Hello all,


I have a problem in olde timey photography that I want to address electronically. I use a large format view camera which allows considerable freedom of movement between the lens and the film. This allows distortion elimination and provides increases in depth-of-field along with providing a lot more ways that the photographer can mess up a shot.


Obtaining sharp focus on various points of the image can be a very time consuming and frustrating. As you shift the various elements of the camera, the focus changes all over the place. What I'd like to do is isolate the autofocus circuit from a digital camera and implement it such that the sensor can be mounted on a puck that I can place on the ground glass and the peak contrast indicated on a meter of some sort. This would allow me to quickly check the sharpness at various point on the image after or while making adjustments.


Has anyone out there worked with autofocus systems or can tell me where to look? I'd consider cannibalizing a digital camera if I knew what to cook and eat.


Thanks,


Captain John

 

There are a few flavors of auto focus. One involves prisms and phase shifts. Another involves contrast detection in the digitized image.

I don't think either will translate easily into your setup. However I'm no expert in photography, so hopefully I'll be proven wrong on this one!

https://en.m.wikipedia.org/wiki/Autofocus

https://www.reikanfocal.com/about-autofocus.html

 

Hello all,


I have a problem in olde timey photography that I want to address electronically. I use a large format view camera which allows considerable freedom of movement between the lens and the film. This allows distortion elimination and provides increases in depth-of-field along with providing a lot more ways that the photographer can mess up a shot.


Obtaining sharp focus on various points of the image can be a very time consuming and frustrating. As you shift the various elements of the camera, the focus changes all over the place. What I'd like to do is isolate the autofocus circuit from a digital camera and implement it such that the sensor can be mounted on a puck that I can place on the ground glass and the peak contrast indicated on a meter of some sort. This would allow me to quickly check the sharpness at various point on the image after or while making adjustments.


Has anyone out there worked with autofocus systems or can tell me where to look? I'd consider cannibalizing a digital camera if I knew what to cook and eat.


Thanks,


Captain John

How are you going to know exactly where the sensor is pointing? A slight error in angle on my iPad means it focused on the wall behind my subject instead of my subject. How will you make sure the image on the back hatch of the camera (film) is receiving a focused image?

You may be better off using a focusing mirror/prism system from a predigital camera. Like the Nikon 8008.

 

How are you going to know exactly where the sensor is pointing? A slight error in angle on my iPad means it focused on the wall behind my subject instead of my subject. How will you make sure the image on the back hatch of the camera (film) is receiving a focused image?

You may be better off using a focusing mirror/prism system from a predigital camera. Like the Nikon 8008.

The view camera projects the image onto a ground glass plate which is at the film plane. I can place the sensor onto the ground glass to take the reading. I normally use a 10X loupe to do the critical focusing so I had in mind a small puck that would hold the sensor at a consistent distance from the glass and parallel to it.

 

There are a few flavors of auto focus. One involves prisms and phase shifts. Another involves contrast detection in the digitized image.

I don't think either will translate easily into your setup. However I'm no expert in photography, so hopefully I'll be proven wrong on this one!

https://en.m.wikipedia.org/wiki/Autofocus

https://www.reikanfocal.com/about-autofocus.html

I saw the one on wiki. It looks like contrast would be my best bet since I'm focusing off on the ground glass at the film plane and the output from the contrast circuit would give a maximum value that I could see on a meter. It works in a similar fashion to what I do when I focus. Maximum contrast normally means sharp focus.
What I hope to find is a circuit diagram which will show me where I can snag the contrast value and display it. After that, it's a simple matter of programming with a soldering iron.

 

I saw the one on wiki. It looks like contrast would be my best bet since I'm focusing off on the ground glass at the film plane and the output from the contrast circuit would give a maximum value that I could see on a meter. It works in a similar fashion to what I do when I focus. Maximum contrast normally means sharp focus.
What I hope to find is a circuit diagram which will show me where I can snag the contrast value and display it. After that, it's a simple matter of programming with a soldering iron.

I think I was picturing your setup wrong on the first read through. I agree that the contrast method, using a sensor pressed against the glass, exactly where the film would be, has a fair chance of success.

I'm curious to do some more reading on this. Seems to me like you're going to need a ccd, a microcontroller, and some code. I found a few articles on code for analyzing contrast. So far, the few ICs I've found all had motor/actuator outputs so that they could control focus, sweeping back and forth until an ideal position was found. They didn't have an output corresponding to degree of focus or level of contrast, which is more like what I think you'd need.

This was an interesting read:
https://cardinalpeak.com/blog/detecting-well-focused-images/

 

The view camera projects the image onto a ground glass plate which is at the film plane. I can place the sensor onto the ground glass to take the reading. I normally use a 10X loupe to do the critical focusing so I had in mind a small puck that would hold the sensor at a consistent distance from the glass and parallel to it.

I wonder if it might be better to put a sensor in each corner of viewing plate so that you don't have to move one sensor arround.

Also, I wonder if some sensor schemes might work better on the inside of the viewing plate.

 

I think I was picturing your setup wrong on the first read through. I agree that the contrast method, using a sensor pressed against the glass, exactly where the film would be, has a fair chance of success.

I'm curious to do some more reading on this. Seems to me like you're going to need a ccd, a microcontroller, and some code. I found a few articles on code for analyzing contrast. So far, the few ICs I've found all had motor/actuator outputs so that they could control focus, sweeping back and forth until an ideal position was found. They didn't have an output corresponding to degree of focus or level of contrast, which is more like what I think you'd need.

This was an interesting read:
https://cardinalpeak.com/blog/detecting-well-focused-images/

Herr Mesander's problem is different from what I'm trying to address. He is analyzing a .jpeg file which contains pixel values for the entire photograph. What he gets is a single value representing the relative contrast between pixels. The higher the value, the higher the overall contrast.
On second thought, this might apply if a .jpeg file was generated in memory for the sensor area and analyzed as per the simple algorithm. That value would be stored. The sample would be repeated and contrast value new will be compared with contrast value old and the difference displayed.
Simpler yet, it just gives the contrast value period. I could turn the focus knob and watch the value max out. Voir la! Focus!
The code is simple. All that I need now is to get a .jpeg file into memory and grind it up ala Mersander. Put it in a loop and I get a continual value that I can represent as a voltage on a meter.
I can handle the code (been doing that for over 50 years). Any suggestions on the hardware?

BTY: thanks for the reply

CJ

 

I wonder if it might be better to put a sensor in each corner of viewing plate so that you don't have to move one sensor arround.

Also, I wonder if some sensor schemes might work better on the inside of the viewing plate.

First last: the sensor on the ground glass will see what I see when I try to eyeball the focus. There are camera backs which allow a photocell on a stick (call it a wand and you can charge hundreds of dollars for it) to get the exposure value. This is necessary since the ground glass attenuates the light that would strike the film. It may be a bit ahead of the film plane, but it's close enough. Not so for focus. The film plane and the ground glass must be at the same distance from the lens.
As far as fixed versus movable sensors, sometimes focus is used in the composition. The subject of interest may be in the center of the image. To isolate the subject visually the background is put out of focus. Also, manipulating the relationship of lens and film plane requires focusing along different planes. You make an adjustment in the FP/Lens relationship and adjust general focus until you get the whole mess back in focus (or as close to it as possible). Where these planes lay varies a LOT.

Thanks for the interest. Keep 'em coming! My omniscience level is notoriously low.

CJ

 

First last: the sensor on the ground glass will see what I see when I try to eyeball the focus.The film plane and the ground glass must be at the same distance from the lens.

I am visualizing that the sensor would have to look at the reflection off the viewing plate. This may or may not have more light than looking at light transmitted through the viewing plate. I am assuming that the viewing plate is only frosted on one side for best focus.

As far as fixed versus movable sensors, sometimes focus is used in the composition. The subject of interest may be in the center of the image. To isolate the subject visually the background is put out of focus. Also, manipulating the relationship of lens and film plane requires focusing along different planes. You make an adjustment in the FP/Lens relationship and adjust general focus until you get the whole mess back in focus (or as close to it as possible). Where these planes lay varies.

Oops. My scheme would very nicely put the background in focus and the subject of the portrait nicely out of focus.

 

Herr Mesander's problem is different from what I'm trying to address. He is analyzing a .jpeg file which contains pixel values for the entire photograph. What he gets is a single value representing the relative contrast between pixels. The higher the value, the higher the overall contrast.
On second thought, this might apply if a .jpeg file was generated in memory for the sensor area and analyzed as per the simple algorithm. That value would be stored. The sample would be repeated and contrast value new will be compared with contrast value old and the difference displayed.
Simpler yet, it just gives the contrast value period. I could turn the focus knob and watch the value max out. Voir la! Focus!
The code is simple. All that I need now is to get a .jpeg file into memory and grind it up ala Mersander. Put it in a loop and I get a continual value that I can represent as a voltage on a meter.
I can handle the code (been doing that for over 50 years). Any suggestions on the hardware?

BTY: thanks for the reply

CJ

How about aiming a computer's camera at the viewing screen and doing the calculation in the computer? The result would be displayed as an image on the computer screen with different qualities of focus displayed as different colors-- say redder for better focus and bluer for poorer focus. This way you can tilt and translate the camera setup for best focus exactly where you want it.

 

How about aiming a computer's camera at the viewing screen and doing the calculation in the computer? The result would be displayed as an image on the computer screen with different qualities of focus displayed as different colors-- say redder for better focus and bluer for poorer focus. This way you can tilt and translate the camera setup for best focus exactly where you want it.

Sounds interesting. The first thing that comes to mind is adding more weight to the already hefty 40 pounds the camera, lenses and tripod account for but it would be pretty neat. Some degree of miniaturization could be looked at for the screen. The format I shoot is 4X5 inches so a screen of that size would be sufficient. The algorithm would be more complex and considerably more interesting. I did something like this in the distant past for an ECM job. But first I'll attack the puck on the principal of crawl first; sprint second.

Thanks for the idea.

CJ

 

Not sure how your idea will work in practice. Digital cameras as I am sure you know have advanced a couple of strategies for auto focus, and offer several auto focus points. They still manage to focus on objects which are not the prime subject at times. My concern would be how to determine which part of the image you want to be in focus without having several sensors across the film plane. If central, every time, then a center mounted sensor could work, but if not, you will need more or some way to move the sensor.

 

Not sure how your idea will work in practice. Digital cameras as I am sure you know have advanced a couple of strategies for auto focus, and offer several auto focus points. They still manage to focus on objects which are not the prime subject at times. My concern would be how to determine which part of the image you want to be in focus without having several sensors across the film plane. If central, every time, then a center mounted sensor could work, but if not, you will need more or some way to move the sensor.

The original idea was to mount the sensor, which I see as a small digital camera ala iPhone, on a small puck which will keep it parallel and at a fixed distance from the ground glass. This should reduce some of the variables mentioned above by GopherT. Instead of looking for the signal that controls the focus, I'm thinking of putting the .jpeg file generated by the fixed focus module into memory and analyzing the file to obtain a contrast value using the method pointed to be ebeowulf17. The value would be used to generate a voltage through an AtoD circuit output to a meter. Put the whole algorithm into a loop so that, as I turn the focus knob, the value changes and I pick the maximum by eyeball. I can then move the puck, clutched in my grubby old paw, to another area of interest and note the value there. This method mimics what I do with a loupe and tired old eyes. If I'm not using camera movements and the film plane and the focal plane are parallel, I don't need to do much more than get the parts of the field of view in focus that I'm interested in (lots of artsy stuff go into this). When I start messing with the focal/film planes things get hairy and I spend a LOT of time getting the focus correct. For a detailed description of what's going on with the various movements, see a treaties on the Scheimpflug effect on line. Wiki has one that more complete than anyone needs: https://en.wikipedia.org/wiki/Scheimpflug_principle .


Scenario 2 was suggested by RichardO which would back the camera out so that it could cover the entire ground glass. The .jpeg file would be analyzed using one of a bunch of algorithms to give the relative contrasts over a field of sampling areas. The output would be output to a small screen overlaying the image to indicate that areas are in focus. There are a LOT of things that have to be worked out here but the problem is similar to a ECM/ESM project I worked on when I was gainfully employed slinging code for the military-industrial world. I'll have to dust off some old brain cells to remember just how I did this and what's applicable.

Since I'm grinding up the .jpeg file rather than relying on an auto focus value generated by the hardware, I don't need to have multiple sensors. Or so it seems. I reasonably sure that a small puck covering a small area of the ground glass so I'll start there. Scene 2 will require a lot more work but looks interesting. It'll give me something to wile away a few hours between taking photos and camping out in the darkroom trying to make decent prints.

Thanks for the interest,

CJ

 

The original idea was to mount the sensor, which I see as a small digital camera ala iPhone, on a small puck which will keep it parallel and at a fixed distance from the ground glass. This should reduce some of the variables mentioned above by GopherT. Instead of looking for the signal that controls the focus, I'm thinking of putting the .jpeg file generated by the fixed focus module into memory and analyzing the file to obtain a contrast value using the method pointed to be ebeowulf17. The value would be used to generate a voltage through an AtoD circuit output to a meter. Put the whole algorithm into a loop so that, as I turn the focus knob, the value changes and I pick the maximum by eyeball. I can then move the puck, clutched in my grubby old paw, to another area of interest and note the value there. This method mimics what I do with a loupe and tired old eyes. If I'm not using camera movements and the film plane and the focal plane are parallel, I don't need to do much more than get the parts of the field of view in focus that I'm interested in (lots of artsy stuff go into this). When I start messing with the focal/film planes things get hairy and I spend a LOT of time getting the focus correct. For a detailed description of what's going on with the various movements, see a treaties on the Scheimpflug effect on line. Wiki has one that more complete than anyone needs: https://en.wikipedia.org/wiki/Scheimpflug_principle .


Scenario 2 was suggested by RichardO which would back the camera out so that it could cover the entire ground glass. The .jpeg file would be analyzed using one of a bunch of algorithms to give the relative contrasts over a field of sampling areas. The output would be output to a small screen overlaying the image to indicate that areas are in focus. There are a LOT of things that have to be worked out here but the problem is similar to a ECM/ESM project I worked on when I was gainfully employed slinging code for the military-industrial world. I'll have to dust off some old brain cells to remember just how I did this and what's applicable.

Since I'm grinding up the .jpeg file rather than relying on an auto focus value generated by the hardware, I don't need to have multiple sensors. Or so it seems. I reasonably sure that a small puck covering a small area of the ground glass so I'll start there. Scene 2 will require a lot more work but looks interesting. It'll give me something to wile away a few hours between taking photos and camping out in the darkroom trying to make decent prints.

Thanks for the interest,

CJ

Sorry I don't have anything more to offer, but my impression is that you've got some promising concepts here. Both scenarios have quickly surpassed my knowledge level, but they sound cool. I'll be following this thread to see where it goes. If you pursue either of these (or some new approach,) I'd love to hear how it develops!