Neato XV-21 Rev 64 (Binky) never finishes cleaning

[nadcicle]

For full disclosure I should start out mentioning that my neato has tumbled down the stairs once early in its life.

I did however, pick it up and grab the charger and just leave it downstairs after that incident.

It did appear to clean fine for a while until the runs got shorter and shorter until it was obvious the battery needed replacing.

I took this as a sign to go ahead and get one of the lithium batteries I hear such good reviews about.

It really didn't seem to want to take these batteries at all at first. Then I read a review that said to jump 7.2v or a little above for 30 seconds or more to get them out of "Sleep" mode. I went ahead and did that and put the batteries back in and it came to life again.

I let it charge overnight and waited until I got home from work (30 hours) to give it a shot and low and behold it set off and cleaned for what seemed like an hour and 45 minutes. When it came time to go back to base it seemed absolutely lost. It didn't go anywhere near the base station at all! It just kept wandering around for about 10 minutes until it just bumped in to stuff enough and gave up against a wall saying "please clear my path". I read about 3ft on either side thing and cleaning off the base unit. I have it on a wall with about 4.5ft on either side and cleaned the base station accordingly.

I went ahead and just put it back on base manually and waited until the next cleaning cycle to see if it was a fluke. Low and behold it does it again. Time after time it keeps doing it without ever getting back to the base.

So I went ahead and just did a spot clean just to test and yet again no base station return.

I have also been noticing that sometimes my robot will go in to dead mode and I have to plug in a usb and hold down the power button frequently to reset and get it to come back on.

I have also monitored it closely and it seems to not negotiate a room like it used to. It just seems to hug against a wall and treat everything like one giant room where I could have swore it would entire a new room and do that kind of look around thing and go about cleaning that room until its done then move on to another room.

I would venture to say it's acting like it has no memory of cleaning at all or where it has been during the cleaning procedure.

Would any of this behavior seems like a C10 capacitor replacement issue. I have never done a replacement yet on this unit and wouldn't be against doing it, but I also see people saying that one of the symptoms is a dim green light. My light has not been dim at all.

I did take the time to hook up to diagnostics and all sensors appear working. I used my camera to verify lidar unit before, but using the diagnostic tool works great. Cliff sensors, lidar, all bumpers, wheels, and dust bin all working.

Any ideas here? I did order a Xiaomi roborock s50, but I do have 2 floors and would like to get this thing working the way it should.

[glnc222]

It might be worth replacing the inexpensive C10 capacitor before replacing the entire system board, just speculating. This capacitor filters the power supply on the cpu and memory, and any defect there can have any number of effects on operation -- and there were some instances reported of failures, suggesting border-line quality. The "dim green" situation is one where the memory is initially inoperative, putting the system in firmware update mode. The present failures appear to involve executing the software at a high level, else it would not work at all. Without the code and elaborate debugging set up, it would seem difficult to identify the precise defect in the system. Another possible action after trying the capacitor, is re-flashing the firmware, discussed in other threads, with files captured by users since Neato stopped on-line updating.

[nadcicle]

Thanks for the reply.

I will go ahead and order the capacitor in this case with the intention of replacing C10.

After doing this I will report back with the results.

[nadcicle]

Alright,

So after receiving the caps I set off replacing the elusive C10 cap.

Everybody always said there were a ton of screws, but I never could have imagined how many until my hands were cramping like crazy.

Anyways, after cracking this thing open it was obvious right away which one C10 was so I was on my way to desoldering the thing in no time. I replaced it and it and put the thing back together and it did something it hasn't done since I got it. It turned on immediately without the need for external usb to help jumpstart it.

So my first test seemed easy enough. I pressed start and scrolled through and chose spot clean because I wanted to see if it would return to the dock.It did! It hasn't done this for some time.

Next day I was stoked to start a full clean cycle. Right off the bat it set off and seemed better than before maneuvering around things and bumping and moving on. THe lidar appeared to be doing its job and slowing down when it approached things and all. With all that out the way came the ability for it to remember things. It isn't doing a very good job remember where it has been. It still seems to wander to certain places and hang around and sort of just stay there like it is lost for a few more minutes than normal. Sometimes it will go about and sometimes it seems like it is just lost and it will simply go over spots it has already done multiple times before. I want to call it the old school roomba effect. This robot used to go in straight lines all the time and never seemed to go over the same spot twice. Now I feel that it forgets things seconds after going there.

The good side is I never see dead mode anymore as it always starts up when i click the orange button.That tells me the C10 cap needed replacing for sure. I guess the next step is to go ahead with flashing the firmware. Maybe .the firmware is corrupt?
SW is 3.1.17844
LDS is V2.6.15295
Board Rev is 64
Board SW is 17624

When i had it open i noticed the board is a rev 4 with manufactur date of 1/02/2012

[glnc222]

I see no reason firmware errors could cause the behavior mentioned, instead of shutting the whole thing down. At the high level of the computer operation it would be difficult to discover what is happening, without the code and special tools. Memory defects? Flashing the firmware runs a risk of permanently ruining, "bricking", the device because the update process itself is performed by the firmware, with data over USB. Everything must be correct. But if you would have to replace the system board anyway....

Failure to move in straight lines (observable in Spot Clean) could suggest something in the tachometers on the the drive wheels, which are monitored to control short distance moves. Magnetic disks on the drive shafts. Or the processor does not handle the data properly. Or other problems with the drive wheels, debris in the motors, whatever.

Over USB it is possible to activate the drive wheels alone, with a long USB cable tether, to move a short distance for testing straight motion capability. Should be apparent from Spot Clean though. After long use, typical to check for motor debris though; blow out insides through any gap in the cases (they usually aren't made to open up though expert mechanics do so, resoldering things and what not, very skilled, and special tools).

There were no significant additions in later firmware versions though they go up to Rev3.4, tweaks to little things.

Most demonstrations of disassembly make use of powered screw drivers. I prefer manual tightening though, to avoid stripping threads in soft plastic. See Mods & Repairs list thread top of forum, for links to tool recommendations; the right driver can be manually twirled quite effectively. Be content there are not plastic locking tabs instead which break with such puzzle boxes. Don't lose any of the screws, metric sheet metal sizes not carried in U.S. stores, only mail order.

See the long video demonstration linked on YouTube by expert engineer refurbishing a Neato.

[nadcicle]

The good thing is I thoroughly went through the bot and it actually didn't look to bad.

The dust and debris was where it was meant to be and that was good.

In spot clean the bot acted normal and went in straight lines and returned to base no problem.

Only when put in full clean mode and approaching corners or furniture did things get a little hairy.

Testing with USB cable was one of the first things I did with Neato Control.

I hooked up to a 20' usb cable and issued commands to do everything successfully. It does go in a straight line.

I am leaning towards going forward with a flash. I have a much shorter usb cable I would use for that.

Before doing the C10 mod I would not feel comfortable doing a firmware upgrade as sometimes the unit simply would not power back on, but now the unit has no issues powering on. Of course that goes without saying, until I attempt a firmware flash and it fails lol

Luckily for me I understand all of the syntax involved and feel comfortable with everything covered in the firmware upgrade guide. So, as long as I get the text commands down and save the file with the correct name and such I hope to not see issues. Right now the bot isn't doing its job and doesn't clean very well so I wouldn't be losing much at this point as you indicated earlier.

[nadcicle]

Well,

I figured I'd report back again.

I did successfully flash v3.4 on without issue. As you suggested. This hasn't resolved any of the main issues.

The bot still seems to clean over already cleaned spots multiple times instead of navigating to new uncleaned parts of the house. It also becomes lost when it is time to go back to base for charge.

The only positive takeaways are that it does vacuum over an hour and 45 minutes and eventually seems like it will clean some parts of the house.

I have been looking over this forum for quite some time looking for similar issues and nothing seems to have ever been resolved. I almost want to crack it back open and go ahead and replace all the other capacitors as well just to see if that has any positive effect on it cleaning correctly.

One thing I have tried a number of times is spot clean successfully. I think I will take the bot away from the base in to another room try spot clean further away from its base to see if this issue persists.

[glnc222]

A known limitation of the lidar is range around 15ft, requiring obstacles in larger spaces in order to map and navigate.
Long hallways without contrasting points for reference are a problem. Suppose laser diode deterioration reduced the effective range so it cannot position properly? Lidar scans with the bot in the room over USB in Neato Control might reveal whether anything is odd with the lidar.

The lidar may also not reflect from dark colored surfaces, some furniture legs and such -- and will be reflected away by polished surfaces, glass, metal furniture etc.

Professional repair would just replace entire components or systems until it works. Not competitive with new unit prices especially if the battery is old. At least Xiaomi sells replacement parts, though has only expensive distribution in the U.S., shipping. Also some limited navigation on carpets with the intake design.

The SLAM (Simultaneous Location and Mapping) software triangulates on contrast identified points in scan, landmarks located by feature as bot position shifts. The lidar works by parallax range finding measuring angles on a camera sensor, and return intensity or brightness -- not signal travel timing like radar, for the short distances. Picks up on corners etc. for landmarks.

[nadcicle]

I can kind of see it being the lidar still.

I want to hold on to hope it's not being that it has never gotten a "my vision is blocked" error.

It gets the occasional "please clear my path".

Here is a video I did of trying to get it to go back to dock after setting it a mere inches away.

You can clearly see the lidar appearing to work. It simply wanders off bumping in to things like it is drunk and forgot its way home. IT does NOT do this during the cleaning cycle. Only when attempting to return to base. It negotiates obstacles just fine when cleaning. It will slow down and go around like normal.

https://www.youtube.com/watch?v=soK5aPvO00k

FYI, the roborock is leaps and bounds better than the xv-21. When set to Max that is! The mapping feature is probably the best thing since sliced bread. The one drawback is not having everything in 100% english, but it is close.

[glnc222]

When the problem is limited to locating and maneuvering into the dock, the patterned reflector in the dock behind the smoke plastic is likely at fault, sometimes missing in manufacture, or obstructed (some cases posted). The lidar recognizes the dock from the striped reflection pattern. Same material as used on bicycle reflectors and signage, returning light from any direction to the source with micro prism coating.
Easy to disassemble the dock.

Other brands use different methods. The Dyson camera sees a black shape pattern on the dock plate, and the Samsung has a coded IR beacon inside on which receivers triangulate. Neato's reflector is economical for use with the lidar.

The limitation of Xiaomi's from what I've read, is the intake design creates high drag on pile carpet limiting use with certain decor -- good for hard floors.

[nadcicle]

Funny thing is it finds the dock fine on every spot clean started from the dock.

Truth is it used to work and now it doesn't.

After 5 successful cleans from the roborock I'm less inclined to even care to get the neato working as I feel I'm exhausting all options.

[glnc222]

While the Xiaomi is attractive on price, at least with more attractive shipping outside the U.S., just as many hardware defects as Neato are reported on German Roboter-Forum, with even more defective lidar's than heard of on Neato (at least Xiaomi supplies replacement lidars fairly reasonable for the case). For higher quality electronics I turned to Samsung -- a major electronics company -- with extremely few defect reports -- and more expensive. Over time, the mechanical lidar parts in Neato and Xiaomi will require replacement parts. The web is full of lidar repairs for these, and other brands including Roomba have switched to solid state camera guidance with advances in electronics over a decade since the cheap lidars were developed (a great advance at the time, but now long in the tooth).
Still, none of the vacuum robots show the reliability typical of much simpler major appliances, or TV's for that matter. They all seem to be pushing the envelope in manufacture and design. The task they perform is challenging for the technology, especially when carpets are involved.

[nadcicle]

After reading through the post here viewtopic.php?f=20&t=19076&start=40

I will be replacing the rest of the capacitors with 35v 220uf ones that i have left over and reporting back again.

Looks like i may just get this thing working after all. *fingers crossed*

[Brett]

Yep - *sometimes* that works...

(-8