another bumper sensor problem

wsngmartin · Posted: June 30, 2008, 7:36 pm Post subject: another bumper sensor problem

One day (literally) after the warranty expired on Strawberry Sundae (a disco/4xxx) she started doing the "backing up" dance. After blowing out the sensors, Resident Roombatech took her apart, cleaned her out, to no avail. He then ran the diagnostics, and found that in Mode 1 the "spot" and "clean" lights stayed on continuously and did not switch off when the bumper was moved. He's thinking the sensors are stuck in the "on" position. He took the bumper off and cleaned the sensors directly w/ compressed air and contact cleaner, but to no avail.

This bot is a replacement for another bot, so we've only had her for *61* days and she is run every 2-3 days, not a whole lot of use. Every other mode of the diagnostics seem to be OK.

Gordon · Posted: June 30, 2008, 9:22 pm Post subject: Re: another bumper sensor problem

wsngmartin wrote: ...a disco/4xxx... started doing the "backing up" dance. ... Resident Roombatech took her apart, cleaned her out, to no avail. He then ran the diagnostics, and found that in Mode 1 the "spot" and "clean" lights stayed on continuously and did not switch off when the bumper was moved. He's thinking the sensors are stuck in the "on" position. One way for them to be stuck, as described and where cleaning has not corrected the fault, is for both left & right LEDs to have weakened (for two LEDs to go dim like that -- due to changes in the semiconductor material -- would be an oddity). Another possibility is for one of the diodes to fail 'open-circuit'. When one of those LEDs fail that way, the partner LED also goes dark -- they are connected in series! But, you also claim... Quote: ...Every other mode of the diagnostics seem to be OK. ... hence, I take it that Test (9-b), in this procedure (which does not refer to each "test" as a "mode"), is giving a PASS indication.

If that is the result that your d/h is seeing, then the Bump-LED conducts. Neither one can be "open", since both Bump-LEDs must be conducting in order to conduct current through the drive-Wheels' rotation sensor-LEDs. All four of those LEDs are in series, and powered by a single control transistor.

A duff LED can conduct current, but not emit useful IR-energy. Viewing both Bump-LEDs with a digital camera may provide some info, however, it is also possible for both to be emitting enough light to be seen in the camera, when only one may be weak. However, the Wheel-LEDs would also be dim, and cause Test (9-b) to fail.

Be aware that the LED is on top on one of Roombas sides, and bottom on the other. Note: To help discrimnate between Bump-LEDs & Bump-PTs, the opposing PTs usually have a splash of opaquing-paint on their exposed end-face, while the LEDs do not.

wsngmartin · Posted: July 1, 2008, 9:43 pm Post subject:

Gordon

Thanks for the help. DH spent some time studying your diagnostics (he'd used the diagnostics on this site under the FAQ). He says he thought the bumper sensors were switches and did not realize they were LED lights so instead of just spraying w/ contact cleaner he may have to take them apart a bit more to clean more directly.

Incidentally, yours truly has to admit that I used carpet freshener (the baking soda blends) on the carpet prior to SS malfunctioning. Dog had had an accident awhile ago in there and it was getting ripe. Perhaps the fine powder attacked the LEDs? I got the eye-rolling on that one....

DH has a couple of ?? for you. First, how do you power up the optical sensors to check them, since the robot will be dismantled by that point?
Also, if it is a bad LED how would you replace them? I haven't seen them on Roomba Exchange.

TQ

Gordon · Posted: July 1, 2008, 11:06 pm Post subject:

wsngmartin wrote: ...DH spent some time studying your diagnostics (he'd used the diagnostics on this site under the FAQ). Both the document which I have been editing, and the one posted at the Swiftracer site can be found in the FAQs section. It is unfortunate that the latter document never got completed, thus it is best to not refer it for use. Quote: He ...did not realize they were LED lights so ... he may have to take them apart a bit more to clean more directly. Solvent cleaning, using squirted-in isopropyl alcohol (while holding the Bumperless robot with shutter-arms vertical, and inner-bumper down) is a good thing to try. Quickly follow with a blast of 60-80 psi air, applied right around the shutter-arms where they enter the black-boxes that serve as mounts for the LEDs & PTs.

I'm having second thoughts about the viability of seeing any IR-radiance with a camera, since only very oblique viewing will be possible! Of course, the shutter will be blocking the view of the LED as well. If DH wants to try the camera-viewing, Roomba's RH-LED should be the upper device (in that black-casting mount); and, at the LH-side, the LED should be the bottom device. However, there could be some diffuse illumination of the rectangular ends of the LEDs (I have never had occasion to test that possibility). Quote: Incidentally, yours truly has to admit that I used carpet freshener (the baking soda blends) on the carpet prior to SS malfunctioning. ... Tsk! Tsk! Those powdered products and Roombas can not coexist in the same house. One has got to go!

Actually, i would be surprised if the powder got into the Bump-switches. Usually, it is the wheel-tach PTs that get coated with dust. Becasue of that, you may have to clean wheel optics sooned then might be expected. Quote: ...DH has a couple of ?? for you. First, how do you power up the optical sensors to check them, since the robot will be dismantled by that point? I think they get powered ON when the Power button is depressed. Otherwise, it would be necessary to put the robot into self-test mode (just ignore results for cliffs & wall checks when the Bumper-shell's connector is de-mated). Quote:
Also, if it is a bad LED how would you replace them? I haven't seen them on Roomba Exchange. ... It is only efficient for Chris to sell you an entire main_PWB assembly; that is why you don't see the piece-parts listed at his site.

EVERLIGHT LEDs have been selected as form, fit & function replacements, and I have listed the P/Ns in my "Scooba Technical" pages. I'll dig out a pointer for you when the need arises. LED-replacement is easy, except all the work of R&R'ing the main_PWB has to also be done!

Gordon · Posted: July 2, 2008, 2:05 pm Post subject:

Wendy, I have evaluated the camera-viewing process further, and think there is little chance of it being productive. I have a 4220's guts laid out on my bench for some months now, so it was fairly easy to access the Bump-SW modules, and try some IR-viewing.

I verified: a) that the Bump-LEDs get powered when Roomba's Power SW is depressed, and b) LEDs remain powered when the Bumper-shell's 7-contact connector is de-mated.

The digital camera appears to be able to indicate a lit LED, but shows quite dimly. A few pix are attached to show the effect. I found that "viewing" the LED's radiance is better done then "photographing it" (for various reasons).

I had no luck in simultaneous viewing. If one wants to verify that one LED is not stronger than the other, then both Bump-SW modules must be viewed together in the same frame! Otherwise, the camera's auto-exposure adjustments will make individual shots appear somewhat the same.

I also investigated doing some electrical checks (voltage measurements across the LEDs) to confirm they both have about the same delta-V drop (dV ~= 1.2V); but, found only the RHS-LED test-points are accessible while the main_PWB is still mounted in its chassis-slot!

Unless some new thinking comes forth, it will be necessary to dismount the main_PWB from the chassis to observe the LEDs, or to make voltage measurements! All connections to the PWB will be left in place, as the PWB comes to rest with its lower edge supported by the chassis.

Once the PWB is clear of the chassis, we would ignore doing any camera-viewing, and move directly to checking signals' swings coming form the two PTs. These data will give a combination measurement of stimuli-levels (LEDs' 'brightness') and PTs' responsivity (detection sensitivity), which is more thorough than just checking the LEDs.

These measurements will require use of three or four hands, so DH will man the voltmeter probes while you simulate the shutter-blade by moving a piece of card-stock (better yet, make a little wand by folding aluminum-foil into a strip about 1/3rd to 1/2 inch wide and a couple inches long) in and out of the slot where the shutter-arm normally fits between the LED & PT.

Each PT's output (collector-signal) ought to swing from a low-level of near zero-volts to some voltage aproaching five volts. A weak LED will not be able to drive the PT hard enough to swing its collector close to zero. I have made some measurements which suggest the threshold is on the order of 2.5Vdc.

When Disco's Bumper is not pushed in, the shutters do not block the path between LED & PT, therefore, the PT's collector-voltage is lowest -- around 0.15Vdc. A weak LED acts the same as a blob of fluff caught in the shutter's aperture, with the result being an increase in collector voltage. The open-path voltage must not be greater than the threshold-value estimated above.

Refer to SHT-2 of Schematic_4 for applicable test-point numbers (I can give DH the coordinates of those pads, as necessary). Note: He might want to get acquainted with my notation scheme by scanning parts of this page.