Gaggia Gelatiera Ice Cream Maker Repair

Last winter I saved an ice cream maker from going to the land fill.

P1020779

This is a Gaggia Gelatiera ice cream maker from the early 80’s. While small in size, it’s a very heavy italian made machine. It’s significant weight (nearly 50lbs) is due to the built-in refrigeration system. In most ice cream makers you need to either pre-freeze a water lined bowl, or use rock salt and ice to cool the ice cream mixture. This machine uses a compressor, condenser, evaporator, and refrigerant to constantly cool the bowl. The advantage being that it is always ready to make batch after batch of ice cream.

After turning it on I soon realized why it was headed for the dump. While the bowl would cool down fine, the mechanism turning the dasher made a horrible grinding noise and wouldn’t turn. While I’m not a huge consumer of ice cream, I do enjoy a good repair challenge!

The machine was remarkably easy to take apart. About 8 Philips head screws and the top cover came off.

With the cover off, the first sign of trouble appeared: a fine black powder spread around the base.

P1020787

The dasher drive motor looked okay and spun freely so it seemed the problem was coming from deeper in the machine.

Getting to the innards required removing pretty much every component from the base. To make it easier to remove all the components I disconnected all the wiring, but not before making a quick diagram.

P1020796

P1020802

With the base removed, the source of the grinding noise was pretty apparent.

P1020803

P1020805

 

One of the bushings on the intermediate shaft had seized and began spinning in its bore. The bore, being made of plastic, was worn into an oblong shape. This caused the intermediate shaft to become misaligned, which caused two problems. First the small metal helical gear was no longer meshing correctly with the large plastic main gear, which is why the dasher wouldn’t turn. Second, the tilted intermediate shaft caused the timing belt to rub against the main gear. You can see in the above picture where the belt cut into the gear and deposited bits of belt all over the inside case.

P1020826

The new drive belt is on the right, the old one on the left. The belt wore about 1/8″ of its width off.

I ordered new parts from my favorite industrial supplier, McMaster. New bearings for the dasher drive shaft, 120XL size timing belt , and food safe grease. I couldn’t find new bushing for the intermediate shaft as they were not a standard size. I inspected the old bushings closely and there wasn’t much wear.

P1020825

Using an x-acto knife I cleared out all the bits of melted plastic from the bushing bore. Fortunately there was enough of the original bore that I could get the bushing back to its original position by pushing it against the undamaged side of the bore. The damaged section was filled with 3M DP810NS, a high viscosity two-part epoxy.

P1020815

P1020834

The black plastic cylinder is the evaporator/bowl unit, which is connected via copper pipes to the compressor and condenser. This made for an unwieldy re-assembly. The copper pipes connecting all the parts are very small, if they become kinked they could leak, or restrict the flow of refrigerant. I found it was easier to assemble the gear train upside down and place the base on top of it.

P1020837

WIth everything back together the machine was much quieter and the dasher spun properly.

No previous post-repair testing has been this delicious (vanilla custard if you’re wondering).

photo (18)

 

Custom Dies for the Mini Molder

Since I decided to debut the mini molder at Maker Faire bay area a custom mold was in order. One of the more iconic images for Maker Faire is the Makey robot figure. It has roughly the correct proportions for a mold-a-rama figure, and a simple enough profile for my basic CNC abilities. I found an image of a pin sold in the maker shed.

I traced the above image in solidworks and created a model of the figure.

Makey_model

The base on the bottom is partly for stability, but also a result of the liquid plastic entry and exit ports. Those ports are not quiet aligned with the legs of the robot, so the flat base ensures the mold cavity is always around the two liquid plastic ports.

Once the figure was modeled I created mold halves using the cavity feature in solidworks and added various holes for screws.

makey_mold

In most injection molding dies, water channels are used to route chilled water around the mold, this helps cool the part faster enabling a faster cycle time. Mold-a-rama dies had a water jacket cast into the back side of every mold die. The mold cylinder mounting plate on the back doubled as a block off for the water jacket.

The mold dies I designed consist of several parts:

  1. Mold face with part cavity
  2. Water channel spacer plate
  3. Rear mount plate

I decided to make the mold die in multiple parts for a few reason. The first is that more of the mold is reusable should I want to make a new mold design. If I had cut the channel for the water directly in the back of the mold, I would have to cut that same feature in the back of every new mold as well. Making a separate water channel spacer means fewer setups on the CNC mill. The second is that cutting a water jacket in the back of an aluminum mold would take an enormous amount of time on my little CNC mill, the much faster material removal rate of delrin made this an easy decision.

What follows is a mostly complete step-by-step of the machining process for the mold spacers.

After rough cutting the delrin stock and squaring the sides, it is tightened down to the tool plate in my mill.

P1020531

Setting the tool height so my mill knows where the top of the stock is relative to the tip of the cutter (center drill in this case)

P1020534

Drilling clearance holes for the 1/4-20 screws that hold all the mold sandwich together.

P1020537 P1020543

1/8″ 2 flute carbide end mill creating the o-ring groove.

P1020551

This center u-shaped section is where the water flows into and out of the mold cooling cavity. The entire center section needs to be cut out. Since I don’t want to cut into my tool plate, I cut the slot to half-depth, then flipped the part over and cut it again. Flipping parts is always tricky, as any small misalignment will show up were the two cuts meet.

P1020557 P1020561

In the first part I made, I removed all the material bit by bit (machinists call this pocketing), this took a long time. In the second part I used a 1/4″ 2 flute carbide end mill to cut a half depth slot around the piece of delrin that was to be removed. Flipping the part over, re-cutting the same path and the whole center section comes out in one piece.

P1020562

The last step on the spacer plate is to drill and tap the NPT threads for the pipe fittings. The drill bit needed to drill this hole was too long to fit in my mill, with its limited z-travel. So I milled out the hole using a 3/16″ end mill.

P1020563 P1020571 P1020576

Taping large threads like 3/8 NPT take a decent amount of torque, so your part should be firmly mounted. I sandwiched the spacer plate between two backer plates, put them in my screwless vise, and then clamped the vise to my shop table.

P1020611

P1020617

Getting a straight start on your tap can be pretty tough without a tap guide, or performing the operation in a mill. I used a small square to align the tap, going slowly and re-checking after each turn.

P1020613

The mold faces started life as a piece of 4″x1″x12″ 6061 aluminum. Like the spacer plate before, they were cut to rough size on a band saw and then squared up with a fly cutter on my sherline mill. I don’t have as many pictures of the milling process, but I did take a time lapse video.

P1020579

P1020581

P1020585

P1020599

P1020600

P1020601

To remove the tooling marks left by the mill cutters, I used 400 grit wet or dry sand paper and some free labor (thanks Ashley!) It was time-consuming, but gave me a greater appreciation for what goes into achieving a mirror like finish on all the plastic products I see.

P1020606

Exploded view of one of the mold dies (minus o-rings).

P1020618

Spacer with its two o-rings. I used plumbers faucet grease as an o-ring lubricant as it was what I had on hand.

P1020628

P1020632

Assembled with pipe fittings installed.

P1020645

P1020642

A Pump, a Check Valve, a Problem

During testing I noticed the molder’s plastic injection pump stopped working after 2-3 consecutive pumps. If the molder sat for a few hours it would work for a few shots, but shoot blanks soon after. The path from plastic melt tank to mold goes like this:

1. The injection cylinder retracts, creating a vacuum in the pump body, which draws plastic through a check valve and into the pump body.

2. With the pump body full of molten plastic, the injection cylinder extends, building pressure inside the pump, closing the check valve and forcing the plastic through the pump outlet

3. Hot plastic travels through a copper tube to the shuttle valve, if the shuttle valve is in the plastic injection position, the plastic pass through the valve and up to the opening in the bottom of the mold.

I checked that the shuttle valve in the pump body was switching and that the injection cylinder was extending fully, both were operating correctly. The next place I could see a problem occurring was at the pump inlet and check valve.

Getting to the check valve means removing the pump body, which means draining the tank. I put a drain port on the tank, but the close proximity to the base make draining a bit of a task. A make shift funnel made from aluminum foil did the trick.

P1020666

After removing the pump body from the molder I inspected the check valve and found a few particles, burnt plastic and bits of cork insulation, but nothing that would clog the inlet completely.

P1020679

P1020681

While I had the check valve out, I inspected the flow path through the valve. The path consisted of some very small holes (~0.050″), the molten plastic having the constancy of honey, this particular check valve could not possibly have flowed enough to recharge the pump when the piston retracted. It’s clear I didn’t consider the viscosity of the working fluid when choosing the check valve.

I looked for a check valve with the largest flow path that would fit on the 3/8″ NPT pump inlet. That happened to be a swing style check valve. The new check valve is much larger than the old one, so much so that it won’t fit in the same spot, coming directly out of the pump. A 90 degree fitting in between the pump body and the check valve put it in the upright position.

P1020684

While I had the pump out, I took this opportunity to fix the threads in the pumps outlet port.

P1020676

 

They had become cross threaded with the fitting that was installed, and was leaking when the pump pressure built up. A few turns of a 3/8 NPT tap and the threads were cleaned up.

Fast forward a few hours and the pump is installed and has a tank of liquid PE wax around it. Cycling the pump’s pneumatic cylinder it was clear the new valve was a huge improvement. You could see a steady stream of plastic flow into the check valve every time the cylinder retracted. Cycling the injection cylinder back and forth yielded a consistent flow from the shuttle valve outlet.

P1020702

 

Sherline 2000 Solid Column Conversion

A while back I finished the CNC conversion of the Sherline 2000 mill I bought off craigslist. Now that I’ve had a chance to use it I can say it fails in a few respects. First is that it is a pain to tram, there are so many axis with movement that it takes forever with a DTI (dial test indicator.) I got a nano tram off eBay, but did not have any luck getting that to work. Checking with a DTI after aligning it showed it was consistently off in both axis by about 0.005-0.008″ or more. The instructions say to use a feeler gauge to fine tune the alignment, but that kind of defeats the point of its supposed dead simple alignment procedure.

The second issue is that the head stock gets knocked out of alignment by taking some pretty light cuts in aluminum (~0.010″ depth of cut). No matter how much I cleaned the mating surfaces and tightened the bolts, the column still shifted.

The adjustable headstock may be useful for certain weird setups, but unless you are cutting exclusively plastics or wood I cannot recommend it.

After ruining one last part on the 2000 mill, I decided to convert it to the solid column of the 5400. Since I reused the existing 14″ base, I ended up with a mill that has 2″ more Y travel than the 5400.

If you want to do this conversion you’ll need to order PN 50050 from Sherline, I paid $48.00 + shipping for it.

The conversion starts by removing the head stock.

P1020490 (Large)

Unbolting the z axis dovetail

P1020491 (Large)

 

And removing the column assembly

P1020493 (Large)

For some reason the bolt pattern on the base of the 2000 mill is different than that of the 5400 mills, so you need to drill two new holes to mount the column. This guy has a clever way of turning the mill on itself to drill said holes. I followed his method, using pieces of 10-32 all thread and a piece of .25″ x 1.00″ aluminum flat bar.

P1020505 (Large)

P1020504 (Large)

Here’s a drawing of the new hole pattern I made. Note that the column bolts are not centered between the dovetail in the base (which is why the circular 2000 column cut out is not even). Basically drill two holes .5″ from the rear edge, spaced 2″ apart and centered between the edges of the base, with a letter F (.257″) drill.

image001

 

I scribed lines into the base and aligned the drill using the pointed end of my center finder.

P1020510 (Large)

And drilled with a center drill followed by a letter F drill bit.

P1020511 (Large)

P1020519 (Large)

The solid column still will not bolt to your 2000 base at this point as there is interference between the column base and those little pointed ends of the dovetail. Rather than machine those off I made a spacer from a piece of 5/8″ cast tool plate I had leftover from the mounting plate I made for the mill.

P1020778 (Large)

Depending on how your mill base is mounted you’ll either need to counter bore the underside of the column mount holes in the base to accept a cap screw, or use a low head 1/4-20 bolt. I chose to do the latter.

One more thing, make sure you remove the z-axis drop down bracket from the z-axis nut. You don’t need this anymore and it will prevent the head stock from raising all the way up, limiting your z-axis travel. The 2000 mill needed it so that the headstock could lower all the way to the table.

Once the bracket is removed, you’ll need to use a shorter screw to attach the z-axis backlash bracket. Or use a spacer like I did.

P1020770 (Large)

 

The left over parts.

P1020766 (Large)

I’ll probably keep these around in case I sell the mill, or if for some reason I come across a setup that requires them.

After the upgrade the difference is night and day while cutting aluminum. Much reduced chatter, more aggressive cuts possible, and the column stays in alignment. It’s like getting a whole new mill for $50.

 

 

Over Center, Under Pressure

One of the ways the mold-a-rama process differs from traditional injection molding is the way the part is ejected from the mold. Usually one of the dies has a set of ejector pins that pop the part out of the mold (if you look at most plastic products you will see a series of small circular or square indentations on the back side, these are small marks left from when the ejector pin pressed the part out of the mold.) In this lego manufacturing video you can see the thin ejector pins stick out right as the part falls out of the mold at the 40sec mark.

To keep the machine simple, the mold-a-rama uses no ejector pins. Instead it relies on the part staying in place while the molds open around it. This is possible because the bottom of mold-a-rama die is actually open. This is a picture of the bottom of my mold dies.

Bottom of Molds

The base of the plastic part sticks to the tank lid (which is why all mold-a-rama toys have some kind of flat base.) Traditional injection molding dies form a sealed cavity when they meet, save for the sprue, the opening where plastic enters the mold. Mold-a-rama dies form a complete cavity by sealing against the melt tank lid (the aluminum square below the molds, which also serves as a cover for the plastic melting tank). As you can imagine, it is very important that the dies press firmly against the melt tank lid, otherwise plastic would leak out between the dies and the tank lid.

Below is a CAD screenshot of the mold cylinder assembly I originally made, the die is the blue rectangle. Can you guess if they functioned correctly?

overcenter_cyl_mount

The answer is no, no they did not. When the dies closed against each other they popped up slightly, instead of being forced downward to seal against the melt tank. The can been seen in the video below. Watch towards the end and you will see the pair of molds raise up slightly.

 

If I were to run this in an injection cycle I would have melted plastic spewing out.  After some head scratching I realized I had the mold cylinders (the grey rectangular part in the CAD screen shot) placed above the pivot point (the grey circular part), this meant that when the molds pressed against each other it tended to rotate the entire mold cylinder and mold counter clockwise about the pivot point, which meant the mold moved in the upwards direction. I tested this theory by flipping the entire mold cylinder and mount upside down. Now the mold cylinder was below the pivot point, which meant the mold cylinder now tended to rotate clockwise about the pivot point, forcing the die downward. This video shows how the molds are forced downward when they meet.

 

Armed with this realization, I redesigned the mold cylinder mounts to match the flipped version I had tested. With the two designs side by side (old on left, new on right) you can see how the mold cylinder location has changed relative to the pivot point.

compare_mounts

I also used this as an opportunity to to redesign the way the dies are constructed and how they interface with the mold cylinder, but that is for another post! Below are some pictures I took while machining the parts for the cylinder mounts.

photo 1

photo 2 (1)

The above part was the first part I CNC cut on my mill and boy was it a learning experience. I probably scrapped 3 or 4 parts before I made the first one correctly. Between fixturing, weird g-code bugs, and getting feeds and speeds right I learned so much on that first part.

The final product. The slotted screw holes allow me to fine tune the position of the die at the extended position. The bolts holding the bracket to the cylinder have serrations under the heads which bite into the aluminum.  This prevents the cylinder from slipping relative to the bracket when the dies meet. The other large cutouts are clearance holes for the cylinders air fittings.

P1020523

 

HP Plotter Take Apart and Analysis

A friend of mine found an old HP plotter from the mid 90’s on the side of the road. Knowing I was into electromechanical objects he asked if he should pick it up for me, and of course I said yes. We initially though about repairing it (its drive belt had desintigrated and it needed new ink), but realized that even after fixing it it would have little resale value.  We decided to take it apart and harvest it for parts, while along the way looking for interesting mechanisms and design features.

Printers are chock full of useful parts like motors, linear slides, encoders, and switches. In fact a very small crude CNC router could be made from the parts of two inkjet printers.

First impression based on getting it out of the car: this thing is heavy and built like a tank; the specification sheet says 95 lbs.

plotter 1

Opening an access door reveals a few MB’s of RAM on standard SIMM cards.

plotter 2

Removing more covers, and now we can see the linear encoder for plotter’s print head. This encoder sends position feedback to the plotter; once homed, the print head’s exact position along it’s travel is known. Barely perceptible scribed lines are in the lower clear portion. The black square in the center of the picture with white text is the read head.

 The silver looking strip is actually a stainless band that is pulled taught by a tensioner at the end. The considerable tension is needed as any sag in the encoder strip would lead to position errors of the print head and messed up prints.

plotter 3

plotter 6

A view of the back, main control board on the left, power supply on the right.

plotter 4

Continue reading