Magnetometer Sensor Noise Std is scaled by a factor of 1.5

[AmolikaSoni]

Describe the bug
The standard deviation provided as input for magnetometer modelling is scaled by a factor of 1.5

To reproduce
Locate line no. 71 in https://github.com/AVSLab/basilisk/blob/develop/src/simulation/sensors/magnetometer/magnetometer.cpp
auto nMatrix = (this->senNoiseStd * 1.5).asDiagonal();

Expected behavior
Not sure on why 1.5 is being multiplied.

Screenshots
Code lines already mentioned above

[schaubh]

Howdy @AmolikaSoni , thanks for the info. I'm not sure either why this 1.5 multiplier was added there in the original code. I'll take a look this weekend and correct or document if needed.

[AmolikaSoni]

Hi @schaubh, any updates on this? I just found the same scaling of 1.5 in the coarse sun sensor modeling as well! https://github.com/AVSLab/basilisk/blob/develop/src/simulation/sensors/coarseSunSensor/coarseSunSensor.cpp

[schaubh]

Howdy @AmolikaSoni , I found the same. I searched the BSK code base for the use of .setNoiseMatrix() and found that only the magnetometer and CSS modules use the 1.5 multiplier. At this point I have a task to make a branch to remove the 1.5x term and document this change in the known issues file.

[sdunlap-afit]

Hi @schaubh, While reviewing the following document, I came across multiple references to scaling standard deviations by 1.5 and remembered this thread. Is the document correct, or should I provide the desired standard deviations without scaling? Thank you for your time.

"where senRotNoiseStd and senTransNoiseStd are input by the user as 1.5x the desired standard
deviation as scalar floats."

https://github.com/AVSLab/basilisk/blob/develop/src/simulation/sensors/imuSensor/_Documentation/BasiliskCorruptions.pdf

[schaubh]

@sjkcarnahan , you wrote this original documentation. Do you recall why this 1.5x factor was put in there? Some modules used it, most didn't?

[schaubh]

I spoke with the original author and he said this 1.5x factor was added to ensure the noise had the expected standard deviation. We will create a test to validate this. If correct after we test this, I don't think the coder should have to set this 1.5 in the BSK module. I'd like to look at a solution where the user sets the expected noise standard deviation and the gaussMarkow library does the required scaling.

[schaubh]

We can verify that the 1.5x factor is required to get the expected noise behavior. We are looking at ideas to avoid having the module need to this easier to setup.

[AmolikaSoni]

@schaubh I read the document shared by @sdunlap-afit. If you see the recommendations section, it does mention that "make it so that the given standard deviation does not have to be scaled by 1.5x" (Ref. section 2.2 ->2.2.1->6)
https://github.com/AVSLab/basilisk/blob/develop/src/simulation/sensors/imuSensor/_Documentation/BasiliskCorruptions.pdf