Skip to main content

Module arts2

Module arts2 

Source
Expand description

Ozark Aerospace’s ARTS2 flight computer’s interpreted data, as its software exports a flight: a # MET Acc Vel iAlt pAlt header, then comma-separated rows.

The ARTS2 records an accelerometer and a barometer, each read by a 10-bit converter, at a rate its owner sets, until its memory is full (ARTS2 User Manual, V1.0.0, Ozark Aerospace, 2007, p9, p16). Its PC software, DataAnalyzer, downloads the flight and works out what it shows: against the mission elapsed time (MET), a barometric altitude, an “accelerometer altitude”, the velocity and the acceleration (p11–12). Flight → Export → Interpreted Data writes them as a CSV file (p14). The manual describes neither the file’s columns nor their units, and its screens show feet, ft/s and ft/s² (p10–11), but the files among the project’s sample logs are not all in those units, as below. Each holds about 16,000 rows, the logger’s memory, one every 0.02 s (50 Hz), from launch:

# MET Acc Vel iAlt pAlt
0.000,0.00,0.00,0.00,0.00
0.020,61.00,0.61,0.01,0.00

Recognition. The first line that isn’t blank is # and then the five names MET, Acc, Vel, iAlt and pAlt, in that order and nothing else, parted by spaces, tabs or commas and compared without regard to case. The ARTS2’s raw data (MET,Acc(HI),Acc/Baro(LO),Baro(HI),State) and AltOS’s CSV, whose header also opens with # (#version,serial,…), name other columns.

Columns. Nothing in the file says whether the software was set to English or metric units, and the sample logs hold both, so every unit is decided from the file itself, by the checks below, each of which the reader runs; a note on the record gives each decision and its numbers.

ColumnQuantityUnit, and from where
METthe clock: seconds from launchthe name (mission elapsed time) and the apogee check (below), which a clock in another unit fails
AccQuantity::Acceleration, along the logger’s axis, gravity taken outstandard gravity (× 9.806 65 m/s²) or Vel’s unit per second, from the boost check (below)
VelQuantity::Speed: the logger’s integral of Accm/s or ft/s, from the apogee check and the step check (below), which must agree
iAltQuantity::InertialAltitude: the logger’s integral of VelVel’s length unit, from the inertial check (below)
pAltQuantity::AltitudeAboveGround: the barometric altitude, the primary trackVel’s length unit, from the step check and the barometric check (below)

The length unit, of Vel and pAlt, rests on two pieces of evidence that don’t depend on each other: how fast Vel falls at the apogee, and how far pAlt moves for one count of the barometer. A file is read only where both give the same unit.

The apogee check decides Vel’s unit, and with it that MET is in seconds. Vel is the logger’s integral of Acc: over the boost, Vel rises by Acc’s integral times one factor in every sample file (the boost check). At the apogee the rocket is slowest, so its drag is least, and the accelerometer reads near zero; with gravity taken out, Acc then reads −1 g, and Vel falls at g in its own unit: 9.81 m/s² or 32.17 ft/s². The reader finds the first sample after Vel’s highest where it reaches zero, fits a straight line to Vel over the second of coast before that zero, to the first sample at or below it, and takes the unit in which the line’s slope is within 25% of −g; m and ft are 3.28 times apart, so at most one unit can be. A clock in milliseconds or minutes would put the slope 1,000 or 60 times off, so the same fit pins MET in seconds. The second after the zero is left out: there the ejection’s jolt can come, as it does in the sample files 1 to 4.5 s after it.

A parachute out before the zero adds its drag to the second before it: one out at 25 m/s on the way up can make Vel fall at 3.7 g in m/s, which is 1.13 g in ft/s, within 25% of g in the wrong unit. So the window is the coast’s last second: where Acc jumps in the 2 s before the zero, the line is fitted to the second before the first jump instead, and a jump earlier in that second starts it after the jump. A row is a jump where it and the row after it are both further from the median of the 3 rows before than half that median, about half a g near the apogee, where Acc reads about −1 g in its unit; one row’s spike is no jump, and the sample files hold single-row spikes of a few g. A canopy out more than 2 s before the zero can make Vel fall at no more than 1.28 g over its last second, so it can’t pass for the other unit: a canopy whose drag grows as the speed squared stops a rising rocket within (v_t / g) π/2, v_t its descent rate, so taking over 2 s means v_t over 4g/π, 12.49 m/s, and then the last second’s fall, v_t tan(g · 1 s / v_t) a second, is at most 1.28 g. As of 2026-10-13, no jump ends the window in the 4 interpreted-data files among the project’s sample logs, and every one’s slope is within 4% of g in the unit its step gives. A file whose Vel never reaches its zero after its highest, whose window holds fewer than 3 samples, or whose slope is within 25% of neither unit, is refused (LogError::Unit).

The step check decides the length unit apart from Vel. pAlt is written to a hundredth but moves by its barometer’s count, a fixed step of pressure. The ARTS2 reads a Motorola MPXA4100A sensor with a 10-bit converter (manual, p16); the sensor’s output is its supply voltage times 0.010 59 P − 0.1518, P in kPa (MPXAZ4100A Series data sheet, Freescale, Rev 1, 2005, p4), so a count, 1/1,024 of the supply, is 1 / (1,024 × 0.010 59) kPa, 92.2 Pa, where the converter is read against the sensor’s supply (the manual doesn’t say). Near the ground that is Δh = Δp / (ρ g): the reader takes the count as 92.2 Pa ± 15% and the ground’s air from sea level at 5% over standard pressure and −20 °C, 1.46 kg/m³, to a 3,000 m site at 5% under its standard pressure (U.S. Standard Atmosphere, 1976) and 45 °C, 0.73 kg/m³, so a count is 5.46 to 14.83 m: written as 5.46 to 14.83 in meters, 17.91 to 48.64 in feet. The two bands don’t meet, so a step falls in one unit’s at most. The step is pAlt’s smallest change between two written values in a row, gaps passed over, that at least 3 changes come within 0.02 of: a single smaller change, as a closing row of zeros makes, isn’t a count. As of 2026-10-13, every sample file’s step is between 8.5 and 9 m, in the band of the unit its slope gives. A pAlt with no step written 3 times, a step in neither band, or a step whose unit isn’t the slope’s, is refused (LogError::Unit). The step is larger aloft, where the air is thinner, so a record whose single counts all lie high above a high site can fall outside the band, and is refused.

The boost check decides Acc’s unit. From the first sample to Vel’s highest, the reader sums Vel’s rise and Acc’s integral (trapezoids) over the steps where both are written: the ratio is Vel’s unit over Acc’s, about 1 when Acc is in Vel’s unit per second and about g in Vel’s unit when Acc is in standard gravities. Within 10% of one of them, that unit is taken; otherwise Acc is kept as written, Quantity::Unrecognized. The software’s own arithmetic sets the ratio, so only the rounding of the written values moves it: as of 2026-10-13 every sample file’s ratio is within 0.3% of the one its unit gives.

The inertial check decides iAlt’s unit, as the boost check does Acc’s: from the first sample to the apogee, iAlt’s rise over Vel’s integral is 1 when they share a length unit. Within 10% of 1, iAlt is read in Vel‘s; otherwise it is kept as written. The sample files’ ratios are 1.000 to 4 digits.

The barometric check holds pAlt to Vel. pAlt is the barometer’s altitude, not an integral, so it can’t match Vel‘s to the digit; but its highest value must be near the climb Vel integrates to by the apogee, which the reader requires within 25% in the length unit. 25% allows for a day’s air that isn’t the standard atmosphere (a few percent), the integration’s drift and the barometer’s step, and is narrow enough that m and ft can’t both be within it. The sample files’ highest pAlt is 0.95 to 0.99 times the climb. A climb under 176 m (20 of the barometer’s steps, below), where one step would use a fifth of the margin, or a pAlt outside the margin, is refused (LogError::Unit).

The primary track. FlightLog::time_s is MET and FlightLog::altitude_m is pAlt: the barometer’s altitude above the ground. The file starts at launch, MET 0, and holds no samples from the pad, so the zero of pAlt is the logger’s own ground reference, not in the file: the average of 16 pressure samples it takes as it enters flight mode on the pad, its “Zero Altitude baseline” (p17). In every sample file pAlt reads 0.00 on the first row; where it reads more than one barometer step (below) off zero there, a note says so. iAlt is kept as a channel: an integral of a single-axis accelerometer drifts, and in every sample file it falls to zero, and stays there, a minute or more before the barometer’s descent ends, while Vel reads hundreds to thousands below zero by the end of the file. So Vel is no vertical speed either: FlightLog::vertical_speed_m_s is None, as for the Blue Raven’s inertial velocity.

The closing zeros. In every sample file the last row, or the last two, write Vel as 0.00 where the row before reads hundreds to thousands below zero: they aren’t an integral of Acc, which moves Vel by under 1 a row there. A closing run of exactly 0 in Vel or iAlt is a gap, with a note, where the value before it lies further from 0 than twice what its rate (Acc times the boost check’s ratio for Vel; Vel for iAlt), held at its largest over the run, changes it by in the run’s time, plus a hundredth for the rounding. Where Acc is kept as written, it sets no rate, and Vel’s closing zeros are kept. In two of the sample files the last row writes 0.00 in every column; its Acc and pAlt are kept as written. The first row’s Acc is 0.00 in every sample file too, also kept.

The barometer’s step. The format’s resolution (LogFormat::altitude_resolution_m) is the coarsest of the sample files’ smallest pAlt steps near the ground, 28.89 ft (8.81 m); all of them are between 8.5 and 9 m. A count is a fixed step of pressure, so the step is larger aloft and at a higher launch site.

Cells. Every row must hold five cells, and its MET a number greater than the row before’s; otherwise the file is refused, naming the line. Any other cell that is empty or isn’t a number is a gap (NaN), counted in a note. A value finite as written that its unit makes infinite, such as 1e308 g, is refused, naming its line, as every reader refuses one. Lines may end in CR, LF or both, blank lines are skipped, and a byte-order mark is dropped. As every row must be five cells wide, the record never holds more values than the file has cells, and a row is read no further than its sixth cell before it is refused.

Functions§

read
Reads a log in this layout.
recognizes
Whether a file’s text is in this layout: its first line that isn’t blank is the export’s header, # MET Acc Vel iAlt pAlt.