Archive for August, 2009
So our whoop-dee-doo Sears Kenmore HE3 clothes washer started emitting horrible scratching & grinding noises, but only every now and then during the high-speed spin cycle or, more rarely, the washing agitation cycle.
In an ordinary washer, you’d suspect the transmission was going bad, but the HE3 has a direct-drive 3-phase motor: no transmission to wear out.
A bit of searching shows HE3 washer drums may fall apart at their welded (?) seams, but that didn’t seem to be the case here. So I press-ganged our daughter into a plumbing job; the machine is something over six years old, so it’s not as if we have any warranty to void.
Move the washer out where you can get to the back without contortions. Pull the plug, turn off the water, unscrew the water supply hoses at the back panel, squash the hose clamp & remove the drain hose.
One obvious, but unrelated, problem appeared when we disconnected the water hoses: plenty of black grit in the hot water inlet. Looks like it’s time to drain & flush the hot water heater again, which will turn into a major project because the anode rod has firmly rust-welded itself in place. That’s a project for another day…
Disassembling the cabinet requires a Torx T-20 bit. There are a lot of screws, so fetch a stable container to hold them all.
Take off the top cover (three screws, then slides back) to reveal the angle brace across the back that holds all the electronics and valves and suchlike. Pull the big vapor vent tube (on the right as you face the rear) out of that bracket, remove the eight (!) screws holding the bracket in place, and move it up out of the way.
Remove the back cover, taking care to not loosen the screws holding the frame crossmembers in place; those are the screws in the U-shaped cutouts along the edges.
Note: you can remove just the back cover by removing the lower screws in the angle brace, all the screws holding the cover, then sliding it down and out at the bottom. It’s easer to see what’s going on if you take the top cover off and moving the angle brace frees up a lot of gimcrackery that gets in your way. Your choice.
Everything connects to the drum housing through exceedingly flexible rubber boots held on by circumferential wire clamps. Grab the ends with Vise-Grip pliers, squash ’em together, and the clamp should slide off the housing onto the boot. Peel the boot off and you can look inside.
Peering in through the pressure sensor opening in the bottom rear of the housing revealed something odd: a loose black cylinder. A bit of deft tweezer and grabber work pulled out a much-the-worse-for-wear ballpoint pen housing, minus the cap, point, and pocket clip. The fiber-fill ink reservoir formed a tuft at one end.
Mmmm, that would account for the blue water in the drain tube…
Combine the number of missing parts with an inability to see any of them in the bottom of the housing: more surgery is indicated.
The housing drains into the ejector pump through that huge black boot clamped onto the bottom of the drum housing. Put a pan underneath the pressure sensor hole, squeeze the boot, and water will bloosh out the sensor hole, generally landing in the pan. There’s a floating-ball check valve inside the boot that prevents backflow to the housing, so you may need some wiggly-jiggly action to work the water around the valve. A few reps will get most of the water out of the drain boot.
Remove the boot from the housing and pump inlet, remove the ball, and admire the innards. We found most of the rest of the pen in the drain boot, plus a generous helping of slime and gunk. Oh, and a hairband and a big chunk of a pencil.
The pump has a juice-can-size reservoir just inside its inlet, which I think is there to collect debris: a barrier keeps most of the big chunks out of the pump impeller. There’s no way to get stuff out other than lying flat on your stomach and sticking your finger into that slimy hole, so get over it. We extracted the remaining pen bits and most of the rest of the pencil.
While you’re in there, roll over onto your back, reach up inside the drum housing and feel around to get anything else out. Your assistant can shine a flashlight down through the drum perforations; you’ll be able to see the shadows cast by any odds & ends that are lying on the housing.
Wipe the slime off the rubber boots, reassemble in reverse order, and you’re done!
However: the bottom of the washer consists of a big metal pan that will, most likely, start to rattle just after you push the washer back in position. Removal isn’t really an option because one of the front screws is under the motor drive box, but you can sort of pry up the edges and stuff thin cardboard, strips of duct tape, or other elastic stuff between the pan and the washer frame. It took me far too many iterations to figure out what was rattling around in there: it is not obvious!
Here’s what we found in the drum housing, drain boot, and pump settling tank: the corpse of one of my favorite Uni-Ball Micro pens, a tiny screw, bits of a pencil that my assistant had been looking for, and one of her hairbands.
That pretty much explains the intermittent grinding sounds: the drum would rotate normally until the swirling water swept the pen housing into contact with the drum, at which point the 800-some-odd RPM rotation would grind the pen against the housing for a few laps. Ditto for the pencil.
Now, the mystery is how that stuff got from inside the drum past the rubber sealing gasket into the space between the drum and the housing. There doesn’t seem to be any way to get a long rigid object through there, but obviously it happens!
After you move the washer back into position, take ten minutes and a generous handful of rags to wipe the abundant collection of mold & mildew from behind the rubber gasket. You can sort of evert the gasket, which simplifies access to the edge of the drum. As you can see, ours has a nice biofilm going on in there; not visible is the gunk growing on the back side of the gasket.
It seems HE3 washers have a reputation for smelling bad, due to that sort of growth in hidden places. Oddly, we don’t have an odor problem, obviously not through any action on our part. Rumor has it that running a pure-bleach hot-water cycle helps, as does a product for dishwashers that removes their stink, but we don’t have any experience with those.
As a mental math exercise, I had my assistant divide the $300 bucks a plumber would have billed for this repair by the $7/hr she’d get for a typical minimum-wage shit job. She’s thinking that becoming a plumber might beat smiling at retail customers in a dying mall… but I think she should concentrate more on her math & science.
[Rant: Not that America seems to value tech jobs much these days, but she has the advantage of being female, so maybe she can still get a tech gig. Don’t get me started, you know how I get.]
Anyhow, the washer runs just as quietly as it ever did, which is to say that like a turbojet engine spooling up during the spin cycle. At least it doesn’t sound like it’s ingesting a bird every now & again…
[Update: And then the spider holding the drum in place failed, as it does with so many of these washers.]
Got a Digital Concepts CH-3988S charger with quartets of AA & AAA cells from buy.com (which no longer sells it, no surprise, but it’s still available elsewhere) on closeout for about 12 bucks delivered, down from the “regular” price of something like $40; anybody who paid that much got well and truly hosed.
I fully expected the cells to be crap and they were: they don’t even bear a manufacturer’s name. Tellingly, they weigh 25 grams each, lighter than the 28-30 grams of more cough reputable brands.
The upper trace (click the graphs for readable pix) is the four AA NiMH after the charger said it was happy with them. The trace drops off the cliff at about 25 mAh. Call it 1% of nominal capacity.
The four lower traces are the individual cells after another trip through the charger. The far-right end of those bottom curves is 70 mAh, with the cell voltage barely over 1 V for the entire discharge.
Fairly obviously, they’re not accepting a charge.
Charging the cells in a known-good 400 mA charger (roughly C/6) brought the best cell up to 160 mAh, with the rest around 100 mAh; the charger was happy with them after far less than 6 hours, so apparently the cells display a much higher terminal voltage than they should.
So I plunked them in a dumb 250 mA slow charger and let ’em cook for the full 8 hours. That should, in principle, give them roughly 2 Ah of charge, no matter what the terminal voltage may be; I measured 1.8 V, which is far too high for that rate.
So, here’s the result…
Crap. Pure, utter, unadulterated crap. The cells supplied 500 mAh, much more than before, but that’s so far below their rating it’s not even funny. There’s obviously one cell in there that’s bad, but the others can’t possibly be far behind.
I didn’t waste any time on the quartet of AAA cells, but I expect they’re pretty much the same.
It’s faintly possible that exercising these turkeys will bring them up to maybe 50% of capacity, but it’s not like that’d make me ecstatic. The reviews you’ll find here and there support the conclusion that something is wrong with these cells.
Here’s the result of the next cycle, after a night in their very own charger. The upper trace is all four of them together, once again failing after 500 mAh.
The four lower traces labeled “Cell x” are the individual cells, tested without recharging. Three of the four have about 700 mAh left in them, which would bring their total capacity to 1200 mAh, roughly half of their nominal capacity.
Cell B, the green trace, is obviously the weak link, as it failed almost instantly. Recharging it on a known-good charger got it back up to 530 mAh (the “Cell B recharge” curve), roughly 25% of its nominal capacity. So much for the idea it’ll get better if you treat it right.
Now, turning to the charger…
The Digital Concepts CH-3988S charger is advertised on its package as a “2 Hour Charger”, but its manual / datasheet indicates that claim is, mmmm, not strictly correct:
Remember that the nominal AA cell capacity is 2.3 Ah, so charging the four AA cells included with it requires three or four hours. Well, OK, only 2.5 hours if you do ’em pairwise, but that’s five hours total.
On the other paw, the charger does (seem to) monitor the cell voltage and cut off automagically, on either negative delta-V or maybe just peak voltage. Unleashing it on a pair of partially discharged Tenergy RTU 2.3 Ah cells indicates that it cooks the piss right out of them, there toward the end.
The charger is (probably) OK for low-rate charging of known-good cells, which is what I got it for; the cells accompanying it are crap. It’s not worth returning for twelve bucks, seeing as how the shipping would eat half of that.
So, anyway, if you ever wondered what a bottom-dollar charger-with-cells offer gets you, now you know.
Some key distances:
- Bezel bottom 3.3 mm thick, excluding depression on bottom surface
- Screw head sticks out of depression 0.9 mm
Some deft work on the bezel installed in the camera, using the blunt end of a transfer punch, a pin vise, and a calculator reveals these protrusions:
- 1.4 mm does not trigger anything
- 2.1 mm triggers the half-pushed focus action
- 2.4 mm reliably triggers the shutter
So the new stem can stick out about 1.4 mm when the button is released and must not stick out more than 2.4 mm with the button fully depressed: a whopping 1 mm of travel!
Eyeballing the shutter release on my DSC-H5, that seems to be about right. I think it has more travel between “released” and “half pressed” than those measurements indicate, but it’s close. And sloppy, too: the H5’s button has a lot of side-to-side wobble, indicating that the stem is not a close fit in the bezel hole.
The screw head is 3 mm dia after being turned down and that’s about the right size for the nut that will adjust the travel distance, as it must fit into the recess in the bezel. The nut sets the protrusion when the shutter button is released: 1.4 mm.
The distance from the shutter button’s bottom to the bezel sets the travel from “released” to “click”: 1 mm, more or less. They’re held apart by the spring, so that’s the default state.
I re-centered the 3-jaw chuck under the spindle, put a 1-72 nut on the turned-down screw, and applied some gentle manual CNC to convert the nut from a hex to a disk. The trick is to approach the nut from the right side (the +X side) and go clockwise around it (climb milling), so that the cutting force tends to jam the nut against the screw head. Do it the other way and the nut will zip downward away from the cutter
Surprisingly, I got that right the first time.
Using a 2 mm end mill and figuring a 2.9 mm final diameter, the radius of the circle to move the end mill around the nut is:
R = (2.9 + 2.0) / 2
So the G-code for one pass looks like:
#<R>=[[2.9+2.0]/2] G1 X#<R> F150 G2 I[0-#<R>]
Now, given the fragility of that setup, you don’t cut it all at once. You start from a diameter of maybe 4 mm and go down by 0.2 mm until you hit 3.0, then make a final pass at 2.9 mm. EMC2’s AXIS MDI mode makes this easy enough: type in the commands for a pass at 4.0 mm, then click on the previous command, change 4.0 to 3.8, and then just clickety-click.
Spindle far too slow at 3000 RPM, feed at 150 mm/min seemed fine. Sissy cuts worked out OK.
After the first few passes, my dim consciousness became aware of the fact that this is how I should have turned down the screw head…
I cleaned up the bezel by putting it in an ultrasonic cleaner to shake the crud off, put it on a warm firewall router overnight to dry it out, then slobbered some Plastruct solvent adhesive into the cracks and clamped it for another night. The bezel was slightly out-of-round from the damage, so I hand-trimmed the bent plastic using a “high speed cutter” (#193, basically an end mill) in a Dremel flexible shaft at about 1/3 max speed until the shutter button bottomed out smoothly within the inner recess. Not a bit of CNC to be seen: hand held all the way.
Then loosen the nut a bit, poke the screw through the bezel, put the spring on, and screw the shutter button in place. Adjust the nut so the screw head is 1.4 – 1.5 mm from the bottom of the bezel with the nut resting in the recess.
Twiddle the shutter button until the screw head protrudes 2.4 mm from the bezel with the button pressed down.
That’s measured with the hole-depth tang of a caliper, sitting atop the screw head. I don’t believe there’s 0.1 mm accuracy in the measurements, but they’re close enough. I did file off a few mold flash bumps from the shutter button & bezel during this adventure.
Mark the screw threads above the button, unscrew it, chop the screw off with a stout diagonal cutter (it’s brass and not very thick, it’s OK), file the end flat, clean up the threads.
The trick seems to be that the button must rest just below the inner ring of the bezel, so that it bottoms out smoothly when pressed. If it’s above the ring, then one side will hang up. The ring depth thus seems to limit the maximum travel, although I can’t say whether this is the way it’s supposed to work or not.
I iterated & filed until the screw was flush with the top of the button with it screwed down to the proper position. It helped to figure out that one turn of the shutter button on the screw changed the “pressed” protrusion by 1/72″ = 0.35 mm.
Urge some low-strength Loctite under the nut and into the shutter button’s hole, reassemble everything, and you’re done.
The fall bent the bezel tabs so they no longer latch firmly in the camera body. I put two dabs of urethane adhesive on the socket in the body. The adhesive expands (foams!) as it cures; I hope it will lock the bezel in place while still allowing it to be removed if needed.
I dabbed off most of the adhesive you see in the picture before installing the bezel; it’s not as awful as it looks!
The final result has slightly less travel than the (undamaged, original) shutter button in my DSC-H5, but it works perfectly: half-press to focus, full press to trigger the shutter.
Having figured out what to do, I started with the button, which is chromed plastic, nothing too fancy, and not at all hard to machine.
A small post turned from an acrylic rod (the gray cylinder) supports the button in the Sherline 3-jaw chuck attached to the mill table; that was the only way to keep it reasonably level. Laser alignment got eyeballometrically close to the middle; it looks a bit off to the right, but the end result was OK.
A 2 mm end-cutting bit chewed off the stem in short order; I set the jog speed to about 100 mm/min and just jogged down until the cutter was flush with the button. Spindle at 4000 rpm, for lack of anything smarter.
I decided to go with a 1-72 brass machine screw, which is slightly larger (1.75 mm) than the original 1.5 mm button stem. That means I must drill out the bezel hole, as well, but the 1.5 mm diameter of the next-smaller 0-80 screws in my assortment was a sloppy fit.
A touch of manual CNC for the drilling, #53 with the spindle at 3000 rpm, Z touched off at the button’s surface:
G81 Z-4 R3 F150
The spindle was slow enough and the feed fast enough to keep from melting the button without applying any coolant.
I tapped the hole 1-72 by simply screwing the tap in with my fingers…
The 3-jaw lathe chuck doesn’t grip a 1-72 screw (no surprise there), so I grabbed the screw in the Sherline’s smallest drill chuck and poked that in the lathe. This doesn’t make for great concentricity, but it was close enough. The right way, as my buddy Eks reminds me, is to slit a nested bunch of brass tubing and use them as collets, but … next time, fer shure.
Anyhow, here’s what the button & screw look like so far. The backside of the screw head looks like it needs some cleanup; there’s nothing like taking a picture to reveal that sort of thing.
The pencil lead is 0.5 mm and the grid in the background has 1 mm squares, just to give you an idea of the scale.
My brother-in-law Tee dropped his Sony DSC-H1 camera, which landed atop its shutter button on the pavement.
- the shutter button broke off
- the bezel popped out
- the teeny little snap ring that held the shutter button stem in the bezel vanished, because…
- the stem broke and the end vanished, too
- apart from some scuffs, the camera still works
- he managed to find the shutter button
- and the button bezel
- and the spring!
A bit of browsing reveals that many, many Sony DSC-Hx (where x is an integer from 1 through 9, inclusive) owners have the same problem, minus the inconvenience & embarrassment of first dropping the camera. Turns out that the shutter button stem breaks at that notch in normal use.
It seems the stem snaps while you’re taking pix, whereupon the spring launches itself and the button cap into the nearest river / drain grate / weedy area, never to be seen again. Tee is exceedingly fortunate to have found all the major pieces!
Here’s the broken end of the stem, with the button cap out of focus in the background. The stem is 1.5 mm in diameter, so the snap ring was surrounding, what, 0.75 mm of plastic? In what alternate universe did this design decision make sense?
I think the snap ring contributed to the problem by eroding the stem in the notch; that little white stub isn’t half of the stem diameter; it may have stretched under impact, but surely not all that much.
Yes, you can buy a replacement button for about 30 bucks direct from Sony, but it seems the new stem is subject to the same failure after a short while. They’re standing by the original design, marginal though it may be.
Now, obviously, this stem failed from abuse, no argument there. Everybody else had their stem fail without provocation, though, so it really isn’t adequate to the task at hand.
Anyhow, there’s also some damage at the bezel socket on the camera body, but nothing major. The dented silver areas on either side of the switch membrane are ESD shields, so that any static discharge from your finger will (most likely) dissipate on the external frame of the camera, rather than burrow into its guts via the switch.
The bezel twist-locks into the camera body, which means that you can remove the bezel if you can get a good grip on it. It turns clockwise to remove.
Peering closer at the membrane switch, it looks as though the button stem did some damage on its way out, although Tee admits to using various pointy objects to trigger the shutter while figuring out what to do with the camera.
More good news: the switch still works correctly, including the focus function with the button half-pressed, That means the switch membrane and contacts are in good shape.
The bezel itself is pretty well graunched, with a nest of cracks underneath that damaged arc to the left of the pictures. I think it’s in good enough condition that I can remove the bent plastic, ooze some solvent adhesive into the damage, and compress it enough to make everything stick together.
Obviously, this calls for some Quality Shop Time!
The overall plan is to remove the remaining stem from the button, drill-and-tap the button head for a miniature brass screw (1-72, I think), reshape the screw head into a membrane-friendly plunger (about 3 mm diameter and flat), then put it all back together with a nut in place of the snap ring.
I should be able to install the bezel (without the button), then insert some drill rod through the hole to figure out how far the screw must protrude to trigger the focus & shutter switches. Perhaps a pin vise will grip the drill rod and bottom out on the bezel’s central ring, so I can do a trial-and-error fitting?
Then I can adjust the screw to that overall length below the bezel with the button pressed, whack off anything that sticks out above the button, adjust the nut to limit the button’s outward travel, slobber Loctite over everything, and put it all together for the last time.
That’s the plan, anyway. As the Yiddish proverb has it, “If you wish to hear G*d laugh, tell him your plans.”
Some useful dimensions…
The rest of the story…
There I was, bandsawing a 24″-diameter circle from 3/8″ plywood when the saw stopped cutting… which is how the bandsaw tells me that the blade joint just came undone. That’s what I get for applying too much torque to the blade, I thought…
As you can see, the joint I described at great length there wasn’t nearly as good as I thought. It has a nice wetted spot in the middle, but the rest of the joint didn’t bond. Alas, the ends of the blade sections had just enough molten solder to look as though they were bonded.
I’ve done other joints with that resistance soldering unit which held up well, but I’ll cheerfully admit I don’t have years of deep experience with it. Given the effort involved in making a bandsaw blade joint (not to mention the fact that I only solder up a new blade when I really need one), this is the first destructive test I’ve seen. It’s not like I’m going to solder up a blade, then tear it apart just to see how it worked, as I did with the nickel strip on those AA cells there.
A change of technique seems in order. The carbon electrode produces enough heat directly under it (which isn’t surprising, that’s how resistance soldering works), so I must cover the full joint area with enough oomph to get a good bond everywhere. That seems to be 1.5 seconds per zap, more or less, repeated as needed.
I have some machinable graphite that I could cut to be one square bandsaw joint at the end, plus a stub from a searchlight electrode that might work with less overall effort. However, a larger electrode probably wouldn’t heat the joint enough to melt the solder everywhere at once.
For what it’s worth, I tried to re-solder the joint without doing full surface preparation by grinding down to bare steel. Didn’t work, natch, which should come as no surprise; wasted half an hour futzing around before I gave up and did it right.
Memo to Self: Use the carbon gouging rod electrode, hit the entire joint, and don’t move until each zap cools!
Much as I expected, there’s just not enough energy in a 5 V / 200 A resistance soldering unit to weld 8-mil nickel strips to AA cells. But the gadgetry needed to contact the cells works fine for resistance soldering.
I took a pair of sacrificial cells, grabbed the positive terminal of the blue one in the pliers, applied a flattened snippet of good old rosin-core tin-lead solder (see below), laid the strip atop it, and held things together with pressure from the tungsten electrode.
About 800 ms of current did the trick; the electrode heated to middling orange by the time the current shut off, which indicates it was the highest-resistance part of the circuit. Eyeballing an ammeter clamped around a secondary lead says the peak current was 250 A, a bit over the nominal 200 A, but close enough.
The obvious dent in the strip over the positive terminal shows that the center of the solder strip melted first; I could feel the tungsten electrode sinking into the strip as it heated.
For the negative terminal, I grabbed both cells in a small vise (resting on insulation below the bottom terminals!), tucked another solder strip under the nickel tab, pressed one jaw of the pliers against the cell, and hit the tab with the tungsten electrode. Lovely fillet, isn’t it?
The joints look good inside, too. I cut the strip, then peeled the joints apart: they’re both fully wetted. You can see some tiny bubbles from the rosin, but I doubt that’s a problem.
Now, you don’t want to solder to AA cells by hand with a soldering iron, because it’s entirely too easy to cook the piss out of the plastic insulators, pressure relief valves, and other internal gadgetry. Yeah, I’ve done that too, and it works most of the time, but it’s not recommended.
A controlled pulse is all over and done with before the rest of the metal case has time to get more than warm. In fact, by the time I put down the electrodes, the nickel strip was cool enough to touch! The copper jaws act as a heat sink for the positive button and the negative terminal is the entire can around the cell, so I think this will be OK.
I’ll do a bit of testing on some sacrificial cells to figure out the minimum time required for a good joint; I think 600 ms will do. I might use a carbon electrode for the positive terminal to get a somewhat larger contact over the whole button and eliminate that unsightly dent.
Solder prep: I flattened about 3 mm of ordinary solder wire by whacking it with a polished brass hammer on a chunk of PCB stock. Flat solder works better than round solder for resistance soldering, as everything stacks up neatly with lots of contact area. The pix there should give you the general idea.
I’m mildly unhappy with the pliers, which must open a bit too far for my paws. A fixture that fits in the bench vise might be in order…