diff --git a/README.md b/README.md index c30ef839e77140a96104eb5420c387949f18b929..f92878b2c756f068cf8d1be67253c1a7a5377c71 100644 --- a/README.md +++ b/README.md @@ -6,10 +6,14 @@ I spent the majority of this week (and last) stuck in debugging world with the e Below is a video of the system working. The motor I am attached to is unpowered, so I have been spinning it with a pair of vice grips, while trying to keep the jaws far enough away from the PCB to create noise. The motor is dual shafted, so on the top is the inductive encoder, and on the bottom is the [CUI AMT102-V](https://www.cuidevices.com/product/resource/amt10.pdf), which is a capacative, quadrature, through-shaft encoder that provides a total of 13 bits of noise free rotary position data (so 0.044 deg of resolution). I also considered just powering on the stepper and sweeping through full steps to normalize my encoder data, but have heard steppers can be trusted to +/-5% (which honestly seems quite conservative considering they're just stamped steel). In the end I also wanted to try and configure a quadrature decoder hardware peripheral on an ST micro, which was exciting and simple to get up and running, so I went with the designated encoder. -With both encoders, I can now plot my dubious encoder data against a known, trustworthy source, and begin to get an idea for how linear/sinusoidal my signal is, how repeatable it is peak to peak (remember this is a 6 pole encoder that was machined on a desktop router), and start to see what I am up against in terms of calibration. The results aren't quite as clean as I was hoping, but all of the makings of a solid encoder are there. +With both encoders, I can now plot my dubious encoder data against a known source, and begin to get an idea for how linear/sinusoidal my signal is, how repeatable it is peak to peak (remember this is a 6 pole encoder that was machined on a desktop router), and start to see what I am up against in terms of calibration. The results aren't quite as clean as I was hoping, but all of the makings of a solid encoder are there.  +If you look at the top (normalzied) and bottom (plotted against time) data you can see the importance of our reference encoder to really understand the accuracy of the system. The bottom plot is just the encoder readout plotted against time, and so there's a stretch at the beginning where I wasn't rotating the shaft + visible changes in pitch from my inadvertent speeding up or slowing down rotation. This is not at all going to be a sin-cos encoder, and instead looks much more like a piecewise triangle wave. I think this is a good thing, considering that more linear the data results in a more constant (and overall higher magnitude) encoder response for a given rotation angle. The caveat is that I won't be able to use the straightforward arctan trick to linearize my two signals, and instead will likely need to go straight to a lookup table off of a calibration. + +[<img src="images/dataCollectionVideo.png" width="800">](https://gitlab.cba.mit.edu/davepreiss/ldcoder/-/blob/master/images/rotorTarget%20.mp4) + ## HTMSTMAA Week 9 - 5/5/21 This week I designed (and ordered) two encoders built around the LDC1101, one is linear and one is rotary. Rather than try and integrate everything with an MCU at this stage, I went ahead and just broke out all of the necessary communication pins, similar to my last board, but this time with the LC tank fully integrated. Not a whole lot to say here until the boards come in and I can test them. In the meantime I have been searching for a suitable ADC replacement for the HX711 back over on the dynamometer project. It would be great to have an infrastructure for high resolution + gain analog signal measurement for a number of projects, this one included potentially. diff --git a/images/dataCollectionVideo.mp4 b/images/dataCollectionVideo.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..6e019d49de4e4c073110a4b53108d782a3dcd34b Binary files /dev/null and b/images/dataCollectionVideo.mp4 differ diff --git a/images/dataCollectionVideo.png b/images/dataCollectionVideo.png new file mode 100644 index 0000000000000000000000000000000000000000..7970e1a351aab47a64e61637a8e337f0bf3490fe Binary files /dev/null and b/images/dataCollectionVideo.png differ