Autopilots are very good and they are servoing to the pressure altitude.
Many pressure altitude encoders used in modern aircraft (for example to drive altitudes that transponders report to SSR radar or via ADS-B) have 25 ft encoding resolution. That 25ft resolution is likely what is being seen here. Other encoders have 10 ft resolution but 25 ft is very common.
edit: trying to improve clarity/correctness but there is too much to cover here.
John Wiseman does great stuff with ADS-B Out data.
Also for pilots/aircraft owners/A&Ps: The FAA PAPR (Public ADS-B Performance Report) https://adsbperformance.faa.gov/PAPRRequest.aspx provide a summary of their aircraft's ADS-B performance, including all the broadcast GPS quality metrics and any reported failure flags etc. The PAPR system will email out the PDF report. The owner/pilot/A&P can reply to that email and request a Google Earth/kmz and Spreadsheet/CSV data for that flight showing all the received ADS-B transmissions including all those accuracy/reliability metrics. Interesting stuff and very useful for diagnosing problems with ADS-B Out installations. So sensitive you'll might see say NACp degrade as an aircraft banks steeply because the GPS antenna now has a view of fewer GPS satellites. Installations in most (non-experimental/non-light sports) aircraft effectively require use of PAPR to formally validate a new installation is working correctly. It's a good thing for owners to also just periodically check their aircraft's ADS-B performance using PAPR. I suggest just before and after each annual inspection for GA/light aircraft.
I once had ATC ask if everything was cool on flight following after a hundred foot drop and I was surprised they were paying that much attention. I had forgotten to put my life jacket on before a water transit and while I was putting it on handed it off to my wife who hadn’t taken lessons yet (she later got her license!). It was interesting to see that their tracking was precise enough for them to chime in.
It would have been cool to use a phone to record a GPS track with altitude and compare them. Pressure != GPS. Also wonder if there would be distinct jumps in the difference if they reset the pressure based altimeter to a different AWOS.
Not sure how it works in big planes, but in little ones you need to set your altimeter based on the local weather. The weather stations measure barometric pressure at their elevation and "correct it to sea level" you get this corrected reading over the radio and set it in your altimeter so your pressure-based altitude reading is corrected for local weather variations. Just going out flying for an hour the altimeter setting when returning to the same place might be off by a few millibar.
Of course, your initial point is still correct: there could be slight variations if using those local settings and getting different values, but you'd only see that below transition altitude.
The idea is all the planes use the same setting so the one at FL35 doesn't hit the one at FL36. But those are not exactly 35000 and 36000 feet above sea level.
There are many EFB (e.g. Foreflight), or log book, or other flight recorders you can use on an iPhone. And some can record the pressure transducer in the iPhone to record an approximate "pressure altitude". e.g. Naviter SeeYou Navigator intended for gliders can do that (but it's not unusual for modern gliders to have an array of sophisticated air data sensors and specialized variometers and flight computers that would feed the app this data over Bluetooth). Popular EFB software Foreflight will not use the iPhone pressure transducer, if you want pressure data there you need to drive that through an external interface like a Sentry ADS-B receiver that has a pressure sensor built into it -- or much better if the aircraft is equipped with ADS-B Out can receive the "own-ship" ADS-B Out broadcast pressure altitude from it's high accuracy encoder). Any in-cabin pressure traducer will be sensitive to the difference between calibrated static pressure and cockpit pressure, things like opening or closing vents, or varying the airspeed significant (and ram air pressure or suction on the cockpit exit vents) can cause observable changes. And when using an iPhone or similar, especially without a great GPS satellite overhead view (e.g. in high wing aircraft) you are likely not to get high-quality GPS altitude data. think best case ~ +/- hundred feet, worse case with little overhead GPS sat view, much worse... but those consumer GPS app is likely to happily display multiple decimal points of precision :-)
These are the instruments we are referring to not the ability of pilots. In fact in RVSM airspace the autopilot must be used.
Instruments must be very accurate given the reduced separation in RVSM airspace. Often on modern aircraft multiple altimeters are compared and voted to provide a single output provided to the displays and autopilot.
If a human can manage to keep it within 100 feet of a desired altitude, an autopilot most certainly can; it didn't require new technology in the 2000s. Autopilots in the 1960s/1970s weren't seesawing all over the skies.
The pressure difference between 5K MSL and 10K MSL at standard conditions is 14.6 kPa.
The pressure difference between 30K MSL and 35K MSL at ISA is 6.3 kPa.
For a given amount of aircraft-to-aircraft variability in their precision altitude sensing equipment, the resulting difference in actual altitude is more than double in RVSM airspace than in the lower altitude range above.
That's the reason for RVSM: there is less change in pressure with change in altitude, coupled with a very busy altitude range (such that controllers would have an operational need to pass traffic overhead with only vertical separation rather than being able to use vectoring to achieve lateral separation between aircraft).
It's not a linear relationship, but if I take an airplane with a 0.75 kPa absolute error in one direction and pass traffic with a 0.75 kPa absolute error in the other direction 1000' indicated above them, at low altitude, that 1.5 kPa total error is a little over 500 feet while IFR-IFR separation is 1000 feet minimum outside of RVSM. (These aircraft would likely be right on the border of passing a non-RVSM static system check.)
If I take those same two aircraft into the mid flight levels and pass one over the other at 30K and 31K feet, the total error is around 1200 feet, which is why non-RVSM aircraft cannot be separated by 1000 feet in RVSM airspace, because you don't know that they'll miss each other.
Improve the accuracy and precision of the static system and improve the examination criteria, making the airplane RVSM-capable, and now you can pass that traffic over each other at 1000' of indicated separation and be sure they'll miss.
[0] - There is a pilot training requirement, which is focused on knowing the rules for RVSM and does not involve a checkride.
You’re talking about getting different aircraft to agree between each other.
The post upthread expressed surprise at an aircraft maintaining a steady altitude to within tens of feet. That’s been a thing for many decades.
For autopilots servo'd to pressure altitude, holding altitude to within 0.02 kPa is more difficult than holding altitude to within 0.05 kPa or to within 0.30 kPa (which is roughly the private pilot checkride standard as-tested).
Modern autopilots are actually better at holding altitude to a very tight tolerance than ancient, analog autopilots. Both can hold standards well within the PPL ACS.
"more" difficult is obviously true, but the difficulty of holding an altitude is only a small part of the overall difficulty of RSVM.
In other words, RSVM is much more about accuracy than precision, and the claim was that planes were "probably fairly precise already". The reason they needed upgrades was to improve the accuracy, not so much to improve the precision.
This is because this way, if a pilot goes 3000 ft for instance, it will be exactly 3000 ft, if another pilot also wants to go 3000 ft on a collision trajectory, it will be a guaranteed collision. When altitudes are not that accurate, there is a higher chance it being just a near miss. The solution, I think, was to simply avoid round numbers. So now, it is 2950 ft, 3050 ft,...
I may have the details wrong, but I am quite sure about that problem being seriously considered.
(Although personally, I agree with the sibling comment that the variability is likely an artifact of the sensor resolution.)
btw: Aren't you the guy who tracks planes flying in circles? I follow you on Twitter. Such a cool project!
GPS altitude is used for vertical guidance for certain types of GPS approaches (i.e. "LPV" approaches[1]) and requires the airplane's avionics to be equipped with a WAAS[2] receiver that provides accurate altitude information.
[1] https://en.wikipedia.org/wiki/Localizer_performance_with_ver...
[2] https://en.wikipedia.org/wiki/Wide_Area_Augmentation_System
At cruise altitude, you're moving along at 500 mph, which is 777 feet per second. So going from +30 feet to -30 feet in a minute is just an adjustment of only about 5 degrees. You'd barely feel it, even walking down the isle. An acceleration of 33 ft/sec per sec is 1 g.
You experience greater changes in vertical motion on any flight you go on.
*edit: units
You would pretty obviously feel a change in pitch of 5° walking down the aisle.
You mixed feet per second and feet per minute. 60 feet of change across 777 feet of run is about 4.5° (inverse sin(60/777)), such as you'd experience if the change was in 1 second instead of in 1 minute.
Calculating 60' change in 777*60 feet, inverse sin (60/(777*60)) is 0.07°, which is why you don't feel that change in inclination of the aisle.