Designing a special wide angle lens

[vinniewryan]

Hello,

I have a specific application where I need a custom lens designed to focus an image onto a very small cmos image sensor. So far, I've figured out the general angles of the lense but I need it to be a perfect design before money is put into prototyping it. Here's the specs:

The image sensor is 19X19 pixels, and measures 0.9mm X 0.9mm
The distance from the cmos to the lens can be no further than 5mm
the distance from the lens to the object (focal length) must be no further than 10mm
the FOV required is 10mm X 10mm (roughly 120 degrees)

The lens I designed has a steep concave spherical indent on the side pointed towards the object, and the side facing the lens is convex with a slight spherical curve. I built a scaled up version and it works perfectly, I can see about a 1ft X 1ft area from about 4 inches away, I also tested the angles with a laser pointer and it all looks fine and dandy, but when it comes to actually focusing an image, I need to be sure it's going to focus correctly and not come in blurry.

Are there any tips or things I can do to figure this out?

Thanks!

 

[vinniewryan]

Basically this is what I have, it seems to direct light correctly but I need someone to confirm whether this is possible while maintaining a sharp image.

http://imageshack.us/m/836/1743/lensdesign.png

The green lines indicate the light focused from the object to the sensor. Does this look okay so far?

 

[Andy Resnick]

That looks... odd. How did you arrive at this design? And, I am assuming this is all in the visible band (you don't provide a chromatic aberration spec.)

Your specifications don't look too strange; the field of view is rather large but I suspect if you use a lens from a camera phone you will be close.

 

[vinniewryan]

Thanks for the reply. I came up with this design by using an optically clear ball filled with water. I used clear plastic blister pack to form the ball in different shapes and using a laser pointer I was able to see and map out the angle which the laser beam exited the ball.

After the laser pointer test, I picked up the lense and looked through it. That's how I found that I could see a 1 foot by 1 foot area from only 4 inches away. I'm just an electronics engineer, but I need to make a lense that can focus a wide view onto a 1mm by 1mm sensor. That's why a cellphone camera lens won't work, the sensors in those are a few milimeters across. I suppose I could copy a cell phone lens and scale it down if that's what it comes to.

image fixed*

 

[DrGreg]

It occurs to me that if your sensor is only 19 × 19 pixels, that's such a low resolution that your lens design needn't be all that accurate. (Did you really mean 19? Just to put that into perspective, my simple avatar to the left of this message is a massive 65 × 65 pixels.)

 

[vinniewryan]

I certainly did mean 19x19, it's being used to track the position on an X,Y grid of a seismic mass. The mass will have 3 points each consuming about a 2x2 area of pixels, these points will move around and we're using the 19x19 grid to track the position of each point. Mind you, each pixel gives us a value from 0 to 64 so if a point moves across 3 pixels, our position coordinate has advanced by 194 values. It needs to be a sharp image because if it's blurry our position tracking will not be accurate.

We're also using these sensors to find printed watermarks for orientation purposes. I have lenses which give me a crisp image, but those lenses zoom into a .2 mm area, where as I now need to see a 10mm area.

 

[AlephZero]

I wonder if you are over-complicating this. Your design looks like a concave lens in front of a convex, except the gap between the two filled with glass.

Why is that different and/or better than using two "standard" lens elements that you could buy ready made and mount one in front of the other?

 

[sophiecentaur]

I certainly did mean 19x19, it's being used to track the position on an X,Y grid of a seismic mass. The mass will have 3 points each consuming about a 2x2 area of pixels, these points will move around and we're using the 19x19 grid to track the position of each point. Mind you, each pixel gives us a value from 0 to 64 so if a point moves across 3 pixels, our position coordinate has advanced by 194 values. It needs to be a sharp image because if it's blurry our position tracking will not be accurate.

We're also using these sensors to find printed watermarks for orientation purposes. I have lenses which give me a crisp image, but those lenses zoom into a .2 mm area, where as I now need to see a 10mm area.

You would need to specify your required "accuracy". Your initial spec seemed to suggest you need a Karl Zeiss lens but now you are talking ultra low res. Where does your implied high acuity come into it?

 

[Andy Resnick]

I thought there would be a trivial solution, like one of these:

http://www.edmundoptics.com/onlinecatalog/displayproduct.cfm?productid=2196

The 1.68 mm focal length lens meets the back focal length spec and the field of view spec, but the working distance is too long- 400mm instead of 10mm. I think your problem is going to be the short working distance, rather than the field of view or back focal length.

Maybe you can get the 8.5mm lens to work:

http://www.edmundoptics.com/onlinecatalog/displayproduct.cfm?productID=2873

 

[vinniewryan]

You would need to specify your required "accuracy". Your initial spec seemed to suggest you need a Karl Zeiss lens but now you are talking ultra low res. Where does your implied high acuity come into it?

The image must be as sharp as possible because we're relying on tracking bright (white) points against a black background. If there is a blur around the points we're tracking, then we will have many shades of gray between the true edges of the points and the black background, and the background will blur into the white foreground and cause our brightest point to be darker, and the points will consume an area larger than 2 x 2 pixels because a blur will expand and scatter the light across the surrounding pixels.

Let me explain with a visual:

http://imageshack.us/m/560/1269/blur.png

Now, the image on the left represents a sharp image projected onto the 19x19 pixel sensor. The 3 points consume only 4 pixels each, and they are one solid bright color against the background.

The image on the right represents a blurred image projected onto the sensor. The 3 points have blurred edges which causes them to consume more than 4 pixels and if they were to move closer to each other, the blurred edges of any two points may overlap and cause them to be seen as one large point by the software.

We need these points to be sharp, in focus, and easy to discriminate against each other.

Also I'm not trying to find a lens that I can order, I'm certain nothing exists for my specific application except maybe in some scientific imaging equipment. I'm planning on prototyping this lens out of a plastic made specifically for lens creation, casted in a highly polished steel mold.

Does this even seem possible? I know there are ways to calculate this stuff but I'm just not confident enough to do that myself.

 

[sophiecentaur]

Those diagrams are ok as far as they go but there are many measurement systems which work fine without the luxury of pixel image resolution. If your signal to noise ratio is good you should be able to determine the centre of the blobs with appropriate spatial filtering. That sort of software would be a lot easier to implement than making your own lens to a high spec. With such a small amount of data you would have no problem implementing an appropriate algorithm.
Don't forget that you may have barrel / pincushion / other geometrical distortion in such a wide angle lens which could introduce gross inaccuracy in establishing the position if you don't take it into account. It's no surprise that you have to pay hundreds of quids for a good lens - the lens designers earn their money.
Astronomers are always working with blobby images and getting good information out of them.

 

[vinniewryan]

If your signal to noise ratio is good you should be able to determine the centre of the blobs with appropriate spatial filtering. That sort of software would be a lot easier to implement than making your own lens to a high spec. With such a small amount of data you would have no problem implementing an appropriate algorithm.

True, but it would be preferred to start out with a sharp image and save the embedded micro processors from the extra work and latency. I'm working within very narrow limits so everything simply has to be as efficient as possible. If the image is distorted I can work with that, if it's slightly blurry I can work with that, but we're relying on obtaining as much information as possible from these pixels and less blur = "more" precise data. If my image does come in blurry, is there a simple focusing lens I can use in unison with my current lens to help focus the light further? I'm learning a lot about light and lenses in this project, I'm sure this can be done.

 

[vinniewryan]

I wonder if you are over-complicating this. Your design looks like a concave lens in front of a convex, except the gap between the two filled with glass.

Why is that different and/or better than using two "standard" lens elements that you could buy ready made and mount one in front of the other?

the cost in china of mounting one lens as opposed to two. It's a plastic lens, no glass here. and believe me, If I could find a 1mm x 1mm concave half sphere lens, I would use that. The closest I've found is used in optical mice but they don't have the focal distance I need. (6 inches as opposed to my required 5-7mm)

 

[sophiecentaur]

True, but it would be preferred to start out with a sharp image and save the embedded micro processors from the extra work and latency. I'm working within very narrow limits so everything simply has to be as efficient as possible. If the image is distorted I can work with that, if it's slightly blurry I can work with that, but we're relying on obtaining as much information as possible from these pixels and less blur = "more" precise data. If my image does come in blurry, is there a simple focusing lens I can use in unison with my current lens to help focus the light further? I'm learning a lot about light and lenses in this project, I'm sure this can be done.

If that's the way you want to go, then ok. But high quality optics is a lot lot harder than a bit of software- innit?

Another lens can only help to modify the focal length of a lens. You can't improve on aberrations by just 'focusing'. The lens should have the right focal length in the first place.

 

[vinniewryan]

That's certainly proving to be true, but software can only do so much with the information received by the sensor that it relies on, and in embedded processes where time is critical it's always more efficient when the hardware does most of the work for you.

 

[sophiecentaur]

If you can make the hardware.

 

[vinniewryan]

If you can make the hardware.

which is exactly the point of this thread :]

I suppose I could always try using two wide-angle lenses one in front of the other to increase my FOV, though I still see one lens needing to be scaled down to give me the fine 1x1mm point of focus needed...

 

[sophiecentaur]

which is exactly the point of this thread :]

I suppose I could always try using two wide-angle lenses one in front of the other to increase my FOV, though I still see one lens needing to be scaled down to give me the fine 1x1mm point of focus needed...

If you think that just sticking one lens in front of another will do what you want then I strongly recommend that you either do a course in optics or learn about two dimensional filtering algorithms - or both.

You are asking if this is do-able and the only responses you have had seem to imply that it isn't. Yet you still want to do it your way. Why did you ask the question in the first place?

 

[sophiecentaur]

Have you heard of vignetting, btw?

 

[vinniewryan]

I asked because this is the kind of thing that can be figured out with math. Math I don't know nor have the time to learn, if it comes down to hiring an optical engineer to design this thing then that's what I'll do, but the only way to find out is by asking. The only responses I've been getting are 'well you could do it this way' but that doesn't help me if it means my focal distance goes from 7mm to 6". I'm working within limits, which is why I provided specs in the initial post. It would be much easier for me to take in if someone knew the boundaries and could tell me "no, it can't be done because..." instead of beating around the bush with ideas that clearly go outside of my boundaries. The point is that this has to be made to work in a very small area.

 

[sophiecentaur]

I wonder whether fibre optics could help with your restricted space. That's a pretty well known technology for looking in small spaces.

 

[Andy Resnick]

I wonder whether fibre optics could help with your restricted space. That's a pretty well known technology for looking in small spaces.

Interesting approach- I like it!. Either a borescope or a tapered fiber bundle may work ok.

 

[vinniewryan]

I don't think fiber optics can be moulded in one piece, and the assembly required for multiple strands would shoot the cost way out of range. Also unless fiber optics can taper from large to small, that method would kill the gradual increase as a point's edge moves over any single pixel. I like the direction though, maybe there is some way I can use the concept of piping the light to each pixel of the sensor.