5 ms·

Modern GNSS can reliably measure altitude at <50cm 2-sigma accuracy using techniques such as RTK and PPP-AR. Modern GNSS is a better source of altitude data tha

Modern GNSS can reliably measure altitude at <50cm 2-sigma accuracy using techniques such as RTK and PPP-AR. Modern GNSS is a better source of altitude data than barometers. Likely SpaceX is using a combination of radar altimeters and GNSS. Source: I do engineering on modern GNSS

cryptoz 7y ago
Super interesting info. Sorry for my annoying questions. Does modern GNSS receive that level of accuracy in cities among buildings, or is that an ideal case like what SpaceX has, being outside? Also, how expensive is a modern GNSS receiver that can get that level of altitude accuracy? Is this something that could make it into phones some day?
- walrus01 7y ago
  To add to the parent comment, super high precision gnss/rtk is often paired with multiple inertial measurement units, not very dissimilar from what you find in a smartphone. Also used in COTS drone flight controllers. Software can combine data from an imu that some vertical acceleration had taken place, with data from gnss sensors that altitude is now 2m higher, for additional level of confidence in what the gnss sensors are reporting.
- njoubert 7y ago
  Not an annoying question whatsoever. Dense urban environments are challenging. The error distribution is more complex there, its hard to definitively say. The numbers I'm quoting is for open sky environments where you have plenty of satellites and little multipath. For the average end-user, these are $500 GPS units (this is the one I work on: https://www.swiftnav.com/piksi-multi https://www.swiftnav.com/piksi-multi). The technology is already moving down the supply chain to ~$10 GNSS units from Broadcom and ST (we're doing this: https://www.swiftnav.com/news/swift-%E2%80%8B%E2%80%8Bnavigation-%E2%80%8B%E2%80%8Bintroduces-starling-gnss-positioning-engine-and-interoperability https://www.swiftnav.com/news/swift-%E2%80%8B%E2%80%8Bnaviga...) The issue with phones is mainly an antenna issue. Your phone antenna is tiny and squeezed in alongside half a dozen other antennas and radios. This degrades the GNSS signal quality a lot. A stopgap here is external GNSS antennas for your phone: you might plug your phone into your car and get lane-accuracy turn-by-turn directions, but it's unclear when we'll have high accuracy GNSS inside phones, but most major phonemakers are indeed working on this.
  - monk_e_boy 7y ago
    ... $10?! Really? Can we hook one up to a Raspberry Pi? This would be a game changer for my sport (kitesurfing)
    - Klathmon 7y ago
      Depending on the update rates and accuracy you need, there are many commercially available GPS breakout boards and small cheap modules under about $20 I've used [1] which can be found for about $18 before on some toy projects. [1] https://www.u-blox.com/en/product/neo-6-series https://www.u-blox.com/en/product/neo-6-series
      - monk_e_boy 7y ago
        GPS doesn’t do height accurately :( 10m error
- TaylorAlexander 7y ago
  I can’t answer all the questions but I can talk about RTK techniques. I recently used two $75 GPS receivers and two raspberry pis to create an RTK GPS system (real time kinematics). With RTK you have one receiver that is fixed - it does not move. The other receiver is on the thing whose position you want to measure, and the two devices are connected through the network. Because the system knows the fixed station is not moving, any drift in measurement represents the real time error of the current satellite constellation. The moving unit subtracts that error from its own measurements to get very precise figures for its position. In the system I built, it seems to have centimeter level repeatability. That is, if I am looking at the read out in real time and I stop moving the receiver, the drift in the measurement is a couple of centimeters. If I move it 20 centimeters that is pretty closely reflected in the read outs. Currently I don’t know the absolute accuracy of my system but the relative accuracy is very high!

cryptoz 7y ago

Super interesting info. Sorry for my annoying questions. Does modern GNSS receive that level of accuracy in cities among buildings, or is that an ideal case like what SpaceX has, being outside? Also, how expensive is a modern GNSS receiver that can get that level of altitude accuracy? Is this something that could make it into phones some day?

walrus01 7y ago
To add to the parent comment, super high precision gnss/rtk is often paired with multiple inertial measurement units, not very dissimilar from what you find in a smartphone. Also used in COTS drone flight controllers. Software can combine data from an imu that some vertical acceleration had taken place, with data from gnss sensors that altitude is now 2m higher, for additional level of confidence in what the gnss sensors are reporting.
njoubert 7y ago
Not an annoying question whatsoever. Dense urban environments are challenging. The error distribution is more complex there, its hard to definitively say. The numbers I'm quoting is for open sky environments where you have plenty of satellites and little multipath. For the average end-user, these are $500 GPS units (this is the one I work on: https://www.swiftnav.com/piksi-multi https://www.swiftnav.com/piksi-multi). The technology is already moving down the supply chain to ~$10 GNSS units from Broadcom and ST (we're doing this: https://www.swiftnav.com/news/swift-%E2%80%8B%E2%80%8Bnavigation-%E2%80%8B%E2%80%8Bintroduces-starling-gnss-positioning-engine-and-interoperability https://www.swiftnav.com/news/swift-%E2%80%8B%E2%80%8Bnaviga...) The issue with phones is mainly an antenna issue. Your phone antenna is tiny and squeezed in alongside half a dozen other antennas and radios. This degrades the GNSS signal quality a lot. A stopgap here is external GNSS antennas for your phone: you might plug your phone into your car and get lane-accuracy turn-by-turn directions, but it's unclear when we'll have high accuracy GNSS inside phones, but most major phonemakers are indeed working on this.
- monk_e_boy 7y ago
  ... $10?! Really? Can we hook one up to a Raspberry Pi? This would be a game changer for my sport (kitesurfing)
  - Klathmon 7y ago
    Depending on the update rates and accuracy you need, there are many commercially available GPS breakout boards and small cheap modules under about $20 I've used [1] which can be found for about $18 before on some toy projects. [1] https://www.u-blox.com/en/product/neo-6-series https://www.u-blox.com/en/product/neo-6-series
    - monk_e_boy 7y ago
      GPS doesn’t do height accurately :( 10m error
TaylorAlexander 7y ago
I can’t answer all the questions but I can talk about RTK techniques. I recently used two $75 GPS receivers and two raspberry pis to create an RTK GPS system (real time kinematics). With RTK you have one receiver that is fixed - it does not move. The other receiver is on the thing whose position you want to measure, and the two devices are connected through the network. Because the system knows the fixed station is not moving, any drift in measurement represents the real time error of the current satellite constellation. The moving unit subtracts that error from its own measurements to get very precise figures for its position. In the system I built, it seems to have centimeter level repeatability. That is, if I am looking at the read out in real time and I stop moving the receiver, the drift in the measurement is a couple of centimeters. If I move it 20 centimeters that is pretty closely reflected in the read outs. Currently I don’t know the absolute accuracy of my system but the relative accuracy is very high!