What would you like to see in a plotting program?

[RalphLambrecht]

Steve

You can add me to your waiting list. As soon as this is available for MacOS, I', buying!

 

[ic-racer]

Cricket Graph?

It was Delta Graph. I did a quick search and it looks like they may have a current product.

 

[RalphLambrecht]

It was Delta Graph. I did a quick search and it looks like they may have a current product.

They sure do. A wonderful piece of software, I've used it extensively for our book, but not applicable for Steve's project.

 

[Stephen Benskin]

They sure do. A wonderful piece of software, I've used it extensively for our book, but not applicable for Steve's project.

I'm curious to know more about its capabilities and its website.

 

[ic-racer]

They sure do. A wonderful piece of software, I've used it extensively for our book, but not applicable for Steve's project.

Yes, its just 'make shift.' I'd love to see Steve's project materialize.

Also, what about the X-axis? I know that if you are "calibrated" you can put the intensity there. What about just having step wedge density flipped (that is how I'm currently plotting)

 

[Stephen Benskin]

What about just having step wedge density flipped (that is how I'm currently plotting)

Could you explain that in more detail.

While the speed readings are done in actual log-H, I thought it would be easier to plot in relative log-H.

 

[RalphLambrecht]

I'm curious to know more about its capabilities and its website.

Steve

Think of it as a wonderful replacement extension to Excel, without the calculation capabilities, but with extremely flexible plotting features and customizable curve smoothing functions.

I still think, it won't help with your project though, but if you want to take a look:

http://www.rockware.com/product/overview.php?id=83

 

[Tim Gray]

Re: graphing software

A very simple graphing utility for Mac is OmniGraphSketcher. You can copy data from an spreadsheet and just paste it in and it will plot it.

For work stuff, I use Python. It's a bit DIY, since it's programmatically driven, but it's very powerful once you install matplotlib (plotting library) and scipy (scientific function library). If you've ever used Matlab, it's similar, though it's free.

 

[Stephen Benskin]

To me, the importance of a good plotting program isn't the plotting of the curve but the ability to interpret it. I have a good example I hope to upload later today.

Ralph,

I was wondering about your idea of a variable speed. Do you want the capibility to assign a base speed point to any density or are you looking for a way to determine a speed from different fixed density points, or is it something else?

ic-racer,

"Third, automatic interpolation to find 0.1 intersection." I think I know what you mean, but could you elaborate just in case?

 

[RalphLambrecht]

Ralph,
I was wondering about your idea of a variable speed. Do you want the capibility to assign a base speed point to any density or are you looking for a way to determine a speed from different fixed density points, or is it something else?

'the capibility to assign a base speed point to any density'

but

'a way to determine a speed from different fixed density points'

is a good idea too!

 

[ic-racer]

Could you explain that in more detail.

While the speed readings are done in actual log-H, I thought it would be easier to plot in relative log-H.

I just use the actual density of my stepwedge on the X axis. So when I plot it out like that the curve goes the wrong way (the toe winds up on the left-hand side).

So I flip the X-axis to go from 3.0d on the left and it goes down to 0 on the right. That way the toe is on the left-hand side. In the 'non-calibrated' world of simple process control strips, the actual density of the step wedge is probably the best independent variable to plot.

 

[ic-racer]

'the capibility to assign a base speed point to any density'

but

'a way to determine a speed from different fixed density points'

is a good idea too!

I agree.

I have seen reference books instructing readers to use density values lower than 0.1 to obtain EI speeds for 35mm hand-held photography. Again, not ANSI/ISO testing, just getting practical, useful personal EI.

Also, someone mentioned transparency film and I think 0.2 below shoulder is a popular exposure index point.

 

[ic-racer]

"Third, automatic interpolation to find 0.1 intersection." I think I know what you mean, but could you elaborate just in case?

This is where I once tried to solve polynomials. So I'd like to see the program do a curve fit on the toe (exponential, spline or polynomial) and mathematically solve for the speed point (0.1 on the Y axis, or what ever value the user wanted) . Thus spitting out the X-axis number via a calculation, rather than manual interpolation.

 

[Stephen Benskin]

'the capibility to assign a base speed point to any density'

but

'a way to determine a speed from different fixed density points'

is a good idea too!

Ralph,

The first way is a good idea and would be easy enough. The second could potentially run into some conceptual problems. Maybe a gradient approach might be applicable or using one of the known points as reference and adjusing the constant in the speed equation. Afterall, that's kind of how it's done now. This would definitely be an interesting topic to further discuss albeit one that would be rather high end.

Steve

 

[Stephen Benskin]

I just use the actual density of my stepwedge on the X axis. So when I plot it out like that the curve goes the wrong way (the toe winds up on the left-hand side).

So I flip the X-axis to go from 3.0d on the left and it goes down to 0 on the right. That way the toe is on the left-hand side. In the 'non-calibrated' world of simple process control strips, the actual density of the step wedge is probably the best independent variable to plot.

The calibrated densities of the step tablet are internal in my program. If I change the step tablet, I just have to enter the new values. I've gone back and forth about having an option to use actual log-H based on the value of the exposure and calibrated step tablet.

My program finds specific points on the curve using the Pset function. It reads the color of a particular X,Y coordinate. I place it in a loop and move it incrementally across the graph until it runs into the curve. It's kind of a cheat, but I'm not great at math.

 

[RalphLambrecht]

This is where I once tried to solve polynomials. So I'd like to see the program do a curve fit on the toe (exponential, spline or polynomial) and mathematically solve for the speed point (0.1 on the Y axis, or what ever value the user wanted) . Thus spitting out the X-axis number via a calculation, rather than manual interpolation.

I gave up on using polynomials to curve-fit film or paper characteristic curves, but had much success with up to 4th order non-linear functions.
Here is an example:

y=(a0+a1*x+a2*x^2+a3*x^3)/(b0+b1*x+b2*x^2+b3*x^3+b4*x^4)

They are great if you want to cover toe and shoulder in one equation, otherwise, 2nd or 3rd order are fine and even simple polynomials will do.

 

[RalphLambrecht]

Ralph,

The first way is a good idea and would be easy enough. The second could potentially run into some conceptual problems. Maybe a gradient approach might be applicable or using one of the known points as reference and adjusing the constant in the speed equation. Afterall, that's kind of how it's done now. This would definitely be an interesting topic to further discuss albeit one that would be rather high end.

Steve

The first is a 'must have' to me. The other would be nice to have, but why don't you keep it for version 2.0.

Seriously, don't make it too complicated. It's better to start with something simple but useful and add on to that, rather than making it too complicated.

 

[RalphLambrecht]

... I place it in a loop and move it incrementally across the graph until it runs into the curve. It's kind of a cheat, but I'm not great at math.

Sounds like the poor man's differential equation. Nothing wrong with that. That's how Phil Davis' math worked.

 

[ic-racer]

Personally I think the "0.3 x G-bar" method of speed is the best, but it came at a time (pre-computer) when it was too hard to really get a good number by manual graphing.

Now that we have supercomputers on our desks (to play graphic internet commericals) I'd think that computing power could be put to some good use and mathematically solve for the "0.3 x G-bar"

For those that don't know (or don't understand my nomenclature) the speed point I am mentioning is the point on the toe where the instantaneous slope equals three-tenths the slope of the main curve. In plain terms, it is the point on the toe where the contrast is three-thenths the contrast of the straight portion.

Anyway this is a WISH LIST (I really don't expect this on a basic first version)
3 ways to obtain speed:
1) 0.1 (or user indicated) above FB+F
2) 0.3 x G-bar
3) intersection of the straight portion of the curve extrapolated in a straigh line toward an intersection with the x-axis.

 

[RalphLambrecht]

...Anyway this is a WISH LIST (I really don't expect this on a basic first version)
3 ways to obtain speed:
1) 0.1 (or user indicated) above FB+F
2) 0.3 x G-bar
3) intersection of the straight portion of the curve extrapolated in a straigh line toward an intersection with the x-axis.

It's on my wish list too, but knowing about Steve's passion and knowledge, when it comes to the subject of measuring film speeds, I'm sure, this will be in his first version.

Steve
If you are serious about this, and I hope you are, I'll help where I can!

 

[Tim Gray]

Most graphing software can fit a cubic spline to a set of data. Since from what I understand, there is no fundamental equation that governs the shape of a film curve, why not just do a spline fit? Then just pick off the point of 0.1+fb?

I'm sure there are open source routines for spline fitting.

 

[ic-racer]

Here is an example of some C++ spline functions. One problem I have had with splines is that in the software I have been using (Deltagraph) it won't give me an equation for the spline. It just uses the spline for the graph output. It gives a nice smooth shape, but won't do any calculations on it. Its a "display only" function. I have used other software like that. So, my question is how do you solve for unknowns with the spline? Maybe its only "PC" software that can do that

 

[Stephen Benskin]

I've attached an example of a couple curves with some analysis as an example of how I was able to answer a question that had been perplexing me for years. As with most people who want to get serious with photography, I began with the Zone System. I learned that a normal negative is one with a negative density range of 1.25 from a scene luminance range of 2.1 (7 stops). Later when I started reading about sensitometry, I learned that a normal negative density range for printing on a grade 2 with a diffusion enlarger is 1.05. Why the discrepancy? And why did both methods produce printable negatives on grade 2?

I got my answer using the program. Both curves are identical, the only difference are the testing parameters. The ISO/ANSI model assumes flare. The discrepancy turns out to be an illusion. One can the validity of the various theories, but the question was answered.

 

[Stephen Benskin]

Anyway this is a WISH LIST (I really don't expect this on a basic first version)
3 ways to obtain speed:
1) 0.1 (or user indicated) above FB+F
2) 0.3 x G-bar
3) intersection of the straight portion of the curve extrapolated in a straigh line toward an intersection with the x-axis.

So, the current ISO/ANSI fixed density, the fractional gradient, and inertia speed methods. I'm sure you've heard me talk about it, the Delta-X Criterion is a way to identify the fractional gradient point and produce speeds equivalent to an updated fractional gradient method. The ISO fixed density point with the ISO parameters is part of the Delta-X criterion. Under that set of conditions, the fractional gradient point will fall 0.296 log units below speed point. So, 1 and 2 are already incorporated.

I've attached an example of a CI/EFS/Delta-X curve showing how the different methods produce different speeds. Both curves are produced from the same set or data.

 

[RalphLambrecht]

Here is an example of some C++ spline functions. One problem I have had with splines is that in the software I have been using (Deltagraph) it won't give me an equation for the spline. It just uses the spline for the graph output. It gives a nice smooth shape, but won't do any calculations on it. Its a "display only" function. I have used other software like that. So, my question is how do you solve for unknowns with the spline? Maybe its only "PC" software that can do that

Same here! What good is a curve fit without equation?