WEBVTT

00:00:15.746 --> 00:00:18.850
So let's start by talking
about 3D printing.

00:00:20.225 --> 00:00:22.173
3D printing is a lot like
normal printing,

00:00:22.173 --> 00:00:24.121
but it's in 3D.

00:00:24.121 --> 00:00:26.069
(Laughter)

00:00:26.069 --> 00:00:28.139
Not that kind of 3D.

00:00:28.237 --> 00:00:29.795
But more like this.

00:00:32.519 --> 00:00:36.123
3D printing refers to additive
manufactoring techniques

00:00:36.639 --> 00:00:38.766
that build objects layer by layer,

00:00:39.891 --> 00:00:40.991
starting from nothing

00:00:40.991 --> 00:00:43.711
and ending up with
a completed physical object.

00:00:43.711 --> 00:00:44.925
A common exageration is

00:00:44.925 --> 00:00:49.120
a 3D printer is just like
a Star Strek replicator,

00:00:49.120 --> 00:00:50.287
you can make anything.

00:00:50.287 --> 00:00:54.195
Although you can make
very complex geometries

00:00:54.195 --> 00:00:56.025
with a wide variety of materials

00:00:56.025 --> 00:00:58.785
like plastics, powders and metals.

00:00:58.785 --> 00:01:01.317
3D printing does have its limitations.

00:01:04.248 --> 00:01:06.531
This is why we have
so many kinds of 3D printers.

00:01:07.156 --> 00:01:10.957
These are a lot of
different varieties that exist,

00:01:10.957 --> 00:01:13.857
of different kinds of additive
manufacturing techniques

00:01:13.857 --> 00:01:16.323
that fall within
the field of 3D printing.

00:01:19.448 --> 00:01:21.393
The true magic of 3D printing

00:01:21.393 --> 00:01:23.338
isn't it being a Star Trek replicator.

00:01:23.338 --> 00:01:25.283
It's how we use it.

00:01:25.853 --> 00:01:28.054
A 3D printer is used by designers

00:01:28.054 --> 00:01:30.700
to generate their parts
in the real world.

00:01:31.119 --> 00:01:32.972
So, you can take a design,

00:01:34.215 --> 00:01:37.061
plug it in the printer
and it'll print it out for you.

00:01:37.061 --> 00:01:38.942
And you can take
that part in your hands,

00:01:38.942 --> 00:01:41.163
make adjustments to it,
change your design

00:01:41.163 --> 00:01:42.126
and print another one.

00:01:42.126 --> 00:01:43.806
So it's used for iterative design,

00:01:44.113 --> 00:01:45.733
and it actually checks parts
with the real world.

00:01:47.233 --> 00:01:49.063
So it's a really useful tool.

00:01:50.819 --> 00:01:54.079
A disadvantage of 3D printing
is that it's actually pretty slow.

00:01:54.741 --> 00:01:57.291
So we have a really nice
little 3D printed cup

00:01:57.291 --> 00:02:00.114
over here on the left 
with an integrated straw.

00:02:00.114 --> 00:02:00.954
Pretty cool!

00:02:01.664 --> 00:02:04.362
That takes about the same 
amount of time to print

00:02:04.362 --> 00:02:09.352
or to manufacture 
as these plastic cups

00:02:09.352 --> 00:02:12.221
or a hundred packs of 50 plastic cups,
so 5,000 plastic cups.

00:02:12.958 --> 00:02:14.619
So it's about the same amount 
of manufacturing time,

00:02:14.619 --> 00:02:16.541
That's low-balling it.

00:02:17.250 --> 00:02:21.381
So, this layer by layer
additive process is pretty slow

00:02:22.201 --> 00:02:25.502
compared to a formative
manufacturing technique.

00:02:27.661 --> 00:02:29.933
So, I started to gain interest 
in 3D printing,

00:02:29.933 --> 00:02:31.635
when I was in 
my senior year at MIT.

00:02:31.635 --> 00:02:34.040
And I wanted to make a printer

00:02:34.040 --> 00:02:38.195
that was really fast and really cheap

00:02:38.195 --> 00:02:40.525
and printing with 
a wide variety of materials.

00:02:41.095 --> 00:02:43.609
So I was a little disappointed 
to find out

00:02:43.609 --> 00:02:45.571
that these goals were kind of

00:02:45.571 --> 00:02:48.123
what the entire 3D printing industry 
was already working on.

00:02:48.123 --> 00:02:49.412
(Laughter)

00:02:49.412 --> 00:02:52.428
So, I decided, I needed to take
a different approach

00:02:52.428 --> 00:02:54.784
if I was going to make 
a big impact in this field.

00:02:56.374 --> 00:02:59.660
So, I kinda looked at the trends
that exist within fabrication tools

00:02:59.825 --> 00:03:03.765
and you can plot them 
on this graph here

00:03:03.765 --> 00:03:08.379
where the flexibility and speed 
of a fabrication process

00:03:08.379 --> 00:03:10.010
are inversely proportional.

00:03:10.010 --> 00:03:14.572
So 3D printing on the left is 
very flexible, but pretty slow,

00:03:14.572 --> 00:03:18.600
and injection molding on the right, 
making legos is very fast,

00:03:18.762 --> 00:03:22.362
but can only make the parts 
the mold is designed to make.

00:03:24.156 --> 00:03:27.546
And I needed something 
that was both fast and flexible.

00:03:27.795 --> 00:03:30.191
Instead of our breakthrough technology

00:03:30.191 --> 00:03:33.567
that jumps out of the curve 
and then I found out

00:03:33.567 --> 00:03:37.385
about a little known field called 
reconfigurable pin tooling,

00:03:37.385 --> 00:03:38.886
probably haven't heard of it.

00:03:38.886 --> 00:03:43.609
Essentially, the idea
is to have a bed of pins

00:03:43.609 --> 00:03:46.068
that are adjustable in height

00:03:46.068 --> 00:03:47.878
and with those pins,

00:03:47.878 --> 00:03:50.636
you can generate a surface
for use in molding

00:03:50.636 --> 00:03:52.064
or for other applications,

00:03:52.064 --> 00:03:54.494
this is from science fiction,
this isn't real.

00:03:54.764 --> 00:03:56.724
(Laughter)

00:03:57.438 --> 00:03:59.798
I was surprised to find out 
interesting facts though.

00:03:59.875 --> 00:04:02.541
This is the first patent
in reconfigurable pin tooling,

00:04:03.411 --> 00:04:08.107
in 1863, that's 150 years ago.

00:04:09.117 --> 00:04:11.055
But in comparison to 3D printing,

00:04:11.223 --> 00:04:14.931
the first pattern in 
3D printing was in 1984,

00:04:14.931 --> 00:04:17.048
that's 29 years ago.

00:04:17.558 --> 00:04:24.848
So, if reconfigurable pin tooling is
so cool and such an old idea, 


00:04:25.442 --> 00:04:28.282
why are there no 
reconfigurable pin tools?

00:04:28.573 --> 00:04:33.223
While so many different 3D printers
exist on the commercial shelves.

00:04:33.403 --> 00:04:36.353
Well, it turns out there are 
just really hard to make.

00:04:37.083 --> 00:04:39.203
So, this is a pin art toy,

00:04:39.327 --> 00:04:41.052
you'll probably be familiar with this.

00:04:41.052 --> 00:04:43.325
This is the most classic example 
of a reconfigurable pin tool.

00:04:43.325 --> 00:04:46.225
And if I were to make this
electronically reconfigurable, 


00:04:46.565 --> 00:04:50.605
I would have to add a motor 
to everyone of these pins, right?

00:04:50.702 --> 00:04:56.012
And there's about a thousand pins
in this sort of cheap desktop toy.

00:04:56.264 --> 00:04:58.394
A thousand motors
is a lot of motors

00:04:58.740 --> 00:05:02.888
and that's a really significant
engineering challenge.

00:05:05.759 --> 00:05:08.088
You probably or you might 
have seen this video


00:05:08.088 --> 00:05:10.255
which actually came out
this last week.

00:05:10.815 --> 00:05:15.451
This is a really cool example 
of a reconfigurable pin display,

00:05:15.451 --> 00:05:18.101
that some of my friends 
made at the MIT media lab.

00:05:19.046 --> 00:05:22.188
And this device
is individually actuated,

00:05:22.188 --> 00:05:26.202
so all the pins have 
a single motor on each one.

00:05:27.306 --> 00:05:30.535
There's 900 pins within 
3 inches resolution,

00:05:30.535 --> 00:05:33.764
and it was used for haptic interface

00:05:33.764 --> 00:05:36.995
and for making experimental services.

00:05:38.197 --> 00:05:41.617
So, if I wanted a surface
that was high resolution to use as mold,

00:05:42.579 --> 00:05:43.829
why can't I do this?

00:05:43.829 --> 00:05:47.537
Why can't I make this surface 
super high resolution?

00:05:48.235 --> 00:05:49.915
Math. That's why.

00:05:50.392 --> 00:05:51.362
(Laughter)

00:05:51.362 --> 00:05:53.552
Math is fighting me on this one.

00:05:55.104 --> 00:05:57.095
When I increase the resolution,

00:05:57.115 --> 00:06:00.486
I get this quadratic scaling 
of the area,

00:06:00.486 --> 00:06:03.177
so length times width is area,

00:06:03.248 --> 00:06:05.578
and that's a nonlinear term.

00:06:06.178 --> 00:06:10.038
So, when we get
to high resolutions,

00:06:10.038 --> 00:06:11.818
this becomes a really big problem.

00:06:11.818 --> 00:06:14.189
We get huge numbers 
of pins to control,

00:06:14.189 --> 00:06:15.730
massive numbers of motors

00:06:15.730 --> 00:06:17.751
and it just becomes 
totally unfeasible,

00:06:17.751 --> 00:06:19.231
and everything falls apart.

00:06:19.727 --> 00:06:21.699
So faced with this hopelesness,

00:06:21.699 --> 00:06:25.331
I decided to do this 
for my PhD and Masters.

00:06:25.331 --> 00:06:26.335
(Laughter)

00:06:26.674 --> 00:06:28.244
And undergraduate thesis.

00:06:30.427 --> 00:06:33.037
And I've been working on it 
for about 3 years now.

00:06:33.684 --> 00:06:36.884
And I've developed
a number of techniques

00:06:36.884 --> 00:06:40.084
to actuate pins and to move pins.

00:06:40.084 --> 00:06:41.766
These are some of the prototypes

00:06:41.766 --> 00:06:43.838
and I actually won
an award for one of them,

00:06:43.838 --> 00:06:45.008
which is the reason I'm here,

00:06:45.008 --> 00:06:46.689
because I got picked up after that.

00:06:48.038 --> 00:06:51.076
I was kinda disappointed
in all of them so far.

00:06:51.076 --> 00:06:52.938
Until recently, and that's kinda of

00:06:52.938 --> 00:06:55.114
what I wanted to talk 
to you about today.

00:06:55.114 --> 00:06:58.554
So, I had an interesting idea

00:06:58.554 --> 00:07:01.024
when I was working
on a different project,

00:07:01.024 --> 00:07:03.096
not the reconfigurable 
pin tooling project,

00:07:03.187 --> 00:07:05.209
but I was working on a machine

00:07:05.209 --> 00:07:07.291
that had a lot of vibrations in it

00:07:07.291 --> 00:07:10.725
and what happened is that
I was attaching a part to it

00:07:11.353 --> 00:07:13.833
and the screws in that part
kept on coming loose.

00:07:13.833 --> 00:07:16.004
And it was really frustating at first, 


00:07:16.004 --> 00:07:19.269
but then I realised that 
I could actually use

00:07:19.269 --> 00:07:22.570
this pattern vibration to turn out screws,

00:07:22.667 --> 00:07:25.497
which is actually a really good way 
of getting linear actuation.

00:07:25.881 --> 00:07:28.941
So moving something along its axis.

00:07:29.530 --> 00:07:34.270
So, what I decided to do is apply this
to reconfigurable pin tooling.

00:07:35.739 --> 00:07:36.719
And here it is.

00:07:37.505 --> 00:07:38.705
It actually works pretty good.

00:07:38.705 --> 00:07:40.506
This an array of screws,

00:07:41.268 --> 00:07:44.508
that has a specific pattern
of vibration applied to it,

00:07:46.023 --> 00:07:49.058
and that causes selective screws 
within the array

00:07:49.058 --> 00:07:53.619
to actually turn out and 
turn them back in as well.

00:07:54.267 --> 00:07:55.579
And it works like this:

00:07:57.276 --> 00:07:59.751
this is a schematic 
of the actuaction here.

00:07:59.751 --> 00:08:03.468
We have dislocations within 
the square array of screws

00:08:03.468 --> 00:08:05.741
and if you dislocate it just right,

00:08:05.777 --> 00:08:08.677
around the screw you want 
to turn and you reset it,

00:08:08.766 --> 00:08:12.466
you get a non linear torque 
applied to one of the screws,

00:08:12.546 --> 00:08:16.046
and you get motion, 
so pretty cool.

00:08:16.334 --> 00:08:19.224
And the coolest thing about this
is that the only actuator you need,

00:08:19.224 --> 00:08:23.258
the only motor you need 
for this array is for the edge pieces.

00:08:23.258 --> 00:08:26.764
So the edges are always 
going to scale linearly

00:08:26.764 --> 00:08:30.191
with the resolution versus 
the number of pins scaling

00:08:30.191 --> 00:08:32.210
this huge quadratic term.

00:08:33.087 --> 00:08:36.070
And all the pins actually 
are just little screws.

00:08:36.070 --> 00:08:37.697
Screws are very cheap,

00:08:37.697 --> 00:08:40.039
and you get can cheap
linear actuators on the edges

00:08:40.039 --> 00:08:41.371
for vibration.

00:08:41.371 --> 00:08:44.520
And this works really well 
at high resolutions

00:08:44.520 --> 00:08:46.564
because that ratio becomes 
higher and higher,

00:08:46.564 --> 00:08:48.708
as you get higher in resolution.

00:08:48.708 --> 00:08:52.113
The ratio between linear and
quadratic terms within the array.

00:08:52.113 --> 00:08:53.273
With me so far?

00:08:53.273 --> 00:08:55.603
(Laughter)

00:08:57.522 --> 00:08:59.990
So, after doing this project,

00:09:01.610 --> 00:09:03.256
I'm actually pretty confident

00:09:03.256 --> 00:09:06.100
now more so than 
I have been in the past,

00:09:06.100 --> 00:09:09.418
that this HD pin surface 
could be a reality,

00:09:09.418 --> 00:09:11.179
and you could see one of these
on your desktop

00:09:11.179 --> 00:09:12.920
and download a file into it

00:09:12.920 --> 00:09:14.399
and have it reconfigure its surface

00:09:14.399 --> 00:09:17.286
into an arbitruary file
that you found online

00:09:17.318 --> 00:09:22.039
and you use it as a design tool
because you could use it as a mold

00:09:23.075 --> 00:09:26.047
instead of just 3D printing objects
layer by layer

00:09:26.047 --> 00:09:28.349
or along with a 3D printer as well.

00:09:28.349 --> 00:09:30.615
So, it's really just 
a close cousin to 3D printing

00:09:30.615 --> 00:09:32.161
versus any sort of replacement.

00:09:32.161 --> 00:09:34.741
And here it is, 
this is kind of the pitch,

00:09:34.741 --> 00:09:38.311
the digital mold as the next tool

00:09:38.311 --> 00:09:44.102
to help form and shape the future
of personal fabrication.

00:09:44.842 --> 00:09:45.673
That's it.

00:09:45.696 --> 00:09:47.536
(Applause)