Hardware build record

daylanKifky

By the way YJBeetle , I'm curious about the actual performance of the HMC5843 in comparison with the LSM9DS1. Do you have both sensors with you by chance? It would be great to perform a simple test which will give quantitative results about the real difference between them. I would go like:

Make recordings of both sensors in different environments, a small room with lots of electronic or ferromagnetic objects (considering also that most modern buildings have metal structures that affect the magnetic field, a big room with few, far away objects, an outside location, etc. Each recording should be similar to the magnetic sensor calibration procedure (rotating it in every possible orientation)
Apply the ellipsoid fitting algorithm to the datasets and compare the results with a target sphere centered at 0,0,0
The difference obtained should represent a good estimate of the performance of each sensor in real scenarios.

It would also be really nice to measure the geomagnetic azimuth accuracy of each sensor before and after the Kalman filtering in both disturbed and clean environments. A quick search in scientific bibliography seems to indicate that after complex compensation steps the mean azimuth is around 3 degrees

YJBeetle

daylanKifky
是的我拥有这两个传感器。
事实上在几个月前我对比了好几个不同的磁场传感器，当时对比的有：HMC5883 QMC5883L HMC5843 LIS2MDL LIS3MDCL。当时得到的结论是HMC5843是最合适的。他的精度比较低，只有12bit，但是他拥有更低的噪音和更好的线性度。
最近这段时间我比较忙碌，在这段时间忙完之后我将会进行你说的测试，但是他不会太快完成，因为我这里的都是装在编带里的裸芯片，我还需要去订购一块PCB焊接上用于测试。

Yes I have both sensors.
In fact a few months ago I compared several different magnetic field sensors: HMC5883 QMC5883L HMC5843 LIS2MDL LIS3MDCL. It was concluded at the time that HMC5843 was the most suitable. His precision is lower, only 12bit, but he has lower noise and better linearity.
I've been busy these days. After this time is over, I will do the test you mentioned, but it won't be done too soon, because all I have here are bare chips in tapes, and I still need Go and order a piece of PCB soldered on for testing.

YJBeetle

About Linearity
HMC5843 :
https://www.sparkfun.com/datasheets/Sensors/Magneto/HMC5843.pdf
in Page 2: Linearity: 0.1 ±% FS

iam not found linearity of LSM9ds1 in datasheet,
but i found linearity of lis3mdl , These chip is designed by ST .
https://www.st.com/resource/en/datasheet/lis3mdl.pdf
in page 8: ±0.12 %FS

daylanKifky

Great! there's no rush, but it will surely provide precious information on future improvements of the magnetometer or the way its data is handled
Thanks!

ZackJiang

最后全部运行了吗？

cubesky

ZackJiang 没有我还没时间重新校准，这个月比较忙

ZackJiang

YJBeetle 我还是喜欢直接贴片厂去做

cubesky

I use the latest notochord, but notochord --scan <IP> says.

pi@notochord:~/notochord/bin $ ./notochord --scan 192.168.204.17
notochord: lib/oscpack_1_1_0/osc/OscOutboundPacketStream.cpp:166: osc::OutboundPacketStream::OutboundPacketStream(char*, std::size_t): Assertion `sizeof(osc::int32) == 4' failed.
Aborted
pi@notochord:~/notochord/bin $

cubesky

cubesky
I found notochord only run on 32bit raspbian.
Blender Addon can only run on Blender 2.93 LTS, not Blender 3.01.

daylanKifky

Hi @cubesky is good to know you are back testing the suit

cubesky I use the latest notochord, but notochord --scan <IP> says.

This can be fixed by modifying the typedefs in lib/ospack_1_1_0/osc/OscTypes.h the osc::int32 should be a 4 bytes integer.

cubesky Blender Addon can only run on Blender 2.93 LTS, not Blender 3.01.

Could you please let us know which error are you getting and in which platform are you running Blender?

cubesky

I recalibrate K-Ceptor in a park. Now, I will test default biped configuation again.

cubesky

Sorry for delay about 1 month, I'm busy on my job this month. And @YJBeetle is also busy for his job too.

This month, we made an PCB for finger sensor test, this PCB is soldered but doesn't test. I will test it these weeks.

And I entrust a PCB designer to design a Hub using CM4, you can see the image in the pervious post (Powering by a lead storage battery). It allowed powered by USB Type-C or a PH2.0-2P connector with 5V to 24V battery and switch between them online. (This port will connect to 18650 battery at last.)

No, I got the same result like pervious video. Rotation and movement not correct.

I recalibrate all K-Ceptor facing south too, but there is no change.

I talk to @ShiroSaki with this problem, she think this is a software issue on blender addon during calculate the quaternion.

@daylanKifky

daylanKifky

cubesky And I entrust a PCB designer to design a Hub using CM4

That's really neat. Would love to know more about this project.

cubesky No, I got the same result like pervious video. Rotation and movement not correct.

I recalibrate all K-Ceptor facing south too, but there is no change.

It's a pose-calibration issue for sure. We are currently working implementing the pose-calibration algorithm in the blender addon and we will be releasing a beta really soon. This new calibration can be performed looking at any cardinal point and it solves many issues with the current one.

In the meantime have you tried doing the T-pose looking other cardinal points? like East or West

cubesky

daylanKifky
I running Blender on Windows, using Blender 3.01. Demo data can not move Test Cube.

cubesky

daylanKifky That's really neat. Would love to know more about this project.

There are a few problem now, I trying to solve it. He designed the PCB in Altium Desiger, I will transfrom it to KiCad if I can.

daylanKifky In the meantime have you tried doing the T-pose looking other cardinal points? like East or West

Yes, all cardinal is tested, divided by 45deg, but no luck. 🙁

cubesky

daylanKifky
Is there any ETA for this addon release?

daylanKifky

cubesky I running Blender on Windows, using Blender 3.01. Demo data can not move Test Cube.

If it's just that there's probably a firewall blocking the incoming osc messages for Blender at port 6565.
Did you get any crashes of error messages from the addon?

cubesky

daylanKifky
I will check the blender console later. I'm using demo data mode in Blender Addon, not osc message, and I already disabled firewall on my PC.

cubesky

daylanKifky
It seems this is a random problem or addon conflict with other addon. I reinstall addon and the issue is gone.

daylanKifky

Most of the code is already implemented in the public repo, branch new_calibration. We are currently fixing some minor details and expect to have a more stable version next week. We'll also do a video showing how to perform this new calibration procedure

cubesky

daylanKifky Oh, That's great. I find that CI on Blender Addon Gitlab Repo says new_calibration branch is failed to compile, so I doesn't test that version. If this new version is release, I will test it ASAP.

« Previous Page Next Page »