Suggestions please.

[Steve GarnettParticipant @stevegarnett62550]

Further clarification on Pixels – I looked up the spec of the scanner I used (Epson 2480), and it makes a distinction between ‘effective pixels’ and the system resolution. So the actual number of cells across the CCD is given as 2400 per inch (so that’s 2400 dots per inch of optical resolution), but by interpolation, this increases to an outrageous 20,400 x 28080. The resolution in the scanning direction though is determined by the stepper motor in the scanner, and this has two distinct physical resolutions, because it can micro-step. Whether it does so is determined as much by the zoom setting you use during scanning as anything.

 

And that’s important. What it means is that to get the best optical resolution, you have to zoom in at the scanning stage to show just the bit you want to measure, because then the micro-stepping is used, and I think that this is pretty general with scanners. I think that in general these days, scanners use pretty reputable CCD devices which use laser-printed layers to construct them, and are remarkably linear – at least as good as a DRO scale, anyway, because the technology to create them is very similar. If there’s going to be errors anywhere, they will be in the transport system, with belts and the ground bar that the scanning head runs along. That’s the axis that really needs to be checked, I think, although the errors appear on mine to be remarkably small, and could easily be caused by the rule I used. I think I might trust the scanner more than that rule! Then ‘all’ you have to do is to make sure that the aspect ratio of the file you are displaying on your measuring instrument (CAD) is correct…

 

But back in the real world, can you grind an angle like this anywhere near as accurately as you appear to be able to measure it? And how long will the tool retain this angle (as Graham has already intimated…)?

 

Sorry to harp on about this, but I think that if you are going to do it this way, then a careful consideration of the potential errors is in order.

[Nicholas FarrParticipant @nicholasfarr14254]

Hi Steve, yes I did read it in the first place, but as these threads progress one sometimes forget what has been said already. Having said that, reading it again it seems to be trying to establish the scanner errors and composating it into the calculations, whereas my thoughts were to be accepting any scanner errors into the final conclusions. I think thats what I mean.

 

Apoligies if I’ve got it wrong.

 

Regards Nick.

[Les Jones 1Participant @lesjones1]

Hi all,

I think there is the potential for no linearity across the width of the scanners I have seen. The CCD sensor on these is not the full width of the page it is only about 25 mm effective width. The reflected light from the item being scanned goes vertically down for about 50 mm where it is turned through 90 degrees by a long mirror set at 45 degrees.

It then travels about another 100 mm to a lens where it is focused on the CCD sensor which is about 25 mm from the lens. The distances I gave will probably vary between scanners. I suspect that most of this non linearity can be taken out by the firmware in the scanner or the driver software.

 

Les.

[Steve GarnettParticipant @stevegarnett62550]

Posted by Graham Meek on 02/04/2011 14:53:18:

A rub with an oilstone on each side of the toolbit has rectified the error such that it now reads zero-zero along the cutting edge at 29 deg.

 

 

Graham, can you scan it again, so we can re-measure it? It might have been only an error of 0.03 degrees (about 2 minutes of arc), but I’d like to see how much it’s changed…

 

As for the scanners, Les – well yes, there is a potential for error in the scan assembly, but as you suggested I think that this is either trimmed out during assembly or compensated for in the firmware. But it’s also worth noting that this only applies to higher quality scanners, not the really cheap ones. The cheaper ones generally use Contact Imaging Sensor technology, and have loads of LEDs and a pickup system spread right across the scanned area. One snag with these is that they don’t have as high a resolution as the more conventional ones, so probably won’t enable you to get such good results. I haven’t got one though, so I can’t test that.

 

Nick, it’s beginning to look very much as though with a decent scanner, you may well be able to ignore the errors – they really don’t amount to very much at all. If Graham can post a new scan of his mechanically optimised cutter, then we’ll know for sure.

 

Regards, Steve.

[Steve GarnettParticipant @stevegarnett62550]

Graham, one day next week would be fine. If all else fails, email it to one of us (or download Firefox!) and we’ll post it.

[mgjParticipant @mgj]

Well we agree about introducing distortion. If the camera back is not parallel to the plane that you need to take a pic of you will introduce distortion. (but then it is terribly simple to dial it out).

 

As for pixels existing.

 

Well not really. As I said they are a wedge of info. Location luminosity colour etc. Its just string of info which says I am so bright and live there. So when you look at a pic on screen at 72 dpi it should be enormous – but it isn’t is it. The same picture using the same number of megapixels at 300 dpi for printing is probably around A4 sized. in the case of my camera A3. So systems include or reject and interpolate up or down according to the requirement, using different algorithms as they do so.

 

So do they exist – well yes, but as a mathematical statement derived from the original capture. You want a picture brighter, you multiply the luminosity values. You want a pic in B&W you drop the colour infor. You want to move a pixel, you select it and change the location value for Xand Y. What you see on your screen is a highly signal processed “thing”

 

As for lenses introducing aberrations, true, but if one is that worried, any decent software allows you to adjust pixel alignment to remove it. Indeed smart cameras nowadys store aberration information about a lens, recognise that lens when it is fitted and make adjustments at the capture stage.

 

The simple truth is that provided certain conditions of parallelism are met, (because it IS better not to rely on subsequent signal processing) any of the photo processes will produce a clean, enlarged distortion free images from which you can gauge this kind of info very accurately.

[Les Jones 1Participant @lesjones1]

Hi Steve,

I have just had a closer look a t the read head assembly that I kept from my old scanner. I wanted to check if the lens was spherical or just curved in one plane. It was spherical. On dismantling the assembly I found it contained a further two flat mirrors which increased the effective distance from the lens to the object being scanned to about 350 mm. (From looking in from the slot at the top I thought it was only about 150 mm.) This would give much less distortion than I originally thought as the angle covered by the lens is so much less.

Les.

[Steve GarnettParticipant @stevegarnett62550]

Posted by mgj on 02/04/2011 19:28:28:

 

The simple truth is that provided certain conditions of parallelism are met, (because it IS better not to rely on subsequent signal processing) any of the photo processes will produce a clean, enlarged distortion free images from which you can gauge this kind of info very accurately.

 

But that condition is very much the issue here – how much effort is it to get your camera back (well, the image sensor anyway) parallel? Quite a bit, I’d say. And as for ‘dial it out’ – how much dialling it out do you do on something you want to measure? How do you know where to stop? That’s two potential additional errors you could well do without, I’d say.

 

On the other hand, neither of these is an issue with a scanner. So I’d still say that the scanner wins hands down in this particular context, as it’s optically set up all the time. The other thing that has become apparent is that scanner measurement technique has been researched quite a bit for use in other fields, and is quite widely used – because it’s pretty accurate.

Edited By Steve Garnett on 02/04/2011 21:41:47

[Sam StonesParticipant @samstones42903]

Hi,

It looks like I missed the boat on a couple of issues. Putting the clocks back last night, may have had something to do with it!!!

I’m quite intrigued by all this (extra) detail about aspect ratios, picture cells, spherical aberrations, and so on, and I fully agree about the difficulty in setting up. Despite my longtime association with photography, setting up a camera is certainly not easy. All I can comment on here (although already mentioned) is, in setting up a camera, use the longest focal length macro lens you have available. However, scanning does seem to have greater set-up merit.

One thought I had Graham (but now too late), was about scanning two images of the tool bit, 90 degrees to each other. The other relates to my own attempts to measure the angle of the previous image via Adobe Photoshop CS3 and my CAD package Keycreator.

Although I measured an angle greater than 29 degrees, the image contrast on the lower side of the tool bit was not sufficient to obtain an accurate alignment. This was due to the shadow on the underside of the image. In CS3, adjusting contrast and what Adobe refer to as `Sharpen’ – `Unsharp Mask’ is readily achievable. However, although I tweaked the image, and in combination with the staircase (ragged) edges, there was still not sufficient `separation’ to accurately show how much greater than 29 degrees, the image was suggesting. I have to accept that being placed on the flat bed of a scanner, my previous comments about tilt, don’t figure.

To digress a little, several years ago, it became necessary for me to determine the profile of a circular flume which was part of a moderately-sized casting, ie. about the size of the bell-housing around a typical car engine clutch. The casting, (which was shaped in such a way that measurements were very difficult to determine), had been hand poured and moulded in black polyester resin. In a geometric sense, the flume (channel) progressively decreased in depth radially as a means of distributing high velocity water evenly through the entire 360 degrees. I needed a low cost method to transfer various details of the moulding into CAD, and which could then be used to design a shape which could be injection moulded. Rather than resort to finding someone with a digitiser, I chose the following approach.

A number of sections were cut around the flume, and were duly `cleaned up and flattened’ with a wood plane. Because parts of the moulding curved away to one side, this part of the sectioned piece became included in the scan. And, being in black material, it proved impossible to obtain a clear outline. Therefore, to increase the contrast, I simply painted the cut edge with white-out. Any excess white-out which had gone over the edge was cleaned away thus improving the edge detail.

This thread has turned out to be most interesting. Well done Graham!!!

Regards to All,

Sam

[Steve GarnettParticipant @stevegarnett62550]

Posted by Sam Stones on 03/04/2011 00:14:36:

Although I measured an angle greater than 29 degrees, the image contrast on the lower side of the tool bit was not sufficient to obtain an accurate alignment. This was due to the shadow on the underside of the image. In CS3, adjusting contrast and what Adobe refer to as `Sharpen’ – `Unsharp Mask’ is readily achievable. However, although I tweaked the image, and in combination with the staircase (ragged) edges, there was still not sufficient `separation’ to accurately show how much greater than 29 degrees, the image was suggesting. I have to accept that being placed on the flat bed of a scanner, my previous comments about tilt, don’t figure.

Sam, that’s a good point about the lower side of the tool. It’s caused simply by the fact that you can’t have the scanner illumination and pickup at an identical point! Doesn’t matter with flat things, because there’s nothing to create a shadow, but with a 3D object it becomes an issue. It should be possible to get around it quite simply though – just align the angles to be measured so that neither is parallel to the light source. That should reveal both edges clearly, I think. Since the measuring method doesn’t rely on any particular alignment of the object re. the scanner, this shouldn’t matter in the slightest.

[mgjParticipant @mgj]

Image/edge contrast on a scanner – fit/use the slide copier and scan that way. You should get a razor sharp high contrast shadow image. Assuming you can blank off the edges from external light. Because that is the problem very often with 3D objects, but htat way round will solve that problem – mostly.

 

As for difficulty setting up the camera- you have an ordinary spirit level and a tripod? You have your centre point marked in the viewfinder with some precision? And some cameras have the spirit level accurate to about 1 pixel in the viewfinder nowadays. A lot of the smart phones have an excellent level built in.

 

People must do as they wish, but it seems to me very much easier to use a system designed to cope with 3 D objects and which is happy with varying contrasts. Frankly I’d just put the camera on a tripod, centre up the tool, use a sprit level on the camera back in 2 planes, because I don’t find a spirit level very difficult to use, and go click. Then I go into Photoshop, straighten out any distortion to limit of 1 pixel if thats necessary, (mine is dialled out by the camera automatically) and use the angle measurer to decide my angles to 2 decimal places – the limit being 1 pixel width on the display.

 

If I don’t have PS then I knock up the biggest print I can, draw lines and use trig.

 

Its not very difficult, but yes, there is likely to be a very small error somewhere, because measuring is like that. If you need greater accuracy than 1 pixel, then probably the National Physical Laboratory are the people to ask, because one is operating well beyond the limits that we can engineer to.

 

 

Edited By mgj on 03/04/2011 10:38:57

[MICHAEL WILLIAMSParticipant @michaelwilliams41215]

Measuring all angles on tooling is easy if you use your milling machine (or even a vertical slide set up ) as a coordinate measuring machine . Mount the tool on the table and a dial gauge on the spindle . Touch your tool to the dial gauge to a known reading (ideally zero) traverse the table and touch the work in a different place to the gauge again to the same reading . The difference in cross traverse readings divided by the difference of the long travel readings is the tangent of the flank angle relative to some (arbitrary)mean line . Repeat for each edge , look up the angles , add them together and get the true tool angle –  all in about 90 seconds . Accuracy about 0.5% for a normal size tool if you read gauges and traverse to !/2 thou .  Do the same with the tool blank edges as well and you can verify that the tool angles are symmetric to the blank .

 

Best type of gauge to use for this is a finger type fitted with a cylindrical tip . Using a cylindrical tip means that you don’t have to worry about third axis (height) variations causing false readings .

 

The general principal of using a milling machine as a coordinate measuring machine is universal and can work on all size work in all three axes .

Edited By MICHAEL WILLIAMS on 03/04/2011 11:05:18

[Ian S CParticipant @iansc]

Steve, the over head projector I was thinking of was the old optical type not the digital ones of today, I dont think the ones I know have a keystone facility, to get the angles right you would have to get the lens at right angles in all planes. Ian S C

[Steve GarnettParticipant @stevegarnett62550]

The old optical ones usually had a tiltable lens at the top of the arm for keystone adjustments – otherwise you had to tilt the projector, which was often a bit of a pain…

 

And mgj, I will try it with the slide scanner and see if it improves the results here. At present, without getting two good edge images, it’s hard to get known angles measured any more accurately than to within 0.15 degrees. With good edges though, it’s better than that, and considerably better than 0.5%.

[Steve GarnettParticipant @stevegarnett62550]

Posted by mgj on 03/04/2011 10:34:14:

Image/edge contrast on a scanner – fit/use the slide copier and scan that way. You should get a razor sharp high contrast shadow image. Assuming you can blank off the edges from external light. Because that is the problem very often with 3D objects, but htat way round will solve that problem – mostly.

 

Okay, tried this now and it does indeed give better results. On page two, Chris mentioned M&W vernier protractors, and I have one of those too… Set the whole thing up under a magnifying glass and on a surface plate to get an accurate zero (these protractors are a bit of a pain like that, but you get a much more accurate result), dialled in 29 degrees dead with the vernier and magnifier again. Set the scanner illumination for slide scanning, increased the zoom to 500% and measured the scanned angle using the two-lines process already mentioned. Answer is 29.054 degrees, and I don’t reckon that the M&W is any more accurate than that.

[TerrydParticipant @terryd72465]

Hi Graham,

 

Thanks for the message, I would just say that when scanning you can leave the lid open and you get a black background (nothing to reflect the light back) and this can give better edge definition on some subjects such as your tool. (that sounds slightly rude )

 

Regards

 

Terry

[Steve GarnettParticipant @stevegarnett62550]

Posted by Graham Meek on 03/04/2011 13:37:28:

Hi Steve,

 

Did you get the scanned images last night after?

Was it a complete failure on my part?

 

Er, no… I don’t think it’s my email system either – I got several other messages.

[Steve GarnettParticipant @stevegarnett62550]

Hmm… I can though!

 

[Steve GarnettParticipant @stevegarnett62550]

The one I measured here comes out at 28.88 degrees. The scan edges aren’t particularly clear when you blow them up, but this is unlikely to be far out.

[Sub MandrelParticipant @submandrel]

Just catching up, but Terry’s comment on leaving the lid open reminded me that there seem to be two types of scanner optics. One is focused more or less on the glass and anything moe than a few mm away goes out of focus. Another type seems to have huge depth of field, I don’t know how the optics work.

 

I now about the second as many, many years ago I accidentally did a scan at work with the lid open. It took ages to figure out what the picture was – roughly a square foot of ceiling in pretty clear detail!

 

Perhaps such a scanner would be better for ME subjects?

 

Neil

[TerrydParticipant @terryd72465]

Hi Steve,

 

On my DeltaCad I make your angle on the two images above as an average of about 28.6 degrees. I hope that helps

 

Best regards

Terry

[Author]

Posts