Hello everyone, we hope you’ve enjoyed the time since our last update as much as we have. We’re getting ready to ship our next batch of Maslow machines any day, and we’re loving watching the community created by the first batch continue to make dramatic improvements to the machine week after week. This week also brings the first Maslow add on because Logan Needham has kits of his linkage system that improve the calibration available now!
Shipment will begin next week! As before, the motors are the final part that we’re waiting on. They were originally scheduled to be delivered this past Monday, October 9th, but an exam at customs put the delivery off by a few days. The most recent document we received showed delivery today, and I’ve asked our contact at Exceed Enterprises to let me know as soon as the motors arrive, but as of 3:30 PST I haven’t heard anything from him. Exceed’s Shipping and Receiving department closes at 3:30 daily, so this may mean they won’t arrive until tomorrow, but it should be any day now!
As soon as Sean, our Exceed contact, reaches out saying the motors have been delivered, we plan to head over to Exceed to do some final instruction and quality control before the mass box-packing begins. All of the shipping addresses have been organized and are ready to be printed into labels, and parts are generally ready to be put into boxes. Hannah will probably spend at least a few days helping to pack and label the boxes. Once your label is printed and your box is sent out, we’ll update your order status to “fulfilled” in our software with a tracking number. You’ll receive an email with this fulfillment confirmation and that tracking number included, so you will know for sure that your box has shipped when that email has arrived and you’ll be able to track your kit from there!
Again, thanks for your patience. We’re so excited to finally have these kits going out!
PID Controller Software Upgrades
Over the last two weeks community member Kevin Keegan has continued to consistently turn out a series of massive contributions to make Maslow better. I’ve been pretty overwhelmed the last couple weeks with rushing to roof the POD before the rains started and getting ready to ship the next batch so it’s not an overstatement to say that he has been fully carrying the software development and doing so phenomenally well. He started last week in version 0.92 by making a series of changes to the way the PID control system works. The new system makes a number of significant improvements. In Kevin Keegan’s own words from the pull request the changes/improvements include:
Removing the running average on the input to the PID velocity calculation (this caused oscillations and damped the control reaction and I think averaging is an incorrect solution in control theory anyways)
Adds a minimum rotation rate for the velocity controller (the motors are unable to turn at this rate, so this only caused integrator windup and noise)
Ensure that position requests are equal to a whole number of encoder steps (prevents oscillations and integrator windup for unobtainable positions)
Fix delay in the straight line loop in the movement loop to be consistent (ensures that the controllers are getting steps of equal length in each cycle)
Fix a few other obscure design issues with the PID controller
The net result of which is that firmware version 0.92 makes the machine faster, smoother, and more accurate as shown in the tests below.
This graph shows the error in the velocity controller from -16RPM to 16RPM in 2RPM increments. The results of the prior system are shown in blue and the new system in red. In an ideal world, this line would be as close to zero as possible.
This graph shows the error in the positional controller for a single move of 5mm, in an ideal world the line would go to zero as fast as possible without overshooting it by much. Again the old system is in blue and the new system in red. This is a significant improvement. The new system gets in the range of zero much faster and never overshoots, the old system had a propensity to overshoot and not return.
The prior graphs are pretty, but they are only analytical tools that can sometimes look good while producing bad real world usage . This last graph show a more accurate representation of real world usage. In this graph, the motor is being sent a new positional target every 10ms. This is the equivalent to cutting a straight line at about 25in/min. Ideally, we want the lines to be as close to zero as possible. Again the old system is in blue the new in red.
When I asked about his background with tuning PID systems, Kevin said he didn’t have any and that he learned how as he was going along. Brilliant. I’m starting to feel less and less like I’m leading the way and more and more excited to see what new improvements each week brings. It’s wonderful.
If you haven’t updated your machine in a few weeks, this would be a good week to do so.
Linkage Kits Available
Logan Needham has made a kit available for his ingenious linkage system which we have discussed in previous updates. You can pick one up from his Etsy store while supplies last: https://www.etsy.com/listing/558743441/maslow-triangulation-linkage-kit
The linkage kit works to simulate the behavior of the chains meeting at the router bit which dramatically reduces the complexity of calibrating the machine, improving accuracy.
He has also put together a beautiful wiki page with instructions for putting the kit together which you can find on the wiki here:
This kit marks the first time someone has taken Maslow’s technology, improved it, and made it available for sale. For Maslow to have the kind of impact that we think it can, we know we can’t do it alone. If we keep seeing the kind of brilliant, well executed improvements that Logan Needham has done with the linkage kit, the future is bright.
Thanks for reading! Until next time, we’ll be filling boxes!
-Hannah and Bar