Optics question: I need parallel illumination for a Contact Image Sensor (CIS)

[rich333]

I am experimenting with a CIS from a document scanner (basically a row of photo sensors).

I want to detect the position of an object as it passes the scanner bar. Initially in one dimension - x along the CIS.

If the object is always a set distance from the light bar, I can provide illumination with a line laser or row of leds and the object will cast a shadow as it passes through the light. The position of the object along the light bar can be detected.

However if the object can be close to the light source or close to the CIS, it's shadow will be cast in different places.

In effect I need a row of light beams such that the object shadow is cast in the correct x direction, regardless of it's y position. Is that even possible?

If I add a second CIS and then find the objects Y position, I guess I can calculate it's X and Y position, even with light sources that are effectively single point?

The ultimate goal is to find the X,Y position of the object as it passes the two CIS scanner bars.

Any ideas most welcome!

(Simple sketch attached)

 

[Drakkith]

In effect I need a row of light beams such that the object shadow is cast in the correct x direction, regardless of it's y position. Is that even possible?

Yes, but it's not easy unless your setup is fairly small. Making a collimated beam of light requires a lens or mirror at least as large as the beam so you'll have to have lenses approximately the diameter of your sensor.

If I add a second CIS and then find the objects Y position, I guess I can calculate it's X and Y position, even with light sources that are effectively single point?

That seems like a good option, as it would avoid adding large lenses or other complications to your setup. Just two light sources and two sensors.

 

[rich333]

I'll have a think on the situation with two sensors at right angles to each other...

I think what I was looking for was some sort of optical device that would create a series of parallel laser beams from one laser source, that doesn't cost a fortune! A line laser certainly doesn't do this...

The sensor has approximately 300 holes along it's length (9 inches), the resolution is actually closer to 3000 pixels, so I assume each hole has a lens in which covers 10 photodiodes.

 

[Drakkith]

I think what I was looking for was some sort of optical device that would create a series of parallel laser beams from one laser source, that doesn't cost a fortune! A line laser certainly doesn't do this...

The simplest setup would be a pinhole light source placed at the focal point of a lens. This would create a collimated beam of light, however, you would need a lens at least as wide as the beam diameter, and a 9-inch diameter lens is probably not what you want. There are various ways to improve this basic method, but none remove the need for a large lens.

I'm sure there are a thousand ways to figure out the position of your object using your linear sensor array, but I'm afraid I can't be of much help as I'm only a student at the moment.

 

[rich333]

The simplest setup would be a pinhole light source placed at the focal point of a lens. This would create a collimated beam of light, however, you would need a lens at least as wide as the beam diameter, and a 9-inch diameter lens is probably not what you want. There are various ways to improve this basic method, but none remove the need for a large lens.

I'm sure there are a thousand ways to figure out the position of your object using your linear sensor array, but I'm afraid I can't be of much help as I'm only a student at the moment.

Thanks for your prompt reply to my initial post! I'm going off to have a think and doodle!

 

[berkeman]

I am experimenting with a CIS from a document scanner (basically a row of photo sensors).

I want to detect the position of an object as it passes the scanner bar. Initially in one dimension - x along the CIS.

If the object is always a set distance from the light bar, I can provide illumination with a line laser or row of leds and the object will cast a shadow as it passes through the light. The position of the object along the light bar can be detected.

However if the object can be close to the light source or close to the CIS, it's shadow will be cast in different places.

In effect I need a row of light beams such that the object shadow is cast in the correct x direction, regardless of it's y position. Is that even possible?

If I add a second CIS and then find the objects Y position, I guess I can calculate it's X and Y position, even with light sources that are effectively single point?

The ultimate goal is to find the X,Y position of the object as it passes the two CIS scanner bars.

Any ideas most welcome!View attachment 219994

(Simple sketch attached)

Welcome to the PF.

It seems like you are using the CIS for something that it is not designed for, and is not optimum for in this application. Is this just a fun project for you to try to see how well you can get a DIY experiment to work? Or is this for an actual product design?

Assuming it's just for a personal project to challenge yourself and learn, can you post the datasheet or specs for the CIS modules you are wanting to use? What is their dynamic range for light sensing at each pixel? Do the CIS modules include the multiplexed digitization of the light levels of each pixel, or do you need to do the shifting of analog charges and do the digitizing in your external circuit?

What accuracy do you need the (x,y) position of the object accurate to? What are the size ranges of the obscuring object compared to the pixel spacing of the CIS modules? What are the ranges of (x,y) values with respect to the position of the CIS module that you want to detect for that object?

Depending on the answers to the above questions, it should be possible to do the (x,y) position sensing with one CIS module and one laser line scanner. Can you take a guess at what I have in mind?

 

[rich333]

Welcome to the PF.

It seems like you are using the CIS for something that it is not designed for, and is not optimum for in this application. Is this just a fun project for you to try to see how well you can get a DIY experiment to work? Or is this for an actual product design?

Assuming it's just for a personal project to challenge yourself and learn, can you post the datasheet or specs for the CIS modules you are wanting to use? What is their dynamic range for light sensing at each pixel? Do the CIS modules include the multiplexed digitization of the light levels of each pixel, or do you need to do the shifting of analog charges and do the digitizing in your external circuit?

What accuracy do you need the (x,y) position of the object accurate to? What are the size ranges of the obscuring object compared to the pixel spacing of the CIS modules? What are the ranges of (x,y) values with respect to the position of the CIS module that you want to detect for that object?

Depending on the answers to the above questions, it should be possible to do the (x,y) position sensing with one CIS module and one laser line scanner. Can you take a guess at what I have in mind?

Ha! Yes a CIS is designed specifically for flatbed scanners and scanning things that are flat and close to the scanner bar. (i.e. a document).

I am using one as a DIY experiment mainly because I was able to buy a complete working printer/scanner combo for £1...

I have no idea what the spec of the CIS is, all I can do is scope it while it's working. Its from an HP scanner.

With no illumination leds on, you can illuminate from directly overhead (or sideways as per my sketch) with a line laser. On the scope you see a full set of set pixels. A shadow cast by a screwdriver shaft is clearly shown on the scope trace.

The CIS works in monochrome. It is just illuminated by different colours with the build in led light bar, which I'm not using. Each pixel is clocked out on one wire serially. It's not a digital output though. The pixel level varies according to illumination level. I will read the CIS either with an AtoD or more likely just run it through a comparitor and get a digital stream since I don't need to know intensity.

A low power laser line is enough to fully 'set' the pixels, i.e. max level. (which is actually a low voltage level output).

 

[berkeman]

With no illumination leds on, you can illuminate from directly overhead (or sideways as per my sketch) with a line laser. On the scope you see a full set of set pixels. A shadow cast by a screwdriver shaft is clearly shown on the scope trace.

Each pixel is clocked out on one wire serially. It's not a digital output though. The pixel level varies according to illumination level. I will read the CIS either with an AtoD or more likely just run it through a comparitor and get a digital stream since I don't need to know intensity.

Hint, hint...

 

[berkeman]

Also, you didn't say if the size of the object that you want to detect is variable, and by how much...

 

[rich333]

Also, you didn't say if the size of the object that you want to detect is variable, and by how much...

It's 2am here, so I am just about awake!

The object is always the same size. 5mm sphere.

 

[berkeman]

It's 2am here, so I am just about awake!

The object is always the same size. 5mm sphere.

Great! Then when you wake up tomorrow morning, think about my hints and how you can use them to make this work. Nighty-night!

 

[rich333]

Great! Then when you wake up tomorrow morning, think about my hints and how you can use them to make this work. Nighty-night!

My only thought was that if the object is closer to the CIS, it's shadow will be narrower. The shadow widens the closer it is to the light source.

The main issue with using that is that the shadow cast by a sphere passing will not be constant. I'd have to be sure to sample a few times as it passes to find width at the widest part.

 

[berkeman]

Yes, the width of the shadow (measured in an analog fashion) was what I had in mind. But overnight I realized that this will work a lot better if you use two line sources (time multiplexed) with the one CIS. Picture a square area with the CIS as the top line, and the two laser line sources at the bottom left and right corners. That will give you better (x,y) position information than you can get with just one source.

 

[rich333]

Yes, the width of the shadow (measured in an analog fashion) was what I had in mind. But overnight I realized that this will work a lot better if you use two line sources (time multiplexed) with the one CIS. Picture a square area with the CIS as the top line, and the two laser line sources at the bottom left and right corners. That will give you better (x,y) position information than you can get with just one source.

Interesting idea! Line lasers are very cheap.

The problem will be duration. I haven't said what speed the object will be traveling at, but suffice to say I think I will only get time to get one scan of the CIS as it passes. (Sorry, should have mentioned that)

I have seen that you can buy linear collimator lenses for LED strip lights, not sure if this would provide what I want...

 

[Hendrik Boom]

Interesting idea! Line lasers are very cheap.

The problem will be duration. I haven't said what speed the object will be traveling at, but suffice to say I think I will only get time to get one scan of the CIS as it passes. (Sorry, should have mentioned that)

I have seen that you can buy linear collimator lenses for LED strip lights, not sure if this would provide what I want...

you might use two scanner bars for x and y coordinates. Or one with mirrors.

 

[Ibix]

But overnight I realized that this will work a lot better if you use two line sources (time multiplexed) with the one CIS.

Time-multiplexed I take it means pulsing the two sources out of phase? Why not use two sources dimmed down so that one source produces 50% activation? Then you can identify fully activated, half activated, or not activated and do a bit of geometry.

 

[berkeman]

Actually, @Ibix brings up an interesting point...

I am experimenting with a CIS from a document scanner (basically a row of photo sensors).

@rich333 -- Is this a 3-color CIS? It it's from a document scanner, it likely is. If so, can you find red and green laser line scanners?