blog; Motor Controller Options for 2015

[Jon Stratis]

Originally Posted by Andrew Schreiber

And the CAN bus was a ticking time bomb on your robot. If one device dropped out the whole bus was screwed due to CAN Timeout exceptions.

Sincerely hope that's been fixed.

The entire CAN subsystem has been re-written for the RoboRio... We've been doing some testing with a CAN robot (6 Jags, PDP and PCM), but we haven't yet tried to drop a jag to see what happens. I'll see if we can give that a try on Saturday.

[marshall]

Originally Posted by Jared Russell

VERY IMPORTANT: Slave mode so that a single sensor can be used to control a multi-motor, multi-Talon mechanism (like the drivetrain). It was never really clear how to do this with Jaguars.

I do not think I would ever run a position control loop on the speed controller, but I can envision making heavy use of velocity control.

According to CRE in St Louis, that is something they were/are working on. Granted, I don't work for them and we have minimal contact with them but this is one of the things we would like to see as well.

900 makes great use of CAN. We've been using it rather successfully (we think) for the last 4 years (ok, 3 but running Java in between was our own fault and we won't be doing that again). LabView was a lot easier to work with for CAN than Java was.

If we have the opportunity to beta test any of this then we most definitely will. We're eager to get started with our testing this Saturday (We're a residential high school and this week has been the first week back for the students).

Also, we've used both position and velocity control. Position is a lot more temperamental about having the correct values so tuning PID is more important, at least from our experience with it.

[philso]

Originally Posted by Monochron

Could someone post some examples of the Anderson powerpoles they have used with success in the past?

The only types that I have used (excluding the battery connectors) have been very shoddy; wires falling out, plastic crumbling, etc. Admittedly, this was years ago so I am hoping that newer versions are much better.

http://2014.discobots.org/node/94

http://lh5.ggpht.com/-4l8yI7w_YQg/Uw...0/P2126448.JPG

The Talon motor controllers all had Anderson connectors on their outputs. This allowed the electronics panel to be built up outside of the robot then "dropped in" and connected to the eight drive motors and the collector motor. There was also a set of Anderson connectors at the end of the collector arm so that the collector motor could be swapped out without having to pull the wiring out of the collector arm. Anderson connectors were also used at bot ends of the wiring for the Spike relay controlling the compressor so that the compressor could be swapped out. They gave absolutely no trouble throughout the severe punishment of two regionals plus 3 off-season events. The polycarb panel only had to be removed for the Inspectors.

Originally Posted by timytamy

The trick to them is that you really need the correct crimper

Absolutely! Doing a pull test on all crimps immediately after making the crimp will decrease the failure rate dramatically. That is what the assembly people at work are taught to do.

[AdamHeard]

Originally Posted by Monochron

Could someone post some examples of the Anderson powerpoles they have used with success in the past?

The only types that I have used (excluding the battery connectors) have been very shoddy; wires falling out, plastic crumbling, etc. Admittedly, this was years ago so I am hoping that newer versions are much better.

You must have been using something else, or counterfeit ones.

We've used Anderson's forever, and they are awesome. The housings are very durable.

[Bryce Paputa]

We've had issues with the Anderson connectors, both with the casings melting and with the terminals falling out of the casings. We use the correct crimping tool and the solder them, any idea what we might be doing wrong?

[philso]

Originally Posted by AdamHeard

You must have been using something else, or counterfeit ones.

We've used Anderson's forever, and they are awesome. The housings are very durable.

We use them at work in some of our industrial UPS' and have never seen them break down like that either.

Originally Posted by Bryce Paputa

We've had issues with the Anderson connectors, both with the casings melting and with the terminals falling out of the casings. We use the correct crimping tool and the solder them, any idea what we might be doing wrong?

It is likely that you may be getting too much heating from the contact resistance and/or sustained high currents. If the terminals are bent, the contact resistance may be high. I have also seen terminals where the silver coating was worn off, possibly leading to higher contact resistance.

[magnets]

Perhaps an EE could help me out, but why go to 15kHz for the switching frequency? Is the 1kHz used on the Victor 888 not smooth enough?

From my small knowledge of EMI, the effective interference from the motors is proportional to switching frequency and current, both of which are very high in this application. Putting so much circuitry right next to the controller (like the CAN bus) makes me a little nervous.

The motors also generate really strong fields and though they drop off quickly, they still have an effect on our control system. Now that our main controller is no longer in a big heavy metal grounded enclosure, I'm slightly concerned about this.

That said, I am not an EE, but I dislike anything that increases electrical interference on the robots.

[s1900ahon]

Originally Posted by Bryce Paputa

We've had issues with the Anderson connectors, both with the casings melting and with the terminals falling out of the casings. We use the correct crimping tool and the solder them, any idea what we might be doing wrong?

I had the pleasure of sitting beside an APP employee (Rich, if I recall correctly-I'm bad with names) at one of the Supplier Summit dinners. He was the guy who created the battery wires+SB50 for the KoP. I mentioned my team that year was crimping and soldering the connections. He discouraged soldering, indicating the potential to damage the terminal if soldered poorly. Since then I've just taken him at his word and my team just crimps with one of their tools. We've never had an issue with the housings deforming, either the ones we've crimped ourselves, the ones we've bought from AM (crimped), or the ones we receive in the KoP (crimped).

[Kevin Sevcik]

First off, I'm delurking after several months of down-time/new baby/work specifically to note that I drooled a little when I saw the tech-specs of these new controllers. We've always been attached to CAN, so we never went with the Talon SRs, and suffered mightily on mounting space for it. I'm really looking forward to packing 4-ish Talon SRXs into the footprint of a single Jaguar in 2015, and I will be sorely disappointed if I'm not allowed to do so. Also, I'm dearly hoping that CTRE will make the source for the controllers a little more open than it was for the Jags. The only thing available for the Jags was the very original non-FRC source, which made it troublesome to figure out what they were doing, why speed feedback wasn't reliable/available, etc.

If we're using this thread to put in feature requests, my request for the Talon SRX is for sensor feedback output regardless of what control mode you're in. One of the problems on the Jag was that you could only get certain feedback if you were in that particular mode. So speed feedback was only available in speed PID control. So no using the Jag as a sensor input for your nicer C++ PID code.

Magnets,
I think the number of teams using Jags/Talon SRs with no EMI problems is testament to the notion that PWM frequency isn't a huge EMI source on the bots. Since the new controllers will use the same sign-magnitude/synchronous rectification control as the Jag/Talon SR, it shouldn't be a problem. Sign-magnitude control toggles between the high/low FETs on the switched leg to switch between driving/recirculating the current. This makes for smaller current ripples and less EMI. You're probably going to get more noise from the brush commutation than the PWM switching, especially on high-speed 550-can motors.

[Ether]

Jags and Talons have always been 15KHz output.

The higher switching frequency reduces ripple current in the motor and motor wires. Ripple current contributes to motor heating and electrical noise.

[apples000]

Originally Posted by magnets

Perhaps an EE could help me out, but why go to 15kHz for the switching frequency? Is the 1kHz used on the Victor 888 not smooth enough?

From my small knowledge of EMI, the effective interference from the motors is proportional to switching frequency and current, both of which are very high in this application. Putting so much circuitry right next to the controller (like the CAN bus) makes me a little nervous.

The motors also generate really strong fields and though they drop off quickly, they still have an effect on our control system. Now that our main controller is no longer in a big heavy metal grounded enclosure, I'm slightly concerned about this.

That said, I am not an EE, but I dislike anything that increases electrical interference on the robots.

I also am not an EE, but have a lot of experience with this sort of stuff. In short, you're right, the higher frequency is proportional to higher interference. That said, I checked the spec sheets, and the old Talons are operating at 15kHz, and people have been fine.

I don't understand the reason to go up to 15kHz for the controllers. Switching up to 100 amps at 15kHz is not something to be taken lightly, as the high current and frequency causes parasitic capacitance to ground. This results in an effect called common mode coupling, which ends up "contaminating" your ground plane. Anything else that uses this ground can be affected by it.

Switching to a higher frequency also makes the system less efficient. Although the more efficient sign-magnitude switching could make up for this loss, you'd see higher efficiency if it switched at 10kHz than 15kHz. The reason is that the faster you switch the FETs, the more time they'll end up spending in their switching state. The transition is not instantaneous, and as you increase frequency, they'll spend more time in between, which is inefficient.

Also, I can't really think of a great reason for them to go this high. The 1kHz frequency was audible, and so will the 15kHz frequency (close to the frequency your CRT television's transformer makes). The talon is 15kHz and the Victor 888 is 1kHz. You really can't tell the difference between the two frequencies, other than the noise they make.

[DampRobot]

Originally Posted by apples000

Switching to a higher frequency also makes the system less efficient. Although the more efficient sign-magnitude switching could make up for this loss, you'd see higher efficiency if it switched at 10kHz than 15kHz. The reason is that the faster you switch the FETs, the more time they'll end up spending in their switching state. The transition is not instantaneous, and as you increase frequency, they'll spend more time in between, which is inefficient.

While I agree that in theory faster switching will decrease efficiency, I'm not sure that it'll be that large of an effect in practice.

At 15 kHz, a MOSFET would switch on and off twice every 1/(15 kHz), which is a switch about every 30,000 ns at 50% duty cycle. The MOSFETs I've seen on the market nowadays all switch faster than 10 ns, and some even faster than 5 ns IIRC. That's about .03% of each state at 50% duty cycle. Of course, as switch your FETs at closer to 0% or 100% duty cycle, the effect you're describing will be more important, but I don't think it'll significantly decrease the efficiency of the controller.

EDIT: For what it's worth, I'm not an EE either. But that could change in the future.

[SoftwareBug2.0]

Originally Posted by magnets

The motors also generate really strong fields and though they drop off quickly, they still have an effect on our control system. Now that our main controller is no longer in a big heavy metal grounded enclosure, I'm slightly concerned about this.

I do not remember this ever being an issue with the old IFI controllers and they were a similar form factor.

[ToddF]

My biggest concern regarding the integrated leads is that installing connectors on the ends of the wires constitutes a modification away from COTS state. This means that use of a particular motor controller on a competition bot will (under current rules) only be legal during the year in which the connectors were attached. Currently, we utilize previously purchased Talons as spares. In a pinch, we have even raided previous year's robots for motor controllers (not at an event). With these new ones, unless the rules change, they have a built in expiration date. If you've already attached connectors, you can't use them on a competition bot after the next kickoff day. And cutting off the connectors and putting on new ones isn't legal because just cutting off the connectors doesn't return the wires to their original length. I think now is a good time to revisit the rules regarding reuse of parts. It was bad enough that we couldn't reuse CIMS, but CIMS are fairly cheap, and don't require swapping out very often. Not being able to reuse motor controllers is going to hurt our pocketbook.

[Gary Dillard]

I'm really torn about this. I'm thrilled with the new form factors and functionality in both of the new controllers. Both VexPro and CTRE build high quality, reasonably priced components that have brought competition to another level. But after being burned by the Talon shortage in 2013 and hearing about the Versachassis shortages in 2014, I have honest concerns with either company's ability to meet the demands of introducing a new product line. I don't have many choices - we can stock up on Talons now and miss an opportunity, try to stock up on new controllers (whenever they're available, it appears I can't even pre-order them yet) and gamble on them being legal, or just hang out and hope for the best.
Anyone at either company want to try to give me a warm fuzzy feeling about being able to buy speed controllers week 1? I'm not a business person but I get the difficulty in estimating demand and controlling supply of specific items. Convince me you've figured it out.