Page 2 of 5

Re: Footsie Bookscanner

Posted: 10 Jun 2015, 07:24
by wmalcolmk
Reflections and Trigonometry

When I built my first scanner, I worked out the position of the lamp by trial and error. However, it should be possible to calculate this, and at the same time gain a better insight into how camera position and glass reflections influence performance.

This diagram shows the key angles...
C is the camera position
F is the lamp position
AB is one side of the glass platen
angle FAB is 50deg (in most scanners)
CDA is 90deg since we must avoid tombstone distortions

The most serious reflections into the camera lens follow the path FBC. That is when the lens can just see the lamp reflected in the edge of the platen. It is probably best to have the lamp a little bit higher, since internal reflections can occur in the camera lens if stray light enters at all. It should be noted that BD is larger than AD, since the platen must be able to handle book covers.

Finding the length AF now reduces to fairly simple trigonometry if we know the lengths CD, BD, DA and the angle FAB.

The steps in the calculation are..

angles DCB, CBE, and FBE are equal and can be calculated, since

tanDCB = BD/DC (both lengths are known, so the angle is known)

If we look at triangle ABF, we know all its angles and the length of side AB, so we can use the law of sines to calculate side AF.
(See for more information)

AF/sinABF = AB/sinBFA

I put these calculations into a spreadsheet, and produced results that agreed with my measurements. (Please correct me if there are errors in my workings, since it is a very long time since I attended a trig lesson).

After drawing this diagram, it became clear that the light source should be as small as possible in the two dimensions here, but could be a strip of light in the third dimension. Although AF (the lamp height) can be reduced by moving the camera out (increasing DC), this can produce problems elsewhere.


Re: Footsie Bookscanner

Posted: 10 Jun 2015, 10:47
by duerig
Great job working out the math. It is good to see these things worked out and not just done empirically. IIRC, the 'ideal' angle for the platen was found to be 105 degrees. Which allowed the lights to come a bit closer to the platen. I think that a lot of people settled on 100 as a 'close enough round number'.

For more discussion of lighting issues, you can take a look at what Daniel wrote here:

In particular, here are my thoughts on things you need to account for moving forward in your design:

Even illumination: The best way to do this is to take your point illumination sources and diffuse or spread the light. This can be with a translucent diffusor. The disadvantage of that is that you lose half the light, so your light source needs to be twice as strong. And heat might be a bigger issue here. But I think the material for diffusors is relatively easy to get.

Another alternative is lenticular sheets. Light dispersal is much more efficient with these, but they only diffuse light in one direction. The Archivist uses two lenticular sheets stacked atop one another in opposite directions. It is harder to get, but I have a bunch of it. And if you want, I can cut some into an arbitrary shape for you. Let me know.

Constraining Light: You will want baffles of some sort to prevent light from hitting parts of the scanner you don't want it to. If it does, it can have unpredictable effects as secondary reflections might cause artifacts in your scan.

Heat: Even LEDs can get hot, especially if there are a lot of them together. Make sure you don't stick them next to flammable things. Fire is bad. :)

CRI: If you are going to be scanning art books or other books where you want a high color fidelity, then pick LED bulbs with as high of a CRI as possible. From what I understand, a high CRI means that the bulb generates light evenly across the wavelengths band it is in.

Color temperature: This doesn't matter so much as long as you have the right settings in your camera to correct for it. 2700K == Tungsten

AC vs DC: You will want your light to run off of DC power rather than direction on AC. Apparently, AC lights can cause artifacts at some camera speeds because they flicker at the AC frequency.

I look forward to seeing what you come up with.


Re: Footsie Bookscanner

Posted: 10 Jun 2015, 16:53
by dpc
IIRC, the 'ideal' angle for the platen was found to be 105 degrees.
Depends on what you consider "ideal". It's a trade-off between being able to get the cameras closer to the platen at the expense of needing to increase the platen angle (so that you don't see the reflection from the page image off of the adjacent platen) which requires that the light be located farther away from the platen "v".

I went through this trig exercise over a year ago and created a spreadsheet where you enter light width, max page size, and platen angle as inputs and it spits out the minimum distances from the platen for the light and cameras necessary to ensure there could be no reflections from either the light source or from the adjacent platen. I wanted to tweak the platen angle to minimize both of these distances to make the scanner as compact as possible and not have my cameras or light positioned farther away than necessary.

As an example, a scanner that supports a 12" page width, has a light width of 5", and utilizing a 105 deg. platen angle would allow the cameras to be positioned at a minimum distance of 22.392" from the platen surface and not see the reflection from the adjacent platen. Locating the camera in this position and zooming the camera to where it would capture the entire 12" page width would require that the light source be located 36.325" above the platen "v" so that it's reflection isn't seen by the camera. That light height might be ideal for some scanner designs, while others may require the light source be closer, with the tradeoff being that the cameras would need to be located farther away from the platens (so that the viewing angle can be reduced by zooming in).

A few lessons learned from playing with these numbers:
1. There's no need to resort to expensive anti-glare glass to eliminate light reflections if the scanner is designed properly.
2. Using multiple narrow lights is better than a big wide one. This allows your light source to be located closer to the page.
3. Ask yourself if you really, really need to be able to scan wide books. 99% of the books in my library that I'm scanning have pages that are less than 8" wide. Changing my scanner requirements to support scanning an 8.5" page width allows reduce camera and light distances which makes the scanner more compact and lighter (good!). I'm now using a copy stand for scanning the large coffee table books I have instead of complicating my scanner.

Re: Footsie Bookscanner

Posted: 11 Jun 2015, 17:52
by dtic
duerig wrote:CRI: If you are going to be scanning art books or other books where you want a high color fidelity, then pick LED bulbs with as high of a CRI as possible. From what I understand, a high CRI means that the bulb generates light evenly across the wavelengths band it is in.
Let me hijack with a question here: do you know if there is a very big difference between LEDs with CRI 95 (the LEDs on the Archivist info page) and LEDs at around CRI 90? I'm asking since all LEDs with printed specs that I've actually seen that have had a CRI around 80. But Now I noticed that IKEA sells a few with a specced "color rendering" of 90, .

Re: Footsie Bookscanner

Posted: 11 Jun 2015, 19:37
by duerig
Let me preface this with an admission that I don't know very much about CRI except that higher is 'better'. I have read about it on Wikipedia and what Daniel wrote.

I think that the difference between 90 and 95 will likely be small. But I am very interested in trying to test this empirically. I have a couple of brands of 95 CRI bulbs right now. In the near future, I plan to get some identical bulbs with various CRI ratings and test them against each other and a known test image. I will print out the test image, then scan it with the different bulbs and put up the results here. This should help us see what the actual trade-off is for the more expensive high-CRI bulbs.

Until I can do more testing, I will stick with 95 CRI bulbs. But I want to get a handle on the actual benefit they provide.


Re: Footsie Bookscanner

Posted: 11 Jun 2015, 20:02
by dpc
I plan to get some identical bulbs with various CRI ratings and test them against each other and a known test image.
Add a shot outside in the daylight too. That should get 100 CRI for comparison purposes.

You might want to get a color card rather than using something coming off of your printer. Search in the archives for work on color calibration done by Misty. While not specific to CRI, you might find her investigations interesting.

Re: Footsie Bookscanner

Posted: 12 Jun 2015, 04:08
by wmalcolmk
I've just been reading a US Department of Energy report on CRI.. ... _index.pdf
I'm not too sure that CRI is too important, since it is used to assess how things look by eye when a lamp illuminates an object.
When a lamp is used in a scanner there are many more factors that could influence the end result. For example in the comparison test that duerig proposes...
..I will print out the test image.. - the printer will alter the colour balance of the image on the page
.. then scan it with the different bulbs..- the camera in your scanner will change the colour balance again
.. and put up the results here...-my computer screen will change the colour balance yet again
Since it is possible to alter colour balance using most common image manipulation programs, perhaps all we can manage to do in a diy scanner, is to ensure that a pleasing colour image can be produced at the end of the process.
The photo process labs return some very dubious colour prints whenever I send them my snaps.

Re: Footsie Bookscanner

Posted: 12 Jun 2015, 07:41
by Marcus Eastty
Quick question about the book support: In the instructables tutorial is looks like a simple hinge connecting two pieces of plywood. I was thinking that wouldn't allow much room for thicker books like textbooks, bibles, etc. Then in this thread I saw what looks kind of like a hammock between the two pieces of plywood. I'm guessing this allows for thicker books to be scanned? How was that put together?


Re: Footsie Bookscanner

Posted: 12 Jun 2015, 09:02
by duerig
Retaining color fidelity is a goal worth pursuing even if it can never be fully achieved. By quantifying what benefit we actually get with high CRI lighting, we can help people decide whether they should pursue it for their own setups. I like the color card idea because that removes a lot of potential confounding factors mentioned here (print quality, screen quality). When I get together the supplies and can experiment, then I will post my results in a new thread. Probably in the R&D section.

I should note that even if you care about color fidelity, there are a lot of contexts in which it doesn't matter that much. If you are scanning black and white printed material, then as long as your white balance setting on the camera is correctly callibrated to your lights then that is all that matters.


Re: Footsie Bookscanner

Posted: 13 Jun 2015, 04:53
by wmalcolmk
Marcus asked how to make a book cradle with a centre section that conforms to the shape of the book spine. The photo below shows my cradle resting upside down on a flat surface. It was made from two plywood sheets, size 250x170x5mm, and two mdf strips 170mmx5x3mm. Hold them together using sticky backed canvas tape, ensuring that there are gaps between the strips. Turn the assembly over and put more tape on the other side, so that it acts as a hinge. The outrigger arms ensure that the rope is guided onto the pulleys. I'm still developing this aspect of the scanner.
Cradle.JPG (21.56 KiB) Viewed 5558 times

I hope this helps to explain it - Malcolm