topic Re: HELP)) Signal deconvolution using JMP in Discussions

HELP)) Signal deconvolution using JMP

Dongjin — Thu, 02 Jan 2020 08:35:49 GMT

Hello, I'm writing because I have a problem to solve.
The company I work for wants to deconvolution the following Original

The Y-axis is the measured Original value.
The X-axis is a kind of energy intensity expressed in measuring equipment.

What I want to do is to separate(deconvolution) the waveform by finding the peak value according to the X value of the Original Signal expressed for the given Y values.

for example, Black line is Original signal,

red line is created signal which has peak at 531ev

blue line is created signal which has peak at 532ev...

I would appreciate your help as we are unable to find a suitable Add-In or method for Discussion.

if you could create an Add-In so that we can do this repeatedly.

Re: HELP)) Signal deconvolution using JMP

Craige_Hales — Thu, 02 Jan 2020 13:49:23 GMT

Are there equation(s) that model the red and blue lines? Do you already know their peak values?

Re: HELP)) Signal deconvolution using JMP

stan_koprowski — Thu, 02 Jan 2020 14:50:00 GMT

Hi @Dongjin,

it seems that you might need to perform FFT analysis. These types of functions can be found in the transcendental functions.

see the following Frequency Analysis and Discovery Summit presentation

to get started.

Others might have more specific examples and use cases on how to proceed.

cheers,

Stan

Re: HELP)) Signal deconvolution using JMP

Dongjin — Thu, 02 Jan 2020 23:43:25 GMT

First of all, I wish you a happy New Year.

I don't know the formula. The problem is found in our lab using the other software, and we think we can use JMP to isolate the signal.

So, the ultimate goal is to create an add-in to get each peak value by each x-value

Re: HELP)) Signal deconvolution using JMP

Dongjin — Thu, 02 Jan 2020 23:52:23 GMT

Thank you for your help.
I've reviewed the data you attached, but in fact, what I'm looking for is not a continuous signal, but just a single peak value based on the x-values we want, so I don't think we can apply it immediately.
I should think more about whether there is any other alternative.

Re: HELP)) Signal deconvolution using JMP

Craige_Hales — Fri, 03 Jan 2020 03:01:25 GMT

Take a look at the distribution platform and see if it does what you need.

I don't know if it is the right answer for the right reason.

I tried it with Y:Binding and Freq:Original Counts and then selected 2nd red triangle->Continuous Fit->Fit Normal 3 mixture.

Two of means look promising

Re: HELP)) Signal deconvolution using JMP

Dongjin — Fri, 03 Jan 2020 04:09:42 GMT

Thank you for your help.
I uploaded a new file.
Deconvolution for 531, 532 x-axis values with existing software
Added value.
Maybe it's using the Lorenz or Gaussian equation, but I don't know the exact formula.

Based on the range of x values, peak values, and Curve, we think we should find a formula to place peak values at x values 531, 532 on the Original Line.

I'd appreciate any further comments.

Re: HELP)) Signal deconvolution using JMP

Craige_Hales — Fri, 03 Jan 2020 12:57:26 GMT

I'm well beyond my comfort zone. Maybe someone else. They probably need to know what the answer will look like:

1: two numbers, like 531, 532

2: a graph, like yours

3: a set of coefficients and exponents for a formula for the red curve, and another set for the blue curve.

The problem, I think, is we don't know what the general form of (3) looks like. It might help to know a little more about how you got the existing results.

Maybe someone with JMP PRO can play with the Functional Data Explorer and see what that can do.

Re: HELP)) Signal deconvolution using JMP

Dongjin — Fri, 03 Jan 2020 13:48:41 GMT

Thank you for your interest in this matter.

I thought it was easy, but it wasn't as easy as I thought.

Considering that there is a separate software that specializes in deconvolution,

it probably won't be easy to solve.

Thank you for your help again, and if find a way, I think it will help those who do similar analysis.

I will continue to find a way and share solutions.

Re: HELP)) Signal deconvolution using JMP

Mark_Bailey — Fri, 03 Jan 2020 13:53:55 GMT

I think that @Craige_Hales solution is good. I was going to suggest it myself, but he beat me to it!

You are estimating the location of the peak height. It seems that you expect whole integer values. Why? Energy in eV can take on any rational value. If you need integers, then round the unbiased estimate.

Re: HELP)) Signal deconvolution using JMP

Dongjin — Fri, 03 Jan 2020 14:51:35 GMT

Thanks @Mark_Bailey, for your advice.

The link below is a page that I found,

http://www.cma4ch.org/chemo/pubbl/peakdeconv.html

but it's similar to the process I've been looking for,

So the file I attached( 'GVisco_deconvoltution1.12_2Gauss.xls') looks similar to the solution I was looking for.
I think someone implemented Excel Deconvolution, but I'm not sure how to apply it to JMP.
I'm leaving a link and an attachment in case who wondering about similar problems

Thanks

Re: HELP)) Signal deconvolution using JMP

Mark_Bailey — Fri, 03 Jan 2020 16:26:55 GMT

As your reference states, PLS is also a classic chemometrics models for spectra. JMP does PLS modeling.

Re: HELP)) Signal deconvolution using JMP

PatrickGiuliano — Wed, 29 Mar 2023 21:23:08 GMT

Since this post was written back in 2020, JMP has made numerous improvements in the direction of solving the Peaks Deconvolution problem that was inquired about here.

In JMP 17 Pro, JMP is much better suited for analyzing peak type data. Basically, there is an alternative way of showing how Wavelet functional data explorer (FDE) and DOE can be used to optimize separation of overlapping peaks for a chromatographic run. For more information, and for anyone currently interested in solving this type with JMP, I recommend the following Developer Tutorial: Modeling Spectral Data Using JMP Pro 17 (available free/ on-demand).

In JMP 18 Pro, there should be more to come in making solving this problem type much easier (more automated). This may include various multivariate curve resolution techniques that will kind of serve as an automated peak identification algorithm. The idea is basically to extract peak widths and to be able to modify those directly.