Power for eight (8) Dynamixel XM-430-W350-R

[jcarlos]

Well. We need a robot on wheels capable of positioning quite accurately on the plane.
The four-wheel drive and independent steering offers omnidirectional behaviour. The design seems "easy", relatively simple, and intuitive. Link

We have decided to build this design. We have planned to work with 8 XM430-W350-R .However the Robotis technical service tell us.:
" If you are running all 8 XM430-W350-R at the same time, I'd recommend to have at least 15A peak current supply."

Effectively. It is very possible that there are times when all 8 engines are running at the same time.

We are not electrical engineers but computer engineers with forgotten electrical notions.

Questions:

1. The Robotis cables(and Dynamixels) support/resist that amperage (15A)?

2. How would you distribute the power? What would be your power design?
a) a unique SMPS 12V 15A (4 traction motor+4 direction motor)
b) two SMPS 12V 10A or so...for two sections 2x(2 traction motor+2 direction motor)
c) a unique SMPS 12V 10A with two outpus... 2x(2 traction motor+2 direction motor)
d) four SMPS 12V 5A 4x(2 traction motor+2 direction motor)
...
?? ) Another options? What is the best option?

3. We will work with cable but we want a autonomous robot.too. The robot may have to move between five and eight kilos. A laptop, metallic structure,... We have 4 LIPO Battery 11.1V 1800mAh LB-012. Are the adequate? Too weak?
What battery do you recommend for this type of application with these requirements and performance? Could you indicate a specific model? What time of autonomy would be expected with a moderate use of the motors?

Thank you very much for the help!

JC
How would you distribute the energy?

[jwatte]

The Robotis wiring is generally 22 or 20 gauge. 15A through a single 22 gauge wire will heat the wire to the point where the insulation will scorch, even if it's high quality silicone insulation. Or the nylon connectors will melt, or the solder in the pins of the connectors will melt. While that COULD happen, what will actually happen if you build a single long string is that the voltage will drop the further out the string you go, when the inner servos draw power, and at some point the servos at the end of the string will detect under-voltage and reset.

Typically, you will build a "string" of 2-3 servos and power each "string" from a low-impedance source. If you want to run all the motors at 2A each (that's pretty heavy load,) then I'd recommend four strands, one for each 2 servos. The strands then plug into some board that can deliver enough amps to the connector, as well as distribute the Dynamixel bus serial connection.
I have at times spliced my own cables, where ground+signal goes to the Dynamixel controller, but ground+power goes to thicker wires that come from my battery.

One 1800 mAh LiPo battery will run for maybe 5 minutes if you run it at 15A, if you're lucky. The actual power rate is only 8C, so most modern LiPo batteries (from 10C to 50C) will be able to deliver that power without too much voltage sag. In fact, I often "float" a LiPo in parallel with my power supply (at 3.9 Volts per cell, so 11.7 Volts for a 3S LiPo) to make the LiPo supply any current surge that the power supply might not be quick enough to keep up with, even when I'm on the bench. Also, the batteries should have a good connector -- Dean's-style T connector, or XT60, are good options. XT30 might be enough. Tamiya or similar aren't up to the task in my experience.

I have used these batteries (depending on whether I'm brave enough to run at 16V or just 12V):
https://hobbyking.com/en_us/turnigy-...5c-w-xt60.html
https://hobbyking.com/en_us/turnigy-...5c-w-xt60.html
Note that the "C" rating multiplies with the "Ah" rating, so a "15C, 4 Ah" battery can deliver up to 60A.
You really want to be getting some connectors and 16-20 gauge stranded silicone covered wires and solder your own harnesses, though. Being able to hook up a controllable power supply while the battery is also hooked up is valuable!

So, if this will be a fixed installation, a single 12V / 15A power supply would be the best option. If this robot will move, then you need a battery that can move for long enough. Two batteries in parallel will run longer, but you have to make sure they are charged to the same voltage when you hook them up, else the higher-charged battery will attempt to load-dump into the lower-charged battery at a VERY HIGH RATE and everything will heat up and potentially catch on fire.
I would recommend a "bench top" power supply rated at 10-20A, because you can dial in voltage and current limitations. I've used these:
https://amzn.to/2jTBC7f
https://amzn.to/2lpMMRD
Especially the B&K is very nice, but beware that the front jacks are only rated to 8A; the back has screw terminals rated to 30A.

To make boards that have sufficient power distribution ability, I design my own in a PCB design package (I use KiCad,) and have the PCBs made at a place like Osh Park or Seeed Studio, and then get the right connectors from Digi-Key or another distributor to solder onto the PCBs to connect the Robotis connectors. The Robotis "hub" boards have not been that great for me, although maybe they are better for the XM cables, which use a different connector than the previous series.

You don't want to use a laptop for the robot. It's way too much weight you don't need. (Unless the robot rolls on wheels, then it might be OK. Maybe.)
Initially I used a "NUC" style Intel-based PC, but that, too, was a poor match to the needs of a mobile robot. I've had better results with OpenCM 9.04, Teensy 3.2/3.5, and Raspberry Pi. The Raspberry Pi 4 at 4 GB might be pretty good for you. You could plug in a display when the robot is docked, but leave the display behind when it moves.
Other options to look into include the Nvidia Jetson Nano (which is basically 1/2 a Jetson TX1) or a Jetson Xavier (which is a lot of compute power, but the heat sink on that is starting to add significant weight...)

[tician]

bleh. I've been encountering lots of people not reading for comprehension recently... one in particular at work last week.

jwatte, they mentioned in the very first sentences: four independently steered wheels on a rover. Whether steered, powered caster like the PR2, or higher speed four wheeled rover design like your money pits, weight will not be much of an issue as long as the chassis can support it.

In the past, there were some robot groups/researchers that used four/eight EX-106 to drive the caster wheel pairs and four RX-64(?) to steer the casters to produce a much less expensive version of the PR2's steered caster system, and I'm thinking it worked quite well up to the tens of kilograms payload. The lower torque servos you are using should be good enough for <10kg on level ground with occasional ADA-compliant ramps. At the nominal/continuous (1/5 peak torque) rating of ~0.8N-m per servo, you should still get ~1.6kg of force at each wheel using 4" diameter wheels. A major potential issue might be broken servo gearsets/cases if you mount the wheels directly to the output horn of the servo without additional bearings to support the wheel and keep the servo providing only the torque to turn the wheel (ideally: mount the wheel on a shaft, mount the shaft in bearing blocks on both sides of the wheel, connect the shaft to the servo horn with a flex coupler).


As for powering the servos, I would use a small hub similar to the U2D2 for each wheel pair (one steering servo and one traction servo) where each of those two servos is individually connected directly to the hub (no daisy-chaining). The four hubs would then be connected to a battery / power buss using 18AWG or larger diameter wire. I had been looking at using DOSA (1/16 or 1/8, like this: https://www.digikey.com/product-deta...503-ND/5878834) brick DC-DC converters for powering servos in Ripley a while back, but other stuff intervened before I could get serious about testing it. Was going to build 8~16S packs out of 16650 cylindrical cells (prefer 3.2V LiFePO4, but had been looking at 3.7V LiPo, as well) to feed the DOSA brick to get a 12V/20A output to power a bunch of AX-12A servos in a modified HROS-1 humanoid.

For reference, the DARwIn-OP is 20 MX-28 servos (1.4A max versus 2.3A max of XM-430-W350) and ~8W single-board computer usually powered by a 12V/5A SMPS or tiny 3S LiPo battery. The HROS-1 (20 AX-12A servos) and most other Trossen kits ship with a 12V/10A SMPS for extra headroom (and allowing the larger HROS-5 to use the same SMPS - i.e. fewer unique items to keep in stock).

[jwatte]

prefer 3.2V LiFePO4, but had been looking at 3.7V LiPo

I agree -- LiFePO4 are nicer in many ways (including number of lifecycles, matching to 12V rail voltages, and fire safety.)
Unfortunately they seem even harder to come by now than a few years back. The lifecycle limitations is being solved by the EV people for LiPo, and the other benefits are really only valuable for "lead acid replacement" use cases in the market.
Which means that one way of getting LiFePO4 cells is to buy LAR batteries and breaking them open ...
Or just use the LAR battery, like this: https://amzn.to/2lwYji3 which claims 5 Ah and 20A discharge capacity.

rover weight

My first iteration of Money Pit used a ITX case Core i5 based computer, and a stereo camera I built with two USB cameras in an aluminum bracket. That, too, ended up being heavier than I wanted. (The latest iteration uses a Jetson AGX Xavier; having NVIDIA is nice because of the StereoLabs Zed camera compability.)

[tician]

batteryspace.com still sells plenty of LiFePO4 cells, and assorted e-bike sites sell them as well.

The 'accurate positioning on a plane' bit says indoor with human interaction not outdoor speed trials, so weight is again not much of an issue if the chassis and suspension can mechanically handle the load without deforming/breaking or the wheels/motors slipping. Darsha with its shitty non-round omni-wheels, surplus gear motors, and cheap arduino motor board clones was still able to move around easily when loaded out with a DARwIn-OP, 3kg 4S/20Ah LiFePO4 battery from batteryspace (https://www.batteryspace.com/lifepo4...assed-dgr.aspx), 17" laptop, kinect, and a couple kilograms of aluminum extrusions.

[jcarlos]

jwatte, tician. Sorry for my English. Thank you for your extensive answers! I get very valuable information from them. And they are interesting answers.


Yes, tician, all your deductions are right :-). It is a four independently steered wheels mobile robot. Here is a sample video.

4 traction servo + 4 steering servo. So 4 wheel pairs (1 traction servo + 1 steering servo)

> The environment will always be indoor with near humans and human interaction. No stairs. :-)
> The speed will be low/modest. Perhaps 1 m/s at full speed.
> The total robot weight will be inferior to 10kg. We have planned 10 kg maximum.
> The laptop is essential. We need it for another task.

We like the proposal:
-> Four strands, one for each 2 servos.
-> Each wheel pair [...] is individually connected directly to the hub (no daisy-chaining)

We want not fire, smoke and despair. Not batteries in parallel. :-) Ok.

Now, let's forget the batteries for now. Let's suppose that the mobile robot is always connected by cable.

As I said our knowledge about electricity is very limited. We are software engineers.

But based on your comments I have made this diagram:




Basically 4 independent SMPS 12V 5V for each pair wheel (1 traction servo + 1 steering servo)
The control data (Data+ and Data-) is connected to all motors (daisy-chain).
Traction servos and Steering servos are all Dynamixel XM430-W350-R

What is your opinion? Thank you for your help!

JC

[jwatte]

That should, theoretically, work fine! RS-485 is supposed to work across isolated ground domains (that being one of the main points.)

[SamQ]

@jcarlos Did you ever complete the build of Turtlebot 3 Friends Conveyor? If, so would you be willing to share the steps necessary to complete the build? I've looked everywhere online to no avail until finding this thread; looks promising.