Introductory notes

13 Oct 2007 stribog had a field test. actually our goal was burbot in the Oka, but it led us a pretty dance and went away. however, stribog data were not so bad.

The stand

We mounted our stand on a car generously provided by M. Grishin. this is its block diagramme:

[stribog boards connected to GPS, odometer and notebook]

Here is a photo of the stand having been mounted:

[aluminium box in the center of the roof, GPS antenna in the corner, the rest under the roof]

The stribog unit inside looks like this:

[main board under a net of teflon-covered wires]

The odometer counter was assembled at a breadboard:

[GPS receiver tied to a board with a couple of IC]

The trace

The whole trace was from Radujny to Colychovo, calibration circles by the Oka and half-way back.

[zig-zag 3km North-South; 700 m East-West]

The session time was about 1000 seconds.


Here is LM74-derived temperature. this sensor is mounted on the main board.

[drift from 29.7 to 30.4 deg. C]

The figure shows that the temperature was almost constant on the observed timespan.


Main board clock frequency drift; compared with PPS from Garmin 25LP.

[drift +/- 0.3 ppm over 1000 sec period]

Clock frequency uncertainty is about 0.05 ppm.

Main board clock drift; compared with PPS and UTC from NMEA-0183 messages (a quadratic polynom of time is subtracted).

[drift -3/+9 us with harsh angles]

Typical clock uncertainty is less than 1 us; the maximum is less than 3 us. the irregularities near 1200 seconds, 1400 seconds and so on were caused by noises on PPS line; at those periods no good PPS was available for many seconds.

[drift is within +/- 300 us over 1000 sec period]

Odometer clock uncertainty is lower than 100 microseconds. it is determined by communications delay between the main board and the odometer board. the UART rate was 9600 baud, so one can hardly expect any higher precision.

Accelerations spectrum

The next figure shows DFT of axial, lateral and vertical accelerometers output.

[the lowest is axial, then lateral; the vertical accelerations increase at higher frequencies]

Currently we can add no comment here.

Magnetometer calibration circles

To get the magnetic sensor calibrated the car moved to and fro by the Oka bank in Colychovo. here is the trace.

[to and fro lines, almost right; the length is about 600 m]

At the same time, HMC1022 sensors were drawing this curve:

[deformed curves around a neat reference circle]

Certainly, we would like to get more regular circles, but the life is hard.

Simple processing

To estimate sensors qualities, primitive mechanisation routines were run. they use two-dimensional motion model; the sensors used are axial (forth-back) accelerometer or odometer to compute path length and XY-magnetic sensors or vertical angular rate sensor to get velocity direction. it was assumed that the vehicle always moves strictly forth. GPS-derived position and velocity were used as reference and initial points of extrapolation.

Thus, we have four mechanised systems: acc+gyro, acc+mag, odo+gyro and odo+mag, where acc is for "accelerometer is used for axial computations", gyro - "angular rate sensor is used to compute the direction" and so on.

The next figure presents axial and lateral errors as functions of extrapolation time.

[many growing curves]

We would note that for small extrapolation times (te) axial and lateral errors show precision of path length and direction sensor respectively, whereas for larger te the lesser error tends to "merge" in the larger one. so, "acc+mag lateral" error growth is greatly influenced by accelerometer errors for te>5 s. we can not explain "odo+gyro axial" error behaviour on te<6 s.

We see that

  1. For te more than 2 seconds, the accelerometer (axial) error is the most one; it is more than 100 m when te>30 s for te=1 s it is about 0.4 m.
  2. Lateral error curves show that the angular rate sensor is better than the magnetic one for te<5 s; the gyroscope-caused error is about 0.2 m for te=1 s, while magnetic sensors show about 0.7 m; however, the gyroscope-derived direction quickly degrades with te growing.
  3. For te>2 s the odometer error is the least; it is not more than 2 m at te=30 s.
  4. odo+mag system is much better than any of the other three systems for te>10 s.

The next figures are examples of extrapolated tracks for extrapolation time equal to 5 seconds: just to show what it looks like.

[GPS points, {acc,odo}+gyro traces]

[GPS points, {acc,odo}+mag traces]

Last modified: 07 Apr 2008

Copyright (C) 2008 Ineiev <>

Verbatim copying and distribution of this entire document is permitted in any medium, provided this notice is preserved.

You may also make and publish an unofficial translation of this document into any language including good English. In this case you should additionally provide a disclamer in the language, into wich you translate, clearly stating that it is not the original text of the author.