Samsung PowerBot Virtual Wall.

Inside the Roomba and Scooba and more, Cool mods, Repair and Upgrades - including the all new iRobot Create Kit. Let's void that warranty baby!

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 11th, 2017, 2:02 pm

Expanded graphics of signals using Irfanview added to prior post. The Guard emission is NOT on TSOP 38khz (I will have to check again with TSOP, just didn't look like it). The Activation signal is modulation of 38khz for TSOP. Long binary code. Don't know if enough to reproduce.

[edit] Just checking voltage on a TSOP 38khz receiver near the bot front will show whether your 9040 emits this signal. If omitted, doubtful the firmware has response to the Guard's signal, but you would have to activate the Guard to tell. Besides, adding your own activation emitter to the bot would seem harder than adding mag strip detection or your own TSOP's.

[edit] Have to check whether I accidentally got the Clean button signal on the remote to run while recording on TV receiver on breadboard. Should use the panel button on robot.

38khz is 26.3 micro second square wave length. Modulation wave about 588 micro seconds or 22 carrier waves.
Guard emission looks like a 1111 Micro second square wave (both phases).
Time blocks 5ms in both graphs.
Interval between signal bursts about 180ms, should not matter to signal detection, just power saving. Seems a second longer interval on Activation, hard to capture.

The proximity detector system might operate in modes, switching to different functions according to TSOP detection and what not (stop obstacle monitoring when Activation signal sent, wait for reply etc.) -- unknown.
More later.

(for full screen click on original in post above; just a link here)
Image
Image
Last edited by glnc222 on October 11th, 2017, 6:12 pm, edited 2 times in total.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 11th, 2017, 5:53 pm

Battery

The higher power 9250 has a 21.6v 3600mah battery and the top 9350 has 4500mah, compared to 1900mah you mentioned 9040 low end.

2-parallel I think a 5 series (for voltage). HIgher capacity for the fan power, no longer run time.
The smalller 7000 bot has similar capacity.
Long life so don't even think about replacing -- you may want a new bot by then. Not the old NiMh.
Unless you got a worn out used one to begin with...
[edit] At this time these models are such new introductions none of them could be very old yet, not much more than a year at most.
Maximum spec seen such cells 800 cycles, will believe when I see it. My dock-based run time clock is to monitor the battery, show when it needed a premature recharge. This can also be seen on the battery charge panel indicator because I only run half an hour, with a charge bar still left at the end. When it goes to zero at the end, capacity has declined with age, and the end is coming...

I expect the pack is 18650 size cells, haven't looked, but consistent with shape and capacity. Thing with batteries is the tabs are spot welded, not soldered (to avoid heat damage -- especially with fire hazards on lithium). Still a skilled German guy used a high quality iron to solder Samsung cells for Neato -- quickly (it's all in the wrist...). I used a 120watt gun on a safer type. Many dealers, like the German Akutiele provides spot welded tab services. I would look at newer LiNMC safer type cells, Panasonic used on Neato projects, newer Samsung's (been a few years). Best is LiFePo4 from electric bikes, solar power, but lower energy density and voltage -- 3 parallel,6 series... inherently safe. Used in Robomow lawn mower bot. 10 times the life... But some newer cells showed up sufficient for a Botvac in 2 parallel. Could be more expensive than other cells, unknown.
Last edited by glnc222 on October 12th, 2017, 3:40 am, edited 1 time in total.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 11th, 2017, 11:35 pm

Ordered Saleae logic analyzer 8ch only $12 ebay delivered -- must not be popular, much higher list price. Ships intl as well. Volume dealer New Jersey.

Confirming the signal emitted by the Guard is NOT 32khz modulated like the activation signal from the Powerbot. Tested with TSOP blocking the bot signal.
Easier to reproduce this signal than try an activate the guard emitter with the complicated 38khz code from the Powerbot.

Added a Point Following 32khz remote signal found in addition to first posted, to post above with first graph.

Mag sensors: there was a Neato test with neodymium magnets but putting them on a steel strip of construction material.
Of course one magnet under the sensor will work -- you just have to cover the entire distance. There is also a particular pole, south I think, not sure -- check parts specs -- needed on top for the Hall Effect sensors. Strips are magnetized with opposite poles on top and bottom, same all along the surface. The refrigerator magnetic materials do not work because alternating, multi-pole pattern (which might stick better?). Xiaomi is also selling mag strips, besides Samsung in parts dept. (Samsung has self-stick kind) -- compare prices (they could all be the same from third party after-market makers; there is a machine which magnetizes such materials, adjustable for different patterns).

Mag Sensor Mounting
Note on the bottom of the Powerbot the intake in front has two triangular forms to left and right of the charging contacts, forming a channel for large debris. The forms press onto the floor and might have empty spaces good for mounting mag sensors inside. How the 9350 with them has not been shown by owners yet.

Note your Powerbot probably has no mag sensors in it. Only added in the 9350 and later U.S. models, different system boards. They also added counter weights in the front. You should be able to see opening the front section, screws under the brush holds the top. They would have to be under the board to be close to the floor, along with the cliff sensors. Half an inch away is too far. Strips placed underneath some carpets do not work. A neodymium magnet will be stronger.

To reproduce the guard activation signal, it might help to first reproduce a button signal from the remote, to get the timing used. I am not sure how the modulation works exactly, the data sheets for the TSOP etc. may need to be consulted.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 12th, 2017, 3:00 am

thresholds

It is sometimes possible for robots to cross a threshold too high by extending the existing threshold molding with a material to make a wider ramp, maybe only a couple inches wider. Possibly not attractive to alter the house, and it depends on what lumber or other hardware supplies are available. Portable, temporary fittings could be made maybe with thin cardboard and tape across the threshold.

Powerbots can climb a 1/4 in. vertical cliff, and some stores carry pre-cut 1/4 in. hobby lumber strips which might be used for a sort of stair step. A beveled molding form useful might be a type used to frame windows. There are also plastic, rubber cord and cable protectors with beveled edges, which can be cut in two. Problem with those is they are flexible and may be harder to fasten in place, besides expensive.

The 9350 top Powerbot can climb an additional half centimeter over other models because of flexible wheel rims adding more traction.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby Bandi94 » October 12th, 2017, 4:36 am

I didn't have time last evening to play with the VW. Good to hear that you decided to go for a logic analyzer surely it will be helpful for other projects to, i could no longer work without one when i have to check PWM output and other logic signals. Also i got finally a response for samsung, the 9040W model is not compatible with VW nor mag strip. Anyway thank you for your support, it was very helpful.

If there are other persons interested in this topic, maybe when you will receive the analyzer you can record 10-15 seconds of data, then we can clearly investigate precise timing as the analyzer will tell the exact period of every edge, then i can replicate them and post some kind of software and circuit diagram to clone the VW if there are interested people to build them.
Bandi94
 
Posts: 14
Joined: October 2nd, 2017, 4:18 pm

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 12th, 2017, 1:20 pm

I will look into adding mag strip boundaries to the Powerbot with a sensor from Digi-Key, as Samsung has sensibly switched to this method from the cumbersome and poorly focused optical system.
For doorways a stick of wood or plastic corner guard is probably adequate for temporary placement. Sometimes furniture has to be moved for best cleaning anyway.
I can forward parts from Digi-Key if wanted; their intl shipping might be expensive fast service.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 13th, 2017, 1:03 am

The fancy solution to your problem is in the top models, some others now, with the persistent mapping and "select-a-room" feature, picking a room to clean on a map kept on the smartphone. What I don't think they have yet is adding a sequence of rooms to clean in order with a single command, completing each room one at a time. A pretty simple extension. Maybe hackers would do it. Maybe it could be done in the Alexa intergration, which has DIY support from Amazon, if it connects to the "select-a-room" feature. The ability in all models to resume cleaning where left off for charging shows how it can go to a location selected on a map, but using that more extensively is a marketing point for the more expensive models (and paying for the software development).
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby Bandi94 » October 13th, 2017, 2:48 am

Hey, i don't think the shipping would worth it, i can buy almost any component here to we have a very good Polish origin supplier across Europe in many countries and they have almost anything you need. As for the mag strip mode i think i will wait a little more as i don't want to loos the warranty. I had seen few reviews about the "select-a-room" feature and many people had returned the powerbot because the feature was not working as it supposed to. Some of them failed to create a map or the created map was not similar to the actual layout, other people complained that after couple of cleaning cycles the powerbot would return after 2min saying that it can't locate the room. Maybe these issues were fixed until now. Anyway i don't see the reason to go buy that expensive model for now, i will stick to some physical object to block the doorway if it will be necessary and when the time will come to replace the powerbot maybe the prices would drop a little bit as the current price is just to high for a vacuum cleaner.
Bandi94
 
Posts: 14
Joined: October 2nd, 2017, 4:18 pm

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 13th, 2017, 1:39 pm

You should lubricate the brush drive bushing maybe annually; just had to open it up when some rattling noise developed after two years:
http://www.robotreviews.com/chat/viewtopic.php?p=145618#p145618

Because my bumper extension made with heavy Gorilla tape without residue, turns out easily pulled off and re-applied. I might add a connector having discovered the lube job needed periodically. There is some fancy tape used by theater stage electricians called Gaffer's tape (gaffers do wire rigging on stage sets, concerts etc.) -- no residue. Maybe electrical rated compared to duct tape.

Even with the wheel extension sensor disabled the bot knows when picked up via the cliff sensors. It runs momentarily then stops. If the cliff sensors are covered, it keeps running until button pressed.

I only paid for higher end model because of new wall-to-wall carpet possibly requiring higher power to clean. Turns out the Max mode is not needed, but the base level appears to be higher on those models, yet I do not have measurements of the lower end ones. The specs are hard to interpret when only a Max mode applies more power used in the rating -- whether the Normal mode turns out the same. The 9250 is cheaper than the 9350 with all but the room thing. I like the washable filter, no wear on it after two years.

I don't see why that Select-A-Room could not be supplied as an update to my model, but they don't sell software except through hardware. Even the "point cleaning" thing seems pointless to me. Some users get the room thing to work, some discussion on Roboter-Forum Germany, but I have seen complaints like you mentioned. Neato has just announced a more thorough "persistent mapping". I expect more Samsung additions -- wait for the "B" version as they say in electronics. Other brands improving, LG and Bosch.

I saw the cliff sensor has four wires compared to common three, will see whether convenient to add a mag sensor when parts arrive. I usually can make additions removable for servicing, even if just desoldering or even clipping a wire; no holes in the case etc. It is also possible to make break-out cable connectors. It looks like a single sensor on the front cliff sensor may suffice. Tape it inside or underneath.
There are some behavior issues, insuring that it does not get trapped by passing the sensor across the boundary turning.

[edit] I saw some four wire sensor components which omit generating and processing the pulsed emission, so have a separate emitter and detector wire, compared to a single processed output wire. Could be what Samsung uses, will see.

Compared to wood sticks or plastic corner guard the mag strips seem to be purely a cosmetic issue, or allowing them to be permanently placed. I only used one temporary across a doorway with the Neato. Common 1in. x 2 in. lumber, 1" x 1" square, or right-angle moldings, by the foot cheap in lumber supply at least here. Home Depot has a saw in the molding aisle to cut.
Last edited by glnc222 on October 15th, 2017, 4:02 am, edited 1 time in total.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 13th, 2017, 2:27 pm

No sooner than opening to lube the brush bearing, it suddenly no longer responds to the dumb Virtual Guard -- having just deciphered this for debugging.

Yet just when restarting after lubing, there was a rare "UPAM" display -- software update over WiFi -- though there is no new version shown through the smartphone data. Maybe here in the U.S. it updated with software for the mag strip models and disabled the Guard software -- I will have to see. Maybe just software damage. Or some wiring inside got pulled. It seems to be running the proximity sensors. Ironic since I don't usually have WiFi turned in the house anyway, started leaving it on for some TV streaming.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 14th, 2017, 12:46 am

For unknown reasons the Guard has stopped responding to an activation signal still being sent from the Powerbot. I have not checked the coding, as nothing could be done about it. The power indicator still blinks. The Guard was used only out of curiosity and because supplied with the bot, but clearly was inconvenient to use with its poor focus etc. I will just retire it. Only one Guard was supplied as well, and other methods can be used in multiple places far less expensively.

I suspect they used it to attract Roomba users with some similarity, with Roomba having a much larger market share than Neato Robotics with mag strips. Finally they realized the engineering and economic considerations. The Guards were also left over from the older Navibot product, so easy to incorporate. It also did not require additional sensors for magnetic fields inside the bot.

Mag sensors will arrive in a few days but a ferrite core from China could take a month, if that proves needed. I will first see if a sensor can be made before examining how it might be interfaced to the Powerbot cliff sensors, and mounted in the bot. Ideally a single one would be used next to the center front cliff sensor getting 5v power from there besides the signal wire. A cliff detection there should produce a behavior keeping that part of the bot away from the cliff preventing possible trap situations. The other side sensors are farther back from the front, probably for cases where when turning it moves the side over a cliff. It still could not cross a strip boundary without running the front sensor over it.

I think a cliff sensor interface might be made of two comparators instead of a comparator and op amp used before, allowing a single dual comparator eight pin IC to suffice, though with a lot of resistors (very small in 1/8W). Quantize the cliff sensor for only Cliff/NoCliff with one comparator, and the other quantizes the mag sensor with output serving as the reference voltage on the Cliff comparator, switching Cliff output Low on mag strip detection. Will see. Tricky arranging hysteresis feedback if needed. I will have to see what the Powerbot sensor does.

[edit] There might be a way to use just a dual opamp IC with an inverter configiuration on the mag sensor input, but requires some research. This would fully reproduce the full range of cliff sensor input without quantizing.
Last edited by glnc222 on October 15th, 2017, 1:53 am, edited 2 times in total.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 14th, 2017, 6:21 pm

Activation Change

The unusual update above not registered on smartphone appears to have changed the activation signal for the Virtual Guard. So nothing damaged inside the Guard or bot. You may not need a Guard to get this signal from the robot, depending on its firmware.

New signal from the robot (only when cleaning, from proximity sensors, on 38khz TV type carrier)

VirtualGaurdActivateNew.jpg
New Guard Activation Signal from Powerbot


Old signal from above (link only; for full screen use post above)

Image

Similar in parts, but does not work now.

Robot still works, "Point Cleaning" still works. Maybe they just made some mistake in servers etc. I am not complaining to Samsung as the Guard system has poor optics to begin with -- good riddance. While not focussed, just the opposite: over each emitter (mounted point up...) is an omni directional reflector molded in the black plastic. Clearly made for some old Navibot with random navigation. Guided nav needs a sharply defined boundary to map. They should not be selling this with Powerbots in Europe. Just some shortcut they took when starting Powerbot, before developing the mag strip versions.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 16th, 2017, 1:50 pm

Example of a portable wood molding barrier cobbled together from bits on hand (note rear support on right keeps it from turning over:
woodbarrier.jpg


For any fittings on baseboards 3M Command Adhesive is removable without damaging paint and has a Picture Hanging type like Velcro but stronger, like a locking type fastener. I would use for small blocks left on permanently to hold barriers. The existing base moldings were available here.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 17th, 2017, 10:41 pm

Mag Strip Interface

Will post separately if completed, not sure how far will go with the expensive unit.
Cliff sensor connectors 4 pin on top of the sensor board visible when opening the front, left, right, and center (from underneath, long cable around the side). These are likely for types with unprocessed emitter and detector lines similar Omron's, allowing varied cpu algorithms. They proved unaffected by and external IR saturating emitter as found on Neato robot project with integral signal processing in the Sharp sensor (which probably compares detection pulse on vs off for background elimination -- zeroed when the detector is always maxed).

Next to the center cliff connector is an unused 3 pin connector labeled HALL which might be for a Hall Effect Mag strip sensor in other models, such as Allegro A1324 used on Neato's. 5v, ground, and signal output. No point using as the software would not be present. There is plenty of room to mount the very tiny sensor packages, aside from any ferrite core, next to the center front cliff sensor, in a corner of that assembly actually.

The cliff sensor connector appears to be 2mm pitch (vs the larger 0.1in 2.5mm common) and a Mini JST 2.0mm PH 4-Pin on ebay or AliExpress might be adapted, though not exactly the same type, to fit for break out -- needed on these close spaced pins.

Possible interface circuits to use with single 8 pin IC or even just a transistor; with the unprocessed detector output it should not be quantized by a comparator and just passed to the cpu as is except grounding on mag sensing:

schematicAltMagSensor.jpg


Awaiting parts.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby Bandi94 » October 18th, 2017, 2:47 am

The circuit looks good, if the CPU expects a grounding i would just simply link the transistor instead of the diode before the 10k resistor, place a 10k resistor on the op-amps output and drive the transistors gate with it. Also that hall-sensor is a linear one with 2.5 mV/G, i would put a more sensitive one like SS496A1, it has a 3.1 mV/G. Anyway you can test it when it arrives, i don't think that the mag-strip has a very strong magnetic field so you might need to amplify the sensors output before comparing it to a reference, you may get maybe only 10-100mV of output change with a weak magnetic field.
Bandi94
 
Posts: 14
Joined: October 2nd, 2017, 4:18 pm

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » October 19th, 2017, 11:05 pm

[edit] Not only is there a 3-pin connector for a mag sensor on the board, there appears to be a mounting bracket for a component included on the frame of the front cliff sensor for a part in front of it, likely a mag sensor. A very small strip board mounted across perpendicular to the bottom, on which a sensor could be facing down at the bottom plugged straight into this board (front face of the sensors are the detection area). 4mm wide sensor.

Top view center cliff sensor with mag strip sensor in front 9000 series:
MagSensorAssy.jpg
cliff sensor with mag strip sensor 9000 series


No ferrite bar used in Neato Robotics appears present.
Image

The Allegro A1324LUA-T mag sensor used by Neato has sensitivity 5mv/Gauss compared to half that on the 1326 or the Honeywell. The weaker ones need to be virtually in contact with the Neato mag strip, while the A1324 detects 3/4 in. away, without any ferrite antenna (couple days yet to arrive), and should be usable.
The 2.5v quiescent output bias and 1/2 volt response signal renders the BC547 switching transistor interface insufficiently sensitive, while the TLC393 low power comparator used allows precision detection at minimum activation and maximum distance (final revisions posted later; use 2M hysteresis, 2.6v reference from 56K/62K divider). The floating open collector output might be connected without a diode but I prefer firm protection of the Cliff Sensor output circuit from any injected voltage, as the Sharp's in Neato's proved delicate. Those floating logic outputs can drive low current things I've found, just not other logic inputs (gives the cliff sensor an asthma attack).

It could be a couple weeks to get the connector needed from Hong Kong to try in the Powerbot in secure fashion.
The small capacitor advised by Allegro might be for the damage some IC's have with instant onset of power from battery supplies compared to transformers, needing some microseconds to stabilize internals; they don't say, but awfully small for filtering (.01uf on Neato sensor; .047uf most common I had). That odd magic number 47 in electronics (compared to the Galactic standard 42).

Interesting that the sensor has a negative response to the opposite side and pole of the mag strip, so systems could be made either way. This YouTube video compared several strip types on the Powerbot and found them all compatible (except the common stick-on material which is multi-pole, alternating, not for sensors).
https://www.youtube.com/watch?v=WnX86fztht4

Ironically, when I used the mag strip with the neato the strip was still fastened to a wood strip to raise it to the carpet level at a hard floor transition.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Re: Samsung PowerBot Virtual Wall.

Postby glnc222 » Today, 7:17 pm

Ferrite Antenna core

A 4mm x 20mm ferrite antenna core rod (labeled "mag slug" by Neato) behind the A1324 (4mm wide), rod end against the back, doubled the detection distance to Neato mag strip from 3/4 in. to 1.5in. outside the robot, or from inside the robot from 1/4 in. below the floor surface to 1/2 in., possibly useful underneath some carpets. Might be better than Samsung's own part.

I still do not have the cliff sensor signals and pin out; one JST connector found did not quite fit the Samsung connector shell, and it may be necessary to remove the JST hollow pins, cover with heat shrink tubing and individually push onto the pins on the Powerbot connector -- just to make it completely removable vs simpler cutting the Samsung cable wires.
glnc222
Robot Master
 
Posts: 4539
Joined: January 23rd, 2012, 9:19 pm
Location: North Carolina, U.S.

Previous

Return to Robotic Hacking

Who is online

Users browsing this forum: abs9986, vic7767 and 403 guests