At its core, LDGraphy is a switched laser, a rotating mirror that moves the dot in a scan-line and a sled that moves the photoresist-coated PCB forward.
Laser scanning is very simple and cheap to do with common components these days. Lasers with a wavelength of 405nm (which is close to the peak-sensitivy of dry-film photoresist) can be found cheaply with around 500mW power.
As for scanning, there are several techniques. A common one is to have a mirror that rotates to deflect a laser. Since it would take a while for a mirror to turn around full 360° after a segment is scanned, they come as a polygon mirror: a polygon that has mirror-reflective edges. A typical mirror that is used in Laser printers 6 sides; you notice that the actual mirror face is not that high - but we only need to deflect a laser dot.
| Top | Side |
|---|---|
 |
 |
With this hexagon, we can project a 120° arc (each mirror face is 60°, so the angle between incident ray and reflected ray is 120°).
Here, we just shine (a fast blinking) laser at an angle from the top, it is
reflected at the rotating mirror and projected downwards.
We can make a scanning device with this simple set-up if we correct for the arc distortion in software; in fact, in an early experiment for LDGraphy I was doing exactly that:
There are advantages for this set-up: we can utilize a large fraction of the laser on-time by covering the majority of the available scan range. The scanning laser is "almost" in focus for the scan range where it hits the board.
This is a simulation of such a scanning in which we use 105° out of 120° (the laser is off only for 15 degrees). We can see that the laser dot is oval in the extreme angles but also even in the center. This is because the actual focal plane of the laser (indicated with a gray disc at the sharpest focus) is hitting the surface at an angle, resulting in an oval exposure.
This set-up is very simple to build, but the oval exposure dot is an issue. Also it is very sensitive to angle variations: just small changes in the shallow angle moves the projected arc-line forward and backwards.
This could be fixed, if we had a mirror that points the laser-arc downwards, so that it hits the resist perpendicularly and is also always in focus. The mirror would need to be a circular segment with a downwards angle -- essentially a segment of an inner slice of a cone. Now we hit the photo resist always at a sharp focus point, even for the wide scan angle of 105 degrees as the geometry keeps the focal plane of the laser exactly on the exposed surface.
However, making custom mirrors in such a shape at home is not easy.
Possible ways:
- With a lathe: this would essentially be a 45° bevel on the inside of a large diameter stainless steel pipe with a wall thickness of about 2-3mm. Then mirror finish polishing this bevel. Stainless steel or aluminum are both nicely polisheable, so these could be materials of choice.
- John M. suggests on his blog to build such a mirror using laser cut parts that, snapped together, defining the shape.
We want to build this machine with minimal accessible tools, so these techniques have to be explored if they can be done with minimal tools. Till then, let's see what else we can do.
We can simplify and use a straight mirror to project downwards to get a line. However, the focus point then is of course not always in the plane of the photo sensitive material. In particular if we cover a large scan angle, the focus variants are significant.
The exposure point now is very large at the fringes. We could alleviate some of it by "plunging" the focus plane below the surface at the center, but it will still be too much of a variant.
The advantages for the home-build is, that this works with a very simple straight mirror (easy to come by with) and does not have the problem of being sensitive to the shallow angle variations like the angled direct arc exposure.
The issue though is the focus, even more pronounced than in the direct arc scan.
In professional scanning applications, a so-called f-theta lens is used to keep a straightly projected laser dot always focused to a plane.
If you take apart a laser printer, you will see a lens set in the laser assembly: that is the f-theta lens.
However, this is also somewhat out of reach for the hobbyist to make (though I still consider to try making one with transparent acrylic). It also does not seem to be available as off-the-shelf part like the Polygon mirrors and the laser does (you can get them, but they are very expensive).
If you find some, it is probably meant for a laser printer and optimized for infrared light not UV light.
What else can we do ?
If we choose to do the straight projection, but only use the 'center piece' and limit ourselves to a smaller angle, the focus variation is not that much and we get an acceptable compromise in exposure dot size.
We don't need any special mirror or lens. The disadvantage is, that we only can use part of the time of the laser: if we use 40° scanning out of 120°, we only use the laser for 1/3 of the time (you see the laser is dark for a long period in the illustration). Also, we need much more space as we need to be further away to cover the same PCB width.
Let's review our options
-
Angled direct arc exposure mirror:
- 👍 fewest parts needed; almost full use of the laser; No light losses.
- 👎 focus plane intersects with exposed board at an angle creating oval dots; susceptible to small angle changes; large build.
-
Cone mirror arc exposure
- 👍 almost full use of laser; no light losses; hitting resist perpendicularly; always in focus; very compact.
- 👎 Requires to build a custom shaped mirror.
-
Straight mirror line exposure with large throw
- 👍 easy to build; very compact.
- 👎 very noticeable focus issues in far angels.
-
F-Theta lens corrected line exposure
- 👍 Almost full use of the laser; hitting resist perpendicularly; always in focus.
- 👎 Requires to build a custom shaped lens; might have optical losses in the material.
-
Straight mirror line exposure with small throw
- 👍 easy to build; focus issues less pronounced than with large throw.
- 👎 need longer exposure time; large build.
Overall, the most promising is the cone mirror solution: it creates sharp focus points, makes good use of the laser time and building a custom mirror is somewhat more feasible than building a f-theta lens.
However, since ease-of-build for everyone with limited access to tools is a priority, we start with the small throw line exposure for this first version of LDGraphy - it doesn't require any manufacturing of a mirror and is a good compromise.
This is how the current LDGraphy device looks like. From the top, we can see all relevant components for the laser scanning:
| Drawing | Reality |
|---|---|
 See full drawing here |
 |
To keep things more compact, the laser is mounted in one corner and the light is redirected onto the polygon mirror. But other than that, this is exactly as discussed above. There is a long mirror on the side that is pointing downwards, a little hard to see from above, so lets look from the back, with some rough indication of the laser path:
When the polygon mirror is rotating, the laser point passes through a slit onto the surface to be exposed.
On the left, you see a fluorescing piece of material that has a photo diode attached. This is our horizontal sync (HSync) to detect when a laser line starts.
There is an issue that I discovered with some cheap polygon mirrors: the six mirror faces have a little variation and not being entirely parallel to the rotating axis, so project lines up and down a little, depending on the face. So what needs to be done is to have another sensor to count mirror faces, and then correct for this after a calibration step in software.
The sled is moved with a stepper motor and a threaded rod. This rod does not need to be CNC quality ball-screw, a simple screw from the hardware store is sufficient. Why ? We only need to move in one direction, so we don't care about backlash issues.
Here is how the base looks like, housing the motor and the threaded rod. The rod is 'mounted' to the shaft of the motor with a heatshrink tubing (and a little expoxy butting these together). A nut is not directly mounted to the sled but to a slab of acrylic. You also see both the limit-switches:
| Just drive | .. with sled |
|---|---|
 |
 |
This allows us to have a separate sled that then sits on top of the moving slab of acrylic. The design detail here is, that it allows to drive the sled forward and backward, but has freedom to move in other directions: this can be used to essentially be immune to 'wobble' from the cheap threaded rod. In the current test version of LDGraphy, I have a horribly wobbling rod that I found in a corner of my workshop.
You see the wobble in the video 🎥; It still produces acceptable quality
Everything is controlled by a BeagleBone Black or Green. It has the nice property that it is a common ARM machine running Linux (so offers the comfortable environment of a full operating system including networking and can do all the work like converting Gerbers and do image processing), but it also has a built-in Programmable Realtime Unit (PRU) - a fast (200Mhz), independent microcontroller essentially to help do the timing-accurate parts interfacing the hardware.
We need this to accurately generate the pulses to control the polygon mirror, switch on the Laser at the right times (we have about 2Mhz pixel clock), and generally keep track of where the laser is. And before you ask: of course, this could be done with an external Cortex M4 or so, connected via SPI to some other computer (it would be more expensive of course, but it works). I just had the BeagleBone lying around and had used the PRU in a previous project.
The laser requires a current source, in our case it needs to be switched in the low Mhz range. Some lasers (that you get from e.g. Aliexepress) already have a "TTL driver", but they are utterly low frequency (50ish kHz), so this is why this project needed its own laser current driver (not perfect yet, but works).
Right now, all the electronics is attached to the outside of the case, for easy hacking and scope-probe access. Eventually, all these parts will go inside the case.
