What's new
What's new

Optical comparator have any advantage over video measurement system other than cost ?

Milacron

Administrator
Staff member
Joined
Dec 15, 2000
Location
SC, USA
And if so, it would seem one could cobble together the camera, X/Y table, software to make one's own video system at low cost such that even the price would be no advantage of the OC unless thinking in terms of buying an older used OC. Thoughts ?

847953.jpg
 
I'm amazed no one is hopping on this subject. Don't tell me I'm the only one that has thought of this...
 
I think a lot will have thought of it but not acted on it, just like me. I have a working OC that I use frequently, but also a 100x digital camera that could be mounted on an x/y table, just havent got around to it. One of my customers has a video measurement system but has reported large measurement uncertainty when the features being meaured are different distances from the lens, something to do with the focal length?
 
Optical comparators are obsolete. Computer vision systems have many advantages, and can take advantage of structured light, interferometry and other phenomena. The secret sauce is the software, and non-trivial optics.

My photo-microscopy project ... the hardware part anyways. (That Starrett is missing its goniometer ;) )

IMG_20130316_181821.jpg
 
Optical comparators are obsolete.

Not a chance. Even the best LCD/cameras wont touch the resolution of a 30" optical system with overlays.
We are getting closer but we have far to go.

On small fields you can go pure digital but you can't do the big field stuff yet.
I've been building vision systems for 30 years now and there are still things my optical compartors do better.

What we need is 100 megapixel cameras, LCDs with higher resolutions and the bandwidth to pump this huge amount of data in real time.

FOV vs resoultion and amount of data, not to mention active senor area in the camera itself. And then there is how you make color and it's killing of resolution.

This is one of those things that look very simple until you get deep into the details and roadblocks.
Bob
 
I've wondered this myself too, but I'm no metrology expert so don't really have much to add. I still see optical comparitors at IMTS so I figure there still must be uses for them. I saw a demo of one that had software on it that would automatically find the edges and do calculations like finding the center of a circle, etc. I'd like to have one at the shop but not sure what brands / models of older machines are still supported, and I'm not willing to buy a new one just yet.
 
One of my customers has a video measurement system but has reported large measurement uncertainty when the features being meaured are different distances from the lens, something to do with the focal length?
Interesting. Can you find out exactly what he has ? Just wondering if it's an "antique" system from the early 90's or something more current.
 
One of my customers has a video measurement system but has reported large measurement uncertainty when the features being meaured are different distances from the lens, something to do with the focal length?

Its a depth of field/magnification ratio problem and newer stuff will not fix basic optics/physics laws that have been known for 100's of years. Newer software may tell you the measurement is worthless by analyzing the edge quality.

One would assumed front lit parts. A back lit system of an optical comparator type would go out of focus before you get the mag/distance problem.

Nice thing about a human in between, they don't try to measure off of "soft" edges.
Things closer to the camera look bigger, obvious to anyone taking pictures with their I-phone.
A big concern when you are measuring things.

Telecentric optics on the front help a lot along with getting rid of the errors with holes off center and are pretty much a must on most measuring machines but are not cheap.
Ideally on a compartor replacement you want to be telecentric on both the light and lens.

In general a 100x video won't get you squat but specifying X is always bad in a video system. A 40 inch monitor gives lots of X without any help.
Most of my vision systems run 500-1000x which is barley good enough to resolve a tenth.

The nicest thing about a pure optical system is that it will pass a almost unlimited amount of data through in parallel at the speed of light.
Bob
 
My .02,

Optical comparators have one advantage - on the shop floor they're invaluable. I used one every day all day for twenty years in the diamond tool industry. So I'm partial to their use and intimate with the speed at which you can set up and acquire measurements during the machining process. Often times this is not clean work, but the comparators are robust in that environment.

In the inspection room the jury's out. However, we did find considerable advantage with digital systems for measuring sub-micron edge waviness of radii and complex edge geometries on monocrystaline diamond tools. The caveat was that the inspected parts needed to be flat and perpendicular to the lens planes, limiting the types of geometries we could inspect that way.

That was nearly 10 years ago, so I'm not familiar with the latest bleeding-edge technologies. Perhaps they have solved some of those problems.

Best Regards,
Bob
 
Its a depth of field/magnification ratio problem and newer stuff will not fix basic optics/physics laws that have been known for 100's of years.

Biological organisms solve this problem with stereo vision.

The nicest thing about a pure optical system is that it will pass a almost unlimited amount of data through in parallel at the speed of light.

Edge detection, and other feature enhancements, can be done directly with Fourier filters in the optical path. In parallel at the speed of light.

Most of my vision systems run 500-1000x which is barley good enough to resolve a tenth.

What kind of features are you trying to detect? At a pixel-pitch of 5μm, with one-to-one magnification, everything else being equal, you should be able to resolve down to 10μm at the Nyquist limit. 10μm is approximately 4 tenths. What do you mean by "500-1000x"? (10/500) micron = 7.87401575 × 10-7 inches. At those magnifications you need a pretty sophisticated positioning system.

Most industrial grade optical measurement systems are 'brute force', and therefore pretty naive.
 
What kind of features are you trying to detect? At a pixel-pitch of 5μm, with one-to-one magnification, everything else being equal, you should be able to resolve down to 10μm at the Nyquist limit. 10μm is approximately 4 tenths.

Since the standard buy-off for a gauge is a six-sigma at 10 times the part tolerance a 4 tenths device is not real useful for measuring.

500-1000x refers to object to display size the way optical comparators are spec'd.

For my in-house use it is measuring cutting tools and the 1/2 micron white light limit is a real problem even when resorting to statistical and gray-scale sub-pixeling.

Stereo vision is one of those things that looks nice until you build a few.
Even with a 90 degree separation of the sensors the system resolution is very disappointing even before the temperature in the room changes, moving your camera mounts and killing everything.
Projecting a grid from 45 or a scanning laser line works better to get the Z info which you can use to change your pixel size based on depth to the feature.

You are absolutely correct that you should do all the pre-processing that you can with the optics to unload the amount of work needed by the software. In our three day "Introduction to Machine Vision" classes the students got two days of pure optics and lighting and I got one day to cover cameras, interfaces, and image processing in the computer.

The world of dimensional measuring using video is full of quicksand that looks like solid ground at first, but it is great fun and sure stretches the ole brain.
Bob
 
but it is great fun and sure stretches the ole brain.
Bob

Agreeing with everything Bob said.

But just wanting to add the true non-obvious advantage to old-skool comparators. They harbor no secrets. This is not true with computer vision systems. They are tightly sealed black boxes. More than 9/10 of what you pay for is the software. Good luck if you want to add a trivial feature to that. Cheap array processors, and parallel programming are just coming of age now. Systems that take good advantage of Nvidia's cuda pipelines (for example) will very soon make last years models obsolete. So buying a computer vision system is catching a falling knife.
 
Sure is an interesting subject!

I'd like to make optical measurement on my T&C grinder. But without removing the work. What I do have now is a lens from Peak with an assortment of replaceable platens with different patterns (thread pitch, angles etc.). I know, crude and in-accurate and not comparable to a comparator.

Often, you do have the problem that the feature you want to measure is deeper than the focal length, so you don't get sharp vision. Means, you need optics that have a distance from say 50 mm for a sharp picture.
When working with digital image sensors, don't be trapped in the assumption that the pixel are evenly spaced. I'd only measure at the same pixel of the sensor. So a crosshair and a movable table. Also, optics will change magnification if you change the focus (I bet extra expensive optics that avoid that problem do exist). You also will have a fish-eye effect. I wouldn't trust any cheap optics to compensate for that.

Some of the optics even allow you to take depth measurement by adjusting the focus.


Nick
 
Nick,
I you haven't already check out Optics, Imaging, and Photonics Technology | Edmund Optics.
Lots of good tech info and just about anything made in optic systems.
Want a microscope with a 500mm or more standoff, they've got it, a backlight for threads for a clean image, they've got it.
Not the cheapest place but top quality and a great selection of ideas.

I'd suggest getting a hard copy of their catalog as you can spend much time looking through it.
Bob
 








 
Back
Top