I seek an object’s footprint but I am getting signal-processing artifacts (or maybe just noise) in the force measurement. This is per firmware 0.14.32.
To ensure that object surface irregularities aren’t the cause, I inflated a balloon and pressed it onto the pad (per photo). That should result in a uniform pressure (and force) across the contact area.
I did this several times with different balloons. The data below is typical and from a single frame. In the 2D plot, the variations are apparent. To help visualize this a bit more, see the 3D plot.
Here is a 45 second video on Dropbox which makes the anomalies very apparent.
The relevant API statements follow:
//Set the frame content to scan force data
senselSetFrameContent(handle, FRAME_CONTENT_PRESSURE_MASK | FRAME_CONTENT_CONTACTS_MASK);
senselSetContactsMask(handle, CONTACT_MASK_ELLIPSE | CONTACT_MASK_BOUNDING_BOX);
I would appreciate any pointers to help reduce the variations for this isotropic sample.
Hi Brad – it sounds like you have some very interesting applications in mind. We’d love to hear more about them, as we may be able to offer more concrete advice.
In general, spreading very light pressure, as with a balloon, across a large area of the sensor in SCAN_DETAIL_HIGH mode will give you a lower signal-to-noise ratio.
To get better results, try switching to SCAN_DETAIL_MEDIUM mode, applying more pressure, or attaching elastic bumps of some sort under larger objects to concentrate the force in a few locations. SCAN_DETAIL_MEDIUM mode will also offer the most linear force response. We primarily provide SCAN_DETAIL_HIGH for detecting the shape of object boundaries, and for resolving very small objects.
I appreciate the interest and would be glad to share. Basically, I am trying to recognize an object from the pressure pattern it makes. For more detail, I would be glad to share under NDA.
I had tried the other scan rates and I attached Low and High below. There seems to be a bug with Medium, at least for my test cases; it results in one or two highly localized spots and not at all like Low or High.
I also attached the sorted force data for Low and High with expectations noted.
Hey there Brad,
Did you end up making any progress with regards to nulling out the unwanted sensor variation?
That’s fascinating that the variation is location dependent. How different were your results when you had the “DynamicBaseline” enabled?
I was able to reduce the variations by training a filter to ‘read’ the variations and subtract them, but it was still a bit too noisy for my needs. I set that problem aside for a while to work on other stuff.
I kinda hope that an improved Morph might have less of that noise due to a fabrication change and/or some firmware calibration.
Hey there Brad,
I’m sorry to hear that it didn’t work out. I was going to suggest that “filtering” method if you hadn’t already tried it.
Best of luck in your projects