Interpreting tau or invtau for exponential fits

Share icon

We are doing automated fitting of a series of traces using built-in functions (either exp_Xoffset or dblexp_Xoffset). We have run into an issue where it seems the time constant is given as Tau for some traces and sometimes as InvTau. Our code simply reads the appropriate coefficient and we need to interpret whether it is tau or invtau for the next stage of analysis. My questions are:

  1. How can I tell the difference (programmatically)? Perhaps parse the text of the wave note of the generated fit wave?
  2. What causes Igor to report tau vs invtau? Is it a poor fit that gives tau?

This should not happen. Are you sure you do not sometimes use the old exp and dblexp function (without _Xoffset) in you fits, which are defined with invtau? If you properly use exp_Xoffset or dblexp_Xoffset then I think you should always get tau. Do you have a small example where this problem occurs?

October 12, 2021 at 09:15 am - Permalink

The built-in exp and dblexp fit functions fit to invtau. This is a historical thing- those functions are over 30 years old, and were written when multiplication was quite a lot faster than division.

The exp_XOffset and dblexp_XOffset fit functions fit tau directly. These functions are much newer- now divisions are almost as fast as multiplications and processors are orders of magnitude faster than they were 30 years ago. Most people want tau without having to do a computation.

A side note- the newest of the exponential fit functions, dblexpPeak, also fits tau directly.

October 12, 2021 at 09:20 am - Permalink

Thank you both for your comments. I thought I had been through the code and changed all CurveFit commands using exp and dblexp to the Xoffset versions, but alas there was one I missed! Hence my confusion.

October 12, 2021 at 12:53 pm - Permalink

Igor's accretion of history over the last 30 years is particularly evident in the CurveFit operation, which is amongst the oldest in the application. Sorry for making it all so confusing!

October 12, 2021 at 02:57 pm - Permalink