LDI, PCBs, And The Propeller (New Photos Uploaded 02/19/2019 - 8:05 PM)
idbruce
Posts: 6,197
Hello Everyone
I am currently in the process of designing and building two new machines, one of them being a rotary plotter and the other being a Laser Direct Imager for PCBs, with the focus of this discussion being on the latter of the two.
I have a large portion of the design and mechanical operation in my head, and I have a large portion of the software already written. Information about the software can be found here: forums.parallax.com/discussion/166415/c-processing-windows-bitmaps-or-dibs-with-the-propeller/p1
Using a raster image file of a PCB and a 405nm laser, I intend to transfer the intended image onto a PCB, which will be covered with Laser Direct Imaging film. A variety of people have made attempts to accomplish this task, utilizing low dollar machines, with some of them, having much better success than others.
As I see it, my best option is to utilize 0.9 degree stepper motors for accuracy, and at which point I can shoot for 600, 1200, or 2400 dpi. My goal is to build a machine, which can easily expose boards having 3.2 in. X 4.0 in. dimensions, which basically matches the maximum dimension available in the freeware of EAGLE CAD. Now in EAGLE CAD, the highest resolution of an image which can be exported is 2400 dpi.
In most instances, whether it be a laser printer or an LDI machine, the laser beam will be spread across the surface of the target, by the use of a polygon mirror. Considering the complexity of this task, I have opted to expose pixel by pixel, however I foresee a potential for gaps within the exposure, because of the starting and stopping of a small round dot, instead of a constantly moving dot. My proposed solution to this problem is dot overlap. For example, let's say that I export an image at 600 dpi, but I actually build the machine to run at 2400 dpi which would give me 4 dots across the pixel and 4 dots down the pixel, so instead of having one dot per pixel, it would have 16, and the dots would overlap from one pixel to another. The ends of a line and the sides of a line might suffer a little because they will not have the same exposure as the center of the pixel, but perhaps with all the bleeding light, it might not make a bit of difference.
Anyhow, that is how I am starting this discussion, if anyone cares to participate.
Bruce
I am currently in the process of designing and building two new machines, one of them being a rotary plotter and the other being a Laser Direct Imager for PCBs, with the focus of this discussion being on the latter of the two.
I have a large portion of the design and mechanical operation in my head, and I have a large portion of the software already written. Information about the software can be found here: forums.parallax.com/discussion/166415/c-processing-windows-bitmaps-or-dibs-with-the-propeller/p1
Using a raster image file of a PCB and a 405nm laser, I intend to transfer the intended image onto a PCB, which will be covered with Laser Direct Imaging film. A variety of people have made attempts to accomplish this task, utilizing low dollar machines, with some of them, having much better success than others.
As I see it, my best option is to utilize 0.9 degree stepper motors for accuracy, and at which point I can shoot for 600, 1200, or 2400 dpi. My goal is to build a machine, which can easily expose boards having 3.2 in. X 4.0 in. dimensions, which basically matches the maximum dimension available in the freeware of EAGLE CAD. Now in EAGLE CAD, the highest resolution of an image which can be exported is 2400 dpi.
In most instances, whether it be a laser printer or an LDI machine, the laser beam will be spread across the surface of the target, by the use of a polygon mirror. Considering the complexity of this task, I have opted to expose pixel by pixel, however I foresee a potential for gaps within the exposure, because of the starting and stopping of a small round dot, instead of a constantly moving dot. My proposed solution to this problem is dot overlap. For example, let's say that I export an image at 600 dpi, but I actually build the machine to run at 2400 dpi which would give me 4 dots across the pixel and 4 dots down the pixel, so instead of having one dot per pixel, it would have 16, and the dots would overlap from one pixel to another. The ends of a line and the sides of a line might suffer a little because they will not have the same exposure as the center of the pixel, but perhaps with all the bleeding light, it might not make a bit of difference.
Anyhow, that is how I am starting this discussion, if anyone cares to participate.
Bruce
Comments
If you really must rasterize, the first thing I would try is a machine resolution of about 3-5 times finer than the dot size. That should provide plenty of overlap with minimal futzing around required, beyond maybe adjusting trace widths to subtract one dot diameter.
Yes, I believe the RS-274X format is for vector type operations, but I believe that is the old way of doing things. With LDI, I believe most boards are now processed with raster images, however I could be wrong. As I mentioned in the first post, by utilizing a rotating polygon mirror, these complex LDI machines can process blanks spots with no problem. Just imagine a laser printer, because it is the same principle, and laser printers do not print using vectors, but rather line by line with a polygon mirror, advancing the page down Y, with no need for X.
The electronic knowledge required for polygon mirror processing is way above my intellect, and as I have said several times before, the Propeller just does not have the open source code available to do serious CNC work or process the RS-274X format. Even though it may take longer to process boards, I believe this method will work, and I believe it can be done with the Propeller, without complex code samples, containing all the necessary algorithms.
Heck, I may not even need overlap, but I am guessing that I will, in order to get some fine detail. In order to achieve overlapping dots, besides altering the image file format that I intend to use, I suppose the only other critical item would be laser focus and microstepping the stepper motors.
Laser focus is a major issue, because at this point, I am unaware of just how fine a dot I can acquire. However I will initially be using a TO-18 405nm laser diode and optics salvaged from a Blu-Ray DVD player. Blu-Ray has the ability of reading a track pitch of 32uM, so depending on my ability to focus the salvaged optics, I should be able to achieve a pretty small dot. Between the salvaged optics, 0.9 step angle stepper motors, and microstepping drives, I should really have some tools to experiment with.
You are asking for comments, so here are some.
I am making some assumptions in your design, which are that the laser dot will be "scanned" in an X (or Y) direction across the surface of the film by a spinning mirror, and that the film will be moved in the other axis by stepping it with another motor. And you are targeting 2400 DPI.
I fear you will be horribly disappointed with this approach. The length that the path the laser beam needs to traverse will vary will vary in a non linear manner as the mirror spins, making the timing control for firing the laser quite complex at high speeds.
A laser diode from a CD reader has a constant distance to its target surface and hence can focus on the "sweet spot" To expect such a diode to be focused to a dot of half a thou, at angular distances varying from say 3 inches through 4 or 5 inches will be near impossible for "home built". In fact you will not even be able to make the mechanical motion of the film to be half a thou, even with microstepping a 0.9 degree motor.
Sure, the theoretical calculation might work out, but in practice that is not achievable without access to a very professional machine shop.
While I applaud your novel thinking, you always seem to skim over and dismiss the salient details. Before spending too much effort on this approach, I suggest you try a simple test by attaching a flat mirror to a rotating motor shaft, without bothering with the other axis for moving the film, and see what results you really get onto actual film.
I think you're on a fool's errand, but that's Just my view of things, and I'd be really happy for you if you would prove me wrong.
Cheers,
Peter (pjv)
I do not intend to use any mirror.
The laser will be moving back and forth across the X axis and the photosensitive film coated PCB will be moving upward on the Y axis. The PCB image begins as a bitmap with scan line rows of white or black pixels. In theory, each pixel of a scan line row is basically exposed individually until the scan line row is complete. When a scan line row is completed, the PCB is moved upward on axis Y by one scan line, and then that row is basically exposed one pixel at a time.
However, as I stated, if I go one pixel at a time, there are bound to be gaps. So in theory, I would image that by taking numerous passes in small increments, I should be able to fill all the gaps, with just a little bit less exposure on the ends of a line and the sides of a line, but I think it will come out nice with practice and experimentation.
Many many years ago I did some direct exposure of film for PCB work using an HP plotter with the pen replaced by a plastic filament fishing line driven by a diode. The plotter moved the film and the horizontal moved the "light". Nowadays one could probably do somewhat better with fiber instead of fishing tackle.
It worked, but probably would not cut it today with the really fine pitch 0.01 lines. Our "dot" was a tad bigger than the diameter of the line. I can't see that working at 2400 DPI either.
Bruce, for those accuracies you will have a bear of a time with the mechanics, even with 0.9 degree steps and microstepping. Microsteps are typically rated at 5% precision (under ideal lab conditions no doubt), and microstepping beyond 8 is pointless.
So, good luck.
Cheers,
Peter (pjv)
Between research, machining, and programming, I probably have about two weeks wrapped up into it. I would imagine there is probably another three weeks worth of time investment, just to see if it is worth pursuing. If it fails, well that is just part of experimenting and hoping for the best.
My intention is to keep the mechanics as tight as possible, because someone else who has been experimenting on a similar project has stated that they had to keep the laser in a fixed position and move the board both X and Y just to get any decent results. However, I am hoping that there mechanics were a little on the loose side.
The project may be a complete failure, but I need to know one way or another.
Peter (pjv)
When I first began this project, I wasted quite a bit of time thinking about the use of lead screws, but then I came to the realization that belts were undoubtedly the only way to go. With the proper timing pulleys, 1200 dpi should be a breeze, and it will be much faster than using lead screws.
I have a lot of faith in the optics of the Blu-Ray player, I just don't have a lot of faith in my ability to focus
I am currently in the process of fabricating the holder for the laser diode and optics. Between the holder and the carriage, I estimate a total weight of about 0.2 lbs. The lens will be manually adjustable by a set screw, and I am hoping that it will be rigid enough to not allow vibration during carriage movement.
Hopefully by the end of the day, or perhaps tomorrow, I will be able to test my Blu-Ray laser
I'm sure you are aware that repeatability and accuracy are totally different animals. A device can be repeatable to a thou, and yet be off by a quarter inch. Just an exaggeration here.
If the hole in your pulley is machined and mounted non-concentric by even half a thou (impossible to achieve without really good equipment) you have already blown the 1200 dpi spec.
This does not mean that the result is still not useful, but one should not believe that the accuracy is what cursory calculations or measurements might indicate.
Measuring anything accurately beyond the capacity of a caliper is really hard.
In some high accuracy equipment (the mill we have is certified to measure and position to one tenth of a thou although I can't prove it) the linear position is laser measured and calibrated into a correction table. This of course can be done for any system, but then you do need that precision laser.
Cheers,
Peter (pjv)
You bring up a very good point, and I must admit that I never thought of that before, but I will have to remember that.
Considering all the possible belt and pulley combinations, with their various pitches, I have yet to find a combination that gives 600 dpi.
All things considered, it never was my intention to add gearing to achieve 600 dpi, and I will not add gearing to achieve 600 dpi.
Here are some known facts for direct belt drive:
1. EAGLE will allow you to set the dpi of an exported bitmap image to a value of 2400 dpi or less.
2. To achieve 600 dpi with a motor shaft of 5MM (NEMA 17), you will need a 0.9 degree step angle stepper motor and a driver driven at least in half step mode, otherwise you will not find a pulley.
Before going any further, let me just say that I used this calculator, to determine my dpi with the various belt and pulley combinations, as compared to determining the outside diameter of the belt and pulley combination, as originally intended. So it is possible that one of the combinations that I checked, may have actually worked for 600 dpi, if measuring the outside diameter of the belt/pulley combination. prusaprinters.org/calculator/#stepspermmbelt
However, I will say that some of the combinations came very close, for example:
Utilizing a GT2 belt with a 2mm pitch, in combination with a 15 tooth pulley, should theoretically put me right at 677.4 dpi, driven by a 0.9 step motor in half step mode
and
Utilizing a GT2 belt with a 2mm pitch, in combination with a 16 tooth pulley, should theoretically put me right at 635 dpi, driven by a 0.9 step motor in half step mode
There are other similar variations, with other types of belts, but no perfect match for 600 dpi, at least with the calculator linked to above.
Of course, if I have to I could easily set the EAGLE exports to multiples of 635 and the same with the machine, but I would never want to go above 600 dpi driven by a 0.9 step motor in half step mode, because then I could never take advantage of the resolution provided EAGLE and microstepping.
As long as I stay at or below 600 dpi in half step mode, when determining my pulley size, then I can increase the resolution through microstepping and a higher resolution of the exported image.
Anyhow, over the years, I must have salvaged parts from about 20 different printers and scanners, but I never kept the drive trains all in tact, so I am now scrambling to find the perfect belt and pulley combination for a true 600 dpi machine.
Does anyone know of the magic combination?
Edit: just saw the "I can use microstepping to..." part of your post. Ignore me.
I invoke the "Brucey was being stupid retraction rule"
Somewhere in the forum, in my discussions pertaining to LDI, I said that I had given up on the idea of using lead screws, because it would take to long for processing. Well.... I may have abandoned ship too soon. Earlier in my research, I discovered that a 5/16 -18 thread could achieve 600 dpi, but only at 6 steps per pixel when using a 1.8 degree stepper motor in FULL step mode. I gave up on the idea of using lead screws, because I thought it would take too long to process the boards and taking six steps for every pixel sounded a little iffy.
As I was drinking my coffee this more morning, a thought occurred to me.... What if I used gearing on a lead screw? So I began to thoroughly investigate and here are my findings.
600 dpi
pixel width: 0.0016666666666666668 in.
lead screw with 5/16-18 thread
travels 0.055555555555555552 in. per revolution
1.8 degree stepper makes 0.00027777777777777778 in. travel per step
1 inch / 0.00027777777777777778 step increment = 3600 step increments
Utilizing a 6:1 gear ratio
3600 (step increments of 0.00027777777777777778) / 600 (1.8 degree steps) = 600 dpi
So in other words, when a 1.8 degree stepper motor makes one FULL step on a 5/16-18 lead screw and utilizing a 6:1 gear ratio, the lead screw would advance (6 * 0.00027777777777777778 = 0.0016666666666666666) 1 pixel.
With 1.8 step angle stepper motor and a 5/16 - 18 thread lead screw, with a 6:1 gear ratio between the stepper and the lead screw:
Stepper driver in HALF step mode 1200 steps = 1200 dpi
Stepper driver in 1/4 step mode 2400 steps = 2400 dpi
Stepper driver in 1/8 step mode 4800 steps = 4800 dpi
So there you have it. As long as I am willing to eat my earlier words, my problem of 600 dpi is resolved.
I am sure you did an extensive google search about the topic.
When I did it, I found a few others that tried and after some experience gave up ...
mayb you can learn from them ...
Yes, I did an extensive search on the subject.
I found one that succeeded and he is now selling his machine.
I did not find anyone who has attempted the same approach that I am attempting.
Haven't found a real world use for this yet, but maybe this might be one...
I suppose I could buy blank circuit boards with photoresist, use bmp output of CAD and laser draw the traces...
Dip circuit board in the etch and I'd have a board. With no holes. Suppose would be cheap and fast for 1 sided board...
I can see many calcs around DPI, but I've not seen mention of the writing energy, and the valid exposure rates ?
Issues I can foresee would be
a) At very high speeds, the exposure energy will not be enough.
b) at low speeds, exposure blooming may be an issue. (basically overexposure fogs the image)
These may vary with photo-vendor, and even from board to board ?
Did you do any test runs yet on under- and over- exposure limits ?
This may need some power control on the laser diode, to vary with linear speeds ?
What power does an engraving laser have ?
hmm, if you could lower the power enough for imaging use, that could work.
Holes could be at least drill-dots, which would need an apply-drill-file pass to remove imaged pixels.
Double sided becomes a careful alignment and carefully avoid-expose of inactive side problem, sounds a pain...
Or you could expose two sides at once ?
I have not done any tests at this point, but I have seen I guy take a 20mW Blue-Ray laser and basically move the laser back and forth across a photosensitive board, followed by developing and etching. Amazingly he had very fine copper traces. So I know that 20mW is enough to do the job, but to do the job properly, I am certain that the proper speed and proper laser output would have to be set appropriately. Additionally, one film manufacturer states that less problems occur with the final product, if the board is processed twice before developing.
Here is what I am going to try....
Top images aligned left to 0, 0 within a square frame
Bottom images aligned right to 0, 0 within a square frame
When exposing top side, the top edge always pressed tight against top edge of frame and then pushed left as far as it will go, meanwhile keeping the top pressed against the top edge of the frame.
When exposing bottom side, the top edge always pressed tight against top edge of frame and then pushed right as far as it will go, meanwhile keeping the top pressed against the top edge of the frame.
So basically you have two 0, 0 locations.... One in the top left corner and the other in the top right corner
I believe that will resolve any registration problems.
I'd be tempted to look at top and bottom lasers, to avoid double handling, and potentially halve the finishing raster time too...
Of course, the use of a safe light would be an absolute necessity, but that is much like working in the dark
IF....... IF I can create a machine that can produce a decent board with one laser, then perhaps I will think about running two lasers simultaneously.
I agree, that would drastically cut down on the processing time, but I don't even know if I will be successful with one laser. Quite a few people have failed with just one laser.
Excellent idea, but instead of top and bottom lasers it might be better to place the board vertically and have left/right lasers. Makes for simpler board placement and mechanics as well as taking up less of the work area.
EDIT: Of course that is just with one laser diode being utilized for processing. Time could be reduced, by using multiple diodes.
Seems to me, producing masks and a good process and fixture for registration would be faster and easier.
I visited the Other Mill people a while back. Those things are sweet, but small area. But fast!
A precision mill mechanism may be less effort. And, it doesn't take much for a on. Make sure it steps at .000x and small mills, .01" are possible to use.
Just saying.
I don't intend to invalidate anything. But, this could be a lot of effort for marginal returns.