This is not a Position Paper on Periodic Pressure Altitude Calibration of FR.
This is a paper on why the above is not needed.
This document will not contain a technical discussion of if, how, why and when the accuracy of pressure sensors in Flight Recorders possibly deviates and whether it then varies with absolute altitude, calendar age, frequency of use or some other variable.
We will not present statistical evidence for either occurrence or non-occurrence of "drifting" of pressure readings. Nobody has these data anyway.
What we will do in this document is to deliberate why we in our sport measure altitude to begin with and what the “good enough” level of accuracy is. From there we will examine how this can be achieved in FAI's legal framework and address possible needs for adjustment of the Sporting Code, if necessary at all.
Spoiler alert: the conclusion will be that unlike regularly visiting a dentist, periodically having to calibrate the pressure sensor in your Flight Recorder is not actually good for you. It is at best an expensive nuisance and if we can get by without this recurring hassle then we all are off for the better.
5. Assessing errors in pressure measurement
Input: Altitude
Pressure at sea level
Output: Pressure at altitude
Input: Pressure measurement error
Output: Altitude with error
Example for offset error:
6. Assessing errors in pressure measurement
Input: Altitude
Pressure at sea level
Output: Pressure at altitude
Input: Pressure measurement error
Output: Altitude with error
Example for gain error:
17. -300
-200
-100
0
100
200
300
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
True altitude (m)
Altitude
error
(m)
?
1% of the pressure value
or
1% of the altitude value
according to 2011 spec
29. https://www.stssensors.com/blog/2020/06/30/the-long-term-stability-of-pressure-sensors/
From opposing position paper, section 3.1
SPAC
Surface Pressure Altitude Correction
• ±3 hPa (offset error)
• ±3 % (gain error)
within acceptable accuracy beyond 10000 m
• ±10 hPa (offset error)
• ±6 % (gain error)
within acceptable accuracy up to 6000 m
They advocate SPAC!
±6 hPa (offset error)
±5 % (gain error)
±8 hPa (offset error)
±5 % (gain error)
30. SPAC
Surface Pressure Altitude Correction
• ±3 hPa (offset error)
• ±3 % (gain error)
within acceptable accuracy beyond 10000 m
• ±10 hPa (offset error)
• ±6 % (gain error)
within acceptable accuracy up to 6000 m
https://www.stssensors.com/blog/2020/06/30/the-long-term-stability-of-pressure-sensors/
From opposing position paper, section 3.2
±6 hPa (offset error)
±5 % (gain error)
±8 hPa (offset error)
±5 % (gain error)
31. SPAC
Surface Pressure Altitude Correction
• ±3 hPa (offset error)
• ±3 % (gain error)
within acceptable accuracy beyond 10000 m
• ±10 hPa (offset error)
• ±6 % (gain error)
within acceptable accuracy up to 6000 m
https://www.amsys-sensor.com/downloads/data/MS5607-02BA03-AMSYS-datasheet.pdf
Looks OK!
±6 hPa (offset error)
±5 % (gain error)
±8 hPa (offset error)
±5 % (gain error)
32. SPAC
Surface Pressure Altitude Correction
• ±3 hPa (offset error)
• ±3 % (gain error)
within acceptable accuracy beyond 10000 m
• ±10 hPa (offset error)
• ±6 % (gain error)
within acceptable accuracy up to 6000 m
Competitions:
• Maximum Start Altitude: 1000-2000 m
• Minimum Finish Altitude: 400-600 m
• Contest Area Altitude Limit: < 3000 m
Silver badge:
• Loss of Height: 1000 m
• Height gain: 1000 m
Gold badge:
• Loss of Height: 1000 m
• Height gain: 3000 m
Diamonds:
• Loss of Height: 1000 m
• Height gain: 5000 m
Distance diplomas:
• Loss of Height: 1000 m
Conclusion:
Are all comfortably covered with SPAC
No need for mandatory periodic calibration
Not covered: Altitude records (typical > 10000m)
(internationally now requiring HAFR)
±6 hPa (offset error)
±5 % (gain error)
±8 hPa (offset error)
±5 % (gain error)
33.
34. Detecting a possibly degraded pressure sensor in need of service
Requirements on any method:
• Simple (no need for fancy tools and programs, perhaps assisted by a website)
• Robust and clear indication
• Should not require a cooperating pilot
• If possible avoid requirement for cooperative OO (for example registering current QNH)
• Enable verification in retrospect, using only the igc file
• If possible avoid the need to look up QNH from an online database (example: Copernicus)
Thoughts:
• GPS Altitude is quite good given sufficient SIU* and FXA** values
• NOT of interest: the absolute difference between GPS Altitude and Pressure Altitude in a single fix
• VERY MUCH of interest: the overall correlation of GPS Altitude and Pressure Altitude in a time window
• Noise exists in all signals and there are statistical means to deal with it
• A linear regression of Pressure Altitude over GPS Altitude ideally results in a slope=1 → perfect correlation without gain error
• We don’t need to know QNH but we need to ensure that it doesn’t change
• Solution: only use the first x minutes after launch, when the glider hasn’t moved much in (lat, long)
• This also makes calculation of the local geoid height unnecessary
*) SIU = Satellites In Use
**) FXA = Fix Accuracy (expressed as Estimated Position Error)
35. Detecting a possibly degraded pressure sensor in need of service
Example:
Only use fixes from the following section
• first 30 minutes after launch
• only 3D fixes
• only fixes with SIU ≥ 8
• only fixes with FXA ≤ 20 m
*) SIU = Satellites In Use
**) FXA = Fix Accuracy (expressed as Estimated Position Error)
slope: 1.007
intercept: -99.2
r_value: 0.99945
p_value: 0.0
std_err: 0.00085
≈1: PA correlates very well with GPS Altitude = no gain error
R2 ≈1: almost 100% of PA explained by GPS Altitude + offset
offset due to QNH ≠ STD (or offset error, we don’t know)
(offset error will be first revealed at higher altitudes or if we know the QNH)
((subtracting geoid height not necessary given sufficiently short time window))
p ≈ 0: statistically highly significant correlation between GPS
Altitude and PA (this is not just random noise)
std deviation ≈ 0: very little spread around the centerline
Conclusion:
No sign of degraded pressure sensor in this FR – no need for service
36. Detecting a possibly degraded pressure sensor in need of service
Example:
Only use fixes from the following section
• first 30 minutes after launch
• only 3D fixes
• only fixes with SIU ≥ 8
• only fixes with FXA ≤ 20 m
*) SIU = Satellites In Use
**) FXA = Fix Accuracy (expressed as Estimated Position Error)
slope: 1.007
intercept: -99.2
r_value: 0.99945
p_value: 0.0
std_err: 0.00085
≈1: PA correlates very well with GPS Altitude = no gain error
Question:
What if an ”unlucky combination”
of offset error and gain error
yields a perfectly linear relationship
between PA and GPS Alt?
Example:
-8 hPa offset error & -6 % gain error
38. -800
-600
-400
-200
0
200
400
600
800
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
True altitude (m)
Altitude
error
(m)
-8 hPa, -6 %
+8 hPa, +6 %
With SPAC
Question:
What if an ”unlucky combination” of offset error and gain error
yields a near-perfect linear relationship between PA and GPS Alt.?
Answer:
It will also yield a near-perfect SPAC correction → not a problem!
+8 hPa, -6 %
-8 hPa, +6 %