Inside the Neato drive wheel

[vic7767]

Several owners have reported that their Neatos suffered from the "stuck drive wheel" LCD display. Further examination revealed some binding within the electric motor itself and had nothing to do with the drive motor gear assembly. Upon dis-assembly and inspection of the internal parts of the motor one can suspect that the epoxy type material used within the motor armature to keep the winding stable has somehow had bits of the material broken away thus causing the armature to bind within the motor case. There is no way for the broken bits to escape the confines of the motor housing. Some owners have forced the motor to turn manually in an attempt to resolve the binding. It is possible that those efforts basically pulverized the bits into smaller bits that then did not block the armature movement against the motor housing. The only way to really clean out the motor is to accomplish a dis=assembly which will require a mini gear puller, soldering station equipped with solder and a small vise.

[mfortuna]

I am confused by that picture. I have never seen epoxy used to secure windings in a brushed motor, there is no need to. The windings get secured by tightly winding them and then securing the ends to the commutator via a mechanical chrimp and solder.

Could it be the epoxy is being used to balance the armature? On many motors I have seen the armature is balanced by drilling out material in the core, maybe here it is being done by adding material?

[vic7767]

I agree Mike, the hard blob of material at the end of one of the copper windings is evidently used for balancing. These are low RPM motors so balance tweaking isn't accomplished using a drill bit and removing metal from the armature but instead a nice size blob of some type of material is added for balance.

That may help explain why there are some motor failures being reported. It's just not known what is causing the material to break off or come loose. But now we know what the motor does when that balance material becomes free to roam within the motor chamber.

[glnc222]

Impressive photos, would all make a great printed manual. There is something exceptionally clear in Vic's photos, the composition as well as the optics.

I wonder how durable this odd gluing method is. How they ever calibrate it I cannot imagine. Take the Neato in for wheel balancing like a car?

[mfortuna]

Armatures are pretty simple to balance by hand using a RC plane prop balancer. Each end of the shaft gets placed on a bearing like surface. Spin the armature, wait for it to stop. Mark the bottom with a sharpie, this is the heavier side. Repeat and see if the marked spot again faces down. At his point you can remove material from the bottom side or add material to the top side. I drill a very small amount of the core on the bottom side. I have never seen the winding slot be the low side. My first pass I just get a 1/8" hole started in the core, i.e. the tapered part of the drill bit doesn't go past the metal surface.

For wheels and props you can add clay or scotch tape to the light side. For an armature I always remove material since clay or scotch tape wouldn't work. I never though to add epoxy. I suppose you could start with clay and replace it with a like amount of epoxy.

Of course a mass producer probably has more advanced equipment.

[vic7767]

I saw different markings and some numbers written with a black felt pen on the different parts of the rotor. Then nothing on the two portions of the armature where the epoxy was applied.

[mfortuna]

Those may have been the heavier sides as determined by the balance check. Not sure why they marked both, maybe before and after the expoxy was applied?

[metabflux]

. . .That may help explain why there are some motor failures being reported. It's just not known what is causing the material to break off or come loose. But now we know what the motor does when that balance material becomes free to roam within the motor chamber.

Great picture. I can confirm that this is the material I observed on the coil and fragments in the casing of a motor that was causing a "wheel stuck" error. Though I only could see the back side of the rotor because I could not remove the gear. Removing the loose material and carefully re-aligning the rotor resolved the error. I am curious whether remanufactured Neatos will have different motors.

[metabflux]

I am curious whether remanufactured Neatos will have different motors.

XV-12 Wheel motors. The motor on the LEFT was from a unit that failed due to a “Stuck Wheel” error (Nov 2012). The motor on the RIGHT is from a remanufactured XV-12 received as a replacement unit. Note obvious differences to (Top) the stylistic and information content of demarcations on the casing, (Middle) the changes to the electronic RPM sensor and (Bottom)the presence of openings in the back of the case. It is unknown whether there are internal changes related to the issues discussed above. - MetabFlux

I'm hopeful that these differences indicate that NEATO addressed the problems causing the "stuck wheel" error.

[vic7767]

Those changes appear to be cost reduction changes, what is not know is if there were changes internal to the motor balancing process and material used.

[MarkProper]

Seriously vic, I could summize that Neato Robotics should either be hiring you or retroactively paying you for all the work you have done investigating and finding fixes for this robot. I don't know what you do for a living, but i bet you'd be an asset to the Neato team if they felt like bringing you on to improve their robots. Although it is individual components which seem to be failing and not the overall design, it gives me hopes that they will have a product which has a usefull working lifetime soon. Once they iron out all the kinks I'll be getting one for sure.

[vic7767]

Thanks MarkProper for the kind words. I'm actually a retired engineer from Bell Telephone Laboratories and now have an interest in home robotics. I've gone through a few different iterations of these Neato models and they have improved a lot. The new XV-21 3.0 I'm running at home is performing quite well presently.

[glnc222]

What if you ran each wheel separately over USB and compared stats for each to see if there is binding? The rpm cannot be set apparently, but a speed and acceleration, and there's a GetMotors with all sorts of detail on control.

[metabflux]

What if you ran each wheel separately over USB and compared stats for each to see if there is binding? The rpm cannot be set apparently, but a speed and acceleration, and there's a GetMotors with all sorts of detail on control.

Figure: Percent loads of wheel motors from an XV12 that gave a "Right wheel stuck" error and the remanufactured XV12 replacement. Data is shown as the average and standard deviation of 3 measurements.

The figure above shows the data from the experiments suggested by glnc222. It appears the new motors are functionally improved. A complete description is below.

Method:
The XV12 that suffered the "Right Wheel Stuck" was running software version 3.0 and the Remanufactured XV12 was running software version 2.6. Each XV12 was connected to a Lenovo windows XP laptop via USB and commands were issued using PuTTy. The command "SetMotor 3000 3000 100" was used to achieve a slow speed over 3 meters. The command "SetMotor 3000 3000 300" was used to achieve a fast speed over 3 meters. During each run, the GetMotor command was issued 3-times and the results were averaged. Tests were performed on a hardwood floor.

The GetMotor command returned slightly different perameters when issued to units running the 3.0 and 2.6 softwares. For example,

Software version 3.0 (defective unit) returned:

LeftWheel_RPM 3
LeftWheel_Load% 46
LeftWheel_PositionInMM 9530
LeftWheel_Speed 93
RightWheel_RPM 3
RightWheel_Load% 45
RightWheel_PositionInMM 24526
RightWheel_Speed 102

Software version 2.6 (remanufactured unit) returned:

LeftWheel_MaxPWM 65536
LeftWheel_PWM 18610
LeftWheel_mVolts -1961
LeftWheel_Encoder 157557
LeftWheel_PositionInMM 22463
LeftWheel_RPM 1740
RightWheel_MaxPWM 65536
RightWheel_PWM 17076
RightWheel_mVolts -2339
RightWheel_Encoder 157554
RightWheel_PositionInMM 22463
RightWheel_RPM 1740

The Wheel_Load% parameter was used directly from the XV12 running 3.0, but since the Wheel_Load% was not directly given in the 2.6 version, it was calculated using the equation:
Wheel_Load% = Wheel_PWM / Wheel_MaxPWM X 100.

Results:

The motors from the XV12 that failed from a "Wheel stuck" error had higher percent loads at both high and low speeds compared to the remanufactured unit. At the high speed, the remanufactured unit achieved an ~80% load compared to over 90-95% for the unit which failed. The right wheel of the failed unit, which was repaired as previously described ( viewtopic.php?f=20&t=16527&start=40#p111933 ), achieved a slightly higher percent load than the wheel which did not fail ( 90 vs 95% at the high speed). At the low speed, the motors from the defective unit achieved a % load of 43-46% compared to only 26-27% generated by the remanufactured replacement unit.

Summary:
It appears that the new motors used in this particular replacement unit generate a lower % load than the motors in the unit that ultimately failed from a " wheel stuck" error. The data might indicate that the motors in the replacement unit are functionally distinct and improved compared to the motor that failed. However, it is not clear whether the way the Wheel_Load% parameter is reported by the 3.0 software is the same as the Wheel_PWM / Wheel_MaxPWM X 100 used to analyse the 2.6 software. Time will tell.

[glnc222]

Nice data. Not sure what they mean by "load per cent", maybe fraction of maximum power used, or current used (voltage is fixed)? So if a wheel is sticking, it takes more power to bring it to a specified rpm under digital control.
Look for deviations between the wheels in the ratio of "load per cent" to rpm, or millimeters speed. Would be handy to have a standard spec range to check for defects.

It seems there is little quality control on these things. They go to all the trouble to balance a rotor but not to check the final product, kind of strange. Mysteries of the orient. Or they only break over time?

How did you get it moving on the floor while running USB? I get an error to unplug the cable -- it runs after the cable is unplugged?

Fascinating it has an acceleration parameter. Reminds me of an ancient daisy wheel printer I once had with all discrete logic chips, which had a function to accelerate and decelerate the printing wheel even though it rotates very fast and it all takes place in milliseconds. No stepper motors there. Like driving across town to bring a letter into position and hammer against the paper.

[metabflux]

Nice data. Not sure what they mean by "load per cent", maybe fraction of maximum power used, or current used (voltage is fixed)?

I'm really not sure, which may make the comparison between the software versions difficult. I interpreted the Load% to be the % of max power being used. Maybe someone here can clarify.

There is also a difference between 2.6 and 3.0 with regard to how it handles speed when sensing resistance. When the set motor command is given while the unit is on its back (I.e. no load on the wheels), 2.6 turns the wheels at the expected RPM, but 3.0 seems to default to some very slow RPM until it senses resistance. If you press softly against the wheel, 3.0 immediately speeds up the motor.

How did you get it moving on the floor while running USB? I get an error to unplug the cable -- it runs after the cable is unplugged?

It didn't give me a problem. I wonder if it is terminal specific. I was chasing behind it with my lap top.

[vic7767]

I think the only accurate reading method to determine the amount of power needed to spin up the motor will be to use a fixed voltage (15 vdc) and measure the amount of current consumed.

[glnc222]

Total current is reported with analog sensors; increments for different parts can be calculated. Only way to get the front brush motor load I think, unless in getmotors. Load may refer to per cent of cycle the pulse modulation is on, 100 being continuous voltage -- fan motor data sheets have that sort of thing. PWM -- pulse width modulation, 100 - maximum width. Average current will be proportional to width.
Sounds like they refined the motor control with added sophistication in 3.0, detecting free spinning. Thought of running pressed to a bicycle wheel.

[mikes60]

Hi to all here. After a month of using my XV-21 I've got same problem with "left whell stuck". Being in Russia I have no way to use my warranty, so I'm going for self-repair. And I've got question regarding the subject.

As Vic wrote it's necessary to have mini gear puller, soldering station equipped with solder and a small vise. I'm out of first and third tools, so I wander if I can detach gold gear (bottom of motor) and black RPM disk (top of the motor) using simplier tools, like pliers or something like that. And if there detailed instruction of how-to disassembly wheel motor, I'll really appreciate the link.

P.S. I have "new type" of motor, so probably improvements of motor doesn't fix stuck problem...

[vic7767]

if there detailed instruction of how-to disassembly wheel motor, I'll really appreciate the link.

Here's a link to how to dis-assemble a Dirt Dog motor, the process is similar for the Neato:
http://www.robot-doc.com/19.html