Category: Stuff

Stuff. Just stuff. Anything goes.

The Maxon DEC 50/5

chorl

Hey Maxon Motors, I’m the #1 Google hit for “Maxon DEC Module”. How many sales have I been responsible for?

What if you told me there exists a brushless DC motor controller about the size of a large postage stamp that was sensor commutated, took an analog 0-5v input, functioned up to 50 volts at a maximum of 10 amps, had built-in current limiting and its own internal logic power supply, and can be easily hacked for more power?

I’d be all like

wants one wants one wants one wants one wants one where can has nao?

Well, that was pretty much my reaction when I was told about the Maxon Motor DEC 50/5 module.

And here it is, in color and real life. A collaboration between Big Shane’s Fresh & Salty Motor Controls and Charles Z. Guan Experimental Vigorous Propulsion Industries Co. Ltd.

It’s everything I just described, and is my new favorite vegetable.

I really despise model airplane controllers. They’re great for their application – spinning a modest, largely invariant and predictable load. This is completely incorrect for an electric vehicle, which has a much more demanding and dynamic load profile. Sensorless commutation is fine for a propeller, where the motor can twitch at will, but it leads to a real chicken and egg problem for a high inertia load like a vehicle. If the motor can’t lightly move to generate a back EMF profile, then it doesn’t know which direction to apply torque in. And low speed performance is often chaotic and jittery.

So for a while, I’ve been praying to Robot Jesus and Co. for something the size of a model airplane controller that had sensored commutation. I was willing to give up significant power handling for sensors, since it takes much less brute motor strength to start a vehicle moving if the motor is actually utilized to its full potential.  Maxon seems to have answered the call with these DEC modules. The 50/5 appears to be the latest in a line of very small and low powered controllers.

Maxon provides a very helpful and thorough datasheet with the DECs, so naturally we had to set up the included basic test circuit.  The module requires minimal support to function – power, ground, three Hall Effect sensor inputs, 3 phase outputs, an analog throttle input, and two configurable digital inputs for setting closed-loop speed control. It provides logic and Hall sensor power internally, and can take an additional analog input for current limiting if your desired current is not “ALL OF IT“.

The form factor is not exactly helpful for breadboard use. It spans more than 1 breadboard’s worth of rails, so it got straddled across two. Next, the dual header had different pin functions on adjacent rows, which means that had to be split between two sides of one breadboard.

Now, how did that happen?

Somewhere in Switzerland, a Maxon engineer is probably crying over this picture.

Yeah. Those are wire-wrap IC sockets cut in half, sanded down, and bent over.

The BWD Scooter’s drive motors are about the most normal BLDC motors you can imagine, so they were selected for testing. It took a few tries to get the 3-sensor-and-3-power wire combinations correct, but here’s some observations about the DEC:

  • It has a built-in throttle and braking ramp that’s noticeable when the motor is freewheeling, but probably immaterial if mechanical inertia is considered. The ramp time seems to be speed difference dependent and is always <1 to 2 seconds.
  • If you try to switch directions while the motor is under power, the DEC will disable itself and then it requires either a power cycle or a toggle of the ENABLE pin to restart.
  • It will stop trying to run the motor after a few seconds if you hooked it up wrong. Or backwards.

I guess the datasheet spells all that out, but I’m just saying that it doesn’t appear to be lying :>

Through some part number trickery, Shane found that the small SMT FETs used are Infineon chips that have an on-state resistance of just under 6 milliohms. They don’t really have the thermal mass (nor the board any real thermally optimal properties) to be pushed significantly harder, but I think they can realistically handle 15 or so amps continuously, with no additional cooling. This is assuming a 100 degree Celsius device temperature, a 7 milliohm resistance at that temperature, and that the stated 50 Kelvins per Watt thermal impedance is accurate.

With additional heatsinking, which ought to be effective due to the thinness of the package, these modules can probably be pushed to 20+ amps. Past that, there’s always the Cheap Shady Chinese Small ESC Manufacturer Hackaround of soldering surface mount FETs on top of eachother. I think the effectiveness of this method just exponentially approaches some asymptotic amp limit.

Either way, to do this, the onboard current sense resistor (which sets the 10 amp built-in limit) needs to be bypassed. It’s a 10 milliohm, 1% tolerance SMT resistor. I normally just neuter my controllers of current sensing completely, but I’ve found that this some times gives unpredictable and fiery operation, so perhaps it’s just better to trick it into thinking it’s putting out 1/n the amps instead, where n is the number of times you parallel the CSR.

Anyways, here’s your daily exercise in absurdism.

Yeah, so I said something about crying engineers… Video of the brushless Etek.

Of course, there are disclaimers. It’s running off a power supply which stops at 5 amps. If I actually tried this with a battery, the DEC would probably have grenaded before it noticed what was going on.

Here’s a video of  That Other Thing We Should Try.  The DEC module sensor inputs are hooked up to the scooter rear wheel, but the drive outputs are connected to the Etek.

I see absolutely no practical purpose for this arrangement. It was late at night.

What’s Next

Deathblades. Deathblades deathblades deathblades deathblades.

This was the controls breakthrough I was seeking for the skates project. They’re small, sensored, and not very high power. They even directly interface with an analog input, so I can skip having an onboard microcontroller – XBEE radios have an “I/O” mode that lets the units be used in simple wireless sensing applications.

With my newly attained Mad Skillz, I’ll craft up a carrier board which has support equipment for two DECs and an xbee. I’ve already made a footprint for the DEC modules, but haven’t actually tried laying a board with them – still getting used to the pin and footprint naming convention of Eagle, so I might modify it for more legitimacy.

is can has bldc motter tiem

Coming Out of the [Course 6] Closet

chorl6 Comments

If I could say that MIT has made anything very painfully obvious to me, it’s that all mechanical engineering projects are spoiled by their electronics components. We can observe the phenomenon in any Course II class which focuses on producing some integrated electomechanical contraption. Invariably, the project is delayed or left unfinished because the electronics or software did not cooperate or had to be rebuilt… over and over and over and over.

This has been the story of many people in 2.007, 2.009, 2.12, just about everything I build, and my entire adventurous spree at the Media Lab.

Now, I still think software sucks (and Arduinos are therefore the best thing to ever happen), but one of the things that I’ve been “meaning” to do for a long time is printed circuit board design. I’ve had a propensity for electronics design for a while, as illustrated by the many Kartrollers, Segfault’s controller, and the disproportional amount of Course VI classes I’ve taken, but the vast majority of my work has been on breadboards, wired point-to-point, or using protoboard. It would be great if I could whip up an idea and have it housed on its own specially-designed PCB, which would have been layout-validated and checked on the computer so I *know* I didn’t put everything in backwards. But the protoboards and breadboards have always been there, and everything I make is a one-off anyway, and PCBs take a long time to make and are expensive if hired out… so it’s been fine.

But I thought that PCB design was the last hurdle to overcome before I’m actually able to take one of said electromechanical integrations and develop it from start to finish. No more being impeded by the other half, especially because now I at least know my way around electronics design and execution, even if I still plug things in backwards and short things with oscilloscope probes. Like countersunk cap screws and chamfered edges, a customized PCB just makes a project that much more professional-looking… and functioning, since there are less of a chance of making airFETs.

I’ve been playing with Eagle for a while, and have even made some halfway decent schematics with it. But what I haven’t done is actually go from schematic to board. I didn’t pay attention to packages, board footprints, or any of that.

So, in a Mountain Dew-fueled rage (Dew-colic)  a few nights ago, I finally sat down and explored the things which make Eagle physically relevant.

it has the worst user interface ever.

But once I got over the fact that essentially no commonly used keyboard shortcuts or mouse motions did what I thought they do, and found where parameters could be changed or adjusted after looking in the most cryptic of places and clicking on the many nondescript and  untagged buttons, it was pretty straightforward.

can i have a copy-paste function that doesn’t take 5 feet of cursor motion and 8 buttons to do?

Also, mouseclicks associate with commands and you have to select another command before it does anything else. So no, clicking furiously at a component because you want to move it but it wants to change the orientation gets nothing done.

Without further ado, I now present the Small Cute Half Bridge.


Shown in roughly 3:1 scale

While I figured a good test would be just jumbling every which component together onto a board, making something I could actually use is better. The SCHB uses the IR21844 (previously explored on this site) and drives a MOSFET half-bridge. Half-bridge units like this form the basis of many switching power supplies and motor controllers.

The actual schematic of the SCHB is the jumbled mess:

Making the SCHB was an exercise in device creation. I had to define a footprint, a schematic symbol, then associate the two together into one device. It’s a clunky system that I think could be streamlined, but the clunkiness probably carries over from industrial engineering rigor. I made a 14-DIP and SOIC-14 version of the IR21844 to cover all my bases. I also made the IRFB3077 FETs. Fortunately, knocking the board footprints from existing libraries meant that I didn’t actually have to draw them.

What’s next? Get the board made!

Many PCB companies have online services where they can make you boards from a design. One example is Advanced Circuits, which seems to be a favorite of MIT students. This is where I think the mechanical engineers and manufacturig industry can step it up – where are the online services for fabrication and machining?

Granted that making 3D parts in metal is much more resource and equipment intensive than PCBs or 2-dimensional machining like waterjetting and LASER cutting, but I contend that it’s worthwhile to look into as an additional market for shops, especially as the DIY trade is growing.

Hopefully, within the next two weeks, there will be SCHBs on my desk and ready to plug in backwards test. It’s incredibly small (1.5 x 1.6″), all through-hole, and the FETs can probably handle 25 to 30 amps with no heatsinking.