LSM6DSV16X SFLP Performance

[mmWaver]

In LSM6DSV16X DS13510-Rev 4 Page 8, Table lists sensor fusion performance:

Question 1: are these values guaranteed over temperature and all angles, or at a particular angle and at room temperature?

Question 2: without a magnetometer or GNSS heading input, what is meant by heading/yaw accuracy? Is it a drift value (over 5 minutes) from whatever value it is given at the beginning of 5 minutes?

Thank you.

[KWine]

With accel and gyro only, yaw/heading is relative only and subject to drift. After any significant motion the heading will change and become essentially arbitrary. It really cannot be relied on for absolute orientation estimation. However, the roll and pitch are likely to be as accurate as quoted or better (~1 degree) with proper accel/gyro calibration and repeatable even after significant motion, which is still very useful.

[Smithson]

The Holy Grail is being able to tell where you are in 3 dimensional space, but that is near impossible below cm accuracy over a short or long time.  The problem is the absolute zero point and hopefully it is not moving.  

All the methods suffer from something.  

I am talking about real human sized objects. 

Plus humans make mistakes in understanding the results.  

[KWine]

Two related problems: accurate 1) absolute orientation estimation and 2) dead reckoning navigation. Both are hard with cheap sensors in small form factors at ultra-low power, but this has been our goal for ten years.

Both require accurate sensors. For us it is the MMC5983MA magnetometer and LSM6DSM accel/gyro, although the LSM6DSV is even better. The LPS22DF baro allows ~10 cm accurate relative height estimation, great for tracking stair climbing, etc. Such a 10 DoF sensor suite should be enough for accurate AHRS and navigation in a perfect world!

By far the most important ingredient is proper calibration for bias, orthonomality erros, cross-axis, cross-sensor alignment errors, scale errors, etc. We have developed proprietary methods for this purpose with some success.

We have achieved the first of our goals with the USFSMAX Module, with which we routinely measure << 0.5 degree absolute heading accuracy.

As for dead reckoning, we would be thrilled with 1 meter accuracy over an hour. Dead reckoning is useful for GNSS gaps in cars, but it is also useful for gaps in magnetic field with no GNSS due to magnetic interference or anomalies. In this case, dead reckoning accuracy is determined mostly by gyro drift. Unfortunately, LSM6DSV gyro drift is not small enough for this task.

One solution is an external 3 ppm TCXO input to the ICM42688 accel/gyro to minimize gyro drift. Reducing the internal clock error from ~50,000 ppm to ~3 ppm means the gyro drift is greatly reduced, and dead reckoning navigation without GNSS or a magnetometer at 1 meter accuracy becomes possible, we just don't know yet for how long.

[Smithson]

We use a tilt meter developed by a small company as part of an accelerometer.   We duplicate some features with ST accelerometers.  

The tilt meter will provide an accuracy

draw a circle of 50 m radius,  put down a 2d nail on the circle, enscribe 100 lines on the head of the nail and the tilt meter will tell you from the center which line it is pointing at, according to a simple calculation. 

But we do large bridges, where tilt is an issue.  See picture. 

So we deal with different problems.  The critical issue is construction damage and large vehicles.  

[mmWaver]

Thank you KWine.

For roll and pitch, why do you think "low dynamic accuracy" is even better than "static accuracy"?

[mmWaver]

That's interesting. Why do you think a precision external clock would help reducing gyro drift?

[KWine]

From the ICM42688 data sheet:

"The CLKIN pin on ICM-42688-P provides the ability to input an external clock. A highly accurate external clock may be used rather than the internal clocks sources, if greater clock accuracy is desired. External clock input supports highly accurate clock input from 31kHz to 50kHz, resulting in improvement of the following:
a) ODR uncertainty due to process, temperature, operating mode (PLL vs. RCOSC), and design limitations. This uncertainty can be as high as ±8% in RCOSC mode and ±1% in PLL mode. The CLKIN, assuming a 50ppm or better 32.768kHz source, will improve the ODR accuracy from ±80,000ppm to ±50ppm in RCOSC mode, or from ±10,000ppm to ±50ppm in PLL mode.
b) System level sensitivity error. Any clock uncertainty directly impacts gyroscope sensitivity at the system level.
Sophisticated systems can estimate ODR inaccuracy to some extent, but not to the extent improved by using CLKIN.
c) System-level clock/sensor synchronization. When using CLKIN, the accelerometer and gyroscope are on the same clock as the host. There is no need to continually re-synchronize the sensor data as the sensor sample points and period are known to be in exact alignment with the common system clock.
d) CLKIN helps EIS (Electronic Image Stabilization) performance by providing:
o Very accurate gyroscope sample points for use during integration to find true angular displacement.
o Automatic time alignment between the motion sensor and the host and potentially the camera system.
e) Other applications that benefit from CLKIN include navigation, gaming, robotics"

[Smithson]

As I was looking at the LSM6DSV16X on the Sensor Tile Box pro.  The timing in the output files shows

A linear regression with reasonable accuracy and an added Fourier Series error.  I have only done one test. 

I should be able to find the FFT results in the overall FFT.  I will look sometime in the future. 

This is 19.1 seconds.

[mmWaver]

I understand the sample alignment issue with whatever (host/camera) that uses the sample. But I don't know how a more accurate clock could reduce the bias in gyro output which is the main culprit of heading error without GNSS/Mag heading correction.