The Smell of Molten Projects in the Morning

Ed Nisley's Blog: Shop notes, electronics, firmware, machinery, 3D printing, laser cuttery, and curiosities. Contents: 100% human thinking, 0% AI slop.

Author: Ed

Sears Kenmore Electric Dryer: New Rear Seal

Our ancient Sears Kenmore electric clothes dryer (which is not matched to the never-sufficiently-to-be-damned HE3 washer) started squeaking again. The last time it did that, I tore it apart and determined that the rear seal between the drum and the back panel needed replacing; I ordered the seal, buttoned up the dryer, and, amazingly, the squeak Went Away.

The box with the new seal arrived a few days later and has been perched atop the dryer for the last few months. Never borrow trouble, sez I.

Unlike the HE3 washer, tearing the dryer down isn’t a big deal. Two screws secure the lint trap enclosure to the top panel; be careful about not dropping them down the chute.

Screws holding lint trap to top

Then push the top forward and pry it off the clips holding it in place. You do not need to remove what looks like clips holding the top to the back panel; they’re sort of hinges that let you tilt the top back. With any luck, you can let the top hang; I rested it on the nearby laundry sink.

Door switch

Two screws hold the entire front door panel in place. Before you remove those, disconnect wires from the door latch switch so you can remove the front panel. The alert reader will note I didn’t do that…

The drum has two sliding seals that bear on the front and rear panels. There is nothing else holding the drum in place, so when you remove the front panel, the drum falls out. It’s helpful to have an assistant holding the drum in place, perhaps with a hand through the open door, while you jockey the front panel out of the way.

Drum belt path through tensioner

Have your assistant continue to hold the drum while you memorize the path of the drive belt around the tensioner and motor pulleys. This is not obvious: you don’t have to take the tensioner pulley off the shaft to remove or install the belt.

There are two sets of slots in the dryer base plate that could hold the tensioner. Only one set will work. Pay attention to the situation in your dryer.

Hint: the drum rotates counterclockwise as you view the front of the dryer. The motor pulls the belt off the drum and the tensioner acts on the slack side of the belt. If you try rotating the drum clockwise, the tensioner and motor make graunching noises that will convince you something has gone terribly wrong. It hasn’t, you’re just turning the drum the wrong way.

With the drum out, this is what the old seal looked like:

Worn seal

I cut the threads at the seam holding the ends of the old seal together and peeled it off the drum. That reveals the dried adhesive all around the drum.

Removing old seal

I applied xylene to soften the adhesive, then used a razor knife and a vast quantity of rags to remove the goo. The key is to get enough xylene on the adhesive to get its attention without slobbering solvent all over the drum; it will soften the paint, which is a Bad Thing. Do this in the garage or outdoors to enhance Family Harmony.

I did a trial fit of the new seal, which showed it’s a snug fit and requires careful alignment. A dozen small clamps held successive parts in place while I got it settled. The trick is to position the center part of the T-shaped seal against the rim of the drum without wrinkles. You probably can’t get it right without a dozen clamps.

To apply the adhesive, I removed two clamps, eased that section of the seal off the rim, and ran two beads of adhesive: one along the rim where the previous adhesive had been and a smaller bead just below the folded metal edge. That pretty well smeared out as I eased the section back in place.

Then remove the next clamp, ease that section off the rim, apply adhesive, and iterate all the way around.

Clamping new seal to drum

I dug a patched bicycle tube out of the drawer and eased it under the clamps around the drum, then pulled it mildly taut all the way around to apply uniform pressure to the seal. Two larger clamps held the slack ends in place.

After supper, we declared the adhesive (which looks & smells a lot like plain old contact cement) to be cured. Off came the clamps and tube and, lo and behold, it’s all good.

Reassembly is in the obvious reverse order. The instructions packed with the seal remind you to ease the loose end of the seal outside the drum where it can ride on the back panel. Make it so.

While your assistant holds the drum in place, reinstall the tensioner and route the belt around it. The belt in our dryer has two possible positions on the pulley (it has ridges), so I made sure it was tracking in the same position as before.

Attach the front panel, rotate the drum a few times to be sure everything is in place and tracking correctly, then slam the top, screw it down, and you’re done!

2010-09-06
Arduino Mega 1280: PWM-to-Timer Assignment

The Arduino Mega has five hardware Timers, each with up to three PWM outputs. Most of the outputs go to headers, but the correspondence between Timer hardware and Arduino PWM number is not obvious.

Herewith…

PWM Hardware

13 OC1C

12 OC1B

11 OC1A

10 OC2A

9 OC2B

8 OC4C

7 OC4B

6 OC4A

5 OC3A

4 OC0B

3 OC3C

2 OC3B

Although the Mega schematic shows PWM0 and PWM1, they don’t really exist; they’re actually the serial I/O bits for the FT232 USB converter.

PWM4 uses Timer 0: OC0B. Don’t mess with Timer 0’s prescaler to get higher speeds; it’ll wreck the millis() function and the firmware’s timekeeping in general.

Timer 5 doesn’t come to a header. If you desperately need three more PWM outputs (that aren’t supported by the Arduino runtime), you could affix three fine wires to pins 38, 39, and 40 of that TQFP. Good luck with that…

Some notes about fiddling with the Arduino PWM setup: changing the frequency and going much faster. The new Timers have different numbers, but the same considerations apply. As before, Thou Shalt Not Mess with Timer 0!

Consult the Fine Manual for the ATmega1280 chip to get all the details.

2010-09-05
Third Eye Hardshell Mirror Repair

Alas, the mirror I installed this spring didn’t survive our bicycling vacation; it succumbed to the second of three stuff-all-the-bikes-in-a-truck schleps arranged by the tour organizers. Being that sort of bear, I had a spare mirror, duct-taped it in place, lashed it down with some cable ties, and we completed the mission.

So.

Back to the Basement Laboratory Plastic Repair Wing.

The strut broke just behind the ball at the mirror, which implies the mirror plate got stuffed against something, rather bending the strut. The ball joint still worked, so I maneuvered the stub perpendicular to the mirror.

Drilling the strut

Normally I’d try to re-glue the joint as-is to get the best fit, but past experience shows that if it breaks once, it’ll break there again. I wanted to put some reinforcement into the strut, not just depend on a solvent glue joint. Some rummaging in the brass tubing stock produced a 1/16-inch diameter aluminum (!) tube about 18 mm long: just what’s needed.

So I filed the deformed plastic flat & perpendicular to the stubs, mounted the strut in the 3-jaw chuck on the Sherline’s table, lined the spindle up with the axis, and poked a 1/16-inch hole into the strut. The alignment looks decidedly off in the picture, but it’s actually spot on: what you’re seeing is some swarf clinging to the far edge. Honest!

Then I grabbed the mirror plate in the 3-jaw, lined up on the stub, and drilled maybe 4 mm down, which was roughly to the middle of the ball. The tubing was a firm push-fit in the hole and I hope it won’t over-stress the plastic into cracking.

Gluing the mirror strut

Run the spindle up, remove the drill, grab the strut in the chuck (actually, I had to swap in the larger chuck first), dab some Plastruct solvent glue on both ends, align the strut with the stub (they’re actually square in that section), run the spindle down to ram the tubing into the strut, then a bit more to apply pressure to the joint. I made the total hole depth about 2 mm longer than the tubing, so as to avoid the embarrassment of having the ends not quite meet in the middle.

No CNC; pure manual Joggy Thing action.

Let it cure overnight.

It’s now back on Mary’s helmet, with a pair of black cable ties ensuring that it won’t pop off, and seems to be working fine. I’m sure the ball joint will fail later this year, although that won’t be due to this repair.

Mirror on helmet again

2010-09-04
CPU Heatsink Fuzz Redux
A friend donated an old Aptiva with an AMD K6 CPU to my collection. It’s too slow & power-hungry to be useful, so I harvested some useful bits and passed the corpse along to the recyclers.

As fate would have it, I have an upcoming project that needs a cooler, so I popped the fan off the top (it’s rotated a quarter-turn: those tabs lock over the edges of the heatsink) to see what’s inside…

Fuzz in AMD K6 CPU Cooler

That accumulation was pretty much invisible from the outside, with most of the fuzz clotted around the periphery of the fan duct. The fan blows downward into the heatsink, which acted (as usual) as a good dust filter.

A bit of vacuum cleaner work and it’ll be just fine.

Memo to Self:
1. The bottom of the heatsink is a 42×78 mm copper block with the heat pipes soldered into notches. Clearance from the block to the step below the widest part of the fins is 18 mm and the fins are 25 mm above the block surface.
2. Fan = 12 @ 70 mA. Reasonably quiet.
3. The small blue heat sensor (at about 8 o’clock in the picture) is upstream of the heatsink and, thus, measures ambient air . It’s essentially open-circuit at room temperature, but a diode test shows 1.4 V in either direction. That suggests it’s not a thermistor or thermocouple, but the CPU is old enough that it’s likely not a fancy IC, either. A puzzlement.
2010-09-03
Arduino Connector & Hole Coordinates: Mega 1280 board
The Arduino Mega uses the ATMega 1280 chip to get more memory and far more analog & digital & PWM I/O pins, but remains more-or-less header-pin-compatible with the older Duemilanove and Diecimila boards (notes on the header coordinates for those boards is there).

Arduino Mega – ATmega1280 chip

Herewith, some useful coordinates for the Mega board in (X,Y) format using the default 0.001 grid: 1 unit = 0.001 inch (a.k.a 1 mil). Values are taken directly from the Eagle PCB layout.

The board outline is bounded by (2100,4000) on the upper right, with (0,0) at the lower left by the power jack. It’s not rectangular, but a conversation with Mr Belt Sander could remove the tab sticking out to the right beyond JP1/JP2 if that were really important.

The header names are not the same as on the old boards. Bolded values seem unusual.
- PWMH 1×8 @ (1300,2000) ← X is not 1290 as before!
- PWML 1×8 @ (2150,2000)
- COMMUNICATION 1×8 @ (3050,2000)
- JP1 2×8 @ (3750,1550)
- JP2 2×8 @ (3750,750)
- POWER 1×6 @ (1550,100)
- ADCL 1×8 @ (2350,100)
- ADCH 1×8 @ (3250,100)
- ICSP 2×3 @ (2555,1100) ← +5 X offset
- Reset switch @ (2920,1100) ← -30 X offset
The PWMH header is 10 mils to the right of its position on the older boards, but still not on the same grid used by the other headers: it’s now offset by a nice, even 50 mils. This probably doesn’t matter for most headers, given the sloppy fit. If you have a finicky board setup, you’re in trouble.

Here’s what the PWMH and PWML headers look like, measured against a Duemilanove board on the top. The offset is not due to perspective!

Arduino Mega PWMH header offset

The Mega board has four 0.125-inch diameter mounting holes (they use 125.984, which is a hard-metric 3.2 mm). The first one is at the same position as on the Duemilanove board.
- (600,2000)
- (600,100)
- (3550,2000)
- (3800,100)
Three fiducials:
- 1 @ (780,2000)
- 2 @ (2319,1603) ← deliberately offset from the grid?
- 3 @ (3800,100)
Memo to Self: As always, verify these numbers before you start drilling!
2010-09-02
Improved Tour Easy Chain Tensioner

A discussion on that post reminded me of this old project: replacing the chain pulleys in the midships chain tensioner on my Tour Easy recumbent.

The problem is that the original pulleys used steel bearings in a plastic race, for reasons that I cannot fathom. They last for a few thousand miles, then get very wobbly and noisy. The solution, as nearly as I can tell, is to replace them with pulleys using cartridge bearings.

This is what one looks like after four years slung below my bike. Surprisingly, the bearings still feel just fine, even though they’re not really sealed against the weather.

Tour Easy – Cartridge Bearing Chain Tensioner

Gotcha: the OEM pulleys are not the same OD / number of teeth as pulleys found in rear derailleurs.

Soooo, after a bit of Quality Shop Time, I had these…

Tour Easy Replacement Idler Pulley

This is where you really want an additive machining process, as I turned most of a big slab of aluminum into swarf while extracting each pulley.

The first step is to drill holes around the perimeter where the chain rollers will fit, plus drill out as much of the center bore as possible. Then mill down to the finished thickness across the roller holes and helix-mill the bore to size.

Side 1

Flip it over and mill the other side to the proper thickness.

Run it through the bandsaw to chop off all the material beyond the outer diameter.

Grab what’s left in the three-jaw chuck and mill around the perimeter to get a nice clean edge.

Side 2

And then it Just Works. I made another for Mary’s bike, but she said it was too noisy (which is why they used plastic rather than aluminum) and I swapped it for a Terracycle idler.

This is from back in the Bad Old Days before EMC2’s version of G-Code supported loops. You don’t need to see that code, trust me on this.

2010-09-01
Monthly Aphorism: On Production Quotes
- We can ship that many, but you must give us time to build the factory.
Once Upon A Time, back in the IBM Video Disk project, a friend (herein known as LLT) built a demodulator for the video data streaming off the disk. This being the development phase of the project, cost was not much of an object, so he used a quartet of high-speed TRW TDC1003J digital multiplier-accumulators running in pipelined parallel to handle the data rate.

While a 175 ns MAC isn’t a big deal these days, it was a state of the art TTL chip back then: a finned-heatsink 64-pin DIP package that cost approximately a bazillion dollars. The board was maybe two feet on a side in classic Wire-Wrap style.

We called them Multifryers: each chip drew 750 mA at 5 V. That board had many cooling fans.

Then, One Fateful Day, came a request to quantify just exactly how much it would cost to build a production version of the player. LLT pointed out that the demodulator board itself would cost more than a really spiffy car, but to no avail: he had to come up with a cost estimate for a fairly large production volume of the as-built hardware.

So LLT calls up TRW and asks for a quote on thus-and-so-many parts, with delivery to-be-determined. There’s a long silence, after which they tell him they’ll have to get back to him on that.

Time passes.

Eventually he gets the quote. He had to tell them to not start building the factory right away, because the project was most likely doomed. Much relief was expressed…
2010-08-31