So, I´ve been getting a lot of PMs and Mails about suitable Cameras for Multitouch Setups.
I had a chance to test some through my University, so I would like to share my findings.
1. Phillips SPC 900NC
The “beginner” Camera.
Specs:
- USB Connection
- Winodws only (more on that later)
- 640×480 resolution max, 90fps max
- Sells for about 58 Euro
Removing the stock Lens:
Removing the lens is really easy, just remove the silver ring that is attached to the Lens.
You can then insert a new, M12 lens without IR filter. Takes about 5min. Detailed instructions
with pictures via NUI Group Member Falcon4Ever: ->link
Performance
Other than advertised, the Phillips doensn´t deliver 90fps at a 640 resolution, it somehow
compresses the video stream, to make it appear that way. Its probably sending each frame
three times or something. Not sure exactly how, there has been quite soem discussion on the forums
about that. Setting the Camera to its highest resolution and maximum framerate uses a lot of
resources, so my fans start to rev up pretty fast. So you will be using a lot of cpu power, which
doesn´t fell right for me. IR sensitivity is good, and it should be enough for most setups, I believe.
2. Unibrain Fire-i
Specs:
- Firewire interface
- 30fps at 640x 480 resoltuion
- about 90 Euros (color), or 129 Euros (B/W version)
Removing the Stock Lens:
No problems here, because you can just screw a M12 Lens right on to it.
So now that we have finished the LED Frame, we will need some kind of
power supply inside the table. Like many other DIYs, I use a standard
PC power supply. They are cheap, available everywhere and can be used
for the LED frame and some fans as well.
Since most modern pc use a lot of power, you can really go for the cheapest,
low power model you can find. They all delivery plenty of power for my use
anyway. Maybe you even have a spare from an old pc, so you might not even
have to buy one at all.
1. This is the cheapest one I could find, still with 400W !
2. If you plug this into a socket, the power supply will atumaticaly turn
itself of, because it isn`t connected to the standard pc power interface.
So we will have to connect two wires, in order to keep this from
happening. So we will have to identify the two wires that we need…
On the image below, you can see the color coding of the wires on
this power supply. We are looking for the one that says “P-ON”, which
will be the green wire (”GRN”). We have to hook this up to a black
wire (the negative), and then the power supply will stop shutting
itself of right away…
3. Find the green wire, cut and connect
You can see the green wire, right on the huge connectorof the power
supply. Now cut the green and one black wire (doesn`t matter which
black one, they all work), remove the shielding and connect them together.
You can solder or use one of these screw connection things, as I did.
4. Identify the wires we need for the LED frame..
Since my led frame uses a 12 V current, we have to identify the wire that
can deliver 12 V. So we take another look at the color coding on the power
supply. As you can see, its not on the overview (2nd picture of post).
For this particular power supply, the current is written right onto the wire.
In my case, its the yellow one. Actually, on most power supplies I used, the
yellow one is 12V. Be sure to check on yours though! To make sure that I
got the correct current, I use a multitmeter as shown below. The yellow one
does deliver 12V, the other end goes into a black one.
Later on, we will also use this power supply for some of the fans and
tamron lens… more on that later
Undoubtedly, one of the most important parts of any MT system is the box. Because
this is not really my field, we asked a fellow industrial design firm in Zurich to assist
us in the design and construction of the box. Plans where made with my inputs, and
finally handed over to a local carpenter for the actual assembly. The box is hand crafted,
painted and delivered to us. Since this was done from an external pertner, I can not
share the exact plans or details. But I can say, that the box is exactly 70cm high, which
is the height most “normal” tables are too.
Some Images of the box, when it is still empty and ready for rigging in the technical stuf
First, the plans where made by plasmadesign, according to my measurements (projector distance etc..)
Outside of the box, notice the wooden platform on which it sits. It has some wheels attached to the
bottom, this allows my to move it around easily. Very handy when instering the hardware..
You can also see the four precut holes on the back. We will use four PC fans to make sure that it doesn`t
get to hot in there.
Top View of the box. You can already see the cutout for the projected image, and a frame on the
far side, where I will insert some speakers and more fans.
Of course, we have a door on the backside, so we can acces the insides while it is running…
I first start with the LED Frame, which is still handbuild by myself.
In retrospect, I would use SMD Leds today, probably already assembled and
ready to go in LED strips. But since I would like to make all tables exactly the
same, I still use normal IR-Leds and a custom build frame. This is probably
the obvious choice for the DIY guy, and its actually quite fun if you only do it once.
In my case this is the 5th Frame I assembled, so the fun is… well not quite as fun
anymore
Material list for the frame:
- Aluminum profiles from local hardware/Home Depot style shop (1cmx1cm)
- Osram SFH-485p Leds (my Setup uses 166 LEDs for a 64×48cm Screen)
- 15 Ohm resitors (24)
- Electrical wires, soldering iron etc…
LED Array Wizard
Using the excellent LED Array Wizard -> link
I calculate the size of resistor I will need.
The following values have to be inserted into the form:
12 = Source voltage (Volts)
1.5 = diode forward voltage (Volts)
100 = diode forward current (mA)
49 = number of LEDs in your Array
Since I use LEDs on all four sides of the Sheet, I use 49 LEDs on the longer sides
and 35 on the shorter sides. By using multiples of seven, I can use the same type
of resistor on both sides.. This is a personal choice, and when you type in the values,
the wizzard gives you some alternatives as well..
See this Screenshot of the result form the wizzard:
The Leds…
///////////////////////////////////////////////////////////////////////////////// Step 1 - Make the profile for the LEDs
Claculate the space between the Leds and mark the holes for drilling…
Drill the holes…
finished profiles..
///////////////////////////////////////////////////////////////////////////////// Step 2- Glue LEDs into the holes with superglue
First off, I secure the profile…
Then I use superglue and aplicate some on the LEDs and holes in the profiles. From my experience,
this does not alter the performance of the LEDs, some tables have been running like this for a year without
negative side effects..
finished profiles with all LEDs glued…
///////////////////////////////////////////////////////////////////////////////// Step 3- Soldering, wiring and resistors
I bend the legs of the LEDs so that I can directly solder the legs together..
now the legs are soldered together and the trim the excess parts..
I then start wiring up the groups at the anode (+) legs with wires..
then the cathode (-) legs, to which I solder the resistors as well….
Unfortunately, you cant see the in the picture, they are under the heat shrink tube…
Before I heat up the shrink tube, I check if the LEDs are working.
I hook up the the frames to a power source and use my mobile camera
(no, does not have to be an iPhone) and see if they are slightly glowing.
If some are not, its still easy to fix them..
In the last step, I use blow-dreyer to heat up the tube and let it shrink around..
I am in the process of building another (the fourth one) IAD MT08 Table, which uses the
DSI (Diffused Surface Illumination) setup. Since I get a lot of question regarding certain
details on hardware, I will now share some infos on how I build one.
I will start of with the LED Frame and will continue to post on other steps on getting the table
complete. Remember that this system was designed over a year ago, so that some parts/ hardware choices
will be outdated. I will comment on these and offer some ideas on how I would do certain things today.
I hope this will answer some questions that arise when building a DSI System.
The main Problem with most Diffused Illumination Setups is getting IR Light to spread evenly across
the screen surface. Normally, you use a couple of IR Illuminators as a IR source, and it takes lots of
effort to get the IR even, while also dealing with possible reflections from the Illuminators. But now
I found a material that make this a lot easier!
Instead of using Illuminators as a IR Source, DSI uses a special acylic to distribute the IR evenly across
the surface. So you basically use your standard FTIR setup with an LED Frame, and just switch to a
special acrylic, that does not frustrate the IR light from the edges. This acrylic uses small particles that
are inside the material, acting like thousands of small mirrors. So when you shine IR light into the
edges of this material, the light gets redirected and spread to the surface of the acrylic.
no need to build a case for the setup (like other DI setups)
fiducial tracking possible
Con
less contrast compared to normal DI setups, because the
plexiglass also redirects the IR towards the camera
potentially more problems with ambient IR because of less contrast
possible size restrictions because of the plexiglass softness
//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// How to make your own setup ////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
First of all, you need to get hold of the “Endlighten” plexiglas. I got my material from
a local acrylic dealer in Zurich. Since he didn´t have the material in stock, I had to order
a pretty big sheet (about 2×3m) which cost about 700 euros. I pretty muched used all
the material in the end, and my university paid for the costs (so I didn´t care..). If
you live in Germany or France, you can order directly from the “evonic” webshop,
otherwise ask you local dealer/distributer for ordering possiblities.
Evonic makes some different types of Endlighten. These vary in their thickness and also
in the amount of particles inside the material. Available thickness ranges between
6-10mm, follwing “L”, “XL” and “XXL” for particle amount. I have tested the 6mm
“L” and 10mm “XL” type. The 6 mm is to flexible for a table setup, the 10 mm works
nicely.
//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// Handling the material ////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
Because the plexiglas is softer than normal plexiglas, you really have to watch out when
working with the material. It scratches quite easily and these scratches show up on the
projecton and during camera tracking. So watch out…
//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// LED Frame
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
You can use your “normal” FTIR LED frame for this setup. My tests showed that shining
light into all four edges of the sheet works better than only through two sides. It works with
two sides, but better with four! I also tried increasing the number of leds, which makes a
brighter image, but no contrast change. So potientally, brighter blobs means better recognition
with ambient IR. For example, on two setups I used about 100 Leds for all four sides on a active
screen size of 64×48cm. This may seem like to much, and thats probably right too! You could
use less, just make sure all four sides have leds.
Like most people, I use the Osram SFH-485 and Osram 485-P. I would recommen the “P” type
because the angle is wider than the normal 485s. The wider angle helps distribute the light on
the edges better so that the distance between leds and the active screen size is kept smaller.
Normal 485´s work as well, but if you can chosse, I would recommend the “485P”.
//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// Screen vs Sheet Dimensions
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
You have to have some extra sheet material around the edges, so the ir from the leds can
spread evenly into the material. This also depends on the angle of the leds, so the 485 P need
less space (because of the wider angle), while the 485 nedd more (narrow angle). I would
recommend about 2-4 cm extra on all sides. So a active screen size of 64×48 cm would need a
sheet of 70×54 cm (with 3cm on all sides). That should be enough for both 485 and 485p leds.
These are just some numbers I found during testing, so other leds could behave differently.
//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// Other Components
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
All other components (Projectors, Cameras, mirrors IR filters, etc…) are the same as other
multitouch setups. Either check Nuigroup.com or my blog for some info. The same projection
screen material that are used for DI can be used on DSI setups. I have tested tracing paper,
lee filter, geriets optitrans and lumin contrast screen. They all work good with DSI, but prices
and image quality vary…
Download the technical information about “EndLighten” from the Evonik website: link German Evonik online shop to order EndLighten: link /// Same in French: link
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
CamCaptures and Video ////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
Some Captures from my Setup with the FireflyMV from PtGrey:
Since I have been recieving quite a few question about DI setups and how to build
them, I wanted to point everybody to a guide from the c-base group. They have build
a really nice table quite some time ago. and took the time to explain their setup
in detail. I habe the chance to meet them, during a visit in Berlin and their table is
pretty nice. So here the link: DI Tutorial by cBase
Just some difference from my setup..
1. They are using Security Glass (4mm thick) and their projection surface/diffuser is
glued to the bottom. This way, you don´t have to worry about reflections from the illuminators.
So you are touching the glass with your fingers, which actually feels pretty nice, although I
still prefer to touch the projection surface myself…
2. They use Gerriets Optitrans as a projection Material, which works really nice. I kind of
looks like rosco grey, and feels like it too. I recieved a sample yesterday and it works very well.
I would advise to get this, instead of the Lee Filter I was using before…
There are lots of discussions regarding what setup is the best.
So I tried to write up some pros/cons, regarding both setups, since
I have build both setups. Please feel free to contact me, I will add
your input to this post..
FTIR:
* blobs are nice and clear
* setup needs an led frame which you have to make yourself (soldering needed)
* only allows touches with fingers
* requires a compliant surface, so that touches also get recognized during dragging operations
I still think that FTIR is a really good way to go, We just need to find a good
compliant surface. Right now, many are using a transparent silicone rubber,
on top of the acrylic, to get good recognition while dragging. Over that goes
the projection surface (tracing paper, rosco etc).. One advantage of the
FTIR setup is that you don´t need to build a box to get it working
(while DI setups need a box..). The other advantage is that the whole IR setup
is in one piece with the surface/screen, so handling is easier when moving the setup.
FTIR sheme:
DI:
* no need for a compliant surface, just an diffuser/projection surface on top
* no need for an acrylic, could use other transparent materials like glass as well
* no frame, allows for some nice box setups..
* no soldering, because you can buy the IR-Illuminators ready to go
* simple setup, difficult to get it working right
* fingers and fiducial marker tracking possible
* blobs are not as clear and sharp, when compared to ftir
Right now, I love DI! Because the dragging is soooooo smooth. I think
this is basically the best thing about DI, because it really makes it so
much more enjoyable. Plus you can track objects as well, using special
fiducial markers, that software can pick up and track them on the screen.
The biggest problem I had, while building my DI setup, was getting the
screen illuminated evenly. This is really the key, to having a good performing
screen. I always had one corner a little bit darker, or some spots (center)
brighter then the rest. But you have to get it even, because when setting up
touchlib you have to tweak the filters for the complete screen. So, if one part
of the screen is illuminated lighter then the rest, the fingertips will be
recognized well, while the in the darker part of the screen, they won´t. So you
can adjust the filters, that the fingertips get recognized in the darker areas as
well. But then, in the lighter areas, the fingertips will be recognized to early,
while not touching the screen. This can get very tricky while using software,
because you get a lot of unintended inputs.. so you have to get it even. It took
me about a week to get this on my setup.I tried so many positions for the
illuminators, there was always a part that didn´t work well. I tried diffusing the
light in front of the diffusors, bouncing of the box walls etc… In my previous
post you can see that I used a silver reflective material for a lighting shop to
get it right. It took a while, but it works now.
DI sheme:
So, who is the winner?
Honestly, I don´t know. Right now, I am focussing on DI, mainly because the ability to track
objects, and because I do not have to use silicone rubber. Since there is no software that can
track fingers and objects at the same time (reactivision tries to do this in 1.4, but I really haven´t
got it to work properly), this seems to be a little pointless right now. Please feel free to discuss this
using comments or via the contact form on the blog.
I was lucky enough to grap a ticket for the exclusive Microsoft Demo of their
Multitouch Table “Surface” here in Zürich. It was presented by Stve Ballmer
and the head of the Surface Group. My thanks goes out to the
Swiss Design Institute for Finance and Banking, which supplied the tickets for
me and Christian Itten from the Tangent Team. I was able to make some low
res videos of the presentation, which I uploaded to youtube:
Part 1
Part 2
Part 3
Well, first of all, their table looks really nice. I have no idea how
they got it so small, they must be using a special, short throw
projector. Secondly, I really didn´t notice to much lag, as others
have posted in various blogs. So maybe they improved that part.
But, as you can see at the beginning from part 2, not everything
is working perfectly. Getting the just shot picture of Ballmer onto
the table, didin´t work right away. This really shows that, even now,
that this is still a prototype. There where other instances in the
presentation, when pictures in the photoapp didn´t get selected
right away and so on… Objects dragged onto the wrong object etc…
I didn´t capture all of it on film, so you have to trust me on this one.
Anyway, after the presentation, me and Christian grabbed the
Surface manager fr a quick talk. We talked briefly about our projects
and asked some questions about Microsoft timetable and plans for Surface.
they have some sort of sdk/api, that their exclusive
partners (4 right now) get, so that they can develop applications
while not being to specific, it seems that this partnership will go
on for the next two years, during which only these partners will
get access to the tables and sdk.
asked about the availabilty for end customers, the answer was pretty
vague, but at least 4-5 years, they said
it also seemed that the always mentioned 10K pricetag is a little low,
while not being to specific unfortunately
So this basically means, that we really don´t have to fear, that microsoft is
going to bring surface to the end user market anytime soon. Of course, this
is only my personal feeling I got after seeing the presentation and talking to
the people.
Jeff Han
Well, suprisingly Jeff Han seems to sell his interactive Wall through the
Neiman Marcus online store for the incredible amount of 100K. Well, I am
not sure what this means, since they offer no info on the hardware, resolution etc..
They did make a new video though, which shows some new demos. You can
even see a stock market kind of app, which I also had in my 180 project
(not saying that jeff stole this idea from me:). I really don´t know if this is
serious, becuase 100K is pretty damm expensive. We will have to wait and see…
Link to the engadget post -> LINK
Link to the store -> LINK
I´ve gotten a few emails asking about my setup, mainly dealing with the hardware used.
projection surface
Lee Filter, 225 Natural Density Frost
The projection surface is actually a foil, that is used with professional lighting for
filmsets, photoshoots, events etc. So its really not intended as a rear projection
surface. Its fairly cheap, diffuses quite well and makes a pretty good projected
image. Its a lot better than tracing paper!
BTW, the 225 Leefilter is a great diffusor, that you can put
right in front of the illuminators. Its pretty heat resistant too, because it was intended
to be used with high power spotlights.
——————————————————————-
silver reflector
Lee Filter, 273 Soft Silver Reflector
The slver reflector was something I tried to get the ir light diffused better. I frist
tried to paint the inside of the box with normal household paint. This also worked,
but not as well as this stuff. Its afordable as well. You can also see, that I don´t
need to put a diffusor in fornt of the illuminators, the diffusion is caused by that
silver reflector. Now you really don´t have to get this stuff, it depends on your box
mainly. I´ve tried so many different positions of the illuminators, and this is how it
worked best for my box.
——————————————————————-
ir-illuminators
2x Monacor IR-10 Led (each 48 Leds, 940nm)
I wouldn´t really recommend these illuminators, because they are too expensive,
they way to bright, and operate in the 940nm wavelength. Cameras, as the
unibrain I am using, are not really sensitive enough in the 940nm range. It would
be better to work in the 850-880nm range, because cameras are about twice as
sensitive, if you comapre the transmission chart! You could also make your own
illuminator by using some ir leds. I was just lazy:) I have bought some cheaper
illuminators that I am going to test next week, I´ll keep you posted.
——————————————————————-
camera
unibrain fire-i, board cam, color
Good firewire camera, works on pc and mac. I´ve heard that I should have gone
with the bw version, because you really don´t need color for ir anyway. I´ll be
getting the bw version next week, and I´ll run a comparison.
——————————————————————-
ir-filter
2 layers of exposed film negative
Cheapdesign gehtto tech filter! Works…
——————————————————————-
software
touchlib beta 1.0, lightboxMT
I used the normal DI Filter setup (background, brightness & contrast, highpass, rectify).
The photoapp is basically whites demo app, with some interface changes, pictures are loaded
from the local drive, not via flicker. I´m still working on it, but I will be releasing it as soon as
possible.
You can get leefilters at some professional lighting shops, check for a local supplier
on leefilters.com
I took some more pictures, and made a quick demo video. You can see that the recognition
is really, really nice. There are some quirks while moving the images. This is probably more
of an flash issue. Cmoore and Jens Franke are working on the flash TouchAPI, which should
be a lot better.
