A Bit about Measuring Beam Diameter.

[mixedgas]

Pl and LPF have review after review of various lasers with rough measurements of beam properties. Watching screen shots of excellent technicians having to resort to using a ruler have convinced me its time to share some old techniques. Well, Old to me. While I do have a 180 foot beam test range in the back yard, there are other ways.

CAMERA BASED TECHNIQUES:

I'm the proud owner of a Spiricon Beam Analyser. Its a PCI card with a rare 10 bit video digitizer. It works by using the fact that the size and spacing of CCD pixels is well known for better cameras. If your the owner of a well documented CCD camera like a JAI Pulnix TM6 or TM7, you can measure to your hearts content. Slide the camera back 100 cm on a rail and you have a very good divergence measurement. So I can also measure Gaussian or Top Hat fit, look at the beam profile in real time, etc. Spiricon does some amazing math on the video sigal. The results are very, very good. But my having a Spiricon does PLers very little. So I'll save Spiricon Screen shots for later in this thread.

The beam does need to land directly on the Camera's CCD Sensor. The CCD does need to be large enough for the whole beam to land on the active pixels.
While a professional might use a lens, it is best to use a raw CCD with no focusing optics.



My intent is to show a few low cost tricks, including using a low cost USB video digitizer and some freeware for a decent measurement that knocks the socks off a ruler.

There are some other techniques I will cover when I'm done with the camera method.

So set back, relax, and ask technical questions. When I have time, I'll try to shoot a video.

Steve

[mixedgas]

So I'm going to start with a simple example. The camera used in this case has only one horizontal line. This is Perkin Elmer linear CCD camera. It has 2048 active pixels. Each pixel is 14 by 14 micrometers. This camera was on Ebay, 40$ used.

A 20 Mhz clock drives out the data from the camera. This works out to one line of video every 104.5 microseconds.


This camera is old, and has a differential output instead of the normal analog video. I'm using a two channel oscilloscope to sum the split video signal and make a clean display. The scope is set to add channel one to channel two. This subtracts out part of the noise from an older second generation LCCD. The digital scope used is ancient, its actually part of a analyzer for fiber optic signals.

A DPSS beam from a nearly TEM00 5 mW lab module is the sample here. A four percent beam is split off using an ordinary microscope slide and is aligned on the sensor. Normally a microscope slide is a bad choice, in the future setups I'll use a glass wedge. This image shows a bit too much signal, as you can see the top of the beam is clipping, which the scope shows as a flat line at the top.

The first picture , called "LINCCDFIRST" shows my measuring two whole line periods from start pulse to start pulse. This is a sanity check to make sure the LCCD is working correctly. 105.6 microseconds is not too bad, for what should be 104 uS. A more careful measurement later showed 104.5 uS.

The second picture shows my measuring the beam. I picked an edge to edge measurement. Normally we pick a level up the side of the beam known as the One over E Squared Point or the 90/10 Percent Power Points. More about that later. You can see I moved cursors, which are the vertical dashed lines, onto the edges of the beam. The scope does some math and reads out the time between the dashed lines. So the second shot shows 12.4 Microseconds from Edge to Edge.

On the scope, in this case, time truly is distance.

Ignoring the very tiny space between pixels, 2048 pixels x 14uM = 28,672 microns give or take a bit. That gives me a 28.67 millimeter scan length.

So calculating the beam size becomes a simple ratio problem. Remember them from 7th or 8th grade?

I multiply the 12.4 uSec times the 28.67 mm. Then I divide by 104.5 uSec time per scan. When all is said and done, that gives me 3.396 millimeters. We'll pay attention to significant figures and call that a 3.4 mm beam.

Yeah, Done!


Most PL users don't have 400$ for a used oscilloscope and 40 to 2500$ for a type of camera that does not exactly grow on trees. I'll cover how to get around that cheaply in the next post! In upcoming posts: I'll cover how to sample the beam without burning the camera, and how to make sure the camera is not saturated.
In the end, you can do this on a laptop for about 100$ with a used camera.

Stay tuned.

[lightlinked]

what about bloom and vertical smear? (or horizontal depending on which way it reads out, CMOS doesn't* have smear because it's not bucket brigaded like a CCD) I presume this has its limits on input power untill you fry the sensor. I suppose you could stick on ND filters but that will be a thing that alters the beam profile, although "by how much" is the question there
at low powers couldn't you just use a traditional dslr with the lens off? it outputs "raw" so no scope needed, plus you could bracket shots to see spill around dim edges. DSLRs and certain digicams have better documented sensors than something like a generic web cam.

[VDX]

... I've measured beam profiles with a simple linear XY-setup moving a 0.03mm thick glass fiber with attached photo-transistor ... should be not more than some ten USD/€, when salvaging old CD/DVD-drives and an Arduino as controller ...

Viktor

[mixedgas]

what about bloom and vertical smear? (or horizontal depending on which way it reads out, CMOS doesn't* have smear because it's not bucket brigaded like a CCD) I presume this has its limits on input power untill you fry the sensor. I suppose you could stick on ND filters but that will be a thing that alters the beam profile, although "by how much" is the question there
at low powers couldn't you just use a traditional dslr with the lens off? it outputs "raw" so no scope needed, plus you could bracket shots to see spill around dim edges. DSLRs and certain digicams have better documented sensors than something like a generic web cam.

You raise some valid points.

I'm not using a generic web cam, I would never even consider such a small sensor. Besides there is no way of ensuring what is inside a web cam.
I've never seen one with the pixel sizes posted in the data sheet. This method depends on having known pixel sizes and no software interpolation of the raw video.
Web Cams often rely heavily on interpolation between pixels.

One danger with some modern still cameras is software interpolation. You need a RAW image free of processing.

A known CCD with documented pixel sizes is self-calibrating which is the beauty of this method.

Bloom and Smear become readily apparent when you look at the raw data. I'll cover the free software I use to do basic tests.

Wasting and risking an expensive DSLR is not needed. Besides using video cameras gives you a real time profile.

The sampling glass wedge launches 3.9% of the beam as long as its at a small angle to the beam, ie... less then 20 degrees. A wedge at is used so the second Fresnel reflection off the other side of the sampler does not hit the CCD. If the beam sampler is not near normal to the beam, Malus' Law shows you have much higher powers coming off the sampler. A thick piece of AR coated glass can also be used, if you have a big enough table space to ensure your not picking the other Fresnel reflection.

Its simple at that point to use crossed polarizers or a coated variable wedge attenuator to reduce to a power that will not damage an older silicon CCD.

There is a human factors issue here when you set the sensitivity of the camera and adjust the attenuator. So there is a bit of subjective error on the thin, outermost wings of a laser diode generated beam. However most of the Energy is between the 90-10 points, which is why even professionals truncate the edges of the measurement.

The line CCD was just to show the concept quickly. While it covers very large beams, you do have the issue of having to scan it across one axis of the beam,.

I have in the past used a galvo and a pinhole detector, however that method is a real pain to set up.

Part two shows how to grab both axis at once and see the whole beam.

Part Two is coming. Both software and the inexpensive hardware.


Steve

[mixedgas]

Here are two pics of camera CCDs . One is a Hitachi KP-120E Security Camera. The other is a JAI Pulnix TM-7EX Lab and Industrial Camera.

The key is finding a camera with a known CCD chip listed on the data sheet. Scientific grade cameras ALWAYS list the pixel sizes in the manual.
Sometimes you can find the data for the B&W security cameras.

Used Black and White, NTSC or PAL Format Cameras abound on Ebay and at local CCTV installers. A decent camera with a 1/2 inch or 2/3rds inch Diagonal CCD can be had for 20-75$, and in a few cases I've gotten them for free.

European PAL format cameras have an extra ~100 vertical lines. This adds more resolution, if your hardware supports the format.

One NEVER wants to try a color camera for this. It might work, but the color pixels are weighed to produce a balanced white. Thus the pixel responses are uneven as you scan across the beam. Besides the color separator filters in the camera are going to reject some portion of your nearly monochromatic light.

The TM-7EX Camera has

8.4 μm x 9.8 μm Pixels

A resolution of
768 H x 494 V (NTSC-EIA)
752 H x 582 V (PAL-CCIR)

So with the NTSC version you have 768 * 8.4 / 1000 = 6.4 mm to work with on the Horizontal Axis
For the Vertical axis you have 494 + 9.8 /1000 = 4.8 mm to work with on the Vertical Axis.

A half inch CCD nicely matches the useful aperture of most scanner sets.

I did not scale the pictures, but rest assured the KP-120E has a 2/3rds inch Diagonal CCD and is much, much bigger.
A 3/4ths inch diagonal CCD from a lab camera or machine vision system would be a real score for this technique.

It is possible but difficult to calibrate a camera with an unknown CCD if you have a precision motion stage of some sort and a HENE with a known beam size. That discussion is far, far, from the scope of this thread. Accuracy of a self calibrated CCD is poor. Thus I'd like to stay focused on CCD CHIPS of known size.

If you cannot get a known camera, by all means still try the technique. There is a wealth of data about the beam besides its size. You can also see the divergence, the profile, and with some careful work, obtain your best collimation of a diode beam.

Note that the beam hits the raw CCD, no LENS is used. Using a lens looses the inherent calibration of the known spacing of the pixels. Lenses also introduce some interesting distortions to the beam profile.

More coming as I have time. Including pictures of an actual setup.

Steve

[CDBEAM]

Please...Please continue this thread !!!! Some time ago you brought this concept up....but now....I have a distinct need to get some accurate FF beam measurements....and if it can be done for a few hundred dollars....with some searching.....that is very valuable !!!! Thank you !! CDBEAM=======>

[mixedgas]

First the PAIN IN THE ARRSE METHOD if you have an oscilloscope:

The attached circuit extracts one VIDEO line from a camera, selectable from lines 15-253. Attach a dip switch and pull-up resistors to the binary inputs.

Not the way I do it now, but included for completeness. This would replace the Linear CCD.

Reference for the Design, the LM1881 Data Sheet. This is directly pasted from the Data Sheet.

I have a love affair with the LM1881, it lets me do field sequential stereo video with two synchronized cameras and LCD goggles. But I digress...

Tomorrow Afternoon, CD Beam...

Steve

[mixedgas]

Next is the critical piece of gear. A high quality means to get the NTSC or PAL video into the laptop or desk top.

Attached are a pic of two of my favorite PCI BUS "Frame Grabbers" . Obviously a scientific or machine vision grade board such as these is the Cat's Meow.

One by IDS, The Falcon
www.ids-imaging.de

The other by Integral Technologies, a FlashBus Lite

Integral was bought out, so if you buy one used, make sure you get the disks.

The two pictured were less then 5$ each at my local surplus store.

However those do not grow on trees. They do show on Ebay...

There are others, like cards by Matrox Inc and Dalsa, Inc. etc.

One reason to consider a Frame Grabber is they often have a sync output for firing strobes, which fires pulsed lasers just fine.

Frame Grabbers usually have a TTL sync input, too. So pulsed laser + delay generator = beam profile.

But Steve, I'm poor and I have no PCI bus:

So the newer, cheaper, slightly less accurate but useful answer is in the second picture.

Not all USB Video Capture Devices are created equal.

Even more critical is making sure the included software lets you grab an uncompressed still.

Steve

[mixedgas]

Next you need some Science Grade software for image processing and pixel counting.

http://imagej.nih.gov/ij/

Image J, formerly Scion Image.

Your Tax dollars at work, and worth EVERY Penny.

Something like 100,000$ worth of image processing programming for the Health Sciences.

A Mac and Linux (gasp!) version is out there some place, too.


Steve