thebestcpu
Fan of Printing
- Joined
- Dec 8, 2024
- Messages
- 101
- Reaction score
- 59
- Points
- 60
- Printer Model
- Epson SC P900
The experiments I shared below were mainly for tuning my editing/printing workflow on the Epson p900. Yet, I thought others might find the results helpful, so I’m sharing my conclusions and experimental data that led me there.
In particular, I focused on grayscale banding differences for various 16 vs 8-bit workflows.
Spoiler Alert: Here are my conclusions for my workflow to minimize banding (your mileage may vary)
--------------------------------------------------
Experiment Setup
For my test, I created a grayscale gradient test image consisting of eight gradient strips spanning the full 16-bit Photoshop range (0 to 32,767). This image was built using Photoshop scripting and sRGB gamma correction. I also created a second version that was converted to 8-bit mode to compare the effects of bit depth.
Prints were only done on Epson Ultra Premium Luster Photo Paper. Albeit limited in nature, it was enough to meet my testing needs.
Here is an image of the test image and a link to the large TIFF data file
Link to 16 bit TIFF test image
-----------------------------------------------------------------
My initial 16 vs. 8-bit test prints were compared visually, and I summarized the results (and not the detailed image results to keep the length of the post more reasonable)
Banding in the 8-bit prints suggested it was widespread, yet the ability to observe the banding depended on many factors, including the luminosity of the image content, the spectral composition of the illumination, the angle of illumination, the lighting intensity, and the viewing angle. Although untested, other potential factors could include the color of a gradient, the observer's eyesight, and more.
These factors may be interconnected with the wide range of online observations/opinions regarding banding, let alone the need or way to mitigate any banding.
I decided to change tack, focus on exposing any amount of banding, and test which printing modes influenced the banding level, and determine if there were any preferred options.
-----------------------------------------------------------------
My test approach was to take the print output, scan it with my Epson V800 scanner in raw mode, add that into Photoshop as a separate Layer above the original 16-bit image, use Difference mode blend, and amplify the difference. All tests were compared against the original 16-bit image with identical amplification for reasonably close comparison.
I did three sets of workflow tests, as shown below. There were two or three tests in each set. I included the data flowchart above the “difference” test result. Here is the legend of the acronyms and numbers for the flowcharts.
PS – Photoshop
CS – ColorSync Utility
D – Printer Driver
16 – Data is in 16-bit (light green background)
8 – Data is in 8-bit (yellow background)
Green to Yellow (and vice versa) gradients represent known bit depth conversion.
A green diamond pattern with a yellow background means I did not have visibility of the bit depth used for the printer driver stages since that is not exposed, and I could not find public documentation.
-----------------------------------------------------
These first three test prints are the 16-bit (far left) and 8-bit (far right) test images sent through the Mac ColorSync Utility in Color Target mode, so there is no additional color management.
The middle test path of starting in 16-bit, converting to 8-bit, and back to 16-bit was an attempt to see if the dithering and halftoning in the Epson Print Driver/Printer would help a 16-bit banded image. I could not see any improvement.
The 16-bit path experienced no banding, while both images on the right had banding.
~~~~~~No Banding~~~~~~~~~~~~~~~~~~~~Banding Detected~~~~~~~~~~~~~~~~~Banding Detected
------------------------------------------------------------------
The following diagram shows three print paths from Photoshop to an Epson printer, each handling color data differently:
Conclusion: To take full advantage of 16-bit printing, ensure “Send 16-bit Data” is enabled in the printer driver.
~~~~~~~~~~~~No Banding~~~~~~~~~~Inconsistent BandingDetected~~~~~~~~~~Banding Detected
------------------------------------------------------------------
This last pair of test images/paths was to determine if banding would be reduced if the conversion to the print color space was computed in 16-bit from within Photoshop instead of within the Print Driver.
Conclusion: No reduction in banding was noticed.
~~~~~~~~~~~~No Banding~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Banding Detected
-----------------------------------------------------
Final thoughts
The conditions that create banding and the ability to observe that banding vary greatly. The need to mitigate and how to mitigate banding is a judgment call.
If image data is created in or changed to 8-bit mode, banding is possible, particularly in gradients. Converting to 16-bit will not help, so masking techniques should be used if needed.
If you create and stay with a 16-bit image (e.g., 16-bit gradients), banding is not as prevalent, so it is one option to consider to avoid/reduce banding.
Final Final Thoughts
I was a bit surprised by how clean it was and the lack of banding when using the full 16-bit path. I would have thought there would be some banding if the printer driver path (or within the printer) dropped down to 8-bit mode.
Again, any constructive feedback/questions are welcome, as it's all a learning process for me.
John Wheeler
In particular, I focused on grayscale banding differences for various 16 vs 8-bit workflows.
Spoiler Alert: Here are my conclusions for my workflow to minimize banding (your mileage may vary)
- If the source image is high-bit depth:
- Edit in 16-bit mode
- Print in 16-bit mode
- If the source image is 8-bit depth (or processed at any point in 8-bit):
- Edit in 16-bit mode (to avoid cumulative rounding errors)
- If banding is “visible” (a judgment call), apply a noise-masking technique.
- Print in 8-bit mode (there is no downside to printing in 16-bit, yet it does not help banding)
--------------------------------------------------
Experiment Setup
For my test, I created a grayscale gradient test image consisting of eight gradient strips spanning the full 16-bit Photoshop range (0 to 32,767). This image was built using Photoshop scripting and sRGB gamma correction. I also created a second version that was converted to 8-bit mode to compare the effects of bit depth.
Prints were only done on Epson Ultra Premium Luster Photo Paper. Albeit limited in nature, it was enough to meet my testing needs.
Here is an image of the test image and a link to the large TIFF data file
Link to 16 bit TIFF test image
-----------------------------------------------------------------
My initial 16 vs. 8-bit test prints were compared visually, and I summarized the results (and not the detailed image results to keep the length of the post more reasonable)
Banding in the 8-bit prints suggested it was widespread, yet the ability to observe the banding depended on many factors, including the luminosity of the image content, the spectral composition of the illumination, the angle of illumination, the lighting intensity, and the viewing angle. Although untested, other potential factors could include the color of a gradient, the observer's eyesight, and more.
These factors may be interconnected with the wide range of online observations/opinions regarding banding, let alone the need or way to mitigate any banding.
I decided to change tack, focus on exposing any amount of banding, and test which printing modes influenced the banding level, and determine if there were any preferred options.
-----------------------------------------------------------------
My test approach was to take the print output, scan it with my Epson V800 scanner in raw mode, add that into Photoshop as a separate Layer above the original 16-bit image, use Difference mode blend, and amplify the difference. All tests were compared against the original 16-bit image with identical amplification for reasonably close comparison.
I did three sets of workflow tests, as shown below. There were two or three tests in each set. I included the data flowchart above the “difference” test result. Here is the legend of the acronyms and numbers for the flowcharts.
PS – Photoshop
CS – ColorSync Utility
D – Printer Driver
16 – Data is in 16-bit (light green background)
8 – Data is in 8-bit (yellow background)
Green to Yellow (and vice versa) gradients represent known bit depth conversion.
A green diamond pattern with a yellow background means I did not have visibility of the bit depth used for the printer driver stages since that is not exposed, and I could not find public documentation.
-----------------------------------------------------
These first three test prints are the 16-bit (far left) and 8-bit (far right) test images sent through the Mac ColorSync Utility in Color Target mode, so there is no additional color management.
The middle test path of starting in 16-bit, converting to 8-bit, and back to 16-bit was an attempt to see if the dithering and halftoning in the Epson Print Driver/Printer would help a 16-bit banded image. I could not see any improvement.
The 16-bit path experienced no banding, while both images on the right had banding.
~~~~~~No Banding~~~~~~~~~~~~~~~~~~~~Banding Detected~~~~~~~~~~~~~~~~~Banding Detected
------------------------------------------------------------------
The following diagram shows three print paths from Photoshop to an Epson printer, each handling color data differently:
- Left (Full 16-bit Path): Photoshop sends 16-bit data to the printer driver, which handles color conversion to the paper profile with black point compensation.
Middle (16-bit File, but “Send 16-bit Data” Off - Right (Full 8-bit Path): The entire workflow uses 8-bit data.
Conclusion: To take full advantage of 16-bit printing, ensure “Send 16-bit Data” is enabled in the printer driver.
~~~~~~~~~~~~No Banding~~~~~~~~~~Inconsistent BandingDetected~~~~~~~~~~Banding Detected
------------------------------------------------------------------
This last pair of test images/paths was to determine if banding would be reduced if the conversion to the print color space was computed in 16-bit from within Photoshop instead of within the Print Driver.
Conclusion: No reduction in banding was noticed.
~~~~~~~~~~~~No Banding~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Banding Detected
-----------------------------------------------------
Final thoughts
The conditions that create banding and the ability to observe that banding vary greatly. The need to mitigate and how to mitigate banding is a judgment call.
If image data is created in or changed to 8-bit mode, banding is possible, particularly in gradients. Converting to 16-bit will not help, so masking techniques should be used if needed.
If you create and stay with a 16-bit image (e.g., 16-bit gradients), banding is not as prevalent, so it is one option to consider to avoid/reduce banding.
Final Final Thoughts
I was a bit surprised by how clean it was and the lack of banding when using the full 16-bit path. I would have thought there would be some banding if the printer driver path (or within the printer) dropped down to 8-bit mode.
Again, any constructive feedback/questions are welcome, as it's all a learning process for me.
John Wheeler
Last edited:
