experiment) = 4.625 +- 0.125Again, I've ignored the standard deviation of the target star measurements for the reason I stated above. Now, lets get back to the SNR question: What is the uncertainty that I may safely quote? The SNR (1000 as you may recall) implies 0.001, so I should be able to say 4.625 +- 0.001, right? Nope. The uncertainty in the calibration star plus semi-proper error propagation leads us to the conclusion that we must quote the larger uncertainty given by error propagation instead of the error from the SNR, therefore the correct magnitude for this measurement is 4.625 +- 0.125.Now, what happens if the SNR for the target star was 100 instead of 1000? We simply chop off a digit because a SNR of 100 places an bound on the error (in this case) of 0.01. Therefore we should quote 4.62 += 0.13 for the magnitude of the star when it is published.I hope to give a talk on uncertainty analysis at the September Citizen Sky meeting because most people with photometry are crunching the numbers anyway, so why not do full error propagation instead of providing uncertainties? Sure, its much more work, but it really shows off your capabilities as a Citizen Scientist!Until the meeting, if you want to know more about uncertainty analysis please do star a thread in the Data Analysis section, or you could get a copy of the book entitled "Error and Uncertainty Analysis" written by Taylor. It's an excellent introduction, especially if you really want to know how the nitty-gritty details work.I hope that answers your question, but not in too much detail.-Brian

While signal to noise ratio gives an idea of how good the signal or count is, it tells little about the quality of the resulting magnitude. What is usually of more interest is how good is the data. Note: This quality criteria would also be an indication of SNR. To do this you should determine a minimum of 3 sets of magnitudes. This means do what you do to get a magnitude, but do it three times (three different sets of observations). Now you have 3 magnitudes. If everything is perfect they should all be the same magnitude (at least of a short period like minutes). Exact magnitudes are highly unlikely, however. Now do a standard deviation (SD) calculation and report that as the data spread or error for the average magnitude (the reported single magnitude). If you don't know how to do a standard deviation (most calculators and software programs have that function built-in) do a GOOGLE on standard deviation.Typical single channel V band report: V= 3.4459 SD= 0.0021Typically single channel V photometry of epsilon Aurigae produces SDs approaching 0.001. CCD/DSLR SDs are typically 0.02 and approach 0.01.Note. visual observers usually can only note the conditions. The visual errors/precision are at best 0.1 magnitudes and typically greater.JeffCounting PhotonsHopkins Phoenix ObservatoryPhoenix, Arizona USAphxjeff@hposoft.com