Aaron Fienberg fienberg@uw.edu

This is a pulsefitting class I built to fit digitizer traces. It is intended to be useful for UW lab studies and test beam analysis. It is intended to be realitively easy to adapt into a template or functional fitter with various types of functions

The fitter works with .json config files. See the example config files in the /configs folder. The fitter is currently set up to work with the exampleSipmConfig.json.

Other than the self explanatory entries, I will give special attention to the following:

• fit_type: The type of fit to be attempted. The current options are 'laser', 'beam', and 'template'.

• clip_cut_high & clip_cut_low: any digitizer sample whose value is not between clip_cut_high and clip_cut_low will not be included in the fit.

• free_parameter: whether this parameter will be varied in the minimization procedure. The second parameter should always be delta t (for double pulse fits). It will always be fixed if you call fitSingle() and it will always be free if you call fitDouble().

• separate_baseline_fit: whether the baseline is fit at the same time as the pulse or fit separately looking at a sample of trace before the defined fit_start of length base_fit_length

• base_fit_length: how many points to include in the separate baseline fit

• base_fit_buffer: how many points to leave between the end of the separate baseline fit region and the start of the pulse fit region

• draw: true means that the fitter will show you a plot of each fit and output the fit results as well as waiting for keyboard input before proceeding.

As of now there is only one constructor and it takes one argument, the path to the config file.

See the src/exampleDriver.cxx for an example on how to use the fitter to fit a trace.

• fitSingle(trace) will fit a trace. Trace can be either an array of floats or an array of unsigned shorts. The array must be at least as long as the trace_length specified in the config file.

• fitSingle(trace, error) will do the same as above but assign each sample an error specified in the function call. The default is one if you don't specify.

• fitDouble(trace, error) is for fitting double pulses. It will free the Delta T parameter and attempt to fit for two pulses.

A couple handy getters:

• getFitStart(): returns the sample fit start that was read from the config file
• getFitLength(): returns the fit length that was read from the config file

After doing a fit, you can call the following functions to access fit results. Do not call any of the following without first doing a fit

• getParName(int i): returns name of i'th parameter.
• getParameter(int i): returns value of i'th parameter determined by the fitter.
• getScale(int i): returns the normalization factor of the pulse
• getTime(): returns time of pulse. Right now this is the same as getParameter(0);
• getBaseline(): returns the baseline of the trace.
• getChi2(): returns chi2/ndf.
• wasValidFit(): returns whether or not fitter converged successfuly
• getIntegral(double start, double length): numerical integral of fit function. This can be slow depending on your function
• getFunctionMaximum(): numerical maximum of fit function. This can be slow.
• getFunctionMinimum(): numerical minimum of fit function. This can be slow.

The fitter also has functionality to do a baseline corrected analogue sum of the trace:

• getSum(int start, int length): does baseline corrected analogue sum of "length" points starting with sample "start". Uses the last baseline estimation.
• getSum(trace, int start, int length): calculates baseline and does baseline corrected analogue sum of "length" points starting with sample "start". Trace can be a float or an unsigned short.
• getMax(int start, int length): finds the largest sample between start and start+length and subtracts the baseline. Uses the last baseline estimation.
• getMax(trace, int start, int length): calculates the baseline and then returns baseline corrected maximum.
• getMin(int start, int length): same as above, except for min.
• getMin(trace, int start, int length): same as above, except for min.

There is no guarantee that any of the function parameters are a good energy proxy for the trace. The integral is usually a good energy proxy, but it is also often slow to numerically integral the pulse. Instead, it is better to determine how to extract a good enery proxy from your parameters. The way to do this depends on your fit function.

In the case of a template fit, the scale ( getScale() ) is a good energy proxy.

If you are not doing a template fit, the normalization of your pulse may be correlated with the other parameters that are also varying. This is a case where you have to be careful. I generally integrate the functional form of the pulse in advance to figure out how it depends on the parameters. This can often be done easily with mathematica.