Open ChucklesOnGitHub opened 1 year ago
Here are more tests with 2 x OE LP headstages v1.1 and 2 x Intan 32ch headstages, using an OE Acq Board that has an Opal Kelly FPGA and GUI v0.6.5, Rhythm plugins v0.2.5.
The processing I did is here. Data folder is here.
I only printed out the data, I didn't analyze anything.
They look reasonable, albeit perhaps a somewhat larger offset on ours. No clipping here, I did not wiggle them hard as I was pretending they were mounted on an animal. Of course the axes don't match in physical position between headstages because one accelerometer is vertical and one is horizontal to the PCB.
Josh looked at this and commented: It seems like changing the input range is not urgent, but it's worth considering increasing the range in a future major update to this plugin. I think slightly reducing the resolution of the AUX channels is a reasonable price to pay for preventing any saturation.
We've run into a scaling issue in accelerometer channels. One of the axis shows clipping in the LFP Viewer and data.
Test using OE LP headstage v1.1 and an Intan 32ch headstage with an OE Acq Board that has an Opal Kelly FPGA and GUI v0.6.5, Rhythm plugins v0.2.5.
Figures correspond to accelerometer axes X, Y, Z from bottom to top.
Josh says: The 16-bit integers output by the AUX channels are scaled by a bitVolts value of 0.0000374, which leads to clipping at 2^15 * 0.0000374 = 1.2255. Options to solve this are: change the scaling factor to increase the available range, or remove an additional DC offset on axis 3, which seems to be the most affected.
Going to check a few more headstages to see if the pattern is consistent, so we can see what to do about it.