Pages: [1]   Go Down

Author Topic: Synthetic Isis targets for measuring optical and patch registration  (Read 1005 times)

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350

The Isis chart scanner is a precision instrument and differs from the I1Pro charts in significant ways. There are no requirements at all for adjacent patch colors to vary. The boundaries between patches are not determined by detecting color change. Instead, the patch locations are determined entirely from an initial calibration of a black bar, 6mm in height and the exact length of the patches. It is set exactly 6mm from the first row of patches. As the scanner goes down each row it also uses little "diamonds" at both ends of each row that are 3mm high and 6mm wide to maintain, and adjust for changes in each row as it moves down the rows.

In order to have a process to verify that the instrument tolerances are sufficient I created a set of charts to determine the portion of each patch, vertically and horizontally, that is used to determine, and report the patch color. I was interested in the mechanical errors associated with the machine, systemic error in forward v reverse motion, and vertical positioning error as the scanning on each row progressed.

Their are two chart types. One with gradually increasing transitions vertically which allow determining the horizontal resolution of the spectro and one with gradually increasing horizontal transitions that do the same for vertical resolution.

For the 6mm by 6mm (default) patch size in US letter Profile there are 29 horizontal patches in 33 rows.

That attached chart shows both horizontal and vertical response. Values are scaled from the black point to white point, 0 to 1. The horizontal scale is in mm. For both the horizontal vertical the each row is plotted on the left, and the average of the rows is plotted on the right.  As can be seen, there is about 1mm margin at both ends where the response is negligible. This allows for significant registration errors.

A point of note is the jagged nature of the horizontal row plots. This is due to the spectro sampling points. The spectro readout is not continuous. It is reset and light accumulated for 5 milliseconds or so then read out. It's windowed so as not to include light near the patch transitions. However, because of the sampling jitter, the window on the horizontal is smaller than the window on the vertical. Jitter on the vertical charts is not due to spectro sampling but variation in the row positioning. The latter is controlled by the accuracy of the stepper motor advance and the optical sensor of the "diamond" locations. The maximum deviation of the row positions is slightly less than .25mm, which, I think, is quite good. Especially as I put the prints in at a slight angle so the stepper motor was shifting the paper tiny amounts as the spectro scanned left to right and back.

Attached are the tif files at 600 DPI which can be used to print and check an Isis should a reader wish to do so. Also, a text file, which can be used for both vertical and horizontal scans, that can be loaded into the Isis as a CGATS patch set. Just set the paper to profile and the dimensions of 6mm by 6mm (defaults) and it should produce a target image that matches (roughly) then go to "Measure" and the patches will be read by the scanner. Save in CGATS, format. Grap the "Y" value, which is linear luminance. and the data can be processed with a bit of work in your favorite analytical app. I use Matlab but Excel can easily be used too.
Logged

smilem

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 272
    • Color Management Services
Re: Synthetic Isis targets for measuring optical and patch registration
« Reply #1 on: June 10, 2018, 02:48:04 AM »

Just would like clarification, you forgot to label the axis of the chart.
horizontal scale is in mm, vertical log scale? is how much variation?

Regarding jagged graphs the digitaldog claims isis measures 100Hz so there should be no jagged edges or isis does not measure at so high frequency imho. I myself never so this specification listed anywhere by Xrite. They told me this is confidential information.
https://youtu.be/TNdR_tIFMME?t=391
« Last Edit: June 10, 2018, 04:08:05 AM by smilem »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Re: Synthetic Isis targets for measuring optical and patch registration
« Reply #2 on: June 10, 2018, 12:00:15 PM »

Just would like clarification, you forgot to label the axis of the chart.
horizontal scale is in mm, vertical log scale? is how much variation?
Yes, the horizontal axis is in mm. The vertical axis is linear and represents the fraction of total sampled light accrued from the patch as measured from the patch edge displacement. The max of 1 is scaled to the white patch "Y" value from the CGATs file. You can see from the charts that about 75% of the patch's reflected light is measured from only about a 2mm portion near the center of the patch. The vertical is a bit broader and measures about 60% of the patch within the centered 2mm.
Quote

Regarding jagged graphs the digitaldog claims isis measures 100Hz so there should be no jagged edges or isis does not measure at so high frequency imho. I myself never so this specification listed anywhere by Xrite. They told me this is confidential information.
https://youtu.be/TNdR_tIFMME?t=391
It is the sampling that causes the jagged edges. The patches are made up of black and white windowed sections starting off 97% black and decreasing to white over the 29 horizontal iSis patches on each row. The graphs on the left are overlaid fractions read from each patch in succession. Since each sample point the graphs are the result of sequential readings of successive patches, the sampling, which I presume is at fixed frequency, occurs at slightly different locations on successive patches resulting in the appearance of jitter, and even slight reversal in the fraction read on successive patches that actually have more "white" area. This is simply a consequence of the fact sampling will not occur on exactly the same physical position on each patch.

Note that the overlaid vertical values are individually fairly smooth. This is because the black white window in the patches designed to measure vertical aperture are not sensitive to the sampling timing. There is, however, displacement variation that is about _/- .2mm. This can be ascribed to the paper feed stepper motor resolution together with variation in the optical sensing accuracy of the vertical registration marks. So their is variation due to vertical registration though quite small.

It should be noted that the vertical error increases when the M0/M1 mode is enabled because the paper is back hitched to read a uV pass after the white LED pass is made. I do not show this but it's about an additional +/- .1mm on a US letter size sheet and probably increases with paper size.
Logged
Pages: [1]   Go Up