Latest / Elon Musk Podcast / SpaceX pitches Starlink as a GPS alternative

Transcript

SpaceX pitches Starlink as a GPS alternative · Elon Musk Podcast

0:00
So SpaceX just went ahead and formally pitched the FCC that
0:04
their Starlink satellites can basically serve as an
0:07
alternative to the US military's Global Positioning System.
0:12
Which is, you know, pretty interesting timing because that
0:14
proposal came right after SpaceX abruptly shut down this sort of
0:19
backdoor software interface. Right, the one the maritime
0:21
operators were using. Exactly.
0:23
Ships were actively using it to navigate through heavily jammed
0:26
environments in the Red Sea. Yeah, they completely closed
0:29
that free loophole right before pitching a formal, presumably
0:33
paid solution. Right.
0:35
So the real question here is how does a commercial broadband
0:38
Internet constellation accidentally become a highly
0:42
precise military grade navigation system?
0:45
Well, to understand why the FCC is even looking for GPS
0:48
alternatives in the 1st place, you really have to look at the
0:50
physical architecture of the current system.
0:52
The physical limitations, really.
0:54
Right. So the Global Positioning System
0:55
operates from medium Earth orbit.
0:57
Those satellites are sitting roughly 20,200 kilometers away
1:01
from the surface of the Earth. Which is just an enormous
1:04
distance to throw a radio signal.
1:05
It is. I mean, by the time there are L
1:07
band microwave signals travel through the atmosphere,
1:10
penetrate the cloud cover and actually reach a device on the
1:13
ground, they are extremely weak. Like, incredibly faint.
1:17
Yeah, they arrive at about -160 decibel Watts.
1:20
Which is a number that means absolutely nothing to most
1:24
people. That's your point, yeah.
1:25
I always think a better way to visualize GPS is to think of it
1:30
like like trying to see the light of a single candle from a
1:33
mile away through heavy fog. That's a good way to put it.
1:37
You can detect it if you have the right equipment and nothing
1:40
else is interfering, but the moment someone turns on a
1:42
flashlight nearby you lose the candle entirely because the
1:46
signal is so faint. Is just easily overwhelmed by
1:49
really cheap terrestrial jammers.
1:51
Oh. Absolutely, you can literally
1:52
buy a jammer online for like 50 bucks, plug it into the
1:55
cigarette layer of a car and completely brown out the GPS
1:58
signal for blocks around you. Which shuts down mapping apps,
2:02
sure, but it also disrupts ATM networks that rely on GPS for
2:07
time stamping transactions. Yeah, and that's just basic
2:10
jamming that's just making noise.
2:11
Right, just yelling over the signal, but.
2:13
Starling operates in low Earth orbit.
2:15
They're sitting at about 550 kilometers away and they use the
2:20
Ku band for their primary transmissions.
2:22
So they're way closer. Much closer, and being that much
2:25
closer to the ground gives them this profound power advantage.
2:30
It's due to the inverse square law of electromagnetic
2:33
radiation. OK, we are definitely going to
2:35
need to translate the inverse square law.
2:38
Right, it's a physics principle. It basically says as a signal
2:41
travels, it spreads out and loses power exponentially.
2:45
So it's not a linear drop off. Exactly.
2:47
If you cut the distance between the transmitter and the receiver
2:50
in half, the signal isn't twice as strong.
2:52
It's four times as strong. Wow.
2:54
OK, so when you move from 20,000 kilometers down to roughly 500
2:58
kilometers, you're getting a signal that orders of magnitude
3:01
louder? So instead of looking for a
3:03
candle a mile away, relying on a Leo constellation is It's like
3:08
having an industrial floodlight bolted to the ceiling right
3:11
above your head. Exactly.
3:13
And a jammer would have to output just an enormous amount
3:15
of power to drown at a floodlight.
3:17
Spot on, but jamming is really only half the problem.
3:20
There's also spoofing. Right, which is much more
3:22
malicious. Yeah, spoofing is when an
3:24
adversary sends a fake localized signal to trick your receiver
3:30
into calculating the wrong location entirely.
3:33
And standard GPS is pretty vulnerable to that.
3:36
Very civilian GPS receivers typically use omnidirectional
3:40
antennas. They just, you know, listen to
3:41
the entire sky at once. Like standing in an open field
3:44
with your eyes wide open, you're taking in light from everywhere.
3:47
So if a scoofer broadcasts a louder fake signal from a nearby
3:51
hill, the omnidirectional antenna just accepts it as the
3:54
truth because it's the loudest thing it hears.
3:56
It has no idea where the signal is actually coming from.
3:59
But startling terminals don't operate like that at all.
4:01
They use phased array antennas. A phased array uses hundreds of
4:05
tiny antenna elements working together to electronically form
4:09
a very narrow beam. The terminal basically stares
4:12
directly at a specific fast moving satellite and tracks it
4:16
precisely across the sky. Which.
4:18
Is more like looking through a narrow paper towel tube.
4:20
Oh exactly. You only see exactly what's at
4:22
the end of the tube and you block out all the peripheral
4:25
information. And that directional tracking
4:27
creates a physical mechanical barrier to spoofing.
4:31
If someone wants to feed false location data to a phased array
4:34
dish, they can't just overpower the signal from some random
4:38
direction on the ground. They have to physically place
4:40
their transmitter directly in the narrow line of sight between
4:44
the dish and the satellite in space.
4:45
Exactly, and they would also have to perfectly mimic the high
4:48
speed Doppler shift and the exact trajectory of a satellite
4:53
moving at roughly 7.6 kilometers per second, right?
4:57
Faking A stationary signal from a hill is a pretty simple trick.
5:01
Faking a moving object? Transmitting an encrypted signal
5:04
into a directional antenna? That's a much more difficult
5:07
engineering problem. So The funny thing is, people
5:09
figured out how to use this hardware for navigation long
5:12
before SpaceX officially pitched it to the government.
5:15
Yeah, out of sheer necessity. The Starling terminal runs a
5:18
local software interface called a GRPCAPI.
5:21
Let's actually pause on GRPCAPI for a second.
5:24
Sure, it was originally just a way for the terminal to
5:28
communicate with local network devices.
5:30
It's essentially a diagnostic port, but users discovered they
5:33
could use this interface to pull very precise round trip time
5:38
measurements. The terminal was measuring
5:40
exactly how long it took a signal to travel from the dish
5:44
to the serving satellite and back.
5:46
Which is basically just a ping. You're measuring latency.
5:48
Yes, exactly. And because the satellite knows
5:51
its own position and velocity, you can use those round trip
5:55
time measurements to calculate the position of the dish on the
5:58
ground. And maritime operators
6:01
completely turn this into a cheat code.
6:03
Oh big. Time because ships moving
6:05
through the Red Sea face constant electronic interference
6:08
from shore based actors. Their standard navigation
6:11
screens would just go completely blank.
6:13
Or worse, yeah. Worse, the spoofing would
6:16
convince the ship's computer that they were sailing over dry
6:19
land in the middle of a desert. Which is obviously a massive
6:22
problem when you're piloting a huge commercial cargo ship
6:25
through a narrow, hostile waterway.
6:27
You really rely on those coordinates to avoid reefs and
6:30
other vessels. So crews started using third
6:32
party telemetry apps connected to that local Starlink API.
6:37
They would literally select an option to use Starlink
6:40
positioning exclusively. When the military grade
6:43
equipment failed, the commercial Internet dish on the roof became
6:47
their only reliable source of navigation.
6:49
And it worked. It worked well enough to keep
6:50
them moving safely through the interference.
6:52
So if this software was actively saving ships in conflict zones,
6:57
I mean, why did SpaceX shut down unauthenticated access to it?
7:01
Well, it comes down to liability.
7:03
When you leave an unauthenticated port open like
7:06
that, you carry liability for the accuracy of the data.
7:09
Oh OK. You also open the network to
7:11
entirely unpredictable use cases.
7:13
If a ship runs aground because they were relying on a
7:15
diagnostic ping that was never intended for navigation in the
7:18
first place, the legal fault gets very complicated.
7:21
So shutting down the free accidental API basically forces
7:25
a shift toward a secure, authenticated, and, you know,
7:28
potentially monetized infrastructure.
7:30
Exactly. By closing the cheat code, they
7:34
transition from offering an incidental feature to preparing
7:37
a professional positioning service.
7:39
You close the free back door so you can eventually open a paid
7:42
front door. I mean, that makes total sense
7:44
from a business perspective. It does, but cutting off the
7:47
local API didn't actually stop people from pulling location
7:50
data. The academic community just
7:52
changed their methods. Right, they just found another
7:54
way in. Yeah, which brings us to the
7:56
whole concept of opportunistic positioning.
7:59
You navigate using signals that were never originally intended
8:03
for navigation, and you do it without the cooperation of the
8:07
company operating the constellation.
8:09
So if you can't ask the dish for its location through an API, you
8:12
reverse engineer the radio waves coming down from space.
8:15
Exactly. And academic researchers at the
8:17
University of Texas and Ohio State University did exactly
8:20
that with the KU band downlink. So they bypass the dish entirely
8:25
and just listen directly to the satellite?
8:27
Yeah. The researchers used a technique
8:29
called blind beacon estimation. Basically, in any communication
8:34
signal, there are repeating patterns used by the receiver to
8:38
synchronize with the transmitter.
8:39
Can you explain blind beacon estimation a bit more?
8:42
Sure. Think of it like listening to a
8:44
radio station that's mostly static.
8:46
You don't know the song playing, but you can hear the repeating
8:49
beat of the drum. If you isolate that drum beat,
8:52
you can figure out the tempo of the song without knowing any of
8:55
the actual notes. Earlier academic efforts mapped
8:58
a tiny fraction of the Starlink synchronization sequence, but by
9:02
applying this blind estimation technique, these researchers
9:06
successfully mapped the entire 240 megahertz orthogonal
9:10
frequency division multiplexing beacon.
9:12
Orthogonal frequency division multiplexing is that is a lot of
9:16
syllables. It's a mouthful, but it's just
9:18
the method the network uses to encode digital data on multiple
9:22
carrier frequencies, right? The full OFDM beacon is
9:25
essentially the complete time and frequency resource grid the
9:29
satellite uses to structure its data.
9:31
It's the scaffolding. By mapping the whole thing
9:34
rather than just the first two symbols, you extract vastly more
9:37
usable data from the signal. And what makes this really
9:40
interesting to me is the hardware they used to do it.
9:43
Oh yeah, it's wild. The researchers didn't require
9:46
$1,000,000 military tracking dishes to read these signals.
9:50
They used standard commercial low noise block down converters.
9:54
The LNBFS. Right, which is the exact same
9:57
cheap, low gain component you find bolted to a standard
10:01
satellite TV dish on the side of a house.
10:03
Literally a $30 part. Yeah, you can buy one at a
10:06
hardware store. And by mapping that full 240
10:09
megahertz beacon, the researchers achieved an 18
10:12
decibel processing gain. And that processing gain is the
10:15
mathematical advantage that allows the cheap receiver to
10:18
separate the Starling signal from the background noise,
10:21
right? It works even when the antenna
10:23
is just pointed straight up at the sky rather than mechanically
10:26
tracking a specific satellite. So you have a $30 part acting as
10:29
a highly sensitive space receiver.
10:31
Yeah. And with that processing gain,
10:34
the receiver can perform Doppler shift and carrier phase
10:37
tracking. When a satellite passes overhead
10:40
at high speed, the frequency of its radio signal compresses as
10:44
it approaches and stretches as it departs.
10:46
Like the sound of an ambulance siren changing pitch as it
10:49
drives past you on the street, it sounds higher as it comes
10:52
towards you and lower as it drives away.
10:54
Exactly the same physics. By listening to the exact rate
10:57
of that frequency shift across multiple satellites, the
11:00
receiver can mathematically deduce its own location on Earth
11:04
purely by observing the physics of the satellite's movement.
11:07
So you don't even need to decrypt the private Internet
11:09
data inside the signal. Not at all.
11:11
You just need to hear the physical structure of the signal
11:13
moving through space. Exactly.
11:15
But navigating by listening to Doppler shift makes sense in
11:18
theory, but applying it to a commercial broadband network has
11:22
some severe technical flaws. Yeah, because Starlink was built
11:25
to move high volumes of data, not to keep perfect time.
11:28
And timing is the core of any navigation system.
11:32
GPS satellites carry ultra precise atomic clocks that
11:36
measure the vibration of atoms to maintain perfect
11:39
synchronization. While Starlink satellites just
11:41
rely on much cheaper base oscillators, right?
11:45
Kind of like the quartz crystal in a standard wristwatch.
11:47
A bit more advanced than that, but yes, the same underlying
11:50
principle. These oscillators exhibit
11:52
nanosecond level jitter and their frequency drift can exceed
11:57
20 parts per million. Wow.
11:59
Because the clocks constantly drift, the Starlink satellites
12:03
have to regularly adjust their internal timing to stay roughly
12:07
aligned with global GPS time. And those adjustments are not
12:10
smooth corrections. No, they are not.
12:13
Researchers analyzing the signal found abrupt, completely
12:16
unpredictable 1 Hertz timing jumps.
12:18
Every second. Every single second, the
12:20
satellite makes a sudden adjustment that can be as large
12:23
as hundreds of nanoseconds and. We should probably contextualize
12:26
what 100 nanoseconds means for physical navigation.
12:29
Yeah, definitely, Because when you're calculating a position
12:32
based on the speed of light, a single nanosecond of timing
12:35
error translates to about a foot of physical distance error.
12:39
So if the satellite's clock is jumping around by hundreds of
12:42
nanoseconds every second, your calculated position on the
12:46
ground is jumping by hundreds of feet every second.
12:48
Which? Really makes you wonder how a
12:50
system with clocks jumping around wildly every single
12:53
second could ever hope to replace the ultra precise atomic
12:57
clocks on GPS satellites, right. If your phone's map application
13:00
jumped 300 feet every second, you would never make your exit
13:03
on the highway. No, you'd be all over the map.
13:06
And to complicate the timing even further, the network has
13:09
another operational quirk called the fixed assignment interval.
13:12
Right, the 15 second thing. Yeah, every 15 seconds, the
13:15
Starlink network completely reconfigures its beam
13:18
assignments to balance the data load among users.
13:21
They're constantly shifting bandwidth capacity to where the
13:24
physical demand is. Exactly.
13:26
But during those 15 second boundaries, the frame timing
13:29
behaviors of the signal completely change.
13:31
The sequence of frames being transmitted to a specific area
13:35
might shift its phase, its rate, or its stability.
13:38
So receivers trying to track the signal for navigation are forced
13:42
to adapt to a completely new timing environment. 4 * a
13:44
minute? Yep, you have clocks drifting
13:47
hundreds of nanoseconds of sudden correction every second,
13:50
and the entire network restructuring its physical beams
13:53
every 15 seconds. It's just a chaotic environment
13:56
for precision measurement. It is extremely chaotic, but
14:00
those erratic 1 Hertz jumps and clock drifts appear to be
14:04
software design choices, not fundamental hardware limits.
14:07
They programmed it to behave that way because it's efficient
14:10
for routing Internet traffic. The network doesn't really care
14:12
about perfect timing, it cares about throughput.
14:15
Exactly. And when researchers analyzed
14:17
the raw jitter of the oscillators between those
14:19
jumping corrections, they found that the hardware is actually
14:22
capable of maintaining the nanosecond level stability
14:25
required for precision navigation.
14:27
OK, so the ray materials are actually good enough?
14:29
Yes, and the network is already improving.
14:32
Researchers noticed that newer satellite generations in the
14:35
constellation exhibit far fewer of these abrupt micro
14:38
corrections. Interesting.
14:39
The code phase of the signal has become much more stable in
14:42
recent deployments, which suggests the software is being
14:45
actively refined to keep the timing smoother.
14:48
But even with smoother software, a crystal oscillator is just not
14:52
an atomic clock. It's always going to drift more
14:55
than the hardware used on traditional navigation
14:57
satellites. True, but that is where the
15:00
concept of fused LEOGNSS comes in to solve the problem.
15:04
OK, see, a traditional GPS satellite needs an atomic clock
15:08
because it's data connection to your receiver is extremely slow.
15:12
It cannot constantly update you on its clock error.
15:15
So the clock on board has to be basically perfect.
15:18
Right, because it takes several minutes to download even a tiny
15:20
text file over a traditional GPS signal.
15:23
Right, but Starlink has immense broadband capacity.
15:27
It doesn't need atomic clocks, it can just constantly broadcast
15:30
low latency clock corrections to users.
15:32
Now that makes sense. The satellite measures its own
15:35
clock drift against a reference station on the ground and uses
15:38
its massive data pipe to tell your receiver exactly how far
15:42
off its clock is in real time. So your receiver just applies
15:45
that correction. Mathematically, it just
15:47
subtracts the error locally. Exactly, giving you atomic level
15:51
decision from a cheaper oscillator.
15:53
You basically fix the hardware limitation with bandwidth.
15:56
And what that means for everyday stuff directly effects consumer
15:59
devices. Yeah, you can see this with the
16:01
direct to cell integration SpaceX is building with
16:03
T-Mobile. Oh, for sure.
16:05
Because the newer satellites driving that service operate in
16:07
the L band. Which is the exact same
16:10
frequency spectrum standard smartphones already used to
16:13
communicate with terrestrial cell towers.
16:15
Right, And because the phone already has the hardware inside
16:18
it to receive those specific L band frequencies, this
16:21
positioning system could theoretically bypass the
16:24
Starlink dish entirely. If the L band signal carries the
16:27
timing data and the clock corrections, the phone in your
16:30
pocket can process it just like a traditional GPS signal.
16:34
The only physical constraint is really just geometry.
16:37
Because you need multiple satellites to triangulate a
16:40
position. Right.
16:41
For a phone to get a fast, accurate location fix, it needs
16:45
to receive signals from multiple satellites simultaneously.
16:48
If your phone only connects to a single direct to sell satellite,
16:52
the positioning calculation will take several minutes.
16:54
Oh wow, yeah, the phone basically has to wait for the
16:57
satellite to move across the sky to gather enough Doppler data
17:00
from different angles. But as the constellation grows
17:03
and your phone can eventually see three or four of these L
17:06
band satellites at once, you achieve instant positioning.
17:10
Exactly. You bypass the vulnerability of
17:13
the weak MEO signals entirely and you process the high power
17:17
Leo signals directly on the handset.
17:19
That's incredible. The phone just leverages those
17:21
fused clock corrections, removing the need for a separate
17:25
phased array dish just for basic navigation.
17:28
So we are really moving away from relying on a single fragile
17:32
satellite system toward this fused network where Internet
17:36
providers act as the primary backup for global navigation.
17:40
The hardware is already in orbit, the signal processing
17:42
techniques are proven by the academic community, and the
17:45
network is actively refining its software to support exact
17:49
timing. I guess the lingering thought
17:50
for me is if a private broadband constellation becomes the
17:54
ultimate fall back for global critical infrastructure, does
17:58
the balance of power over global navigation shift from military
18:02
operators to a single commercial telecom company?
18:06
If you're not subscribed yet, take a second and hit follow on
18:08
whatever app you're using. It helps us keep making this.
18:10
We appreciate you being here. Also, check out our YouTube
18:13
channel for more business and tech updates.
18:15
There's a link in the description.