Hey all,

Didn't have a lot of time today to play but I didn try a couple of different camera positions and laser angles.  So far, I seem to get better results being lower/close to my subject with the camera and the laser raised slightly 4"  above the camera.

Not much else to report on this front but I'm gonna jump over to the MACH CLOUD thread as I have questions that should probably be asked over there to keep things straight!

Sid

Quick question so I can understand a bit clearer.

From what I can make out from the picks Art's setup (Using Rainea's convention of Y being across the gantry and X being along the table) has the laser in the XZ plane (with tilt) and Sid has it in the YZ. Also looking at the data it appears that both of your field of views is wide enough to see your whole hand so you are scanning it in one pass. Am I right so far? Art is still working on "tileing" to do wider subjects with successive passes or strips which are merged later into one mesh.


The setup I am invisaging will have a much tighter FOV to increase the measurable Z resolution. This will mean that I will have to tile to even scan my hand.

I am curious what resolution in Z you are getting with your current setups. Ie scan a smooth vertical wedge and  figure out how many steps it is interpolating in the scan.

I posted in another thread that I think the ideal resolution would be from 1:1 to 2:1 the cutting resolution of the machine. Does that sound right? If it does then on a camera with 700 odd pixels in the horizontal (the security b/w cam I'm testing) the max FOV would be something around 35mm to 70mm  for a machine that can cut to 0.1mm. This would obviously get better if you had say a 6Mega pixel camera with 3000 pixels in the horizontal. What size cameras are you guys using?

Thanks

Mark

Quick question so I can understand a bit clearer.

From what I can make out from the picks Art's setup (Using Rainea's convention of Y being across the gantry and X being along the table) has the laser in the XZ plane (with tilt) and Sid has it in the YZ. Also looking at the data it appears that both of your field of views is wide enough to see your whole hand so you are scanning it in one pass. Am I right so far? Art is still working on "tileing" to do wider subjects with successive passes or strips which are merged later into one mesh.

It's kinda early and I need more cofee but check out the drawing below to get a better idea of my setup.  I thought that I did it the same as Art?  Maybe not??   Yes, FOV is wide enough to see my whole hand.  "Tileing does work", it's just kinda funky- check out the 32" wrench I scanned yesterday- it's posted on the Mach Cloud Thread

I am curious what resolution in Z you are getting with your current setups. Ie scan a smooth vertical wedge and  figure out how many steps it is interpolating in the scan.

Not a clue about my resolution.  What do you mean 'scan a smooth vertical wedge'-   I guess I could try to scan something to figure it out but there are a lot of variables that you need to provide (what size wedge- a drawing would be helpful, how is this wedge oriented on the table, what do you want the step to be)?

I posted in another thread that I think the ideal resolution would be from 1:1 to 2:1 the cutting resolution of the machine. Does that sound right? If it does then on a camera with 700 odd pixels in the horizontal (the security b/w cam I'm testing) the max FOV would be something around 35mm to 70mm  for a machine that can cut to 0.1mm. This would obviously get better if you had say a 6Mega pixel camera with 3000 pixels in the horizontal. What size cameras are you guys using?

Maybe the coffee still needs to kick in but I really don't understand what you mean by 1:1 or 2:1.  Maybe you could explain.  I am using a cheapie webcam rightnow and I don't know the resolution but I will check.  I think is set at 640x480 and who knows the megapixels (actually it's probably not even MEGA).  Sorry I can't tell you more.

I can say that I have taken some measurements and my camera is approximately 10" above the Z0.  I am using a FOV setting of around 42 and getting some OK P.C. acquisitions.  But I am beginning to see what you guys are talking about in terms of position, alignment and distortion and I think that I may have to try some camera laser posititions to get less distorted scans

Thanks,
Sid

Hi Sid

Yeah I could do with a coffee too It's 1am here.... only a couple of hours to go. First off this is just my theory and I could be way off the mark.

My idea for a wedge is that it would be used instead of the cube Art now prescribes. This is so one can calibrate the Z axis measurments which will be non linear due to the optics and alignment causing distortion. So imagine a ramp that starts at the bed (Z=0) and raises to whatever you can measure with your setup (say Z= -3") and is orinetated to be along the Y axis in your case. Each scan should produce a Z measurment for the part of the ramp being measured and if the wedge is smooth and linear the steps should be equal. At present the way i understand it Art calibrates on 2 measurements, the table and the top of his cube.
 What I mean about 1:1 or 1:2 is that if I want to reproduce something to the tolerance of my machining ability I really need to scan it at twice that resolution or at least a minimum of 1:1. so if you can machine a surface to say 0.1mm (0.004") then you need to be able to measure to at least that if you want a true replica of the original.

Thats where the number of pixels in your CCD comes in (when they talk about mega pixel they are refering to the H times the V resolution of the CCD... I think they sometimes fudge this number some what by talking about "effective" resolution)

If when you are looking at the image of the laser on your camera of the table (Z=0) I assume you have a line either on one side or the other of your screen (depending on orientation) and when you are reading the max Z value (Z= -? so make a wedge bigger than you can read to fine that point) it draws a line at the other side of the screen.

If you are using the maximum width of the CCD (ie laser line is shown vertical not horizontal in the video window as I've seen in all the pics) then you have the horizontal pixel count (you mention 640 x 480) as the devisor of you max Z reading. So if you can just read 2" in Z then your resolution would be 2/640 or  0.003".
If you are reading the laser the other way (the way I think it looks as if Art has it set0 the it would be 2/480 or 0.004"

All that is in a perfect world and assuming the laser is only one pixel wide and the CCD can resolve each pixel.

I think the ideal scan setup would be one that can measure to the resolution of the ability to cut and also to the max depth of cut so say 150mm (4") . I may be wrong but I would think the greatest use of this scanner would be to reproduce objects that one does not have cad files on. I think it could do a lot more too,  like auto tool height setting and non contact edge finding.

It's pretty cool stuff Art is turning out .. too cool in fact as its distracting me from getting on with building my machine.



Keep working out the bugs guys... It gives me time to finish my machine and have it all sorted by then

Cheers

Mark

Not a bad idea...  Ill giev some thought to a better way of calibrating..

Art

Mark,

I think I understand what you mean regarding calibtrating to a wedge instead of a cube but 'a picture is worth a thousand words".  Could you possible sketch up a picture and post it.  Then I will try and let you know  what I discover?

Thanks,
Sid

A ramped or stepped surface may not be necessary if you choose to scan a volume (x, y, and z) to acquire data for a cube.

I think that using a sphere for a calibration target would be perfect but spheres are hard to do a least squares fit for. A ping pong ball has a standard diameter and a great surface for scattering light.

Tom Hubin
thubin@earthlink.net

Hello Art,

I would like to post some drawings and sketches. I use Vellum which has a VLM suffix. There is a free downloadable reader for VLM files but you don't list VLM files as attachable.

JPG is not one of my VLM conversion options.

I can convert VLM files to Windows BMP but this is also not listed as attachable.

Can you add VLM and BMP to your list of acceptable attachments?

Thanx,
Tom Hubin
thubin@earthlink.net

I Just sent a PM to Benny asking him if that was possible.  Are you runing Windows?  If so, there is a utility called 'Paint'  You can open .bmps and save as .jpgs  I know it's an extra step but it should work until we hear from Benny.

HTH,
Sid

P.S.  I look foward to the PIX!

If I allow BMPs others will use them too. A Jpg is 10% the file size as a BMP for the same quality result. Its just too much server usage to allow BMPs
Use this free 1 Meg program to do your conversions. www.irfanview.com
The other option is to zip your work and upload the zip file. With zipping, you can essentially upload any file.