Solved and fixed! Strange Neato D5 wheel and LIDAR problem

[HiTec]

First of all, i bought this bot used, with stuck left wheel problem. No warranty, no returns.

I diassembled it, cleaned it and swapped drive wheel assemblies to other way around.

Robot moves out from the dock correctly and also returns to dock, but when it starts cleaning, only the left wheel (when looked from the front of the robot) works. What in earth causes that it doesnt work when in cleaning mode, only when driving of the dock and to the dock? Edit: Also when i pull robot out from the dock, it wiggles back to it correctly. Firmware version 4.0. Connects to WiFi normally, cleaning summary shows small lidar map. Only error i sometimes get is "Please return me to dock, after it tries to drive with one wheel only".

Edit2: I somehow got it to start cleaning after couple of tries. Cleaned well for about 10 minutes, after that the speed of both wheels somehow lowered and strange sound appeared, like and electricity cracking sound? After that the the other wheel died completely. Very strange. Wheel still worked when the robot was but back to dock mode? Drive wheel brushes? Transistor that drives the wheel?

[HiTec]

Tested the motor and swapped it again. Definitely not a motor problem. Robot starts cleaning correctly, but loses left side motor in about 5 minutes. Sometimes it comes back, but its very, very slow. Does someone have any idea where to start looking? Where is the H-Bridge/Motor driver circuit in this robot? Medtech? Anyone?

Edit:

Updated to new firmware (4.2.3). Fault still there.

Took the light and magnifying glass and started to look around. Found two small chips right next two the drive motor connector (between the connector and solid cap), where it leaves the motherboard and goes to the lidar. Both chips are marked DRV8800, so they are motor drivers with H-Bridge capabilities. Im pretty sure the problem is on those chips. Search continues and i will keep you updated.

[glnc222]

Unless you can make measurements of the motor driver input and outputs to identify a fault, the normal procedure is to replace parts until it works. Besides power transistor driver IC's, there are often large capacitor cans for supporting the power supply for motors. Commercially the whole system board is replaced to minimize labor cost. One can also swap system boards with another of your many robots on hand...
If the model has a USB port the motors can be run separately with firmware and the Neato Control Program -- not sure whether the WiFi models also have a USB port. There may be a serial port connection edge connector on the system board in its absence. Up to the experts or engineers to examine. I doubt anyone has done this already for you on newer models.

[HiTec]

Thanks for the response. I only have Botvac Connected and Samsung Powerbot (Storm Trooper edition), and D5 motherboards are impossible to find right now.

More careful examination revealed some kind of water/juice spillage on that corner of the board. I first thought it was some flux residue, or protectant laquer, but then i saw corroded legs of those driver IC:s.

I cleaned them up, applied flux, swapped 0.2mm tip to my soldering iron and used 0.5mm solder to refresh the driver ic:s legs. Result was no go, left wheel doesnt work.

I already studied DRV8800 datasheet. I should be able to measure enable input (0.8v-2.0v), which enables the H-Bridge and also the Phase input (0.8v-2.0v), which controls the direction. If those work, then the earlier circuitry is ok and the fault is in the driver IC circuitry. I just need laptop, USB-cable and then hope that the Neatocontrol is working with D5.

Logic inputs pull-up resistors need to be checked also and charge pump ceramic caps. They are all located around the driver IC:s, as well as Load Supply pins bigger solid cap.

If the driver IC(s) are bad, then i need to buy hot air rework station, as those chips cant be removed, or correctly installed without it. They have heat dissipation GND pad beneath the chip, thats soldered to the motherboard.

[glnc222]

I switched to Powerbot some years ago for cleaning new carpet. Just replacing battery after 400 cycles now. See hacking forum if your's does not support mag strip boundaries, I added.

I've seen YouTube videos on the hot air IC replacement, not sure about heat sink pads. Well, they assemble by baking in an oven over solder paste. I think this has been used to replace the Maxim Fuel Gauge IC, not sure.

One component which has caused problems on old Neato's is damaged diodes from insufficient heat sinking and under specification, leaking current and causing intermittent problems.
With Neato Control (or terminal emulator) which might work on D5 since there is a USB port, the motor can be switched on and off for testing while tracing the signals.
The D3,5, etc. WiFi and Lithium battery models are very new and may have different system boards than other Botvac's, which were never traced by engineers anyway -- no longer a new thing like the XV introduction years ago.

[glnc222]

A possible factor is the wide range of loads from drive wheels on the motor driver IC. There is about a 1.5 amp load from drive wheels, depending on flooring, less on hard floors, normally -- at least as estimated from about 1 amp on the vacuum and 1/2 amp on the brush, and 3 amps measured over all. However, climbing and maneuvering certain ways, bumping things, can raise drive wheel loads momentarily up to 4 amps. Stalled motors can create spike loads until limited by driving software and hardware (why digital PWM control is wanted for one reason; I put a polyfuse on a side brush motor I added once). I just wonder if the kind of failures transistors are subject to could make them work under low loads but lose their normal capacity for high loads, if damaged by some short or heat, something.
Point is you may not be able to easily observe the failure of the driver IC under simple conditions.

[HiTec]

Very good point. This D5 struggles on carpet, whereas older Connected drives without problems.

DRV8800 maximum peak output current is 2.8A.

I got it to run again, but it only cleans 5-10 minutes. After that the left motor dies. If i raise the robot little bit on that side, the wheel barely spins. It could be that the driver IC dies or goes to the protect mode after some current peak. I havent measured temperatures, but if i recall correctly, the temp limit for protect mode was very high. Edit: 160c triggers nFault pin and somewhere around 170c outputs are disabled.

I think it would be more dangerous to try to measure so small legs on those chips than replace them.

I will order new driver IC:s.

[HiTec]

Just a quick update. I have started to study the driving circuit, with comparing the circuit board connections and DRV8800 datasheet so i could isolate the problem.

C328 solid cap (100uF/35V) is driver IC:s VBB pin cap.

When looking at the front of the robot, left side DRV8800 controls the left wheel (left wheel when looking at the back of the robot). DRV8800 outputs are filtered via coils or some kind of ferrite components, maybe for preventing some electronic interference.

L8 is left wheel out +
L9 is left wheel out -
L10 is right wheel out +
L11 is right wheel out -

From drivers IC:s, the outputs goes to the"L" components and then to the motor connector and from there through the LIDAR PCB and then to the wheels.

This helps a bit, because now i know not to touch on the right side driving circuit components.

[glnc222]

I should have thought of this sooner, but another factor is the tachometer signal from the drive wheel motor, important in motion guidance and other things with the motors, like detecting stalling. Tach's on motors have been defective before, though maybe not on drive wheels. But if a motor stalls as when bumping an obstacle, the drive might be cut by software to prevent over loading the motor, as these permanent magnet motors sort of short out when not rotating with applied power (why I put a polyfuse on a straight, non-pulsed drive motor). As there is no bumper sensor in the rear, wheel stalls are even used as substitutes for bump sensing to trigger diversion behavior. The robot won't run at all without some sort of tach signal there, so whether there can be some intermittent loss is unknown.
Mechanical problems with the tach magnetic disk could be involved, or the mag sensor on it.

(edit) I suppose swapping the drive motors left to right with no change in location of the problem would rule out defects in the tachometers.

The diagnostic problem is getting telemetry from all these signals when the robot is running and encountering the problem, the issue of reproducing the problem and being able to observe in detail when it happens. It is possible to rig up a tether with meter lines, and fancier things could be done with radio links not worth the trouble -- more expensive than a robot... One might try monitoring the motor signals over USB. You can also mount the robot on blocks with the wheel suspension pushed in to suppress the "off the floor" sense -- resting wheel arms on blocks -- and activate the wheels for driving. Robots with manual steering remotes can be used to engage drive when set up that way, besides diagnostic control over USB of individual motors. Of course, if the motor cuts out for one reason, a good tach signal will cut out as a consequence. So how you would sort that out is a question. A high speed Logic Analyzer might be needed on all the signals (inexpensive). Lacking the schematics and firmware code limits diagnostic ability -- and commercially it is never worth the labor compared to replacing entire boards. One can wonder if the mfg has specially made diagnostic equipment to work on these things in development -- a big supply of boards might be cheaper.

[HiTec]

Quick update, nothing fixed yet, but i noticed that when the left wheel dies after 5 minutes, it comes back when side brush turns off. Then when it comes back on, the wheel dies. Im now running a test without side brush motor and its encoder connected. (Theres side brush spin sensor in these new D-series) connected. Robot still runs without motor, so its little strange why the encoder is there, maybe to turn motor off when side brush is stuck so it doesnt wear the belt.

BTW. Interesting find. In D3 motherboards, there are side brush motor and encoder connectors below the motherboard even though the D3 doesnt have the side brush.

I think this is because easier manufacturing process between the 2 lower models. It would be interesting to see what happens when D3 owner connects side brush motor and encoder. If there still is a side brush code in a firmware, it should work (if the motherboard has other components ofc).

Edit. The side brush and left wheel relation was just a coincidence. Without side brush the behavior was the same.

[glnc222]

The Vorwerk V100 XV model had a side brush absent on the Neato XV models, later added on the Botvac. The firmware was shared and configured for the equipment with the unpublished USB command SetConfig, also used to specify lithium or NiMh battery type etc. where Vorwerk was first to use lithium batteries.
The same system board could have been used, even if not fully populated with components on the Neato -- no pictures of VR100 boards posted to compare (Vorwerk not sold in the U.S.). The firmware flashing procedure is a terminal emulator script including both memory flashing commands and configuration commands. The driver package installs a special terminal emulator "updater" program along with the driver. How the Botvac was done is not published yet, and the new WiFi models can have a lot of differences, with updating over WiFi etc.
If the D5 runs Neato Control over USB, you should be able to turn on and off the side brush there. The posted command manual may omit the side brush commands but I think they are mentioned here somewhere, a tedious search. Will see if I can find it time permitting, I just seem to recall this being posted sometime maybe from Russia. Might confuse with the configuration commands, not sure. Maybe something like SetMotor SideBrush ON/OFF.

Incidentally side brushes do nothing on carpet where dirt is embedded and needs the main brush. I have never liked side brushes since having them on a Roomba. They break off against sharp edges, wear out easily, pull on electrical cords etc. One reason I got a Powerbot lacking them. Each to his own.

When I added a side brush to the original Neato XV, I powered a small gear motor off the vacuum power line with high capacity, as the load varies widely over different filter clogging states (so much the load serves as a proxy for air flow measurement).

[HiTec]

I do like the idea of the powerbots auto shutter and adaptive suction power. What i dont like is that it seems to be idiot for the most of the time. Navigation is nowhere near what lidar is capable. My powerbot "drifts" along the wall, doesnt follow it on straight line. Powerbot also gets stuck to places where Neato doesnt.

I do like the IR navigation in front of the robot, very accurate on avoiding obstacles without bumping to them (if they are not dark/black).

Well, i mostly bought it for a testing purpose and because it was Limited Star Wars Edition.

If we could combine these two robots it would be perfect.

[HiTec]

Newest (3.3) NeatoController works fine with D5. Latest test: Run the robot normally until left wheel dies, then connect it to PC and run left wheel manually with neatocontroller. Works perfectly, disconnect and start cleaning -> left wheel dead?

[glnc222]

Like all the robots, I had to add things to the Powerbot to keep it from getting stuck. It seems to become a matter of which one is easiest to modify. The Neato would get trapped spinning around underneath cabinets too dumb to find its way out. Vacuum software seems to have gaps.
Roomba, Samsung, Dyson and others now use the camera more durable than mechanical lidar, and unaffected by dark materials, glass and reflective surfaces (including polished stone). While the Neato travels along walls, the Samsung runs the front up to them, driving back and forth between walls. More than one way to clean it seems.
Samsung has yet to offer the virtual boundaries, "no go lines", replacing physical boundaries, mag strips etc.

[glnc222]

That the motor seems to work suspicion might turn to some sensor input directing the cpu to stop or turn. The old XV's had a log file on the cpu which could show an error condition, but was omitted in the new OS used in the Botvacs (the command for that in the manual is not implemented on Botvac).

[glnc222]

With wheel recovering when side brush stops, there is a drop in system power load, which might reveal some defect in the logic power supply causing some borderline failure condition. Or load on something shared by the side brush and the wheel motor, driver or what not. Power supply problems can cause any number of weird failures throughout logic systems.
If possible run the wheel motor over USB while turning on and off the side brush.

The game is to reproduce the problem in a controlled way it can be analyzed.

[HiTec]

I suspect that the running time and temperature are irrelevant. After you drive the robot to that point that the left wheel doesnt work, it doesnt come back even if you reboot/reset the robot. The wheel works when driving with NeatoControl.

After recharge it works again in normal cleaning program. If i remember correctly, neatocontrol showed something over 16 volts as a battery voltage, when i connected it to pc after wheel failure. Very strange. So its something voltage related.

Edit. I will test more tomorrow.

[glnc222]

16v is a full battery level for the lithium batteries, which can fall a couple volts under load, after discharge. 16.8v is the maximum for a four series of lithium cells. Worn out batteries will still charge to this level but fall more rapidly than usual under load -- accelerated discharge curve, reduced run time. The computer alone is a small load which will not show the effects of cleaning loads or bad batteries. You can load the system about 1 amp by running the fan over USB. The front brush and wheels can add another two amps, sometimes more, when running, but 1 amp may show condition of the battery. If the vacuum is mounted on blocks, with the wheel arms depressed as though on the floor, all the motors can be turned on. The wheel load will be minimal without drag on the wheels.

Diagnosing electronics requires a known good battery to eliminate power supply problems.

The system has 5v logic component supplies and 3v cpu supplies from step-down regulator circuits which can potentially be defective. 5v runs the IR proximity and mag strip sensors. Noise in the power from inadequate filtering of motor supplies -- using PWM oscillations -- can disrupt cpu operation. Advanced test instruments can detect such things, oscilloscopes and such.

See the Mods & Repairs wiki thread top of forum for links to long video by engineer trouble shooting a Neato XV circuit board with instruments, kinds of things they do.

When on the charging base the system is powered off the charger.

[glnc222]

Telemetry trick: technically it is possible to run the robot with a USB tether to a notebook PC, arranging to keep the wires clear of lidar, out the back. When a cleaning run is initiated over USB, or from the panel with USB connected, the system requests the cable be disconnected to allow free movement. A Windows utility mentioned in the Mods & Repairs index thread allows turning off the USB port on the PC to simulate disconnection, and after the cleaning starts the USB can be restarted, and used while cleaning. Practical if the problem can be reproduced fairly early while cleaning, where it is possible to follow around with a tether. If an error condition is reported by the cpu, the error message will be sent over USB, in terminal emulator connection if not in Neato Control, not sure how that works.

If the motor is running under a cpu input to the driver and a driver fault stops the motor, the tach input to the cpu will show a navigation obstruction causing the cpu to stop the motor, either with a navigation error condition or some evasive behavior -- and when such fails for the driver fault, then a nav error condition. The tether method could allow seeing whether the motor stops from a cpu input change or from a driver fault. Neato Control data logging features might be used.

I once measured the running loads of the XV model using an ammeter tether to the battery.

[HiTec]

Thanks for the responses!

Battery is good, 36 charge/discharge cycles. Tested also with my Botvac Connected battery which has over 400 cycles.

I havent had time to run more tests, it will take couple of days. Interesting idea to keep usb logging active when in run.
Does robotreviews users btw have interest of detailed D5 disassembly pics? I can take some next time when i open the bot.

Edit. Quick test with lidar and main brush and wheels on and i was able to reproduce the problem with neatocontrol, where i can see cpu error codes?

The left wheel spins same speed if theres no load, if i force it, it will stall with slight touch. If i stop the main brush, the wheel takes more force to stall. And then if i stop the lidar it will take even more force to stall the wheel, but still less than right wheel.

It just feels like the left wheel doesnt get enough current. I will make some measurements with multimeter in a couple of days. DRV8800 has an nfault pin, but i havent studied it more closely to see if it needs special equipment (oscilloscope) to see what the fault is, or is it just and indicator for any possible fault (pin low/high for fault).