Last weekend I installed a Motobrain on one of my motorcycles. It squeezed in behind battery and was velcroed to the side of battery box. The installation is pretty trivial with just a single output activated and one input setup but it is nice to be rolling around with a Motobrain. The single input is connected to the high beam and that signals the trail lights to turn on with the high beams. I also tested the over-the-air firmware update capability while actually mounted inside the vehicle. It all works great.
Removing the Motobrain from the mold this morning was a mixed bag. I successfully extricated it from its prison but failed to do it without the use of a hammer, saw, and vice. At the cost of $550 for the mold I’d say this is a failure. I am going to need a better strategy for casting these things. Unfortunately, this is not my core competency so if I try to go it alone I am bound to run into more issues like this. The good news is that all my bashing didn’t appear to cause any damage. Unfortunately, this means I cannot make any more potted Motobrains until I get disposable molds made. I think the smart move is to find someone whose business this is and inquire about copying a plastic version of the mold in high temp wax or plaster. Alternately I need to find a potting box that has no bottom. I want to have a high thermally conductive surface on the bottom of this so it can be mounted on something that offers good heat dissipation.
In the photo below you can see the Motobrain. It ought to be jet black but the paint used to give the mold a smooth finish stuck to it. This means that the fancy $130 a jar mold release didn’t work. I like to blame the tool but I’m certain it was the craftsman. Notice the new screw down terminals for the signal inputs. Those are looking pretty sexy I think.
Below the Motobrain is the mold, in pieces.
You may recall a few days ago I mentioned a programming and QC jig I had fabricated. Well, it arrived in the mail this evening. After dinner, I went about building it with my new Pick and Place machine. 😀
If you are wondering what happened to the Motobrain I put in the mold last night, I realized this morning that it needed 24 hours to cure so I left it the oven all day at 60C and will try to pry it open in the morning.
While many people were enjoying a day off work, I was enjoying a day working on Motobrain. I got all the firmware for the new board written and tested. I also got this new Motobrain in the mold. Tomorrow morning I’ll pop it out and see how it looks.
I also got that pick and place machine built this evening. I am basically pleased with it. The Z axis bearing is a little loose which makes it less precise than I’d like but I suspect it is better than nothing. Next to it in the photo is a USB microscope which acts as a third eye for me which should help reduce fatigue.
The new board testing was rough this weekend. There was a bad chip or bad installation of the chip which caused the new Motobrain to not function. This was the first time I had a board fail because of this sort of issue and it took all day yesterday to sort out. On days like that, I remind myself why I started my engineering career in software. I lost a day to troubleshooting and repair which stings a bit. The good news is that it is working now, the bad chip has been found and replaced and everything looks good again. A quick perusal of all the functions shows that the new board is behaving nominally. Success!
I expect the parts for the pick and place machine I am building to arrive today, 5 days late (damn blizzard). I’ll start building it up tonight and hope to build up a couple more Motobrains with it to ship out by weeks end. I’m going to Arizona this weekend to install a unit on my G650xC (which I keep out there for desert riding).
I am also preparing for larger scale Motobrain production. I have designed and have had fabricated a programming and QA/QC jig so I can quickly program and test all the Motobrain units once I have had them manufactured (rendering below). It can test all the inputs, outputs, voltage measurements, and current measurements. It can also program both microcontrollers and the flash memory chip which contains the fall back code in case of OTA update failure. This will allow you to “unbrick” a Motobrain should I put out a bad update. Likewise, I may be convinced to write custom firmwares for certain customers and having this will allow them to return to “factory” code if they want to do something different with their Motobrains.
The new boards just arrived. They were delayed by Atlanta’s second blizzard in as many weeks. I spent 4 hours yesterday shoveling snow off my entire street so that FedEx would not have an excuse to not deliver them today. I will work on building one up tomorrow. Assuming testing goes well, I will be building up several more for immediate shipment to some prospective customers. At this point I am late on my test unit shipments. I promised one person a unit in January but my molds were delayed two weeks and then since then I decided that the inputs locations were really not acceptable to me and I did not want to send a version of Motobrain out with those awkward inputs. Now I am feeling the pressure of being behind schedule which I really don’t like.
The boards in the photo below are resting on some MDF which will become a custom made manual pick and place machine. The hand building of Motobrain is by far the worst part of the entire process and I hope this little gizmo will make it pleasant enough that I can hand make a couple dozen of these guys before I have them manufactured in bulk.
On the current version of Motobrain, I have the inputs terminated with 2×4 headers sticking out of the epoxy. I didn’t love this design but I wanted to test some things and I didn’t want to delay my testing while I tried to find the “perfect” solution for the inputs. I created a little “disposable” break out board so I could attach wires to the board with standard tools and attach that board to the Motobrain headers. After seeing and feeling this setup, I recognized it just will not be sufficient for our needs. Much too fragile! I’ve redesigned the board with terminals directly mounted to the PCB which will extend out of the epoxy. I ordered these new boards with an express three day turn around. Likewise, I redesigned the mold and also ordered it to be delivered post haste. I also took this opportunity to add some temperature compensation (in hardware) to the inputs. I found on the current version that the voltage sensed would drift a few volts when we approached the far end of its operating temperature envelope. The new amplifiers I’ve included will not drift significantly over the temperature range specified for Motobrain (-20C to 80C). Lastly, I decided to include a thermal cutoff circuit to the device. If the device reaches 112C, it will shut off. This is hardware enforced, so there is no getting around this. As such, it should so a good job protecting itself from would-be Motobrain killers.
Photos of the current board with break out board and renderings of the new board and mold are below.
Out with the Old
In with the New
I ran a test of the encapsulated Motobrain to confirm the thermal characteristics in its new potted state. I started the test by turning it up to 50A for 30 minutes. I took a measurement and then turned it up 10A and set a timer for 10 minutes. After 10 minutes I took a measurement, reset the timer, waited, took another measurement and then turned it up 10A. So it ran for 30 minutes @ 50A, 20 minutes @ 60A, 20 minutes @ 70A, 20 minutes at 80A, 20 minutes @ 90A. The next step was to turn it back down to 50A and check it every 10 minutes for 40 minutes to see how long it takes to get back to a similar temp as before. The ambient temp in the lab was 19C when I began and was 26C when I finished the 90A test because the power supply and DC load are really hot creating and dissipating all that current (so I can add 7C to the first 50A reading and say I achieved a baseline reading which I failed to achieve in 40 minutes). I then let it run with no load for 10 minutes and called the test complete. Results are below.
I took a few internal temp sensor readings along the way. The internal temp sensor is on the far side of a second PCB from the power transistors and the potting compound is not a great conductor of heat so I expect a large temperature delta. When I pot one of these in thermally conductive epoxy I expect to see much greater correlation between the two numbers. When no current is flowing, I would expect the external and internal temps to be roughly the same.
One edge of Motobrain is bolted to a table with most of it hanging in still air. See the last post for a photo of its location in space.
50A 30 minutes (internal temp sensor was 25C)
60A 10 minutes
60A 10 minutes
70A 10 minutes
70A 10 minutes
80A 10 minutes (internal temp sensor was 41C)
80A 10 minutes
90A 10 minutes
90A 10 minutes (internal temp sensor was 51C)
50A 10 minutes
50A 10 minutes (internal temp sensor was 41C)
50A 10 minutes (internal temp sensor was 39C)
50A 10 minutes (internal temp sensor was 39C)
0A 10 minutes (internal temp sensor was 35C)
Test completed. The results are suspiciously like my previous unencapsulated tests where 80A was in the low 60s and 90A was in the low 70s. This tells us that the epoxy is not significantly worse than nothing. The final 0A measurement shows that the internal temp sensor and external temp sensor are well correlated to each other and that there is a pretty significant thermal gradient inside Motobrain when it is running. When I get the fancy (and expensive) thermally conductive epoxy we should be in good shape!