SparkFun Forums

[pizza]

tried the self test for positive and negative fields. see attached

[pizza]

all right remember the anomaly(bending of elliipse) i posted earlier? it was due to placing a brushless outrunner nearby. the motor is not moving but the sensor is.

i've discovered an easier way to obtain data manually. instead of rotating/turning/tilting it on a wooden table, place it on your swivel chair. it's not flat though, hence inaccurate. best would be to have a turntable.

it's very much calibrated into a perfect circle. but when i slide it across the table with one axis parallel to the edge, heading increases increases. this is very weird. is the pc causing some magnetic field?

[esklar81]

best would be to have a turntable.

How about the one in your microwave oven?
Eric

[pizza]

best would be to have a turntable.

How about the one in your microwave oven?
Eric

hahaha! nice suggestion!
but well my quadrotor can't fit in and i don't have one in the lab.

i think the magnetic field that is causing more offsets is the hewlard packard oscilloscope on my desk.

by the way, how do u calibrate for z axis scale factor and offset? mount the sensor perpendicular?

mounting a sensor perpendicular is sort of like weird to me because i do not think i can do it accurately.. the same goes for finding the bias of accelerometer z

[pizza]

hey guys, i mounted my sensorboard perpendicular on a cardbox(pretty desperate eh) to calibrate the z axis..
hence, x and z axes are not on flat surface (2d) ready to draw a circle when rotated.

please note that x and y have been calibrated for scale factor and offset while z is not yet.

please see attached plot.

holy sh*t! why is x against z being rotated?(soft iron distortion)? whereas x and y is not. pretty weird eh!!!
i did not add any soft iron to it.

[sebmadgwick]

This article explains things very well, particularly soft and hard iron calibration.

http://www.sensorsmag.com/sensors/motio ... fects-6475

[fax8]

Hi guys, I'm moving my first steps with the HMC5843 magnetometer and I stumbled upon this little Arduino library for the HMC5843 and with that I've been able to plot the attached graph.. Which doesn't look bad to me.

So, I'm now wondering.. is the calibration process of this chip (moving it around all the axes to get the correct offsets) is really needed or the calibration routine in that library is enough? I'm experiencing with MARG sensors arrays for using them in HCI (tangible interfaces) so I don't think that I need a super precise readings..

What do you think?

Thanks,

Fabio Varesano

[fax8]

Just made a 3d version of my graph above .. you'll find it on my blog with all the gnuplot and python sources needed to plot the graphs, just in case anyone is interested.

Of course they looks like an ellipsoid...

[pvmellor]

All,
I'm have a few questions re: the HMC5843...

1) is the chip doing a set/reset strap pulse each reading?
2) for cross axis calc (i.e. tilting of the ellipse) are the sensors arranged with primary/secondary axes of x-y, y-x, and z-y?
3) i was surprised to read that the full offset bias if considerably bigger than the actual reading. Several ways are given for this to be corrected in the app note. I am assuming that this part uses the "digital subtraction" method, and that it obtains the offset from the set/reset pulse. Anyone know how often this proceedure is performed? Is it a one-off or done each reading (if the latter then i would not expect any other offset correction or temp correction would be needed).
4) The 1043 sensor in this part seems to have two offset straps, one for xy and one for z. Any thoughts on the possibility then that these may vary electrically and that this may mean gains calculated via the offset strap won't be accurate xy to z?
5) i have the old 100uf electrolytic reservoir cap in place to make it work - i'm worrying that the time for the recharge after a set/reset pulse might have gone up a bit with this, and the noise could affect the readings. Anyone seen anything useful on a scope re: this?
6) if you keep part still and take a series of readings - what variance/std dev do you typically get?

Thanks,
Paul

[fax8]

@pvmellor hi there, I can't help you with most of your questions but if you have some code for computing 6 I would happily run it on my HMC5843..

EDIT: nope... code don't needed .. I'm creating a simple octave script to compute those statistics.. talk to you soon..

[fax8]

See bottom of this page: http://www.varesano.net/blog/fabio/firs ... ts-results

You'll find my STD and VAR result as well as the Arduino code used, the Octave script and my dump.

[UhClem]

1) is the chip doing a set/reset strap pulse each reading?
2) for cross axis calc (i.e. tilting of the ellipse) are the sensors arranged with primary/secondary axes of x-y, y-x, and z-y?
3) i was surprised to read that the full offset bias if considerably bigger than the actual reading. Several ways are given for this to be corrected in the app note. I am assuming that this part uses the "digital subtraction" method, and that it obtains the offset from the set/reset pulse. Anyone know how often this proceedure is performed? Is it a one-off or done each reading (if the latter then i would not expect any other offset correction or temp correction would be needed).
4) The 1043 sensor in this part seems to have two offset straps, one for xy and one for z. Any thoughts on the possibility then that these may vary electrically and that this may mean gains calculated via the offset strap won't be accurate xy to z?
5) i have the old 100uf electrolytic reservoir cap in place to make it work - i'm worrying that the time for the recharge after a set/reset pulse might have gone up a bit with this, and the noise could affect the readings. Anyone seen anything useful on a scope re: this?
6) if you keep part still and take a series of readings - what variance/std dev do you typically get?

1) The data sheet doesn't say but I suspect it does a set and reset pulse for each reading.
5) Replace the 100uF electrolytic with a 4.7uF ceramic. Low ESR is what is important here.
6)

allan.png

[UhClem]

5) i have the old 100uf electrolytic reservoir cap in place to make it work - i'm worrying that the time for the recharge after a set/reset pulse might have gone up a bit with this, and the noise could affect the readings.

I guess I should mention that a fixed amount of charge (approximately) is being transferred from this capacitor to the 0.22uF set/reset capacitor. Voltage drop with the reference design values of 4.7uF and 0.22uF is about 154mV.

Remember that this is the digital supply so ripple has no impact on device operation. Unless it gets large enough to drop below the minimum operating voltage anyway.

[fax8]

David, I'm wondering how you managed to measure that.. could you explain?

[newbee]

Hi 2 all,

I also took the readings of my HMC5843, and the x vs y readings form a near perfect circle. See the plot.
The offset is around 63 in the X axis and 20 in the Y axis. However, the offset in the Z axis is very large.
I read a value of -640 when the device lies flat on the table.

Does someone also find such a large offset, or can it be a defect of the device?

BTW the offsets one finds with the pos and neg bias in the device do NOT AT ALL
correspond to the real world offsets, as they were all around +13 in the strap cal.

Thanks for reporting on your Z axis readings!!

Andreas