Schlieren Images for the masses

[pezman]

Ok, here's a way that anyone can make an image of flows with a simple digital camera and a PC.

(1) Make a "mask" image. This is a picture of the device under test against a patterned background. The device should be in a quiescent state (cool and no significant flow present). In this case, I have an unlit candle sitting on my laptop keyboard and the screen of the laptop is providing the "schlieren" pattern. I used a random pattern of black and white dots for the pattern.

(2) Do not move anything. Make an image of the device under active conditions. In this case, I lit the candle.

(3) Convert the images into 24-bit color BMP files and xor the bits of the images. The changes in density of the air will move the dots of the mask a little to the left or a little to the right and when xored, it causes the dots to become darker or lighter.

No expensive optics required. Very simple, very cheap. I have attached an image and a zipped copy of the VB progam that I used to process the two images. As written, the program will put the image in the file "C:\SyntheticSchlieren.bmp" if you are handy with VB, you won't have any problem modifying it. The program is very crude - I just made a quick hack of a BMP parsing program that I found on the web.

Here is a link to a paper the describes the synthetic schlieren technique (I actually came up with the idea of a synthetic schlieren tchnique and then did a litle research and found this). http://taylor.math.ualberta.ca/~bruce/p ... _style.pdf

Attached another image (taken using black & white mode with my camera using a strobe and a paper mask).

[pezman]

I just want to mention that it should be easy to apply this technique in the field. The back-ground just needs to be textured. For example, grass or sand could be used as the background pattern as long as the wind is not moving it around.

This means that folks at the PJ meet could probably take Schlieren images. Just set up a tripod, photograph the jet while it is cool and then photograph the jet again while it is running. By subjecting the two images to the masking process, you should end up with a pretty good image of the engine's exhaust flow.

I would be happy to process any photos if anyone wants to give this a try. Try to make the photos in some lossless format if possible (TIFF, PNG etc.), make sure that the camera doesn't move and pick a high-contrast background with lots of texture (grass, sand, gravel, a mesh grille on a truck etc.).

[pezman]

http://www.damtp.cam.ac.uk/lab/digiflow/

Hard to beat the price ...

[Mike Everman]

Oh, that is so cool, pez. Must dive deeper. Wish me luck.

[pezman]

Ok, here's an image generated by DigiFlow. It is the same as mask and photo as SyntheticSchlierenSmall.JPG above (http://www.pulse-jets.com/phpbb2/download.php?id=3413). This is a gray-scale image of the density gradients in the X-axis.

DigiFlow has lots of fancy options, so I'm not sure how best to use it, but it sure is cool ...

[mk]

Very cool, pez, very cool!

I'm going to try it out when I'm back at home.

[pezman]

DigiFlow was an easy download and install and has a nicely written user's manual. I loaded it on three machines and it ran on two of them. It gets a "virtual memory allocation error" on my laptop (running an old version of XP pro). Runs just fine on my machine at work (Win2k) and the kid's PC (a cheesy E-Machines w/ 256m ram and a 1.something GHz Celeron).

It is insanely slow on the celeron, but reasonably peppy on the machine at work (a 2.something GHz P4 with 1GB of ram).

It is incredibly well written, does a lot more than synthetic Schlieren images and has excellent documentation.

[steve]

(3) Convert the images into 24-bit color BMP files and xor the bits of the images. The changes in density of the air will move the dots of the mask a little to the left or a little to the right and when xored, it causes the dots to become darker or lighter.

I am trying to generate an image showing the shockwaves produced by my aerospike rocket nozzle running at 110 psi and sofar have had no sucess with either of the programs. I obviously have no idea what I'm doing- there is no sense in denying that.

could you explain what is ment by "xor"

alternatively could someone give me a quick rundown on how digiflow is used/ do I need to get the full version?

I could also post the original pictures if that would help.

[pezman]

Steve,

Use DigiFlow for sure -- my program is a quick and dirty experiment.

The unlicensed version of Digiflow will work fine.

Please feel free to post the photos as well -- I'll see what I can do. Basically, you want a photo w/ just the textured backround and then another with the flow occurring on front of the background.

This'll be cool.

[Rossco]

Hey,
This sure is cool.
BUT, also not having a clue what im doing, its not giving me fast results.

I have done about the same. Speckled background (chair back) and then a lit candle (scrunched up piece of paper in an office coffee mug).

Im not getting good results tho?
Im running it as a "Pattern match" under "Analyse\Synthetic Schlieren"?
Is this right? Im just getting the output as the "candle" all else white!

When i do a "Qualitative Preview" i dont get any idea of a shape for the heat flow, under any settings?
Do you think that its just my images? I don't have a remote trigger either! This may be causing some movement in the shots?

Anyways, looks cool, and doesnt take long at all to download and start using. This is a MUST! for those that want to get further into flow patterns in the reall world. I just cant wait to get it onto a jar. (many hours need to be set asside, this will take time to get the right shot i think)

Rossco

[steve]

here are pictures from one of my attempts.

[Rossco]

Whats the difference steve?

How do you run that nozzle? Just with compressed air?

ill have a play with your images if you dont mind, ill see what i can see with schlieren thingy

Rossco

[Anders Troberg]

could you explain what is ment by "xor"

It's a bitwise logical operator, short for exclusive or. Basically, if one and only one of the bits is a 1, then the output is a 1, otherwise it's a 0.

Lets say we have the following color samples (8 bit grey scale):

Code: Select all

Pic1  0 1 1 0 1 0 1 1
Pic2  0 1 1 1 0 0 1 0
----------------------
Res   0 0 0 1 1 0 0 1

In other words, only the bits that differ between the images will be set, providing a map of the changes.

I assumed grey scale just to make the example shorter, it works just as well on a 24 bit color image (or any other color depth).

One important thing for this to work: Do not use JPG or any other lossy image format. JPG tends to create an unpredictable slight "haze" around sharp edges, which will mess up these images as it will look a lot like the heat haze we are looking for.

[Rossco]

HA! thankyou!
JPG!
Of corse! stupid me.

No wonder i get a nice fuzzy mess. There is much, much to much change!
This isnt my sole problem, but its led me right to it!

Stupid point and shoot. I didnt even think of it. With its caotic post processing its a wonder the image comes out looking right as a jpg, let alone trying to do some intricate data gathering from changes in individual pixles! Shoot a cat, and it might come out as a dog the other end!

Dissapointed.... will get hold of better camera!

Rossco

[Anders Troberg]

Before you run out and buy a new camera (unless you are looking for a reason to buy a new camera...), check the settings of the camera. Most cameras can take uncompressed images.