Manufacturing – Process Changes and Progress
We’ve been putting out some smaller updates to make sure you’re in the loop with what’s going on here, but we thought it was time to take a step back and look at things from a bit of a higher level.
In case you missed it, we put out a new video series that shows how to use the Palette. Check out the videos, under the resource tab of our website here. Link
To be completely up front – we have not been hitting our production goals, and will be pushing back the delivery timelines. During pre-production we were working to get our operation up and running, now, we’re working on implementing new, more efficient processes, and improving the systems we already have in place. In this update, we’re going to talk you through some of the improvements we’re going to make, and what we expect the impact to be.
Drives – A.K.A the bane(s) of our existence
A few updates ago we went into the concepts behind our quality control tests. We covered how we have both ICTs (in-circuit tests) as well as FATs (Final Acceptance Tests).
It is very costly when a component passes an ICT, but fails a FAT. This cost is due to the replacement time it takes to remove the broken component, and replace it with a new (working) component. In some cases, this takes hours to complete – a process which has taken up too much of our time.
This has been most evident in our drive systems. We’d like to dig in a bit into what the process is behind our drives, and give you an understanding of what we’re doing to get things back on track.
You may remember back in a previous update, we were receiving out of spec drive gears from our supplier. Because of this, every gear that we receive in house is measured, and categorized into different groups. One of the groups is discarded (out of spec) and the other moves into production.
Next, the gear is fastened to the stepper motor shaft. From there, the motor/gear combo is bolted into our laser cut acrylic drive housing, and it moves on to the drives ICT.
The ICT involves each individual drive gear pulling filament for 1.5 meters…..except the filament is attached to a weight. If the drive cannot lift the weight up over the standard test distance then we are able to identify that drive as being defective. This is usually an indicator of something being out of spec or of an assembly error.
Despite this test, we were getting a substantial number of failed drives in our full Palette quality control tests. This is an issue for two major reasons. First, the percentage of drives that fail is above our acceptable limit. Secondly the fact that the weight lift test will yield a pass result for a drive that fails the FAT means that this test needs to be improved.
Although we are confident that the final acceptance tests will catch drive failures, it is too costly to discover these failures at this stage and thus a change in the process is required. We have been working diligently on finding ways to better tests drives earlier in the assembly process so that we can avoid time fixing and spend more time building.
Measuring Drives and Predicting Failure:
When you get your Palette there will be a small QC card inside the box. This QC card will have three values on it – “steps per count”, “splice distance”, and “opto to cutter”. These values are recorded and determined during our QC and calibration process – we’ll focus on steps per count, as it is the most important when it comes to drive failure.
When we’re determining steps per count, we’ll run 45000 steps on a stepper motor, and record how many pulses result in the paired scroll wheel. The number of pulses is typically between 14100 and 14400, depending on the scroll wheel assembly, the drive gear, and a few other factors.
We will run this 45000 step test twice with each motor, and record both values.
Now – we started noticing some interesting trends with these numbers. We noticed trends in two specific areas – the difference between test 1 and test 2 on any given drive, as well as the maximum difference between the counts on all drives (max value (minus) min value).
Now, it’s very difficult to predict things without a large enough data set to base your prediction off of; however, we believe we are seeing the start of a trend.
After looking at our database of Palette drive readings, we noticed that drives that had a difference of >100 counts between test one and two were likely to be in Palettes that had a much higher failure rate of the FAT.
The fact that these measurements correlate to drive failures means that we can begin to use this data to help predict drive failures well before they are installed into a Palette.
What this means in terms of the production process is that we will need to quantify drive performance during the drive assembly cell of our production line. Our engineering team is already working on a set up to automate the data collection and analysis portion of this. This ICT will still involve lifting a weight but rather than looking for a binary failure (motor slip or motor skip), the test will quantify the performance of the drive giving us a much better chance of finding the bad eggs.
By changing our QC process for the Palette’s drives, we expect to substantially increase our throughput, and be moving more Palettes each week than we previously were. (We’re going to cover a few more things and we’ll get to exact numbers at the end).
Getting our manufacturing facility up and running was a challenge on its own – now the challenge is optimizing it, creating a line, and copying those lines to increase throughput.
We’ve expanded our assembly team, and these team members are going to begin specializing, working on a set schedule, and aiming for a specific number of sub-assemblies per day. Altogether, this is going to make our assembly much more efficient, and more measurable.
By specializing we’ll be able to better track the time it takes to make each sub-assembly, set more measurable, achievable goals, and be able to work out any inefficiencies in the process.
We believe that the new processes, and assembly team members will allow us to ramp up to 25/week by the end of May.
The final acceptance test pass rates remain an issue that we are working diligently to get to the bottom of. After focusing on our electronics failures, we’ve seen a huge drop in the failure rates of our boards over the past few weeks. We believe that by targeting the repeatability of the drive systems the same will prove true for this subcomponent.
We’ve delivered 15 units so far – we have another 20 sitting in our QC quarantine zone. We are aiming to deliver 75-100 in May, 100 in June, and 120 monthly from July onward.
This means that backers 1-100 should have Palettes by the end of May, 101-200 in June, 201-260 in July, and then we will be finished fulfilling our campaign.
We apologize for another delay – it has been really great to have feedback from our initial users. There’s been a good deal of chatting going on via the Slack channel recently – if you haven’t already, you can sign up here! <<>>>
Also - we will be exhibiting at the Bay Area MakerFaire in San Francisco from May 20-22. If you’re in the area please come by, we’d love meet you!
And as always, please let us know if you have any questions