Scooba 5900 pump problem

[Jeff Carver]

I'm trying to get a used Scooba 5900 working. The problem is that the pump stops working erratically. I have used the method suggested (somewhere) for freeing the pump by thumping the Scooba on its side. That got it going for a while, and twice again when the pump quit (each time after a few minutes of running). Right now I can't get it to pump at all.

Sometimes I get the Check Tank light, sometimes not. But what is consistent is that I don't hear the pump running when I first power up. (Or, sometimes, it works correctly for a few minutes--and then stops pumping liquid, but keeps running as though on the dry cycle. However, when that has happened, there's been no change in the lights--just the power and the clean lights lit.)

I've read a lot of troubleshooting posts here on the Check Tank light, but haven't seen much about what to do when there's no sound of the pump running.

Oh--I've tried it with full clean tank, and half full clean tank. With vinegar, and with Scooba solution. And I tried it with a tank from a working Scooba. None of those things made any difference.

Ideas, anyone?

Thanks.

Jeff

[vic7767]

Head over to the READ FIRST Sticky and locate the procedures to perform the built in diagnostics on your Scooba. It may be that you just need to run the pump for a little while. That may get it going. There are know issues with the pump not running and it will be difficult to isolate to whether it is the pump motor or the electronics on the PCB. Either way if you want parts to play with your Scooba, go here and get some great parts from Chris at www.RoombaExchange.com

[Gordon]

I'm trying to get a used Scooba 5900 working. The problem is that the pump stops working erratically. ...Ideas, anyone?...

Read through this old thread. It may provide some guidance. If you want more reading, just do a RR Advanced Forum Search for 'pumpmotor' (no space between words).

Doing the Self-Tests is essential. You can confirm that the motor and its controller / driver, are functional.

Then concentrate on verifying that the solution inlet lance is doing a proper job of keeping the tank's poppet-valve open all the while the tank is engaged with the robot. If the tip of the lance has ever been crushed, the magnitude of valve opening will be reduced, and solution flow may get pinched off. When that happens, Scooba halts the pump and goes into the drying mode.

One member became convinced that the pump-motor may develop a dead-zone. I can't find that post, and I'm not sure it was ever affirmed.

[Jeff Carver]

I finally got to doing the self-diagnostic tests on the Scooba. The pump runs as long as I want it to on the pump test. But then I put it down on the floor on a cleaning cycle, and after a couple of minutes, it stops pumping fluid. (No indicator lights.)

When I return to diagnostic tests, I might have to thump it or shake it, and then it runs again, as long as I like. (Ah--but although the pump keeps running on the diagnostic, when I lift it up, I see the actual squirting tapers off. Doesn't quite stop, but becomes pretty minimal.)

The inlet lance appears fine. But here's something that puzzles me: when I hold the tank up and insert the little handle of a brush into the tank poppet valve, the liquid runs for a couple of seconds, then tapers off and stops. If I shake or tip the tank and try again, I once more get a flow, but only for a few seconds. Is that normal?

I thought maybe it was a vacuum lock, if air wasn't getting back up into the tank, but it seems to happen too fast for much of a vacuum to build up. (I don't know what the mechanism is for air returning to the tank as it empties.)

Anyway, I'm wondering if the flow from the tank should be continuous on that manual test, and if maybe a problem with that is causing the unit to go into drying mode when the fluid flow stops.

(That was a long shot, and I think it was wrong. I just tested a known good tank the same way, and got the same result.)

[Gordon]

More details are needed.

... doing self-... tests ... The pump runs as long as I want it to on the pump test.

Test-12, I presume.

But ... on a cleaning cycle, and after a couple of minutes, it stops pumping fluid. (No indicator lights.)

That ceasing of fluid ejection, coupled with "no indicator lights" and this next statement tends to imply:

When I return to diagnostic tests, I might have to thump it or shake it, and then it runs again,...

a) solution is still wetting the fluid-sensor (located just below the lance), and
b) the pump-motor is no longer turning.

Since the bot is in Test-12 at that point, you ought to hear a motor running. Did you?

...and then it runs again, as long as I like. (Ah--but although the pump keeps running on the diagnostic, when I lift it up, I see the actual squirting tapers off. Doesn't quite stop, but becomes pretty minimal.)

Test-12 does slow the pump-speed, starting at a high-rate, and throttling to a slower speed. I have no more detail about that function.

The inlet lance appears fine. But ... when {I manually open} the tank poppet valve, the liquid runs for a couple of seconds, then tapers off and stops. If I shake or tip the tank and try again, I once more get a flow, but only for a few seconds. Is that normal?

Apparently so, if the Tank has not run a recent mission, and even to some extent if it has! I just tried the draining operation on my Tank, which has been drying for about a week, and found the quick curtailment of flow. The Clean chamber had been filled to the max with plain water.

I then blew lung-air into the Venting valve (this valve is illustrated in Figure-4 as Item #7, in this section of "Scooba Technical"), and that break-away purge allowed water to flow out for another couple tens of seconds. To get a definite, continuous-flow of water, I had to un-stopper the filling port to allow as much air into the chamber as it wanted! I can't explain why the Vent vents satisfactorily while doing normal cleaning; the pump is barely capable of pulling a suction on the Clean chamber!

Just be aware that the Vent-valve may do a little sticking of its own when given a chance to dry thoroughly. It wouldn't hurt to wet it as part of the Tank preparation routine. {080324: Deleted comparison of valve construction to a sport-ball bladder-valve. Inadequate space for that type!}

...Anyway, I'm wondering if the flow from the tank should be continuous on that manual test, and if maybe a problem with that is causing the unit to go into drying mode when the fluid flow stops. ...

I think flow should continue, as long as you keep the bot in Test-12.

[Jeff Carver]

More details are needed.
Jeff Carver wrote:
... doing self-... tests ... The pump runs as long as I want it to on the pump test.
Test-12, I presume.

Yes, on Tests 12 and 13, the motor runs continuously as long as I allow the test to go--starting fast, then slowing. However, sometimes it takes a thump to get it started, suggesting that something is sticking somewhere. And as I mentioned above, the squirting is strong at first, then drops to a bare trickle, or no flow at all.

When I run it on cleaning cycle, the wetting starts strong, then tapers off and stops. Can't tell when the pump motor stops running, because the sound is drowned out, but it must.

Just before reading your reply, I did the same thing of retrying the manual test on the tank valve--and discovering that only by opening the filler port could I get a continuous stream (as you just reported). I was wondering where the air inlet was, and following your pointer to that other page, I have now done exactly what you did, and blown through the air vent port by mouth. My results were exactly like yours.

Right now I have the unit disassembled, with the pump pulled loose far enough for manual testing. Moving the pump by hand, I get good flow through the system. (I couldn't figure out a simulated gravity feed but I have a length of aquarium hose attached to the inlet lance, and a big baby syringe gently feeding water to the hose.)

I haven't figured out how to test the pump motor in the disassembled state. (I've seen the jig diagram elsewhere here, but don't have parts on hand to hook up a 9-volt battery. I'm also not too keen on opening up the circuit board compartment--though it may come to that.)

I think I may reassemble and try the self-diag tests with my syringe feed instead of the tank.

[Jeff Carver]

Forget what I just said. As soon as I'd posted, I decided what the hey, and opened the circuit board compartment far enough to unplug the pump motor. I hooked it up to a 9-volt battery and ran it right inside the machine, feeding fluid through the aquarium tubing. Ran it a lot, stopping and starting, everything worked fine.

So then I put it all back together and repeated the self-diag test #12, again feeding fluid through the aquarium tubing. Worked fine. Tried it with the tank. Worked fine.

It is now wetly mopping our bathroom floor. :) :)

So what'd I fix? I'm still mulling that. I think blowing out the air relief vent may have helped prevent vacuum lock from killing the pumping action once it was started. (Which might explain why there were no lights indicating dry cycle--if the inlet still had liquid present, but not moving.) (I had a similar problem once with a car that kept dying from vacuum lock in the gas tank.)

That doesn't explain why sometimes I had to whack it to get the pump started. (I have just performed at least a dozen start-stops, with no problem starting the pump.) Maybe my disassembling and reassembling the pump assembly fixed a bad connection, or a bad alignment. Who knows?

I'll keep using awhile and see how it does.

[red57]

I've been doing the prime method for a while and now I have the problem where it starts putting down solution and then stops about 15 min later. Sometimes when I prime the pump with the normal method it doesn't work. Maybe I need to get a tighter seal and use more volume?

Anyways, I took the scooba apart, hooked up tubing and water supply to the port and manual primed the pump on both sides. I do that and it seems to work. This has worked on 2 cycles and then I end up manually repriming again. Probably need a new pump

[Jeff Carver]

And the answer is...it worked for maybe half an hour, then stopped pumping. I terminated the cycle, and started a new cycle, and no pump sound.

Blast.

[Gordon]

And the answer is...it worked for maybe half an hour, then stopped pumping. I terminated the cycle, and started a new cycle, and no pump sound. ...

That was a lot of work!

It would be nice to know whether voltage still exists across the motor terminals after the pump has halted. If voltage exists, that would confirm another member's assertion that the motor has a normal dead-spot somewhere in its revolution (I don't subscribe to that for a 3-pole motor as used here).

If voltage across the motor has gone to zero, that would indicate the controller has terminated drive to the pump-motor--possibly due to an over-current condition.

To gain such knowledge about the applied motor-voltage, you would have to connect a two-wire test cable to the motor's terminals, thread the cable though some handy slits/vias of the robot's assembly, so the cable could connect to a small DVM taped to Scooba's Tank!

[Jeff Carver]

It would be nice to know whether voltage still exists across the motor terminals after the pump has halted.

That would indeed be nice to know.

If voltage exists, that would confirm another member's assertion that the motor has a normal dead-spot somewhere in its revolution (I don't subscribe to that for a 3-pole motor as used here).

I'm also wondering if it might be overheating and soft-seizing.

If voltage across the motor has gone to zero, that would indicate the controller has terminated drive to the pump-motor--possibly due to an over-current condition.

Or an intermittent bad connection in the wiring? Or something flaky on the board?

To gain such knowledge about the applied motor-voltage, you would have to connect a two-wire test cable to the motor's terminals, thread the cable though some handy slits/vias of the robot's assembly, so the cable could connect to a small DVM taped to Scooba's Tank!

Hm. I have an old-fashioned multimeter that I suppose could ride on top of the machine. But how to connect to the motors' terminals? I guess I'd have to solder leads on? (Ulp.)

Ironically, I am currently trying to track a somewhat similar problem on my Calypso washing machine! A motor controller board has twice blown a fuse, and I'm trying to figure out if the problem is in the motor or on the board. :(

[Gordon]

I have a number of things to write about, and don't have a well thought out sequence for their presentation, hence I will just palaver on!

Not many Scooba owners know that during a cleaning mission Scooba halts solution-pumping momentarily in response to certain events. After Roger Crier received his 5900 in early Spring 2006, he set it to cleaning his glass-topped dining table! He got under the glass to watch what Scooba was doing. He noticed that after bumping an object, and when linear forward travel halted so the bot could do a body-centered rotation, the jets ceased emitting fluid.

I now think that those brief stops, i.e., stopping the pump motor, give opportunity for the pump-motor to NOT come alive afterward! The cause of that non-operation is still unknown. But, iRobot is sufficiently convinced that its source is mechanical; a claim supported by the Company issuing formal instructions for sufferers of the fault to overcome it by impacting the periphery of their Scooba, in a specified direction, by holding Scooba in a specific orientation, and then allowing it to fall through a specified distance to the floor.

So, if we depend on those instructions as truth (and, in all probability are truth, since I think those instructions were derived by the iRobot trouble-shooting team that deduced and confirmed the source of the problem), we should be able to rule out several fault causes that simply don't relate well to mechanical conditions. Those to rule out would be things like: a) no motor-ON command issued by the MCU, b) motor's power-switch intermittent, and c) motor-armature sitting on a dead-spot. There may be others to add to this list. All linear, mechanical-shock inputs intended to rotate the motor-armature should not apply, since the directed impacts are parallel to the shaft-axis; and, even if applied orthogonal to the motor-shaft, would not work unless the armature was sufficiently unbalanced--which it is not.

In recent years we have learned that the motor-speed is controllable by Scooba's MCU, and at least two speeds are used. The slow speed might be used when Scooba is spiraling. I think there is less potential for motor-halting to occur as a result of running the motor at a slower speed, but, this two-speed (at least) operation also tells us something about the motor's driver circuit: It very likely uses a PWM (pulse-width modulated) signal to operate the motor's power switch.

The Mabuchi motor used for driving the pump is the same frame-size as Roomba uses to drive its vacuum-impeller. That comparison coupled with other observations about many similarities of component groupings found on those two robot's main_elex boards strongly imply that similar circuits have been used wherever practical. IOW: Scooba designs copied Roomba-Discovery's designs to a practical extent. Simply good business practice! So long as the bases are strong.

I then claim that the Scooba Pump-Driver Schematic (if anyone gets around to rev-engrg the Scooba board!) will look very much like Schematic_7, which we now have for Discovery's Impeller-Motor Driver. That claim is partially confirmed by noting both circuits use the same type TO-92 cased transistor, an S8050 (Q35 in Roomba, Q13 Scooba, just right of the J9 PUMP jack); current shunt-resistors placed between emitter and SIG_RTN, uses one ohm in Roomba (drawing about 5X as much current as the Scooba motor), vs. 4.7 ohm power resistor (a SMD on bottom of board) for Scooba; and, flyback diodes strapped around motors (Roomba D20, Scooba D1). Red-leads of both motors directly connect to raw battery voltage. The major difference will be in the signal train from the MCU. In the Roomba, the ON/OFF signal is simple bi-state, HI/LOW, while Scooba must use one of the MCU's PWM outputs in order to vary pump-motor speed.

That information may help you see where your voltmeter will be connected. An old, analog, meter may work better than a digital meter when measuring a pulse-train--as I expect Scooba's motor-voltage to be.

Thanks to a raft of Scooba sub-assemblies donated to me by glo69, I have a loose Pump Asm to play with.

I loaded my operational Scooba Tank with water, plugged into it a loose Solution-Inlet Lance (which was tube-connected to the Pump's inlet port), and dc-powered the motor via a bench supply. I used two voltage levels, 15Vdc and 12Vdc, to emulate possible speeds which the MCU might be using. When powered with 15Vdc, motor-current measured 50mA. With 12Vdc applied, current fell very little to ~ 47mA.

You can see that not much heating occurs with 0.05A * 15V = 0.75W.

While running with 15Vdc applied, I estimated the cam's rotation rate to be ~240 RPM.

I had also run the motor-gearhead sub-asm after dismounting the Pump sub-asm from it. No-load current under that mechanical condition was only 40mA (15Vdc applied). As the motor ran, I gripped the cam's periphery with three finger tips and applied force in an attempt to stop it. It did not stop, but the motor slowed and current was driven up to 0.25A to 0.3A in the process.

Since the Pump sub-asm was free of the drive mechanism, I decided to measure the volume of each jet's squirt. I also demonstrated the pump is capable of establishing a siphon (pump held lower than water-level in a jar; Lance-end submerged, with tubing leading up and over the jar's rim; the rocker-arm was manually pushed back and forth to expel air from the pump and draw water to it). I hand-pumped 300 squirts, 150 rocker-cycles, of water into a dish, them poured the contents into a 100 mL graduate, and read a volume of ~97.5 mL. Thus, each jet-squirt emits approx 0.32 mL, (0.065 tsp) per squirt. If anything, I think this volume is toward the high-side, because I was pushing the rocker-arm fully to limits of travel; however, we know that the cam does not push quite that far.

Adding to these manual-actuations trials, I asked myself what level of force would be needed to move the rocker-arm if a jet got stopped up with some dirt. To check that, I simply closed off an exit port with a finger-tip, and used the other hand to move the rocker-arm in the direction that should pump water out that port. The rocker was very hard to move! Pump valves sealed well enough that one is basically trying to compress a liquid--always a hard thing to do.

If Scooba had quit ejecting solution because a jet was plugged in that manner, and the owner applied the iRobot recipe of shock-treatment, those shocks would only serve to move the plug back toward the bellows, rather than out of Scooba. The plug would then be able to re-seal the exit port. Of course, the mass of such a tiny plug would also be tiny, and would require very great shock forces to kick it out of a jet's orifice (not outside, but back inside). This is not a likely plugging scenario, yet if it occurred, the rocker's resistance to rotation could be great enough to cause the motor to draw too much current, and that would force the MCU to shut off power to the motor.

But, even that over-current scenario is flaky, since I think its fair to assume Scooba's safety shut-downs to be similar to Roomba's. When Roomba detects an over-load, and acts on it, all motors are shut off. We know that is not happening with Scooba, because it is reported at top of this thread: Scooba continues on the mission!

What?!! After reading all that, there is no answer? No bottom line?

Nope, just Scooba-info and ideas to help support additional thinking and trouble-shooting.

[Jeff Carver]

Whoa. This is a great forum.

That's all very interesting information. I'm no electrical engineer, so some of it went by me, but I do get the sense that you've provisionally ruled out more than you've ruled in, to explain the pump motor stopping. (Or not restarting. I was interested to read that under some circumstances the pump is shut off momentarily.) I guess we still don't know why thumping the Scooba causes the pump to restart.

I'm thinking when I next have time to work on this, I would like to try the voltmeter attached to pump motor experiment. However...

Red-leads of both motors directly connect to raw battery voltage. The major difference will be in the signal train from the MCU. In the Roomba, the ON/OFF signal is simple bi-state, HI/LOW, while Scooba must use one of the MCU's PWM outputs in order to vary pump-motor speed.

That information may help you see where your voltmeter will be connected.

...um, no, not really. If one of you with electrical engineering knowhow can figure out and explain exactly where to attach voltmeter leads on the motor, I'd be willing to give it a try.

As a side note, one thing you said got me thinking about my other problem (my washing machine--bad pump or bad controller). Here's a basic question you probably know the answer to, but I don't: if you interfere with the operation of a motor (say, by blocking the action of a pump or otherwise increasing the load) does that cause the motor to draw more current? Or to put it another way, could a sock in a pump cause the pump to overload its circuit?

[vic7767]

could a sock in a pump cause the pump to overload its circuit?

Yes.

[Gordon]

... I would like to try the voltmeter attached to pump motor experiment. However... If one of you ... can ... explain exactly where to attach voltmeter leads on the motor, I'd be willing to give it a try. ...

You may have noticed there are four wires in the Pump-Motor cable. All four terminate on a row of solder-pads on the motor's EMI_PWB. BLK attaches to terminal #1, and RED to #2. Those two terminals constitute option-1 for your test-cable's connections. The terminals are small and close together, requiring caution to not end up bridging the two. You would be soldering to the up-turned ends of the cable-wires.

Don't be concerned with the pair of yellow wires. They were intended to connect to a reed-switch fitted to the oval slot in the PCB, and to become a motor-rotation indicator. But, it was never implemented.

Option-2, for test-cable connections, is to solder the wires directly to the motor's terminals. A 35 to 45 watt iron will be needed to heat those large terminals. A trace from the RED-wire's pad arches across the upper portion of the PCB and reaches a pad that has a large '+' sign next to it. Solder a wire to that pad, and label it for connection to your voltmeter's positive input.

The motor's minus-terminal is located close to terminal #1, and to a device labeled FB2 (which would conventionally be either an inductor, or a low value of resistance; the "FB" label is unfamiliar to me). Solder your meter's negative-input wire to the pad through which the motor's terminal passes. Try to avoid harming FB2 or its solder junction(s).

You will want to use fairly small wire-size, say AWG-#24 largest, or AWG-#26. Also, remove the conformal-coating material from the terminal you intend to solder to, and for a few mm extent around the terminal, before heating the terminal. Later, after the wires are not required, you can de-solder them and re-form the moisture seal over those terminals by applying a thin layer of RTV-Silastic (Dow Corning) Aquarium adhesive.

[Jeff Carver]

Thanks for the detailed instructions.

remove the conformal-coating material from the terminal you intend to solder to,

File? Sandpaper? Solvent?

Everything else I get. I'm only moderately skilled with a soldering iron, and have no idea what power mine is--all I know is, it's small and was cheap. :)

It'll be a week or so before I'll have time to try it, but I'll report back when I can.

Thanks.

And Vic, thanks for the answer on the motor.

[Gordon]

...

remove the conformal-coating material from the terminal you intend to solder to,

File? Sandpaper? Solvent?

Slip the keen edge of a thin knife-blade under the conformal-coating, glide the blade along the surface of the PCB a few mm, then lift the blade to pull out a chunk of coating. It is very tender stuff, and does not adhere well to the parts. Repeat as required to create a coating-free crater large enough for the tip of the iron to work in. Any one of the various X-acto blade-types, or equivalent, should work OK.

... I'm only moderately skilled with a soldering iron, and have no idea what power mine is--

You will need some rosin-core solder-wire (0.032" dia), and rosin-flux. Optical magnification may also be of help.

...It'll be a week or so before I'll have time to try it, but I'll report back when I can. ...

We'll be waiting, tic, tic, ...

[Jeff Carver]

Well, I've soldered leads to the motor and run them out to an old-fashioned analog multimeter that's presently riding on its back as it scoobas around the bathroom floor. At the moment, everything is running perfectly. It's registering somewhere around 12 volts (hard to read exactly) as it moves forward; it drops to near zero when it gets into a corner and starts maneuvering around. Once it's out of the corner, or moves even an inch or so straight ahead, the voltage goes back up. So far, it's wet behind the robot.

I'll be back with more, as the cycle proceeds.

----------

Curse it! After 25 minutes, it's still working perfectly! What's the world coming to? Won't anything malfunction when you want it to?

----------

Auugggghhhhh! Forty minutes in, it switched to its drying cycle--blinking Clean light--and is now drying the floor. (Volt-meter dropped to zero on the pump.) For the first time ever, it's gotten through the whole cycle without a malfunction. Oops--there's its happy song. It's finished.

My wife says, maybe it just needs to have a voltmeter riding on its back to work right.

[Gordon]

...My wife says, maybe it just needs to have a voltmeter riding on its back to work right.

She is possibly right! What else could it be?

Actually, if you can't obtain a fault after about ten cleaning missions, you might consider replacing the meter with an approximate equivalent resistance!

I just looked at my wife's Micronta knock-off of the old Simpson multi-meter to see what its effective input resistance is. 100,000 ohms per volt is printed on the meter's face. The Micronta would require you to use its 50Vdc F.S. setting, hence input resistance should be 50V * 0.1M ohm/V = 5M ohms. You culd probably get away with subbing in anything between that, and down to 100k to emulate the meter's presence.

[vic7767]

Now to add a water pressure gauge, a water temperature gauge a tachometer gauge and an amp-meter gauge and possibly a Hurst 4 speed shifter. I see another mod coming.