Wolverine-Hawkeye Telecine

[Bruce Davis]

Hi Stan, my BenQ_PD3200U monitor is not a true 10bit, it is 8-bit + FRC dithering and is 100% sRGB and Rec. 709.
What effect would Binning 1x2 or 2x2 have on image pixel saturation, maybe not much? from what I have read it only affects the SNR and resolution but does nothing for dynamic range (bit-depth).
From your description due to our limited 8 bit viewing platforms (monitors and TV's) we will not benefit from any of this increase to say 12 bit imaging.

Really do not know where to from here.
Maybe the IMX178LQJ larger pixel size, clutching at straws.

Re- lenses :-
(12 Mpix)
Pixels: 4000 x 3000
Pixel Size: 1.85μm x 1.85μm
Sensor Diagonal: 9.5mm
Nyquist Frequency: 270 lp/mm
2/3 of Nyquist: 180 lp/mm

Nyquist Frequency (line pairs/mm) = 1000 / [2 x pixel size (μm)]
Pixel size = 1.85μm
Nyquist Frequency = 1000 / (2 x 1.85) = 270 lp/mm
2/3 of Nyquist: 180 lp/mm

I assume a lens with 180 to 200 lines per mm is required (David's 16mm lens).
ACH1656MAC
Resolution: 10MP 200lp/mm
https://www.aico-lens.com/product/16...ns-ach1622mac/

The 12mm lens currently used TBL 12-2 C 5MP from IS only 5 Mpixel.
None of this probably makes any difference.

Regards - Bruce

[Stan Jelavic]

Good points Bruce. In summary (from my point of view)
- 16mm David's lens main advantage is less flare and better contrast.
- combined with 37UX226 camera for better alignment (S8,N8) and higher resolution
These are the main advantages. But really concerned about that 3.1 USB driver. Not really usable like this.

Let's wait for David for his tests.

For now I am planning to return the camera back to TIS.
BTW - tested it on my Linux machine. Not much different. IC capture is missing some features.
https://photos.app.goo.gl/ajobCm2buaTXEh1GA

[Bruce Davis]

Stan if "The-IS" are willing to accept the return of 37UX226 camera would it be worth testing the IMX178LQJ (no pressure, if you feel it is not worth the effort then don't bother). If the driver is bad (I assume IS have acknowledged this, are they going to fix it) then presumably the same driver is used for the IMX178LQJ.

Many of the better sensors (just from reading) seem to have pixel sizes of around 3.3 - 5.5 µm (APS 3.9 µm).

Looked at your images from the MT9P006 and 37UX226 in Photoshop.
The skin colours of the 37UX226 are better & better overall, more true to life, whereas the MT9P006 is darker (face of woman on left in distance looks a darker grey with just a bit of skin colour), and the overall image has more of a blue tint (this may just be settings).

Thanks for spending the time on this exercise.

Regards - Bruce

[Stan Jelavic]

We will see what TIS say. Will ask them about IMX178LQJ. My main problem is the USB driver. Could be my laptop but David is having similar issues.
Yes I see some fine detail improvement.
Thanks for the suggestions Bruce
Stan

[David Brown]

Stan, you rock. I would never have thought there could be a wiring change with the trigger circuit.

I re-read the Hawkeye user manual looking for a diagnostic checks.
Nothing helped with the trigger, but I found this: Make sure the fan is off. It could sometimes cause 5V startup issue.

Now I know why there is a switch on the fan! I have experienced that startup problem and thought that something was loose or shorting. Makes sense now.

TIP OF THE DAY. Save an IC config file with your captured frames. If you need to recapture some frames to repair part of the whole scene, the exposure and color will match.

This is strange, and I have not determined how it happened. Several minutes of captures would go in and out of focus for only 2 or 3 frames. It might not be noticed when played at speed, but watching at capture speed of 2fps, it was easily seen.I re-captured at 0.5fps and again at 2fps. It did not repeat. Still a good idea to save your configuration file with the capture folder.

Here are 4 contiguous frames. Notice the entire frame goes very soft, frame lines and film grain.


Stan & Bruce. I attempted to try 2 pixel binning today. No binning available on the 37UX. Yes there are some features missing, CONTRAST & BINNING.

This sensor is large enough to possibly use binning. The new driver has settled down. Must have been a conflict with something else I was running.

It will also run the 72BUC or at least I plugged in the camera and captured a still frame.

I haven't used the color/contrast menu much, more focused on the lens. This part may be new to this driver, COLOR MATRIX. Nothing in the help menu. RR, RG, RB ????? I guess this is color fine tuning with the second R= highlights, the second G= midtones, and the second B=shadows?

There is also no information on WDR (WDF on my menu!) I plugged in Stan's values from the Hawkeye manual. Works good. I would not capture anything without HDR turned on.

My limited 2 days with this leave me using- 1 PUSH color balance with an un-cropped frame; 4000x3000. Then crop as close as possible (ROI). No BRIGHTNESS nor GAIN. GAMMA at .75 (varies).

I don't trust the histogram. It is bad or is something different than I'm used to. I am using my monitor for tonal range. That said, it looks pretty good so far.

I can't be the cheeleader for this pairing when I don't know if the expense will equal the final experience.

It does for me.

There is no way to improve on the sensor, that I can find. You can't have a 4k final image from a 4k sensor. Lenses jump from 16mm to 25mm and few have aperture choices. In the end my image will be a 2k at best. If I put this on DVD it will drop down to 720P (1.3k).

NEVER, I'm a registered pixel peeper!

The 12mm lens Stan is using is rated at 5MP, so all we see is a larger physical image with little or no resolution improvement.

I've not seen any footage yet where the WDR has not helped me. If IC is saving files in 8 bit then it might be possible to preserve the details in the highlights and shadows by capturing with very low contrast and saturation, then boosting everything in post when the video file has more bandwidth.

Stan, I also got Nick's capture to work, but He is missing functions that IC currently has. Might be an older version he modeled.

[Stan Jelavic]

Thanks for the report David. That was lots of work on your part. The startup issues with the fan on can be resolved by changing the resistor divider values in the startup circuit.
Not sure why some frames would go soft but here are a few thoughts:
- lens not tight
- mounting screws for the board and or camera left loose
- some issues with the gate - film too loose
I also noticed the additional color settings and WDR with the new camera. WDR does help. I think it is to do with the fact that we use 8 bit monitors. The camera 12 bits have to be fit into 8 bits. I tried capturing the image by reducing the exposure until I can see clouds and then expanded the histogram in the post. The resulting image had all the details but loos flat and dull because there are less bits in the mod range (less contrast).

[David Brown]

Hello Kamel

Thanks again for the test film strip. I have finished installing my lens and every time I tried to focus on the film's grain or sprocket holes, I'd put in the test film in the gate and find it was not completely focused!

I don't know if you have continued to change the switch panel or have had any time to work on it.

I had to close up the Hawkeye. I have been working on the frame skeleton with the switch panel held in place with magnets and alligator clips.

I used one of the older panels you designed, and I had added some recessed labeling before printing it long ago.

It works though the switches are snug and REV had to be moved. It's how I want leave this. It will hold me for now.

Interested to see what you have designed. I can wait if you are not ready.

[Bruce Davis]

Hi David, I am also great-full for your report, disappointing that the DFM 37UX226-ML does not have binning (for testing).

When you stated "The new driver has settled down" does this mean you are able achieve the spec'd frame rates like 8-Bit 30fps, 12-Bit 20fps, 16-Bit 15fps. Many of the security/industrial camera manufactures using this sensor chip are rating their cameras between 5 to 10 FPS (using 1000BASE-T RJ45 connection).

Copied from document - basler-sensor-comparison-are-all-imx-sensors-equal :-
The sensors – and particularly the 12 MP IMX226
– were only developed for operation with a full or nearly
full (4k) resolution. This makes it impossible to achieve
increased speed with a simultaneous reduction of the
read ROI that is otherwise customary in industrial image
processing.

Are you having any problems with time delay, in transferring a 64 bit depth file from the camera to the HDD, considering the file sizes of 15MBytes plus? Normal TIFF file size with the 32-Bit DFM 72BUC02-ML camera was around 4MBytes.

Quote "In the end my image will be a 2k at best" do you mean 2048×1080 or 2560x1440 pixels (Quad HD).

I think DVD for me (as you stated) would be a poor choice, but blu-ray resolution is still good for disc archiving.

Regards - Bruce

[David Brown]

Help Stan! I've done something bad again. Once Hawkeye is in the case and closed, I won't be able to cause more problems!

When I plug in power, the takeup is running full speed. It's the same as rewind. No switch combination will stop it from running. REVERSE will reverse the direction.

Switching on RUN, the takeup will change to increment. Changing speeds increases the increment like it should for 1 or 2fps capture.

I've checked the plugs and looked for broken wires. The same happens if the takeup motor is plugged into the original 12V on the power supply board.
Thoughts?

Bruce, You ask hard questions!

The driver is faster, some things were slow to respond. I thought the runaway value selector stopped, but for Exposure and others it keeps incrementing while the cursor is over top.

2048x1536 is my final crop size.

I've had to revise this as I go. I just realized that ony the specs given for IC Imaging Control matter. We don't use USB3 Vision or uvcvideo Linux.

The Device Driver for The Imaging Source USB 33U, 37U and 38U Cameras simplifies this variety of possible pixel formats and offers two video formats instead: RGB64 and Y16.
The RGB64 format results from the driver's automatic debayering the raw image data, while the Y16 format contains the raw data which is reinterpreted as monochrome.

So we should always use the RGB64 prefix? Is T16 B&W or a RAW file that can be decoded? 24 and 32 do capture smaller files.

Photoshop only recognizes rgb_64_2048-1536 as 16bit color, with 24 & 32 as 8 bit color.

The files I captured suggest thru file size different color depths or container sizes.

8,824,134 rgb24_2048-1536.tiff
9,459,268 rgb32_2048-1536.tiff
22,464,452 rgb64_2048-1536.tiff


I have completely ignored framerates since we are not capturing video. FPS sets the stream bandwidth. Maybe it can be back converted to 'time to save a single frame'. I have not seen any unsaved frames or damaged images.

The shutter speed must be faster than the capture framerate of 0.5 (<2s), 1 (<1s), or 2fps(<1/30s) of the Hawkeye or you'll see blurred frames.

Framerate is what is on the FPS list in the menu.

The available framerates in the IC list vary with the "video size and color format".
RGB 4000-3000 @24&32 are 1-2.91 fps. @64 are 1-1.94 fps.
RGB 2048-1536 @24&32 are 1-11.12 fps. @64 are 1-7.41 fps.
RGB 640-480 @24&32 are 1-30.67 fps. @64 are 1-20.45 fps.

There are multiple performance specs given, we only have access to the IC Imaging Control. Speed can be altered by image size.

[Stan Jelavic]

Hi David. With the run switch off the takeup should be off. It is possible that you have something loose in the takeup switch bringing the 3.3V onto the MSP input pin 8. The takeup switch one side goes to the speed connector pin4. The other side is wired to pin 1 (gnd) of the speed connector through the front panel wiring, Check the voltage on MSP pin8. Should be low when takeup switch is off (and run switch off). If high, trace it through the connector and wiring with a voltmeter. Use a needle or a pin to access the connector pins.

[Stan Jelavic]

I am in touch with TIS and running some tests that they suggested. The frame rates that I am getting are much slower than what the reference doc says. I will try their USB3 cable and see if that helps.
David, my experience is that the frame rate at 2FPS speed has to be around 20 for very dark images. For normally lit images it can be lower.

[Bruce Davis]

Thanks David for the detailed information, there is a newer version of IC_Capture available "IC Capture_2.5.1525.3931_May 11, 2020" not sure if it is any better. Sorry did not mean to create any difficulty.

2048x1536 is about 3 mega pixel that's very good, Stans capture with the 12mm lens is about 2 mega pixel, not sure if one could pull any more detail from the S8 frame by increasing resolution even further. File size may also be a consideration (storage space).

Far from being an expert and I do not know the formula for calculating lenses that could be used, so I use these two calculators, there may be better, but have not found any :-
http://scorpion.tordivel.no/LensCalculator.htm
https://www.flir.com.au/iis/machine-...ns-calculator/

Test sensor (1/1.7) IMX226CQJ with 16mm lens used by David:-
2048 x 1536 @ 0.00185mm pixel size = 3.7888mm x 2.8416mm (full sensor size = 7.5332mm x 5.6351mm).
That's utilising about 1/4 of the sensor area (3.146:12.403).
That's about the same size as a 1/4" sensor type area (actually fractionally larger).

Currently used sensor (1/2.5) MT9P006 with 12mm lens:-
1256 x 942 @ 0.0022mm pixel size = 2.7632mm x 2.0724 (full sensor size = 5.7024mm x 4.2768mm).
Very crudely that's utilising about 1/4 of the sensor area (1.183:5.039).

Just added some useless information to show the small area of these sensors being used.

Regards - Bruce

[Kamel Ikhlef]

Hello everybody,
David,
I haven't touched up the front panel yet. because I'm waiting for V12 (not yet received).
As soon as I have it, I will modify the panel front cover accordingly.
If you have ideas of modifications according to your tests and experiences,
make me a small diagram in pencil and I will make the 3d design.
It will also serve others.

[Stan Jelavic]

You were on a pretty good start Bruce. I would like to add to it.
Check this interesting article:
https://static1.squarespace.com/stat...apixellens.pdf


Expanding your analysis I used the reasoning in the article and calculated the requirements for of the current and test sensor and the two lenses that we use. The summary is that the two lenses are very close and good for the current sensor but not as good for the test sensor.
But, we have to keep in mind that we operate way beyond the specified range for these lenses.

16mm spec

Suitable Distance:

0.1 ~ 0.5m

12mm
Not specified

Here is the detailed calc:

TBL 12-2 C 5MP Lens
5MP
1/2"


a - horizontal width im mm
b - vertical width in mm
d - density - pixels per mm

a = 6.4mm

a*d*b*d = 5MP
a*b*d^2 = 5MP

b/a = 0.75
b = 0.75*a

a*(0.75*a)*d^2 = 5MP
d^2 = 5MP/(0.75a^2)
d^2 = 5MP/(0.75*6.4^2) = 5/(0.75*41) = 162,601
d = 403 P/mm

LP/mm = Pixels per horiz line/(line width*2) = 403/2 = 202


Currently used sensor (1/2.5) MT9P006:-
LP/mm = 2592 Pix / (5.7mm*2) = 227


Test sensor (1/1.7) IMX226CQJ
LP/mm = 4072 / (7.5332*2) = 270

16mm David's lens
10MP
2/3"
a = 8.8 mm

d^2 = 10MP/(0.75a^2) = 10MP / (0.75*8.8^2) = 10MP/58 = 172,413
d = 415 pixels per mm
LP/mm = Pixels per horiz line/(line width*2) = 415/2 = 208

[Bruce Davis]

Hi Stan, thanks for the link and reply, I can only go by the information I have read, as I have not been educated in any of this.
https://www.visiononline.org/userAss...uartSinger.pdf MEGA PIXELS starts at page 65.

As always I put my faith in your knowledge of this subject.

The test sensor calculation comes comes up with the same answer of 270 lp/mm:-
Nyquist Frequency (line pairs/mm) = 1000 / [2 x pixel size (μm)]
Pixel size = 1.85μm
Nyquist Frequency = 1000 / (2 x 1.85) = 270 lp/mm
But not sure why the recommended lp/mm lens is 2 thirds of the 270
2/3 of Nyquist: 180 lp/mm

Quote from your linked document "Ideally the lens resolution should be greater than the camera resolution".

Testing the FPS with the TIS USB3.1 cable will be interesting.

Thanks again Stan, Regards - Bruce