Epson V800 Calibration Area

[IanBarber]

When the scanner is turned on or a preview is performed in Vuescan, the scanner appears to do a self calibration routine. Dows anyone know what it is calibrating.

[Ted Baker]

As far as I can tell, Each sensor is being "reset" so that "1" equals the maximum intensity, so the output is a "digital white" 1,1,1.

Using vuescan you skip this step entirely.

[IanBarber]

As far as I can tell, Each sensor is being "reset" so that "1" equals the maximum intensity, so the output is a "digital white" 1,1,1.

Using vuescan you skip this step entirely.

How can you skip the step please Ted

[Ted Baker]

How can you skip the step please Ted

Just skip the preview button and just press scan only, that is the most compelling feature of viewscan IMHO. I have a profile for each type of film i scan. So if a have 20 sheets of HP5. I can scan that in 10 passes of the sensor which should take 80 minutes... Everything including the exposure I want for the FP5 film base, cropping, resolution is already in profile. So no preview required of anything.

The per pass calibration still occurs BTW, just to correct parameters loaded by vuescan.

Its impossible to figure out exactly what the scanner does, but by looking at the source code of the driver, etc and how it responds some reasonable assumptions can be made.

[IanBarber]

Ted, I assume that you are using a holder which always ensures the negative is in the same place for your profile to correctly place the boundry box around the negative.

I personally place the negative on the wet mount holder but do scan it dry. I probably could make a template or something showing me where to put the negative and then like you create a fixed profile

[Ted Baker]

Ted, I assume that you are using a holder which always ensures the negative is in the same place for your profile to correctly place the boundry box around the negative.

I personally place the negative on the wet mount holder but do scan it dry. I probably could make a template or something showing me where to put the negative and then like you create a fixed profile

or just select the whole area, or a large enough area to be sure. The scanner always scans the whole width, if I have two 4x5 images I still just scan in one complete pass. I wrote my own software for 35mm and 120 to digitally cut the film. For 4x5 just scan the whole area and cut it "photoshop" much less fiddling around. My software would work for 4x5 if I adjusted it.

You just need the exposure to be set to suit the film base, or (you skip that consider the film base to be a density of 0),

I can scan 6 rolls of 35mm in 1 hour at 2400dpi, on my old 4990

The only problem with this approach is 4Gig tiff file limit imposed my most if not all scanning apps (including vuescan) This might be a problem with colour, if like to scan beyond 2400dpi.

[IanBarber]

Ted, are you scanning for a RAW output or just a regular tiff

[Ted Baker]

Ted, are you scanning for a RAW output or just a regular tiff

Always RAW, however depending on what options you actually choose there isn't much difference, but it it depends on what options you actually choose... That is actually why I like vuescan, as I understand all the options I can and I DO them myself when appropriate OUTSIDE of vuuscan. If you already have epson scan you can achieve exactly the same thing with the following exceptions, what you get with vuescan is the ability for:

1. No need for preview (save time)
2. Can adjust the exposure to optimum.
3. Can easily do multiexposure (useful for old Kodachromes)

These are are hardware related.

You can get back to the RAW from epson scan if that is what you want?

[IanBarber]

Ted, you mentioned...

2. Can adjust the exposure to optimum.

Am I right in thinking that in Vuescan, you press the "Lock Exposure" checkbox which then reveals the RGB exposure slider. Moving the slider to a higher number to decrease the exposure and visa versa. Is this controlling the physical brightness of the lamp or just a software adjustment.

Are you changing this value looking for a specific value for the film edge.

Ian

[Ted Baker]

Ted, you mentioned...

2. Can adjust the exposure to optimum.

Am I right in thinking that in Vuescan, you press the "Lock Exposure" checkbox which then reveals the RGB exposure slider. Moving the slider to a higher number to decrease the exposure and visa versa. Is this controlling the physical brightness of the lamp or just a software adjustment.

Ian

Neither, it adjusts the timing of the logic gates that clear the sensor, and gates that take charge from the sensor to the A/D to give a number. A bit like adjusting the shutter speed on a FP shutter does not speed anything up at all it just changes the delay between the curtains... The curtains always travel at the same speed.



Are you changing this value looking for a specific value for the film edge.


Yep to ensure that the film base is a close to "white" or about 90% in linear measurements. This also speeds up further software steps later, as I don't need to make another measurement.

[IanBarber]

Neither, it adjusts the timin of the logic gates that clear the sensor, and gates that take charge from the sensor to A/D to give a number.

So briefly how are you using it your workflow and was i correct about pressing the Lock Exposure to reveal the slider

[IanBarber]

Once you have your scanned RAW file in Tiff format, are you then doing the conversion in ColorPerfect, Negfix8 or another way

[Ted Baker]

So briefly how are you using it your workflow and was i correct about pressing the Lock Exposure to reveal the slider

Yes, removing the lock exposure reveals the slider, and then you can manually control the exposure. If you don't do that it will use the value calculated from the last preview scan. This is downside of vuescan is that while you CAN choose exactly what you want if you choose the wrong option, you might not get what you wanted...

How I use it?

I have all the options set I want for a particular type of film. I scan the entire platen area in one pass.

These options are:

1. RAW, i.e. take the digital values that come out of the buffer after the A/D has converted them and pixel shift algorithm has completed, and put them in a file untouched.
2. Have the exposure set so that the RGB values come close to 90% linear values (can see on the raw graph) for a particular film stock. For BW you only need one color...

Nothing else. Very simple and fast, or at least as fast as epson is !

I then use software details here https://www.largeformatphotography.info/forum/showthread.php?147672-scantools-a-few-tools-to-help-with-film-scanning

though I use a more updated version.

1. To digital cut and frame each image
2. I then invert using the same (at least similar) method as the Kodak Cineon system, though I am still building my calibration matrices... (that is not in the github release)

I have something similar working for my D700 (DSLR).

Ultimately I hope to make something useful that will probably work best with A DSLR, but my old 4990 is still useful. Plus I like to some photography at the same time :-)

[IanBarber]

Thanks Ted.

Just to clarify one thing...
You mentioned that you adjust the exposure slider to give you white or 90% of the film edge. Would the film edge not give you black as it is clear or am I missing some step

[Ted Baker]

Thanks Ted.

Just to clarify one thing...
You mentioned that you adjust the exposure slider to give you white or 90% of the film edge. Would the film edge not give you black as it is clear or am I missing some step

No the film base being the most transparent part of the image will let the most light through, this will give the highest value on the sensor. i.e. close to "white" or "1" if your using normalised numbers. I would expect to see normalised values approximately 0.9 in the file i.e. 0.9 x 2^16 = 58982. If you put the Raw graph on it vuescan it will show a graph of these values against a linear scale. That is good enough, it just the consistency that matters 0.85, is just as good a 0.95. But you do not want to go over 1.

[IanBarber]

I have just done a preview of some film edge. Would you agree this doesn't look right. The graph is also set to RAW view

194368

[Ted Baker]

I have just done a preview of some film edge. Would you agree this doesn't look right. The graph is also set to RAW view

194368

No it looks good, if you look at the raw file outputted it should look white or have values close to white, it might vary a bit depending on what program you open it with.

If you scan RAW it should look like film, if may look dark at first if you imaging tool does not understand that file is saved with a gamma of 1. i.e. without any gamma encoding. If your shooting HP5 it should look like a negative.

If you tick the lock exposure butting what factor is displayed?

[IanBarber]

The outputted value of the film edge in Photoshop shows an L value of 97

I presume by scanning the entire bed, you then manually crop the image. Doe scanning the entire bed not affect the final image size?

[Ted Baker]

The outputted value of the film edge in Photoshop shows an L value of 97

Using L in photoshop in THIS instance will probably just confuse you if you don't understand the underlying formula.



I presume by scanning the entire bed, you then manually crop the image.

No I as mentioned earlier I wrote some software to do that, but for 4x5 it is not really necessary



Does scanning the entire bed not affect the final image size? Yes, until it is cropped, but that doesn't really matter.

[IanBarber]

Using L in photoshop in THIS instance will probably just confuse you if you don't understand the underlying formula.

Is there a method you can suggest to measure the value of the white or is an understanding of the formula required for this

[Ted Baker]

Is there a method you can suggest to measure the value of the white or is an understanding of the formula required for this

Lots of simple photo tools, even a text editor if you use an image with consistent density, e.g. clear film, Photoshop probably as well but I never use it, if I measured your graph at 92%. Then in photoshop if you took the RGB value of and I am guessing this is monochrome image, and then expressed it as a percentage it should either be 92% or 83% depending on if you told photoshop that the file was not gamma encoded. 83% is 92% ^ 2.2

Ultimately if the real value is 92% then you want to all your calculations with 92%, all behind the scenes calculations need to be done in a linear colourspace, otherwise they don't work... How you display that to a user is something different entirely. Our perceptual model is not linear (mathematically) at all that is why we have things like the L in photoshop in the first place.

If you post the file that is probably the easiest :-)

[IanBarber]

Ted,
I have found this a very interesting discussion. I understand nearly all of what you have explained, just a little confused with the latter about measuring the white. I use Photoshop mainly.

My initial thought pattern was that if the RAW output gave me an L value of 97 after the file was assigned a gray gamma of 2.2 then this would have represented 97% brightness. I just need to work more on this to get a grasp of what you was explaining.

[Ted Baker]

Ted,
I have found this a very interesting discussion. I understand nearly all of what you have explained, just a little confused with the latter about measuring the white. I use Photoshop mainly.

My initial thought pattern was that if the RAW output gave me an L value of 97 after the file was assigned a gray gamma of 2.2 then this would have represented 97% brightness. I just need to work more on this to get a grasp of what you was explaining.

the L is very complicated formula that attempts to model human perception and includes colour in that percetpion, but the input is the real values that can measures with instrumentation. In the case WE ARE discussing is to get the real values used in the INPUT. Plus you never know if photoshop does it correctly... Basically cut to the chase and don't confuse the issue.

I have attached a tiff file that is 100x100 pixel it contains all pixels that have a value of 92% of 2^16 i.e. 92% of the biggest number you can store. If you open it in Photoshop hopefully there is something that shows it 92%,or the number 60292, or 235 if you using an 8bit scale. I put the number 60292 directly. If your doing digital image processing you need to use a real linear scale, i.e. 1,2,3,4,5,6,8,9,10 and you want to use the real numbers that get from the sensor, or you use some calibrated number and you are aware its calibrated.

Its a complex topic, which I tried to simplify...

194370

Actually I can't seem to add a tiff file so I will attach a jpeg with the same data. (actually 92.16%)

[IanBarber]

Thank Ted.

In Photoshop using the 8bit RGB sampler, it does indeed show 235

194371

[Ted Baker]

Thank Ted.

In Photoshop using the 8bit RGB sampler, it does indeed show 235

194371

Ok so take that value divide by 255 and you have the raw value, without gamma applied its 92.16%.

If we imagine that value represents the light hitting the sensor in a give time by the hardware, and you put a piece of film in between the light and sensor that lets 18% of the light through. It would look to the human eye about half of the light was removed, but the sensor would record 18% of 92% or 18% of 235 which is 42. The sensor is not perfectly linear but it not bad through its sweet spot.

42 is 16.4% yet most people would describe this perceptually as around 44% give or take. For image processing your are usually interested in the former value. Real people are more interested in perceptual scale that makes some sense, and there is a disconnect. In analogue photography we use a real values but we use logarithms so we can use ratios to deal with the real linear measurements underneath.

I.e. a char curve has the real values that are measured by a linear sensor, plotted as log values.

[IanBarber]

Do you have any RGB exposures in vuescan you have recorded for say FP4 FomaPan or HP5, just curios to see how they relate to what Im testing at the moment or would that be a false comparrison

[Ted Baker]

Do you have any RGB exposures in vuescan you have recorded for say FP4 FomaPan or HP5, just curios to see how they relate to what Im testing at the moment or would that be a false comparrison

I will try and find some test shots of grey scales that I am sure I created with HP5. I do my testing and software dev with color film, and real photography in black and white. If your shooting in B&W a lot of things are simplified. Often as a starting point you want to try and mimic the resultant tonal scale of grade 2 paper when used with normal b&w film.

I think color perfect offers that, I have done something similar.

For B&W and 16bit you can always recreate anything with a curve, that is NOT the case in colour.

[IanBarber]

Before I retire for the night, something just occurred to me.

Because the film edge is very narrow, when placing the rectangle around it in VueScan to establish the exposure for nearly white, under the crop Tab in Vuescan is a Buffer % which ignores inside the crop box when calculating the exposure. I am wondering whether this needs turning down from the default value of 8%

[Ted Baker]

Before I retire for the night, something just occurred to me.

Because the film edge is very narrow, when placing the rectangle around it in VueScan to establish the exposure for nearly white, under the crop Tab in Vuescan is a Buffer % which ignores inside the crop box when calculating the exposure. I am wondering whether this needs turning down from the default value of 8%

Possibly, I just used the film leader, and for 4x5 I just had many spare sheets that I forgot to press the shutter, before I developed the film... A long as you get around 80-90% that is fine. Its the consistency that I am mainly after.

[IanBarber]

Ignore this comment, user error