here's another posting to let you know how things are going.
We now have all the parts made for the full batch, including the electronics, cases and steelwork, and these are now ready for assembly. To optimise the assembly process we've broken the scanner into several sub-assemblies which can be built in parallel, and a number of these are ongoing as we speak including all the scanning heads (the most complex part of our assembly, consisting of cameras, PCBs and printed masks).
We had planned to start shipping last week, but during testing found that there was an issue with the paint used on the scanner cases. We've worked out the cause and the fix for this, but this has held us up a little longer. It's possible that we'll have these first few units assembled this week and ready for shipping, but with factories in the UK shutting down for Christmas at the end of this week it's not certain.
Software-wise we've got the high resolution scanning working well and have decided to add in a 'single shot' mode too. This allows you to capture a single 3D image, meaning that you can get a snapshot of any object up close at the highest resolution. You can take multiple snapshots too and if required these can be merged afterwards using third party software (such as the wonderful and free Meshlab). We think this will be a really useful feature for many of you who have slightly off-beat requirements or just want to experiment to find the best settings for your project.
And the high resolution data? Well, here's a comparison of low resolution and high resolution scans of the same object, a 3D printed bust.
In these snapshots you can see the sparser low resolution data on the upper left and right sides and the much denser high resolution data on the lower left and right sides, as well as the main image. The low resolution can have up to 480,000 colour vertices (3D points) per snapshot, while the high resolution can have over 5 million.
We plan to have another update at the end of this week and, in the meantime, hope this one is again of interest.
We had a few hitches last week relating to EMC which meant we had to source a few additional components to improve the RF shielding, but these have now been addressed. We're now moving on with the manufacture of the first few units, and you can see a picture of the completed housings for these below.
These are now at the assembly facility and the first few units should be ready next week, with the rest to follow over the coming weeks.
The system software is also making good progress, and while the main focus is raw scanning performance we're also making a few small optimisations to improve the system. One of these has been to speed up data acquisition, and we've reduced the capture time to only a few minutes for a full low resolution scan. We've also addressed some memory management issues with the high resolution data capture and processing (we have a lot of very large colour images that need manipulating efficiently - each one is around 10mb) and have sped things up here considerably too.
Meanwhile here's a preview of the software interface which we're keeping as simple as possible for now. The shot on the left shows a preview of the object being captured, while the shot on the right shows the user settings.
Finally, it's worth reiterating that this isn't a build to print design but the very first version of a product built around completely new technology. While we are doing our very best to deliver this as planned this does mean that unforeseen issues occasionally arise, usually requiring additional work to resolve them. Every one of you here has been brilliantly understanding, and we truly appreciate this. It's your support that keeps us going, and we're now almost there.
a short update on progress with manufacture.
Last week we completed a pilot build and this week will be at the factory building a small run of the first scanners. The aim is to optimise the assembly process prior to the building the majority of pieces the following week. We've had a few last minute hitches but these are mostly in hand.
We'll post some pictures from the factory once full production starts, but here's what the tooling for the outer case looks like in the meantime.
While the units are being built we'll be finalising the software and performing quality testing of the full system. This is absolutely necessary to ensure you receive a fully working product, and while the units may be physically built we won't start shipping until the complete system is ready and fully tested.
Each scanner is being shipped directly from the factory, using UPS in most cases, so once we're ready we can get the scanner to you very quickly (next day for UK deliveries, 3-4 days for international shipping). We'll send you a tracking number once your scanner is ready and we'll keep you notified of progress through some further updates.
So, we're almost there and are ready for the final push!
this update will concentrate on the invention at the heart of the Cubik 3D Scanner, the DepthPhase® scanning technology.
A little bit of background first. We setup CADScan nearly three years ago with the idea of bringing a low-cost, easy-to-use desktop scanner to the market. Back then we hadn't settled on any particular technology although it was clear to us that structured light methods that used pattern projection had numerous advantages over other approaches as a result of their ability to quickly build up sets of overlapping depth images. It was also clear that there were barriers too, though, the main one being the need for a decent projector, its associated expense and inflexibility.
So we set about trying to design out the expensive projector, looking at different ways to illuminate an object with a pattern using simple, readily available components. Around about two years ago we came up with a new way of doing this, using a camera, a set of LEDs and a pattern printed on transparent film. Of course, the maths and geometry worked great in theory but developing a practical, real-world version was to take us some time and involved many extensions to the basic idea to make it work. We managed to develop a working prototype with the support of the UK Technology Strategy Board which led us to the point where we were able to engage with Kickstarter - and all of you good people - to develop this into a product.
You've already seen the images of the DepthPhase® scanning heads themselves (update #16), but what about the output? Well, the good news is that we managed to address the issues with the latest projector design and were able to show the first results at the 3D Print Show last week. This is what we'll present here too with imagery from the pre-production unit.
The first stage of the system is to project a pattern onto the object, so that you get a set of images similar to the one below. By illuminating different LEDs we change the position of pattern so that each image is a little different from the others and this gives us the information we need to build a 3d image.
We then process this set of images to determine the amount by which every pixel has changed, giving us a phase shift map for the entire scene viewed by the camera. You can see the raw output from this process in the next image. Note that this is a low resolution capture (800 x 600 pixels) with no pre- or post- smoothing. You'll also see that because we don't use triangulation to calculate the depth, measurements are available for every pixel in the scene. Using triangulation requires an offset between the light source (laser or projector) and the camera, leading to some parts of the object being occluded.
As the name suggests, the amount of phase shift is a function of distance, and we can then generate a depth map of the scene. By moving the object on the turntable we can do this from many different viewpoints, giving us a set of overlapping depth images. This is very useful because it gives us multiple measurements of the same parts of the object illuminated from different positions, reducing errors associated with surface reflections. The more images that are collected the richer the data and the better the output. Here's a set of 10 turns, but you can have more (for better quality) or fewer (for speed) - this set took about 10 minutes to capture and process.
Once we have these depth maps we're then able to combine these into a fully meshed point cloud. The overlapping data allows us to use a probabilistic method to select the best data at each point on the surface from however many samples are available. The key to accurate reconstruction is having high-quality raw data from the outset, of course, and this is where we have concentrated our effort.
We're finalising the porting of the full reconstruction code into the new platform-independent framework, and will show some downloadable examples from the production code in the next update. In the meantime here are a few snapshots of some 3D scans of single face point clouds (we've turned them to show some depth).
You'll note that the point data is very dense. In fact the cameras have three resolution settings, with the highest mode being 3.24 times higher than these images (2592 x 1944 pixels). This increases the depth resolution as well as providing much more surface detail. In the software you'll receive there are a few options you can set to trade off quality and speed. In addition to choosing low, medium or high resolution, you can also select the number of turns and whether or not to apply noise reduction. A simple point cloud / mesh viewer is also included and you can save the data in several portable formats. We'll cover this in another update along with some screenshots from the final release.
Well, we hope this rather long and detailed update has been of interest. We've had great feedback on the progress so far and would like to thank you once more for your continued support.
we'd like to thank everyone we met at the 3D Print Show over the last few days for the interesting discussions we had and the great feedback on the scanner. Here's a picture of the stand and the new case (this one a prototype) for those who missed out.
This week has already seen the delivery of the tooling for the case to the manufacturer and our main focus is still on getting units delivered. We've also got some great colour scans taken at the show from the new scanning heads which we'll be posting shortly, along with a further update about the software. Thanks once more for your patience!