12 August 2013

EtekChopper: Using the Sevcon Gen4 Motor Controller

The Sevcon Gen4 is both a truly great and reliable controller and a royal pain of finicky finickage. But it can do this: 

On one hand, it's well built, infinitely customizable, has lots of safety features, and is truly a 21st century controller that's years ahead of Kelly Controls in terms of features. I'm really not worried about dropping it a couple feet, the thing is built solid. It has internal pre-charge monitoring, and when you give it logic power it precharges automatically until it's ready to activate main battery power, and then trips a contactor relay, and then has internal economization for keeping that relay held efficiently. 
Getting down to the nitty-gritty, I can change EVERYTHING. Perceived motor inductance, max stator current, max RPM, PID gains on speed or torque controllers for both drive and regen braking. I can have 10 of them on a CAN network and drive them each individually. I can use a single throttle to drive two separate motors for two-wheel independent torque-vector drive. Or I can put one on my motorcycle and make it go VROOM VROOM (Okay, more like whirrrrrrr). 

But on the other hand, if you mess up any of the hundreds of variables, it will have many faults on it and won't work until they are dealt with. Once they are dealt with, there's still a strong possibility it will just dump current into the motor but not do anything. Or it'll dump current into a different motor lead and spin uncontrollably, threaten to fly off the clamps that are holding it down, at the slightest touch of the throttle, leading me to have to disconnect the main power. Not sure if that's awesome or scary. But I've gotten it right. I got my motor spinning properly and controllably. This is my journey, documented as best I can for reference to the EV community, because many are ready tame the Sevcon, and this may just help them get a handle of it. Because it's supposed to be more a reference to other than to me, I will be rehashing some of the stuff I've written in previous posts into this one. 

Let's begin our journey, then!

First things first is selecting the proper Sevcon Gen4 for you. I didn't have much choice, seeing as I was just gifted the $925.00 Sevcon Gen 4 Size 4, the 80 volt 350 Amp peak version. (I know this because it has a 355 Amp fuse on it, and the manual states the 350A peak version needs it on page 33. And the general brochure states the 350A version is meant ofr a 80 Volt system.)

As for picking a good motor for you controller setup, I guess it's a matter of the max motor current and voltage, and continuous current. My Brushless Etek motor can deal with 300+ Amps peak, 150 Amps continuous, at 5000 RPM (probably a bearing rating, and I can probably push it quite bit harder if necessary). As I recently found out (will detail this in a later post), my special snowflake special edition Etek motor has a Kv of 60 RPM/Volt. Nominally, this means 3.3 Volts per series cell * 24 cells in series * 60 RPM/Volt = 4752 RPM. So it's almost perfect for this controller. 

Now to figure out the wiring. Above is my setup. Note that my motor is clamped down to the table in multiple spaces, so it doesn't go flying across the room from the 50-some NewtonMeters of torque, dragging the controller and batteries with it. 

For a good reference, look at the diagram on page 56 and the table on page 35 of the manual.  The Sevcon Gen4 needs the following to function bare-minimum: 

  • A Brushless Motor of some sort, with its main U, V, and W terminals tied to M1, M2, and M3 on the controller. 
    • Use thick gauge wire for this to prevent melting. I used 2AWG. 
  • A big line contactor with the coil (switch) leads plugged into pins 2 and 3 and the contactor leads between the battery + and the + side of the controller. 
  • A Battery whose voltage range lies within the voltage range of your controller. Tie the battery ground to B- and the battery high voltage to one end of the contactor. 
    • I have a 79.2V nominal battery pack on an 80V nominal controller. 
    • Again, use thick wire for these connections. I used 2AWG. 
  • Logic Power Switch at Battery Voltage into pin 1, 6, or 10. 
    • This is the main "On" switch for the controller, and allows internal capacitor precharging to occur.
  • Throttle, with the high side plugged into pin 34, the wiper plugged into pin 22, and the ground tied to battery ground. (I find it very annoying that there are no battery ground lines anywhere in the 35-pin connector)
    • I use a 0-5k potentiometer Magura twist throttle from EV Drives
  • An encoder of some sort. I use Hall Sensors that uses pins 5, 15, 17, 26, and 29. 
  • Enable switches: Foot Switch (FS1), Forward, and Seat Switch are the bare minimum needed to enable drive, and I had them tied to pins 18, 30, and 31. 
    • You can configure these to be at any of the reserved digital switch pins. 
    • They are considered "On" when tied to battery ground. 
  • CAN bus leads for programming, with CAN high tied to pin 13, CAN low tied to pin 24, and pin 2 (CAN Termination) tied to pin 24 if this is only controller in your set, which it is for me.
    • May also need a separate wire to system ground.  
Yeah, that's a lot to deal with, and there's more involved in the setup to get regen braking or other features enabled, but you'll soon see why every bit of it is necessary. 

In order to start connecting things to the 35-pin header, we need the proper individual female connectors and the large 35-pin housing. I ordered these female connectors and this AMPSeal housing from Digikey. Something ended up being incorrect, however, because when I tried to insert the connectors into the AMPSeal housing and then insert the entire housing into the socket on the controller, there was no connection made between the female and male connectors. 

And so I ended up just putting extra electrical insulation on the female connectors and plugging them directly into the controller. I'll deal with how I'll properly isolate these from each other and the elements later. 

After plugging the logic power into the connector and switching 72V into the logic power input, THE LED TURNED ON! LOOKIT! IT'S ALIVE! I tried plugging everything else in (Contactor, hall sensors, throttle. No CAN or enable switches yet) to see if it would trigger the contactor.

Then it blinked 11 times, clearly a fault of some sort. Page 110 of the manual specifies it's an encoder fault, which could result from the encoders not being attached to the controller correctly. 

Now, here's my issue. I don't know which of the bajillion of internal configurable variables are set, and at what values they set to. Odds are, either an incremental optical encoder or a sin/cos encoder (or a combination) is set to be the default for the controller to look for, and those require different pins, hence the thrown fault. 

The only way to see what's going on in the software is to purchase a USB to CAN converter. However, according to this forum thread, and this one, not any old USB to CAN converter for under $90.00 will do. The Sevcon Gen4 requires a specific $300.00 adapter made by IXXAT which has the required proprietary software libraries for the Sevcon to be happy. 

According to this Sevcon program setup document, the amount of configuration power gained by getting this thing is well worth the $300.00, so I called IXXAT directly (They're closeby in New Hampshire) and purchased one, and I got a small academic discount for doing this project through a student club (MITERS). Carla, their sales representative, was a great help and was able to send me one quckly. 

And here it is. Even came with a free pen! 

Okay. One side is a USB plug, the other is a D9 connector, which goes to the controller.

The manual it came with specifies these 3 wires from the male D9 connector. 

I found a female D9 connector lying around the shop, and soldered 3 leads to it at the appropriate pins. I also soldered a jumper from CAN Low to CAN Termination, necessary for the last device of a CAN bus. 

Here is the default screen of the Sevcon DVT Software. To acquire the software, I called technical support at Sevcon, and a specialist named Joseph was kind enough to send me access to the software, as well as offer some tips for getting the controller working. I would later contact him again to acquire an updated firmware for the controller and a parameter configuration file for the ME0907 motor. 

I later found out you can acquire the software by following the instructions on this forum post, but I believe the software is older and possibly outdated. And after calling up Sevcon they can send you a special download link for free, so there's no reason not to just call them up and get the latest software. 

So, once you got the software running, just follow these DVT instructions. You can see there's an area to change the baud rate for the CAN communication. 

After I set it to the proper baud rate, IT CAME TO LIFE! GREEN LIGHTS!

The controller started speaking, too!

By pressing the "H" Helper button at the top of the screen, I am able to modify controller parameters that pertain to function. The first thing to do is set the controller to Preoperational mode. That will turn off the contactor (if it isn't off already) and enable you to modify the internal settings. I first set up motor parameters based on the ME0907 datasheet, but there were so many parameters that I just left some as they were. 

By looking at the datasheet for the Kilovac EV200 contactor I am able to set the contactor pull and hold voltages. 

I set my throttle ranges using the specified instructions. 

And I am able to test them in the Status tab, and I can see the current perceived throttle voltage as I move the throttle. Cool! 

In addition, I went to the Input/Output tab (Where switch inputs like Forward and the Foot Switch and throttle configurations are determined) and ensured that the first selection was Left Wheel Drive (Joe from Sevcon let me know that is required for drive, in addition to a Forward, FS1, and Seat switch). 

So after doing ALL THAT and going operational, the following faults come up -_-. I played with some settings, and I'm not quite sure what exactly I did, frankly, but I got it to the following uncontrollable state: 

What's happening here is hardcore integrator windup in the controller's torque control system. I can modifythe P and I gains, but I really don't want to mess with them too much because I don't understand the internal transfer function of the controller. I'd rather just call up my friendly neighborhood Sevcon tech support specialist and see what he suggests. 

Joe, the technician who had originally sent me the DVT software was able to send me an updated firmware for the controller as well as a .DCF configuration file for the ME0907 motor. With this config file, I don't have to go about modifying every parameter, but I can load this default one (which includes proper P and I torque control gains) and modify smaller parameters as needed. Remember, while my motor is a little more powerful than the ME0907, it's roughly equivalent, and this will serve as a proper starting point. I'll tweak max current and other settings as needed, as well as later add regenerative braking. 

After flashing the updated firmware and sending the new .DCF file, the following awesomeness ensued: 

Sure, the motor is spinning in the opposite direction it should be, but that's only a matter of playing the sensor game. I have to either shift the sensor leads

And there you have it! In a nutshell, to use the Sevcon Gen 4, you must: 
  1. Have a Sevcon Gen 4, a brushless motor, a contactor, logic wire (18AWG), power wire (2AWG for 350A peak), and a power supply/battery pack
  2. RTFM
  3. Buy the right pin connectors
  4. Buy/rent an IXXAT USB to CAN Compact
  5. RTF other M
  6. Call Sevcon for the DVT software and motor configuration file
  7. Do some finagling, and win!

07 August 2013

Getting the Dyna Myte 1007 CNC Mill Working


Here's how it happened:

Towards the end of spring 2013 MIT Prof. Ian Hunter, my lecturer for the class 2.671 (Measurement and Instrumentation Laboratory) and director of the Bio-Instrumentation Laboratory donated to MITERS a 3-axis CNC mill. 

Why, you might ask, would someone donate something as expensive as a car to a humble student shop? Well, it appears five years ago someone replaced the CMOS battery, but in doing so inadvertently reset the computer's BIOS settings. As a result, the software could not detect the machine, and it sat unusable for five years. Instead of getting it working, they bought a HAAS instead, leaving the task of getting the machine back up and running to us. 

It's a Dyna-Myte 1007, manufactured by an old Californian company Dyna Mechntronics. It's named 1007 for having 10" of travel in X and 7" of travel in Y (There are also 10" of travel in Z), so it has a decent build area. 

Other than the 10"x7"x10" build area, this machine sports many other great features like an external flood coolant system and filter, a pneumatic automatic toolchanger with a tool carousel that can house up to 6 tools, an extra servo drive for commanding a fourth rotary axis (Prof. Hunter wanted the original rotary axis, which came from his personal home machine shop, back) and a built-in computer running DOS with decent proprietary Dyna software. 

Now if we can only get it working...

Behold, the computer system of the Ancients! Sporting a whopping 250MHz Pentium CPU, this machine is weaker and takes up more room than a $35 Raspberry Pi. The built-in software running on DOS was not bad, however. It seems capable of doing everything your standard CNC software (like Bridgeport EZTrak, which I am accustomed to) can do with some additional features. 

Upon powering on and loading the Dyna4M software, we were greeted by an instant "Error 465: No Response From Driver." 

Well, let's take a look at the inside of that computer and make sure everything it wired correctly. The interface card between the computer and the large driver board behind the machine is an ISA card, which we later found out takes up the RAM space of COM port 2.

After a couple days of digging through the Dyna 4M software to see if we can expose the internal software configuration, we came upon this. It would appear IRQ 3 is the hardware interrupt resource that the Dyna 4M software looks for to talk to the ISA interface card. Issue is, IRQ3 is currently occupied by Serial Port COM 2. 

To alleviate this, we went into the BIOS and changed all the settings involving IRQ 3. We made sure to disable COM 2's attachment to IRQ 3, as well as any other setting that may occupy the IRQ 3 memory space. 

And it worked! Freeing up IRQ 3 allowed the ISA Card to subscrible the to the proper memory space, and the Dyna 4M software is able to communicate with it! 

YES! We can move axes now! 

However, when we tried to do anything beyond that, particularly involving the pneumatic toolchanger or spindle, the program threw errors. The former was to be expected: we didn't have air hooked up to the toolchanger to allow it to move!

We hooked up an air compressor set to 90PSI to the machine, but it turns out there was a severely leaky solenoid valve preventing the system from working properly. We ordered a new set to alleviate the issue. After replacing the set and hooking up all the hoses to their proper spaces, the fault was cleared!

But led to another one. Great. Now the spindle wasn't working. 

Taking a look in the back of the machine at the spindle driver, it seems there are lots of things that can go wrong. In fact, I count 18 DIP switches, all of which could be incorrectly placed. 

It also seems some of the wires may have come loose from... overzealous rewiring on our part, in an attempt to fix the machine. 

Looking into the back of the machine, I see the spindle driver is a Glentek SMA8115 Brushless servo driver. A quick Google search, and I am browsing the manual for DIP switch settings. Turns out, the default hall sensor configuration for this driver is 120/240 degrees, but the spindle motor has 60/300 degree hall sensors. That got rid of the hall sensor fault, but there was still a drive enable fault. A bit more digging through the manual of the spindle driver and the axis drivers (also by Glentek, but smaller models), and I find that the spindle driver has an active disable (turn a pin to HIGH to deactivate), rather than the axis drivers and their active enable (turn a pin to HIGH to activate). 

As a result, the wiring is EVER so slightly different. Clever girl... see that subtle swap between the green and orange wires? That lets the regular axis interface card output talk to the special snowflake spindle driver. After changing the cable to the proper one, IT WORKS! IT WORKS IT WORKS! NO FAULTS! TIME TO CNC ALL THE THINGS!!!!!!!!

The following will be a set of photos with no dialogue. Just imagine this song playing in the background as you browse through the following images. It's almost like art.