Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:02
There's nothing that could possibly go wrong. Well,
0:05
at least it's a backup, right? Hello
0:08
and welcome to the Hackaday podcast. I'm
0:10
Elliott Williams. And I'm Al Williams. This
0:13
is episode 250. We'll
0:15
be talking trains, RC planes,
0:18
and EEPROMs in flames. This
0:21
week in the news, I
0:23
am off at Chaos Communication
0:26
Congress and it is absolutely
0:28
insane. 14,000 hackers
0:30
all together in the Hamburg Convention
0:32
Center, which is actually kind of
0:34
fun because the Congress used to
0:36
be here and then they had
0:38
to renovate the building. And so we had
0:40
to go somewhere else for a while. And
0:43
then there was a pandemic and now it's
0:45
back in the old building. And it's kind
0:47
of fun. It feels feels like old times.
0:49
If you're here, you already know what's going
0:51
on. But if you're not here, I've got
0:53
a link. I'll throw in the show notes
0:56
for you. But also we'll put some links
0:58
to some of the really amazing talks that
1:00
are going on. In fact, I wrote one
1:02
of them up. We'll cover that in a
1:04
minute. The really nice thing about
1:06
the CCC is that they have video of
1:08
the talks up as soon as the next
1:10
day. So if you're not here, you can
1:13
also follow along by heading over to the
1:15
website and checking out anything that looks interesting
1:17
to you. I don't know. I
1:19
looked through, there's at least 15, maybe 20
1:22
talks I want to go see. And I've
1:24
managed to see three so far. But
1:27
it's always like this when you go
1:29
to conventions. You can never see as
1:31
much as you wish you could. On
1:33
the other hand, I was able to
1:35
get us a CCC related. What's
1:37
that sound? This is a journey into sound. What's that
1:39
sound? What's that sound? So
1:42
you want to give it a listen?
2:00
I was heading for something solenoid
2:02
related at the beginning but the
2:04
end didn't make me think that's
2:07
correct So I don't know.
2:09
I'm just gonna take a random stab and
2:11
say it's a paper tape punch, but that's
2:13
not right I'll give you one hint and
2:15
that is that there are two different versions
2:18
of the same thing so
2:20
those were two like distinct sounds from
2:22
two different kinds of a Blank
2:25
I'll tune in next week. That
2:28
sounds good. I'll probably tell you
2:30
before next week But
2:32
if you're listening at home and have a good
2:34
guess as to what these things are head on
2:36
over to hackaday.com Slash podcast fill
2:39
in your best guess give us your handle
2:41
and we'll see who guesses it right next
2:43
week All right. So with that let's
2:45
head on off to the hacks I'd
2:47
like to start off with an amazing
2:50
story from the Chaos Communication Communication Congress
2:52
where I am this is a talk
2:54
given by Redford Q 3k and mr.
2:57
Tick Unbreaking
3:00
trains uncovering shady behavior. This is
3:02
just one of those Incredible
3:05
hacker stories that you kind of have to
3:07
see to believe you you really should all
3:09
go watch the talk So the link is
3:11
over on the hackaday page The
3:13
long and the short story is that
3:15
a company in Poland that has
3:18
a contract to repair and service
3:20
locomotives couldn't get them fixed
3:22
and working and Wanted
3:25
to know why and so according
3:27
to them Anyway, the company googled
3:29
Polish hackers and came up with
3:32
Redford Q 3k of mr. Tick's
3:34
names Called them up and
3:36
said hey We think
3:38
there's something with the firmware
3:40
on this train like ECU
3:43
controller thing Can you poke
3:45
at it and see if you can make these
3:47
trains run again? And they said Yeah,
3:50
we'll give it a shot and then they're like,
3:52
oh by the way, you have two weeks to
3:54
do it In fact, they managed to get the
3:56
trains running about 43 minutes
3:59
before the deadline they had been
4:01
given. So the trains are basically
4:03
run entirely over CAN bus and
4:05
are programmed in this kind of
4:07
weird archaic programming language. They managed
4:09
to a sniff the CAN bus
4:11
and find out that it was
4:13
sending some zeros essentially where it
4:15
should have been sending ones. That
4:17
is the controller was telling the
4:19
motors not to fire up when
4:21
they should have been spooling up
4:23
even though the you know even
4:25
though they were throwing the lever
4:27
up front. And the reason for
4:29
this is that it
4:32
looks like from reverse engineering
4:34
the software the software was
4:36
intentionally bricking itself. The software
4:38
was intentionally failing
4:40
after a certain period of time
4:43
and they ended up having access
4:45
to I think 30 individual locomotives
4:47
and dumping the firmware from all
4:50
of these reverse engineering it
4:52
looking at it. Not only could they make
4:54
them run but they found
4:56
the code that made them not run
4:59
which included things like if it's
5:01
the 21st of the month and
5:03
the train hasn't run in four
5:05
days don't fire up the motors.
5:08
The absolute best as far as I
5:10
can tell which made this very clearly
5:12
kind of an active I
5:15
can't say it's an active sabotage this
5:17
is an ongoing legal case. Malfeasance perhaps.
5:19
A parent malfeasance. A
5:23
legit malfeasance is. Is
5:25
that the manufacturer put things
5:27
in there like geofencing against
5:29
a bunch of their rival
5:31
companies that were servicing the
5:33
trains to make them not
5:35
work when they were in
5:37
their stockyards but able to
5:40
work when they are elsewhere.
5:42
And in fact they had
5:44
a flag that made it not work
5:46
when it was in their own stockyards
5:48
but they also were able to turn
5:51
it on and off so they had
5:53
tested this geofencing thing apparently and
5:56
then implemented it in actual
5:58
real trains. And so
6:01
this is just a phenomenal talk
6:03
and it kind of brings home
6:05
the best of the reasons why
6:07
we hack. And that is some
6:09
companies will do technologically bad things.
6:11
And if you can reverse engineer
6:13
their firmware, you can figure out
6:16
that they're doing this and
6:18
hold them accountable. And that's exactly what happened
6:20
here. But go watch the talk. Trust me
6:22
on this one. So I
6:24
did think the geofencing out your
6:26
competitors' repair yards was particularly egregious.
6:29
That's awesome. You know, because up till
6:31
then I was kind of like, well,
6:33
let's try to be open-minded. Maybe they
6:35
have a reason, but that's really that
6:37
seems over the top. And I think
6:39
I always have a resistance to getting
6:41
the courts involved in anything like this
6:43
because that's never good. Right. It's just
6:46
usually the courts do not make great decisions
6:49
about technical stuff. You know, going all the
6:51
way back to the, well, you can't see
6:53
the copyright message in the ROM. So therefore
6:55
it's not copyrighted. You know, that was back
6:57
in the 70s or whatever. But, you know,
7:00
this does seem like a big deal. And
7:02
I could not help but think of the
7:04
McDonald's ice cream machine fiasco, right? Which has
7:06
a lot of these same sort of shadows
7:09
in it where, you know, you say, well,
7:11
we did publish the codes, but nobody bothered
7:13
to read them. And so when somebody made
7:15
a raspberry pie to read them, we didn't
7:17
like that. So then we started, you know,
7:19
saying, oh, it's not safe and it might
7:21
be dangerous. And, you know, I don't think
7:23
it's dangerous to have a raspberry pie connected to
7:25
your ice cream machine. So it did kind of
7:27
remind me of that. And it's the same kind
7:29
of thing. Once it devolves into a legal battle,
7:32
it just seems like nobody really wins except perhaps
7:34
the lawyers. But I do
7:36
agree this seemed to be especially
7:38
egregious and maybe even
7:40
more mean-spirited than the ice cream
7:42
hack. So they looked at
7:44
all of these trains and they all had
7:46
different triggers in them. Some of them stopped
7:49
working after the odometer got
7:51
to a million kilometers. Some
7:54
of them stopped working after the train was
7:56
inactive for more than 21 hours. Some
7:59
of them stopped. working on the 21st
8:01
of the month. And that one,
8:03
there was actually a bug in
8:06
the code and tongue-in-cheek Q3K suggested
8:08
that the 21st be a
8:10
Polish national holiday, the day the trains
8:13
don't run. And I thought that was
8:15
pretty sweet. Oh man, there's so
8:17
many crazy details in this talk. One of
8:20
the things that I really love also is
8:22
that they had access to one of these
8:24
units, they dumped the ROMs and then they
8:26
sent it back to the manufacturer and told
8:29
them a little bit of what they had
8:31
found and what the problem was. The
8:34
manufacturer sends it back, working of
8:36
course, with a new firmware that has,
8:38
as you mentioned, strangely
8:40
enough, copyright notices in the
8:42
ROM and says, you know,
8:44
if you're dumping this ROM now, you're
8:47
breaking the law or whatever, right?
8:49
And the funny thing is they
8:51
changed a number of other things.
8:53
It's probably illegal to change the
8:56
ROMs on the ECU of a
8:59
train without getting it recertified. So
9:01
it's very likely that not only
9:03
did they do something that is,
9:06
you know, trying to cover their
9:08
tracks and protect them legally from
9:10
this incredible storm that's about to
9:13
hit them, but they probably also
9:15
put trains back on the rails
9:17
that did not have properly vetted firmware in
9:19
them. I'm not sure what the safety regulations
9:22
for trains are, but you would think that'd
9:24
be something, you know, certainly like an airplane,
9:26
you couldn't do that. You can't just say,
9:28
oh, here's some new firmware we slapped together.
9:30
Let's throw it on an airplane. They hid
9:32
cheat codes essentially in the firmware that reactivated
9:35
the motors and stuff like that. I mean,
9:37
because they didn't want to do tons of
9:39
work, they didn't want to have to reflash
9:41
the firmware to get the trains running again.
9:43
So there's actually a two
9:46
lever command sequence you can give
9:48
the train and then it'll start running
9:50
again. And that's actually what
9:52
these guys found out about all these
9:55
trains. They actually didn't end up changing
9:57
any of the firmware on the trains.
10:00
did was they read it, dumped
10:02
it, reverse engineered it, discovered the cheat
10:04
codes that were embedded in the firmware,
10:06
and then just replayed them, you know,
10:09
push this lever here while pressing this
10:11
button and hitting the brakes. And
10:13
then that made the train run again. Yeah.
10:16
Well, and it kind of reminded me too of
10:18
some of the old copy protection schemes that you
10:20
used to see on like games, you know, and
10:22
if you dig down far enough, there was always
10:24
some piece of code that said, did
10:26
I see my copy protection? Yes, no. And you
10:28
could always at least reverse the sense of that,
10:30
right? So then it
10:33
wouldn't work with the hardware dongle or whatever
10:35
the protection device was, but it would work
10:37
without it. You know, for a
10:39
long time, that was a big emphasis on people
10:41
hacking was finding ways around the copy protection for
10:43
games. They kind of reminded me of that a
10:46
little bit. Glad those
10:48
days are gone, but that kind
10:50
of behavior is absolutely unacceptable in
10:52
large public sector
10:54
infrastructure. So I waxed
10:56
nostalgic for my first one this
10:59
week, and it was actually a
11:01
piece that Chloe Abenesu, I
11:03
think at PC magazine wrote, which was
11:05
a tech obituary for all the tech
11:08
gadgets and services that died in 2023.
11:11
You know, and most of them were
11:14
really consumer oriented. You know, Google obviously
11:16
had killed 20 things like they do
11:18
every year, including I was
11:20
amused to see Google glass died
11:23
again, because they had a commercial version
11:25
of glass that nobody wanted either. You know,
11:27
that's one of those things where man sitting
11:29
in a room somewhere, you would assume that
11:31
would just sell like gangbusters. And it just
11:33
doesn't. And I don't know, but there's probably
11:35
there must be some larger reason for that.
11:38
But it ranged to other things to
11:40
Netflix, apparently you stopped shipping DVDs, which
11:42
I didn't know they still were shipping
11:44
DVDs, and just a lot of
11:46
different things like that. But it got me
11:48
to thinking, you know, those were all really
11:50
kind of consumer oriented. And we're all kind
11:52
of interested in that. But it really made
11:54
me think about what have we lost in
11:56
our sector, right in the hackerspace. So
11:59
what products. have kind of come and gone.
12:01
You know, some of the 3d printing stuff
12:03
has started to fall out. So XYZ
12:06
printing, you know, they were
12:08
infamous for having the closed
12:10
filament. They're gone. I
12:12
guess the sculpt your marketplace was gone. And
12:14
I was hoping in the comments, maybe people
12:16
would know about some other little obscure things
12:18
that had come and gone, but surprisingly, I
12:20
didn't see a lot that just made me
12:22
think, Oh yeah, you know, that's a big
12:25
thing, but I think there's a lot of
12:27
changes in the industry and maybe we're due
12:29
for a shake out. So maybe 2024,
12:31
there will be more things
12:33
shaken out. Uh, and I think part of
12:35
it is too, is we've got so many
12:37
big players now that maybe the things that
12:39
kind of came and went, you didn't even
12:41
notice. So it's one thing when
12:44
Google shutters something down because they're very visible,
12:46
it's another thing when, you know, the Williams
12:48
is a start their garage company and it
12:50
just doesn't go anywhere and they go back
12:52
to writing for Hackaday. Right. So, you know,
12:54
I think that's maybe less noticeable, but I
12:56
imagine there's a lot of that going on
12:58
as well. I think the other thing
13:00
that was interesting to me about this
13:02
is the fact that, you know, we
13:04
are seeing a lot of
13:06
stuff that subscription based seemed to go
13:09
away. So, you know, Google
13:11
that back on Google, since they kill everything,
13:13
they had offered the deal where you say,
13:15
Oh, if you pay us so much a
13:17
year, every couple of years, we'll give you
13:19
a new pixel phone and you'll get free
13:21
access and, you know, all this stuff. And
13:23
I guess if you signed up for that
13:25
tough because they kind of canceled it, that's
13:27
interesting too, because I fear the, you know,
13:29
we won't sell you anything. We're going to give
13:31
it to you as a service, some because of
13:33
what we just talked about with the trains, right?
13:35
Because once you own my phone and I don't,
13:38
you can justify getting away with a lot more naughty
13:40
things on my phone than if I own it. But
13:43
I, it seems like maybe the market is rejecting
13:45
that. So that was kind of a cool takeaway
13:47
from that. I think I'm looking down this list
13:49
on PC magazine and you know, if, if I
13:52
get to be hardware guy versus software guy, I
13:54
will note that these are like 90 something
13:57
percent software that God
13:59
discontinued. or like you said,
14:01
they're software services that got discontinued.
14:03
So few of them are actually
14:06
physical things out there in the world.
14:08
It's just like we thought we would
14:10
sell you this software and now we're
14:12
not anymore. And yeah, well, easy come,
14:14
easy go, right? Well, I mean, Google
14:16
Glass certainly was a piece of hardware,
14:18
but you're right. It's in the minority.
14:21
I don't know if their business model was...
14:23
I guess they sold it to you for
14:26
outrageous sums of money. So they probably were
14:28
counting on it being the hardware pushing the
14:30
device. Well, not that I'm in any position
14:32
to give Google advice, but it
14:34
always strikes me that they don't
14:37
have the lesson we learned back with
14:39
the CPM computers, which is that no
14:41
one wants to buy computers. They want
14:43
to buy applications, right? So you didn't
14:46
really buy a computer. You bought Wordstar
14:48
and you bought VisiCalc or whatever the
14:50
early spreadsheet program was. And so it
14:52
seems to me that's what is missing here
14:54
is you don't need the, hey, isn't that
14:57
cool, there's glasses and you can do a
14:59
web search on them. You need the killer
15:01
application that says, hey, I'm a surgeon and
15:03
I can get a tactical view of my
15:05
patient laying on the table through these glasses
15:07
or whatever. I don't know what the killer
15:09
app is because if I was, I'd probably
15:11
be doing that instead of this. But that's
15:14
what will sell it is when there's the
15:16
app, you just can't live without it and
15:18
you'll pay anything for the glasses. Just
15:20
saying I made the glasses, only
15:23
we get excited about stuff like that. I
15:26
think you're right. You're
15:28
saying app like software application, the
15:30
killer app, but I'm saying what's
15:32
the real physical world application, but
15:35
we're both saying the same thing.
15:37
Right. And I mean, it's just like
15:39
if you say, look at the early computers when it
15:41
was just the weird people that had them and you
15:43
said, oh, we did my bio rhythms and I kept
15:45
my recipes on them. And I don't
15:48
have to say, well, why do I make this recipe for
15:50
six when it's for four? I just tell it I want
15:52
six people. No one cared about any of that. So
15:54
that was there, but no one cared. And what
15:56
people really cared about at first were the
15:58
spreadsheets and the word. processors and eventually
16:00
the real surprise was is they were
16:03
interested in music, media and talking to
16:05
each other, right? And that turned out
16:07
the killer app was not just having
16:09
a computer but having your computer talk
16:12
to everybody else's computer. Well it's wintertime
16:14
here and that means the fall semester
16:16
is over and that means Cornell's ECE
16:18
4760 class. This is a micro controller
16:24
class where the students all have
16:26
to, through the course of the semester,
16:28
put together a final project and
16:30
show it off on the internet.
16:32
This class used to be run
16:35
by Bruce Land, it's now run
16:37
by Hunter Adams, but these student
16:39
projects are no less awesome. So
16:41
expect to see a bunch of
16:43
them coming out in the next
16:45
few weeks. The one I wanted
16:47
to cover is Raspberry Pi Pico
16:49
Mandelbrot Computation. This is a really
16:51
cool project done by Ryan Colm
16:53
and Ignacio de Jesus Romojimenez in
16:55
which they set up a parallel
16:58
computer that they're calling the Computron
17:00
to calculate the Mandelbrot set and display
17:02
it on a VGA monitor.
17:05
The Computron is a parallel computer
17:08
in the sense that they have
17:10
one Pico that's responsible for doing
17:12
the graphics and projecting that onto
17:14
the VGA monitor and then they
17:16
can plug in a whole bunch
17:18
of other Pico's with kind of
17:21
compute software on them and spread
17:23
up the work amongst them. They
17:25
all talk over an I2C bus
17:27
which provides a simple kind of
17:30
robust data backbone for them and then
17:32
the main computer prints it out to
17:34
the screen. Their write-up is superb and
17:37
it ends up coming to the conclusion
17:39
that a lot of people come to
17:41
when they're like, I'll just take this
17:43
problem and parallelize it and that is
17:46
that you really don't get a end
17:49
time speed up if you throw end
17:51
computers at the problem and they find
17:54
that in spades here. They get kind
17:56
of you know an 80% increase in speed
18:00
when they go from one Pico to
18:02
two Pico's working on it. But they
18:04
have a beautiful graph where they show
18:06
it taper off. And in fact, they
18:08
have the really strangest thing, which is
18:10
that sometimes they'll have little jumps in
18:12
their performance where you would expect the
18:14
performance to keep increasing as you throw
18:17
more processors in it, but it doesn't
18:19
or it stalls. And what they think
18:21
is going on here is when you
18:23
divide the problem by two, it's pretty
18:25
evenly split among the two. But when
18:27
you divide it by eight, one
18:30
of the processors gets the hardest problem
18:32
every time through as it goes through
18:34
this kind of round robin, you take
18:36
the slice, you take the slice, you
18:39
take the slice procedure. And that poor
18:41
one processor, long story short, it's
18:43
so tempting to just throw massive parallelism at
18:45
a problem when you have a problem like
18:47
this, especially when you have like cheap devices.
18:49
You're like, I need it to go twice
18:51
as fast. I'll just use two of them.
18:53
No, you'll use three of them. And then
18:56
it's like, I need it to go eight
18:58
times as fast. I'll just use eight of
19:00
them. And the answer there I think is
19:02
you're gonna have to use 32 of them
19:04
or step up to a bigger
19:06
processor. You know, the I2C bus
19:08
almost had to be some factor to that.
19:11
You normally see some sort of higher
19:13
end fabric or shared memory, you know,
19:15
but this was such a great effort
19:18
for a student project. And it's
19:20
always one of the high points of the year
19:22
when Cornell sends over their batch of student projects
19:24
because it's always amazing. And this one certainly was
19:26
very interesting. You know, they sent it over in
19:29
a block and I think there was like 47,
19:31
I wanna say, there
19:33
was some huge number of projects, but a
19:35
lot of them had pretty interesting titles, you
19:37
know, gesture-based scheme or something. And some of
19:39
them would say like, Computron, and you're like,
19:42
well, I don't know what that is. So
19:44
part of our job was to kind of
19:46
dig through and go, oh, this is where
19:48
they used a bunch of Raspberry Pis and
19:50
a parallel processor. And then that's what Computron
19:52
is. So, you know, that might be part
19:54
of the student work there is to make
19:56
sure we understand what you're doing. Oh
19:58
man, you give me a project called- Computron and
20:00
I am digging into it. I'll be
20:02
like, that's bait for me. I'm like,
20:04
what the heck is a Computron? What
20:07
I really love about this Cornell class
20:10
is that all the students who take
20:12
it are graded on both the success
20:14
of their project, but the quality of
20:16
their documentation around it. And that's, I
20:19
think, why we've been covering these projects
20:21
for so long. I'd be really stoked
20:23
to see other engineering programs follow suit.
20:26
I mean, I'm sure that other schools
20:28
out there must be doing things
20:30
like this, where they give them an
20:32
end of the semester capstone project. And
20:35
it's like, how cool can you do
20:37
this? But by requiring the students to
20:39
document their project and make it available
20:41
to everyone else, it not only helps
20:44
each successive year of students do
20:47
cooler and cooler things, which we've
20:49
totally seen in the kind of
20:51
whatever, 10 years, 15 years that
20:54
Hackaday's been covering this ECE class
20:56
here. So the students get to
20:58
benefit from each other's work, but
21:01
then we get to write it
21:03
up too. And then everybody gets
21:05
to benefit from the students' work.
21:07
And I think that's really what's
21:09
super cool about this particular ECE
21:11
class. And why are other
21:14
professors out there not requiring their students
21:16
to document their work well and put
21:18
it up on the internet? If
21:20
you do, it offers them a
21:22
fantastic carrot, right? If
21:24
they want to put that extra effort
21:27
into it and make it look really
21:29
good or put more work into it
21:31
and document that super well, it just
21:33
makes such a huge reward for them
21:35
in terms of the ability
21:38
for them to get recognition for their hard
21:40
work over and above just getting an A
21:42
in the class. Yeah, if
21:45
you're an engineering professor who has
21:47
your students do a practical class
21:49
like this, give them a
21:51
documentation requirement and send us the links. We're
21:53
dying to see them. But I don't know
21:56
about you. If I'm interviewing somebody
21:58
and they said, oh, yeah, I've had three projects. features
22:00
on Hackaday, I'm hiring them, right? Oh, I
22:02
swear. I mean, look, featured
22:04
on Hackaday or not, going on the
22:06
job market can point to this work
22:09
and say, here, look at this. And
22:11
that's like... Absolutely. Builds
22:13
your portfolio. It's an interesting thing that you've
22:15
done. It shows you're competent and well-rounded, and
22:17
who knows what else. Like, it's a win-win
22:19
for everybody. It's like seven wins. Okay, my
22:21
next pick for the week was from Keep
22:24
Making. You know, it's one of these things
22:26
after you read about it, you think, why
22:28
didn't I think of that? You
22:30
know, you've owned for a long time that if you're going
22:33
to 3D print some stuff, it makes sense to fill up
22:35
your build plate and try to print as much as you
22:37
can in one pass. And you're
22:39
thinking in X and Y, but this article
22:41
is talking about actually filling up the build
22:43
volume in the Z axis as well. And
22:46
that's apparently not a super new idea, but
22:48
you don't see it very often. The
22:51
idea is kind of to print
22:53
pieces that are separated by just
22:55
a single-layer gap with ironed top
22:58
surfaces. And it's almost like
23:00
printing support pieces, except instead of the
23:02
support just ripping off, you essentially rip
23:04
off finished pieces and then rip off
23:06
more finished pieces and rip off more
23:08
finished pieces. The videos were
23:10
really well done in that
23:12
you could see how they were kind of
23:14
stuck together. He'd take a screw and just
23:16
use them to pop open the little pieces
23:18
where it was still stuck together. Obviously, one
23:21
side of the piece isn't going to be as
23:23
well finished as the other. In this particular case,
23:25
he's making things that hang on the wall, so
23:28
you really don't care what the wall-facing side looks
23:30
like as long as it's not too grody. It
23:33
didn't look like it was that bad to begin with. So
23:36
I'm kind of interested to try that out
23:38
myself. I don't always do large volumes of
23:40
prints where I'm printing 50 things at one
23:42
time, but that gives you
23:44
a lot of options for producing more things
23:47
than just going to the right, the left,
23:49
up and down on the bed. Of
23:51
course, I don't know if you could do
23:53
this with resin printing. That has its own
23:55
advantages because then filling up the X and
23:57
Y on a resin printer is easy because...
24:00
It doesn't take any more time to print an
24:02
empty bed than a full bed. But obviously, this
24:04
would add time in that case if you're going
24:06
up in the z-axis. But I don't know if
24:08
those would separate quite as well or not, but
24:10
it'd be something interesting to experiment with. I
24:12
guess it only works for things that
24:14
are kind of relatively flat in a
24:17
way. Cuz if you have too much
24:19
kind of vertical topology, it
24:21
means you'll have to think about putting fill
24:23
in between the different object layers. It's
24:26
a really neat kind of switch in your
24:29
thinking to think about multiplying the things, not
24:31
just left and right, but also up and
24:34
down. It made me wonder if the
24:36
slicers couldn't support something like this, where
24:38
you could actually build little
24:40
support pillars in between the layers.
24:43
Just like you do now for tree support,
24:45
for example, that could all be done automatically
24:47
and you could get nice stacks even if
24:49
they weren't flat, like you suggest. That
24:52
would be super clever. And if you add holes
24:54
in it, you could even do, yeah, you could
24:56
run supports up the vertical holes and
24:58
tubes and stuff. Oh, that would be really neat.
25:00
Yeah, a lot of fun things to think about
25:02
there. All right, bringing up the
25:05
rear for me, absorbing traffic noise
25:07
with bricks using Helmholtz resonators. And
25:09
I think if I had my
25:11
life to live over again, one
25:13
of my crazy dream jobs has
25:15
always been kind of an acoustic
25:17
architect. There are people who get
25:20
paid good money to design symphony
25:22
halls. But when I walk into
25:24
a number of other spaces, I'm
25:26
always kind of like, sometimes they
25:28
have really harsh echoes and sometimes
25:30
spaces are so loud, you can't
25:32
even hear the person you're talking to. And
25:36
traffic noise can be overwhelming. And if
25:38
you have big flat
25:40
walls, it can kind of get seemingly
25:42
amplified. That's kind of the project
25:44
that is being undertaken here, is
25:47
how to beat that down. And
25:49
in particular, he's using cleverly
25:52
designed stackable bricks to absorb
25:54
some of the energy out
25:56
of the waves and
25:58
put a notch into the level of. noise
26:00
in an environment. In this demo
26:03
he's using clay as the material
26:05
and making kind of bricks that
26:07
have cavities in them that want
26:10
to resonate at certain frequencies but
26:12
then because the bricks aren't you
26:14
know perfectly hard they absorb all
26:17
of the energy in those frequencies. So
26:20
you can think of it you know
26:22
like kind of any tuned system except
26:24
a tuned system where there's a lot
26:26
of resistance so it dissipates the energy
26:28
out randomly. And it's a
26:31
really neat design here. He starts off
26:33
by doing 3D printing to
26:35
make these kind of extremely complicated
26:37
clay brick shapes that have exactly
26:40
the right interior volume to absorb
26:43
kind of low frequency road noise.
26:45
And then they end
26:47
up cracking. There's a whole bunch of
26:49
reasons and he actually goes through three
26:51
or four kind of failed iterations here.
26:53
He ends up slip casting into 3D
26:56
printed forms to make the
26:58
bricks and that ends up working. The thing
27:01
that is really clever about the
27:03
design he ends up settling on
27:05
and this is why you'd want
27:07
to do it using modern design
27:09
techniques is that they have not
27:11
only a resonant cavity inside the
27:13
brick but they're shaped on the
27:15
outside so that the spaces in
27:18
between them also makes
27:20
a resonant cavity. And
27:22
because of this kind of irregular shape
27:24
you can only stack them in exactly
27:26
the right way to make these shapes
27:29
both the interior and the exterior
27:31
one. Just it's a really sweet
27:33
design and if you've never thought
27:35
about how you would go about
27:38
making hollow arbitrary shaped
27:40
funny bricks like that's
27:42
a cool problem to have and to have
27:44
to solve and he goes through his work
27:47
in that direction here too. Really just a fun
27:49
talk you should check it out. Well
27:51
I'm always surprised at how poorly I
27:54
can predict what comments will be on
27:56
any particular hackaday article. If you read
27:58
the comments on this one... almost
28:00
to a one saying, well that's great
28:02
until the possums nest in those hollow
28:05
cavities, right? Or the
28:07
insects, birds, pigeons. That probably has some
28:09
merit to it. I'm not saying it's
28:11
insurmountable, but yeah, it seems like they
28:13
would be a maintenance issue. But
28:16
yeah, what an interesting exercise and
28:18
shape. And I don't know,
28:20
I'm not as talented as you
28:22
are with hearing the echoes, but I will
28:24
say a room with good acoustics is a
28:26
thing of joy. I don't always understand why
28:28
it's the way it is, but certainly makes
28:30
a big difference. Acoustic architecture, I'm
28:32
sure it's a field. I'm sure there are people who
28:34
make their living doing this. In
28:36
some alternate universe, I'm one of them.
28:39
My last one this week is from
28:41
Veritasium. I have to say, this is
28:43
probably something that won't surprise some people,
28:45
but for me, I have
28:48
had this lifelong obsession with how in the world
28:50
is a sewing machine work, and I just have
28:52
never taken the time to figure that out. It's
28:55
one of those things, I think, hackaday readers
28:57
in general, we like to think we know
28:59
how the world works. So why does your
29:02
cell phone work? Why does air conditioner work?
29:04
And we usually know those things. But
29:06
I was never able to really understand how
29:09
a sewing machine takes a needle, jams it
29:11
through some fabric, and stitches it together. What
29:13
was interesting about this video is it was
29:16
very accessible, but also he had built
29:18
a large scale demonstrator with a
29:20
needle the size of a garden
29:22
stake, I guess, and big giant
29:24
belts and pulleys and
29:27
the thread looks more like
29:29
twine and is able
29:31
to demonstrate all the different working pieces.
29:34
It's amazing when you think about it is, if
29:37
you said today, let's make a machine that binds
29:39
fabric together, well, you got a lot of options,
29:41
right? You could say, well, we'll get a stepper
29:44
motor, we'll have some little arm that goes and
29:46
pulls this here and there. All this
29:48
was back in the 1800s and earlier
29:50
where you didn't have all that. So
29:52
it's all these clever contraptions with cams
29:55
and hooks. And I just
29:57
thought it was a super interesting piece. And
29:59
I'm so. I'm certain I'm in the minority of people
30:01
who didn't know how a sewing machine works, but
30:04
I was really happy to run across this.
30:06
It's something that's been in the back of
30:08
my mind for years. I'm certain you're not
30:10
in the minority. It's a pretty crazy mechanism.
30:13
And the rotating claw that pulls
30:15
the loop through is just, I
30:18
would not have thought of that a priori. Well, you
30:20
got to wonder how many iterations it took to get
30:22
to that too, right? That's right. How
30:25
many times you said, oh, that didn't work. Let me try this.
30:28
You couldn't 3D print them in those days. So
30:30
that was probably a lot of work. And the
30:32
advantage to being able to make a working one
30:34
was so huge. Like
30:36
before, when everybody had to sew by hand,
30:38
you'd be king of the world if you
30:40
could make a machine that would sew. Well
30:43
that was actually one of the things that was in
30:45
the video was he was saying, and I don't know
30:47
what the source for this is. He was saying that
30:49
in 1900, a family would spend 15% of
30:52
their income on clothes. But in 2003, that number is
30:54
less than 4%. And
30:58
yet you will have way more clothes. The
31:00
family will have more clothes in 2003 than they had in 1900. But
31:05
yeah, people made fortunes. And especially, it
31:07
was kind of interesting. Singer, who you
31:09
associate with sewing machines, he didn't actually
31:11
invent any of it. He just saw
31:13
the value of it and bought up
31:15
patents and commercialized cheap sewing machines. And
31:18
that's where Singer came from. Not
31:20
to mention 50,000 year old needles. Good
31:22
God. Who was thinking that up 50,000 years ago?
31:25
That's amazing. People have been sewing for a long time.
31:33
All right, my first quick hack this
31:35
week is veteran SpaceX booster lost to
31:37
rough seas. And this
31:39
is a piece of news coming
31:41
from SpaceX naturally, written up by
31:43
our own space man, Tom Nardi.
31:45
What happened is they landed
31:48
one of their boosters as they
31:50
often do, except this time
31:52
due to rough seas, it fell
31:54
over and got destroyed. But
31:57
the booster in particular is one that had
31:59
made 15 launches, including
32:01
some historical ones. So
32:03
Tom wrote up a brief eulogy for
32:05
B1058 here. Give
32:09
it a read, it's an interesting view
32:11
into kind of the new space situation.
32:13
But speaking of old space, Lift
32:16
Off, The Origin of the Countdown,
32:18
is a long form piece written
32:21
by Cristina Panos. I'm mentioning it
32:23
here because of the super cool
32:25
tidbit that the countdown may have
32:28
been invented for drama in
32:30
a film before it got
32:32
used for launching actual rockets. Turns
32:34
out to be useful, of course.
32:36
Gets everyone in sync on the
32:38
same page, but this is a
32:40
case of life imitating art. And
32:43
last up, Tom Stanton, building
32:45
a rad supercapacitor RC plane.
32:47
This is really fun, despite
32:49
the fact that supercapacitors have
32:51
a lot less energy density
32:54
than a comparable lithium-ion battery.
32:56
Tom wants to make one
32:58
because he already has a
33:00
hand-cranked charger generator to fire
33:02
it up with. And supercapacitors
33:04
charge and discharge very easily, very
33:07
readily, and can put out large
33:09
current, of course. Here he's turning
33:11
it into a radio-controlled plane that he
33:14
can just go zip, zip, zip, charge it up at the
33:16
field, and then give it a fly. He
33:18
also deals with the real downside
33:20
of supercapacitors, and that is that
33:23
their voltage basically starts dropping from
33:25
the minute they start discharging. They
33:28
have a capacitor-like voltage roll-off, and
33:30
how you deal with that and
33:32
get the plane home safe when
33:35
the motors can't turn anymore is
33:37
an interesting problem that he looks
33:39
into here. My three
33:41
quick hacks for the week. Danny Lewis
33:44
had an interesting post about
33:46
how to build a guitar for $30 using parts
33:48
from Amazon. So
33:51
all my cool friends build guitars. I have
33:53
no musical talent, so I don't, but I
33:55
always suspected it would take more than $30 and
33:58
this article shows that that's not necessary. necessarily the
34:00
case. My second one
34:02
was another one of Hunter's students'
34:05
projects from Cornell. Giacomo
34:07
Cuomo and Sophia Lynn had a
34:09
talking ohm meter that spits out
34:11
color bands for you. And
34:14
I kind of described it as a
34:16
multi-mode ohm meter in that
34:18
it measures resistance and it's able to
34:20
actually speak the resistance, but it also
34:22
has like a giant model resistor with
34:24
multicolor LEDs on it. And it actually
34:27
shows you the color bands on
34:29
the fake resistor of what the real resistor
34:32
value is. It's one of those projects that's
34:34
probably not very practical but very cool. Last
34:37
but not least, Diode Gone Wild, always
34:39
loved that name. It
34:41
had an interesting video on how
34:43
wire changes resistance by temperature and
34:46
how you can use that
34:48
to measure temperature in circuits.
34:50
So obviously that takes
34:52
a lot of wire and or a
34:54
very precise resistance measurement, but if you
34:57
ever wondered, you know, how hot is
34:59
that transformer getting? That's
35:01
one way to do it is to measure the change
35:03
in resistance in the wire in the winding. All
35:11
right, that brings us to our Can't
35:14
Miss articles. These are long form pieces
35:16
written by our fantastic Haggaday writing staff.
35:18
This week I wanted to pick Dan
35:20
Maloney's Keeping Watch Over the Oceans with
35:23
Data Buoys. And this is another
35:25
one in the series that
35:27
at least among us Haggaday writers,
35:29
he's calling remotely interesting, which
35:31
I love. I wish he'd used
35:34
that in print because these are
35:36
all remote measurement devices and how
35:38
they work. In this case, he's talking
35:40
about ocean buoys that both measure
35:43
temperature, wind speed, wave height,
35:45
all sorts of things and help give us
35:48
a better picture of what the
35:50
globe's weather is like, especially out
35:52
over the seas where it's otherwise
35:54
harder to measure. The thing that
35:56
I learned from this that I
35:58
hadn't thought about before. was how
36:01
hard it is to connect
36:03
to more deep sea buoys
36:05
to the ground. And
36:07
in particular, there are all
36:10
these multiple material, multi-section cables
36:12
that you have to use
36:15
if you're going down really deep. Because if you think
36:17
about it, naively I'm like,
36:19
oh, well, they just must have a chain
36:21
that goes down to the ocean floor. And
36:24
yeah, that's fine if the ocean floor is
36:26
like 30 feet down. But
36:29
if you've got like miles and miles or
36:31
even just hundreds, thousands of meters
36:34
of steel cable down there,
36:36
that's clearly not going to do
36:38
it. And so they end up
36:41
using these fun mixes of nylon
36:43
and polypropylene lines that float
36:45
better than nylon lines and all
36:47
of these with floats on them.
36:50
There's a whole engineering problem in
36:53
designing the mooring systems that keep
36:55
these buoys basically in the same
36:57
place. I thought that was really
37:00
neat. Of course, there's all the standard
37:02
stuff like how you design the buoys
37:04
to be rugged enough that they can
37:06
withstand high seas. And Dan
37:09
actually points it out in one of these.
37:11
There's a buoy with a cage on it
37:13
that prevents seals from climbing up on top
37:15
of it and maybe messing
37:17
up the instrumentation. These
37:20
things are put out in the
37:22
roughest environment for electrical devices,
37:25
right? Take the middle of
37:27
the ocean with nasty waves.
37:30
And then they're supposed to be
37:32
tethered to the ocean floor, stay
37:35
basically put, and they need enough
37:37
electricity, enough generating capacity, whether gas
37:39
or solar, to keep
37:41
running and keep sending their data
37:43
back to us here on the
37:45
land. So it's a really neat
37:47
problem. And designing
37:50
scientific deep ocean buoys is
37:52
just one more of those
37:55
kind of remotely interesting sensor
37:57
things. And I'm happy to read this
37:59
one. Yeah, this got me nostalgic
38:01
because one of my first jobs was at
38:04
the same place where NOAA had a
38:06
big operation for data buoys. I
38:09
didn't have any direct involvement in it. But
38:11
back in the late 70s, early 80s, this
38:13
was way more difficult than it is now.
38:17
And they had all the problems with
38:19
sending data, so you had very precise time
38:21
measurement. Now, we didn't have GPS, right?
38:23
So you had to figure out the time
38:26
so that when your one minute of
38:28
data on the satellite came up, you
38:30
could squirt that data up over the satellite
38:32
and not interfere with everybody else who
38:34
was going to be using a different minute,
38:37
things like that. So my guess is things
38:39
are a lot easier now. So I know
38:41
if they're close to shore, sometimes they use
38:43
cell modems. And that was all unheard of
38:45
in those days. They do have
38:47
a lot of free floating buoys as well.
38:50
So even though you're right, it's amazing what
38:52
all they do to anchor something in 6,000
38:55
years of ocean, some
38:57
of them are actually just floating around. And
39:01
one of the things that I don't think Dan
39:03
mentioned, but one of the reasons I think there's
39:05
a lot of money put into these is at
39:07
least at a certain point in time, they were
39:09
very interested in the ocean conditions for where the
39:11
submarines might be coming over. And
39:14
that was a big part of this as
39:16
well, was picking up data about
39:18
ocean salinity, thermoclines, things like
39:20
that. So I feel like
39:22
there was probably some other
39:24
non-civilian use of some of this data, at
39:26
least over part of the time, that it
39:29
was operational. But it's an amazing
39:31
network. And yeah, Dan did a
39:33
great job of writing it up. He did mention,
39:35
by the way, I was also
39:37
dismayed he didn't use remotely interesting
39:39
in the title, but he does
39:41
say towards the end, he said
39:43
something about other remotely interesting articles.
39:46
So I guess he's kind of got that
39:48
as a series, but it's not actually in the
39:50
title. So we ought to fix that. We ought
39:52
at least make it the series name. Well he's
39:54
actually tagged them all as a series retroactively and
39:56
put them all down at the bottom of this.
39:59
Yeah. If you're interested in following him out,
40:01
you actually can find them in
40:03
this one or any of his others in
40:05
the series. I think, too, I just wanted
40:07
to shout out real quick where his no-snow
40:10
monitoring snowpack with the
40:13
SnowTel network because that's
40:15
a particularly cool device. And
40:18
then feeling the heat, railway
40:20
defect detection system where
40:22
they basically have heat cameras that look at
40:24
train wheels as they go by and see
40:26
if they're getting too hot, which would indicate
40:29
that a bearing is busted or something.
40:31
Both of these are really, really
40:33
cool and they're kind of, what,
40:35
pieces of essential remote sensing infrastructure
40:38
that keep the world running smoothly
40:40
but are somehow, you know, I
40:42
hadn't known about any of them
40:44
before reading these articles. So to
40:47
that extent, at least, you know,
40:49
invisible to our normal everyday experience.
40:52
Yeah, there's a lot of strange stuff out there. Yeah,
40:54
I can't wait to see what he comes up with
40:56
next. Well, my pick this week
40:58
was Lew and Day. How do
41:00
you test if an EPROM can hold data for 100 years?
41:03
This is always an interesting topic to me. I
41:05
spent time working for Motorola where we did this
41:08
kind of stuff. And, you know, you say, well,
41:10
you know, this flash memory will hold data for
41:12
100 years. Well, we haven't
41:14
had flash memory for 100 years, so obviously
41:16
somebody's making a guess. And
41:18
how do they do that? You
41:20
know, Lew and does a great job
41:22
of explaining failure mechanisms and say, well,
41:24
okay, you know, as the electrons bleed
41:27
out of this or the
41:29
charge bleeds out of this gate, you
41:31
know, that's going to get worse over
41:33
time, but also as a temperature or
41:36
the humidity or whatever. And so you
41:38
wind up designing these tests where you
41:40
say, okay, me heating this up to,
41:43
you know, 150 degrees is going
41:45
to be the same as if it was at room temperature for 20
41:47
years. So now if I
41:49
can get it to last five years, I can safely
41:51
say it's, you know, five times 20, 500 or 100
41:53
years. And
41:57
whether or not that's really true or not, I
41:59
don't know because I don't know if anybody everybody's
42:01
ever actually going back to see if any of
42:03
that works. We barely have enough time on some
42:05
of these devices that maybe you could start actually
42:07
looking at were those predictions accurate or not. It's
42:10
kind of interesting though, they use
42:12
that for a lot of things, which I
42:14
didn't realize. They use it for archival paper
42:16
and inks, which I learned that reading this.
42:19
But at the end of the day, it's
42:21
kind of an estimate or a model, if
42:23
you will. When
42:25
you see the science fiction show where
42:27
700 years from now, somebody pulls
42:30
out something and it's still working and you say, that
42:33
flash drive wouldn't have lasted that long. Maybe it
42:35
would. We don't really know.
42:37
But certainly the data suggests that it
42:39
wouldn't. How
42:42
you actually got there is what Luin
42:44
kind of exposes. It's well worth the
42:46
read if you've ever wondered how those
42:48
numbers arrived at. What I like about
42:50
this is it is kind of based
42:52
in fundamental science at the end. He
42:54
says in this article, and I guess
42:57
I hadn't thought about it, that like
42:59
bit rot degradation of memory cells is
43:01
basically because of chemical reactions that
43:04
are going on inside. That
43:06
gets you the answer already
43:09
because chemical reactions increase with
43:11
heat as described by the
43:13
Arrhenius equation. You
43:15
can figure this out, right? You
43:17
increase the temperature by 10 degrees,
43:19
the reaction speed goes up by this
43:22
much, et cetera, et cetera. This is
43:24
true for chemical reactions. That
43:27
means it's true for how quickly
43:29
materials oxidize or things like
43:31
that. I don't know
43:34
if it's necessarily true for
43:36
how quickly trapped electrons diffuse
43:38
through a silicon barrier.
43:41
But actually, now that I think about it, it's
43:44
got to be, right? This is chemistry again. It
43:47
probably makes a whole ton of sense
43:49
to think about things this way. I
43:51
guess you could also think about whether
43:53
some of the bit rot is due
43:55
to cosmic rays, and then you'd
43:57
want to subject them to something.
44:00
some x-ray bombardment or something for what
44:02
do I know, right? But
44:04
the basic science here I think is really
44:06
sound in the sense that they're saying, look,
44:09
this has got to be a chemical
44:12
reaction that's causing the damage here
44:14
and so we know at
44:16
what rate chemical reactions speed up given different
44:18
temperatures. Let's just crank the heat up on
44:21
it. And of course that only
44:23
works till the thing melts. Yeah, you've got to
44:25
wonder, you know, and maybe this is a good
44:27
science fiction story, in a thousand years what
44:29
the archaeologists are going to make of all
44:31
this because presumably all of our storage media
44:34
will be pretty much impossible to read then.
44:36
So it's like, what is this? I don't
44:38
know. It must have had some religious significance,
44:40
you know. No, it's a CD-ROM, but
44:42
no one knows how to read it so you can't
44:44
see the Panda movie that was on it or whatever.
44:47
Maybe not. And it makes you wonder too,
44:50
just like we have techniques now where we
44:52
can pull writing off of things that they
44:54
used to think were lost forever, you know,
44:56
maybe they'll have some way to do something we
44:59
don't know about and figure out, oh yeah, this
45:01
is what used to be stored in that flash
45:03
memory or on that piece of magnetic tape. Who
45:05
knows what they'll develop. So if you want it to last
45:08
for 10,000 years, write it down
45:10
on paper in some good ink, though. I
45:12
have a whole bunch of, I have
45:15
a whole closet full of like CD-Rs
45:17
that are just rotting away right now.
45:19
I don't even dare look.
45:23
That wraps it up for this
45:25
week's Hackaday podcast. Thanks very much
45:27
for listening. If you'd like to
45:29
follow the links, head on over
45:31
to hackaday.com/podcast and if
45:33
you see anything cool or do
45:35
anything cool, write us. Tips at
45:37
hackaday.com. And until next week, keep
45:39
those hacks coming. What's
45:42
the German word for a piece of crap software?
45:45
Crap, why did I do that? Give
45:48
me one sec. I
45:50
knew you were gonna do that. I should have looked it up for
45:52
you. I
45:55
should have looked it up for you. Well,
45:57
apparently my phone's off. the
46:00
audio.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More