flatbed scanner spectrometer idea

[rgbphil]

I just had the idea that the photodiode array in a flatbed scanner could be used to make a spectrometer. The light would go through a prism or diffraction grating onto the array, once calibrated the intensity along the array should be the spectrum.

Has this been done before, don't want to reinvent the wheel?
Phil

 

[ks_physicist]

That is an excellent idea. And I have a spare scanner head sitting in my junk box.

Jim

 

[rgbphil]

I think the scanner part may be OK....but I did some testing with an old prism from a pair of binoculars......not nearly enough dispersion. What sort of dispersive element do they use in commercial spectrometers?? Diffraction gratings? Purpose built prisms??

 

[ks_physicist]

I have seen both diffraction gratings and prisms in commercial spectrometers. However, the vast majority of them I've seen (lab environment) are grating based.

Jim

 

[winny]

Sounds like a very fun idea and I don't want to stop you but keep in mind that a flatbed scanner uses an florescent tube as light source which haven't got a nice spectrum. Therefore, it has been calibrated in order to compensate for the FL tubes low CRI. It's fully doable, but you need to figure out how to recalibrate your readings.

Good luck and keep us posted! :wave:

 

[rgbphil]

Well....it's been a while but I finally had a look at implementing this idea....it looks very feasible.

I took apart an old Canon N670U flatbed scanner with 'Lide' technology, which basically means it uses a Contact Image Sensor (CIS) type scanhead. This sort of scanner has a bunch of smaller sensor chips inside all lined up in a row. Instead of complex optics, there is a simple linear waveguide, which transmits light from the sensor top, to individual photodiodes.

The advantage of this is that the sensor by it's very nature has to be very close to the thing being scanned (thus the 'Contact' part of the CIS name), and an RGB LED with a suitable spreader can supply the light source. Scanners made from CIS modules can therefore be USB powered....which the N670U is.

Because the thing is so simple, I found at one end a USB connector and on the moving (scanning part) was a stepper motor assembly which climbs up and down a cable, the CIS sensor and the electronics.

The electronics has in it a LM9833, googling that datasheet I found it was a special purpose USB appliance, which can connect to a variety of CIS/CCD linear sensors via a simple interface. To summerise, you send a reset on one pin, clock another and on each clock the analog value of the light falling on that sensor in the array appears at another pin. Additional pins control the LED light source. All very encouraging, as this is the sort of thing a micro can do....it doesn't need to scan a whole page to work out a spectrum, and we don't need to worry too much about throughput either.

The CIS sensor has 12 pins, unfortunately I couldn't beep back into the LM9833 to find out which pin goes to which signal, because of some buffer circuitry which is a bit tedious to sort out.
The CIS sensor has a Canon part sticker on it with these markings:
Under the Canon logo, the numbers 82408
Under the barcode, the numbers 58587C75F
an additional number by itself FH7-7576
If there is anyone who knows (or can find out) what the manufacturer/MFPartNumber of the sensor in the scanner is, or better still a datasheet, it will not be difficult to get started on this. (you could also help save me the effort of beeping out the buffer circuits).

Phil

 

[rgbphil]

Another post...this one on some ideas for the actual spectrometer.

It occured to me that having a dispersing element that will fit the entire span (about 200mm) of the sensor will necessarily mean less sensitivity because the light is spread out so much. Also a prism or diffraction grating with that sort of dispersal angle will not be very efficient and probably expensive. So a diffraction grating that puts out mulitples of the spectrum looks more likely and affordable.

A diffraction grating will project onto a line a repeating set of spectra, at high angles the spectra are spread out further but fainter. At low angles the spectra are bright but narrow.

So a simple spectrum would be read by picking a length of sensor module, and outputting the light reading along that length. The device could be calibrated using a red and a green laser, or some sort of other known light source (or reading from a lab spectrometer....maybe even an LED datasheet).

Running the required software in my head, I would have the choice of a low resolution but sensitive spectrum at the shallow angle multiples, or a high resolution low sensitivity spectrum at the higher angle multiples....cool.
Then the question of whether I could integrate the low resolution spectrum with the high res one, thereby utilising the entire sensor width and reaching a compromise on resolution and sensitivity (some error added).

What does anyone think of adding the power of the low angle multiples to the high res, high angle multiples, using some sort of interpolation?

Phil

 

[rgbphil]

I've done some beeping and backtracing on the dismembered N670U scanner bits and I think I've got a pinout:
Pin 1 - scanner element analog output, 0-5V value of the light at the sensor
Pin 2 - GND
Pin 3 - GND
Pin 4 - V+, I measured 4.78V
Pin 5 - VBias, I measured 1.385V
Pin 6 - scan line reset signal, positive pulse
Pin 7 - scan line clock signal
Pin 8 - Anode to LEDs in CIS sensor
Pin 9 - Blue cathode
Pin 10 - Green cathode
Pin 11 - Red cathode
Pin 12 - GND

These pins are a little different from any CIS sensors I've been able to get a datasheet of, but were measured with the bits connected to the PC. Fortunately for some reason the controller sends reset/clock signals out even when the scanner isn't scanning, so using a CRO I was able to see the actual line scan analog signal. I could move light sources in front of, or occlude parts of the scanner and see a corresponding signal....very cool.

Either way, the electrical signals are very simple, so next step is a cct to talk to it. I was suprised at the 5V operation of the thing, however if that's what canon was feeding the CIS sensor, then that just makes my job a little easier.

Phil

 

[rgbphil]

Hi,
Things are progressing nicely with the cct. I have a 18F4550 QFP set to generate the CIS signals and I'll be using a nice little 128x128 colour LCD module I had lying around (as you do) as the primary interface. USB will allow downloading of spectra to a PC.
Whiles there is a bit of work to do still on the cct, I'd like to start on the optics.
For others working on something similiar I might suggest use of the CIS element in a portable business card reader...a little less bulky than the long CIS sensor from the N670U scanner.
Any ideas on a suitable way to gather, colliminate and make a slit for the incoming light? Would an optical fiber (starting with a cheap plastic one, maybe a better one later) that has a lens focussed on it, then poked into the diffraction grating be a better approach than some more elaborate optical setup?
Also, does anyone know of a place in Sydney where I can get a cheapish diffraction grating....if things work out, I can replace it with a better quality one later.
Phil