I am able to make a binary on the ESP32 which can program the FPGA using Fomuflash and initialize the stepper motors via the TMCStepper library. The controller.py class of Hexastorm can be imported in Micropython. Procedure to make binary is here.
Most people here are probably waiting for an exposure with a PCB motor.
I decided, however, to focus on finalizing the PCBs and porting the software to the ESP32.
I created a main board. The board is built around an ESP32-S3-WROOM-2. I use a USB C connector, but pull 12 volt from a different source.
I removed the micro-SD card, I used in an earlier design. The ESP32-S3-WROOM-2 has 32 MB onboard memory which should suffice. Steps for the stepper motor are generated via the FPGA from the laserhead. The laserhead connector moved from 15 pins to 20 pins.
The main board is shown below
The laser head did not really change. I decided to make the screws optional and the plates can be soldered together via pads. I optimized the tracks a bit an moved around some components.
I made a new PCB motor with the code from atomic14. This uses 12 coils and should work better than the archimedean spirals.
I spoke to a couple of mechanical engineers about my sliding bearing. They did not see any issues with the bearing and thought it would be "very" accurate because it was made on a rotary mill. I aim to do measurement on its flatness at a later stage.
My goal now is to finish the code in Micropython for the ESP32. I hope I can find my earlier prototype which must be somewhere :-).
Laser scanning via mirrors is of relevance to prism scanning, being an alternative. For mirrors it is assumed angle in is angle out (law of reflection). Still, this is not always valid. The following article on arxiv goes into depth on the latteral shift and angular shift known as goos hanchen effect and imbert federov effect. It is seems to be simple consequence of Fresnel equations.
Aberration is small but you would have to account for distance to substrate and also compound both effect. Details in article
Last july, an article was published in Nature for a new high viscosity vat printing method, https://doi.org/10.1038/s41467-023-39913-4 This method is depicted below, the figure is copied from the article.
It uses four rollers to apply a new layer to 3d print a part. The article comes with some cool videos and equations describing the mechanism.
The method, and this is not mentioned in Nature is not entirely new, see patent EP2272653A1. The only difference i can see from a legal perspective is recoater number 2. I also remember it is not a true vat method. Resin went to a waste box, see number 23. I think the Admatec machine is a spinoff from this concept. It also uses a a rotating foil, coated with a blade and the remainder is put into a left over box.
Furthermore, in my white paper on reprap, i claim the usage of a foil with my Prism scanner. I made two drawings to protect up and down projection.
I think the authors of the nature article should consider the following two patents; US8777602B2 (recoater patent)
Loophole might be not using recoater.
US9939633B2 (scanlab, EOS subsidiary patent, reflecting lens)
Their reflecting lens seems very similar to scanlab. The authors mention the 3SP patent, so I did not include it.
I use a sliding bearing. This keeps costs low and ensures the prism is kept in a given plane. A challenge is that the prism falls if this sliding bearing is flipped 180 degrees. This is fixed by adding a ferrite sheet on the back of my PCB.
Materials used flexible ferrite sheet MHLL5040-200 Laird, Magnets MKSA-8x5-ZW-N45 i.e. 8x5 mm, neodium 45 with a black epoxy coating. I think that ferrite instead of neodium is possible if the whole contraption was made better. Ferrite is typically 2-7 times less strong. If the distance would be decreased a bit, which is still possible, ferrite should work.
I made a presentation on how laser prism scanning might use Bessel like beams. I outline how they are created and how they might be used in combination with laser prism scanning. This due to recent interest to startups like Inphocal.
- I claim the use of two lenses which in combination focus light into a structured optical beam. Between these lenses there is a prism which can rotate so that a scanline can be formed. The first lens focuses the light parallel to the scanline and the second lens focuses the light orthogonal to the scanline.
Claim 2:
- The optical system of claim one, wherein the focal length of both lenses are different such that they circularize the light source.
Claim 3:
- The optical system of claim one, wherein each of these lenses is similar to an axicon. The difference with a normal axicon, is that each individual axicon focuses the light in one dimension and not two as a normal axicon. The light is just propagated and not diverted along the other axis.
Only the combination focuses the light along two axes, one orthogonal and one parallel to the scanline.
In the laser prism scanner, so far Gaussian beams are used. An alternative would be to use a Bessel beam. An advantage of a Bessel beam is that they have a longer depth of field as compared to a Gaussian beams for a given focus size. A true Bessel beam is non diffracting. True Bessel-beams are not possible as they require infinite energy. Bessel beams can be approximated and then called structured light. These structured light beams are less diffractive than Gaussian beams.
A Gaussian beam is created by focusing light into a point. Structured light is created by focusing light along a line. This is denoted by Bessel beam in the figure below.
Recently, companies such as ASML, Europe's most valuable tech firm according to BBC, started to support laserscan startups such as Inphocal which use structured light. Inphocal is based on intellectual property from Cern (WO2019211391A1)
Inphocal received funding from HighTechXL, NWO, DeeptechXL and the European Union.
CERNs contributions to Kicad are highly appreciated. WO2019211391A1 feels incomplete.
In short it outlines an improved version of the axicon. This new lens produces a beam which better approximates a Bessel beam and diffracts less.
I can understand this due to the greater design freedom as the shape seems mathematically more general to me, but still I would be interested to see the actual differences backed up by optical simulations. The authors from CERN do not provide this.
CERN furthermore does not really goes into the difficulties which arise by scanning a laser bundle. I can imagine Inphocal uses a galvo scanner from scanlabs and does not have a challenge with cross scan errors. Galvo scanner are slower than polygon scanners.
In specific, the authors do not outline how the movement orthogonal to the scanline can be mitigated if a Bessel beam is scanned by a prism. Given my experience with cylindrical lenses and Gaussian beams. I propose a similar solution for Bessel beams.
For Gaussian beams, my solution is outlined below;
Light diverges from a laser diode and is collimated by an aspherical lens. The first cylindrical lens focuses the light parallel to the scanline and the second cylinder lens focuses the light orthogonal to the scan lens. As the cylindrical lenses have a different focal length. The elliptical laser spot is circularized. Furthermore, the second lens removes part of the cross scan errors created by not mounting the prism orthogonal to the rotation axis.
Light will still be parallel after leaving the prism if the prism is perfect and as such be focused into the same point.
Perfect means here that opposing faces of the prism are perfectly parallel. If this where not the case and it would be tapered, the outgoing rays would not be parallel to the incoming rays of the prism.
Please recap what happens. You could focus the light directly after it leaves the aspherical lens as a collimated bundle.
The lens is split into two components and each component only focus along one axis. The light is not altered along the other axis by each component.
My bet is that this trick for Gaussian beams can also be repeated for Bessel beams.
For Bessel beams we repeat this trick, here the cylindrical lenses are replaced by a pair of axicons. The difference being that normal axicons focus light along two axes where i focus light along one axis, similar to the Gaussian beams.
Here again the focal length can be chosen so that the beam is circularized. If the first lens focuses light along the scanline and the second orthogonal to the scanline, part of the cross scan error can be removed.
This can again be repeated for the lens outlined by CERN in WO2019211391A1. This lens is split it into two lenses and a prism is placed between the lens. Again the beam can be circularized and the cross scan error can be removed. The patent from CERN is pending and its final claims uncertain. I added a presentation on the topic.
Hexastorm
It uses four rollers to apply a new layer to 3d print a part. The article comes with some cool videos and equations describing the mechanism.
