I'm building a machine to produce pressed pennies. It's similar in many ways to the ones you might find at a zoo, airport, or museum, with a few important differences. While most penny crushers let you pick between four possible images, this one can make eighteen unique images. It also produces double sided coins, so there are nine front side images and nine reverse side images.

Before I started on this project I built a simpler prototype penny crusher that used interchangeable die plates. That allowed me to validate my some fundamental elements of the design. These elements include the sizing and choice of bearings, die design and fabrication, and the ratio of the drive reduction gearing. For this project, I took the lessons from that prototype, and used them to design a more complex machine that could produce a much wider array of images without manually changing die plates.

If you're curious about why I'm building penny crushers, you might be interested in a talk I gave on that subject or the slides from that talk. My collaborator on this project, Shaun Slifer, put together a great project summary page as well.

Step 1: Penny Crusher Fundamentals

The most basic constraints on the design of a penny crusher begin with the process for crushing a penny. The penny is rolled between two wheels with a gap between them that is slightly smaller than the thickness of the penny. This mechanism is an instance of a rolling mill, a common and widely used device for reducing the thickness of metal sheets. In industry, rolling mills can be quite large, and a large body of theoretical and practical knowledge exists to support their design and operation. I applied only the most basic models of rolling mill operation in designing my penny crusher, and added large safety factors where possible to accommodate any inaccuracies in my model.

I started by asking some basic questions including:

How much force will it take to crush a penny?
How much torque will be needed to turn the rollers?

It turns out that the answers to those questions depend on a few factors, some of which are within the designer's control. The two most important free parameters that determine the force and torque needed to press a penny are the final thickness of the crushed penny and the diameter of the roller wheels. Important fixed parameters outside of my control that are important to the calculation include the material the penny is made of and the thickness of the penny before crushing.

The physical process that the penny undergoes when it passes through the rollers is called plastic deformation. Plastic deformation in zinc and copper, the metals US pennies are made from (post and pre 1982, respectively), is defined by two properties called the Compressive Yield Strength and the Strain Hardening Exponent of the metal. Yield Strength is a measurement of the pressure required to cause the metal to flow, rather than bend like a spring. In Zinc alloys, this is about 30,000-40,000 pounds per square inch. The Strain Hardening Exponent defines how the effective yield strength increases as the metal is deformed, and varies between zero and one. This reflects a property of many metals called work hardening - as the metal experiences strain its Yield Strength increases, until eventually the yield strength exceeds the ultimate strength and the metal fractures. Work hardening is why you can bend a piece of metal back and forth several times and then easily break it off. Knowing only the starting and ending thicknesses of the penny, we can calculate the average pressure applied over the contact area to produce the thickness reduction based on these two numbers.

In my case, a penny is about .060 inches thick, and I wanted to reduce it to a little more than .030 inches thick. I arrived at .030 inches after measuring a few pennies in my collection of pressed pennies. Based on these values I calculated the average flow stress over the contact region to be roughly 30 Kpsi.

If you look at the diagram of the rolling mill above, it's clear that only a small section of the penny is actively deformed at any instant. Specifically, the part in contact with the rollers is changing thickness, but the part that's already exited the rollers, and the part that's not yet entered them, are not. This means that we only need to apply that 30,000 PSI of pressure over a small area of the penny.

Ignoring frictional effects, the total force applied by the rollers is equal to this pressure integrated over the surface in contact with the penny. Roughly, this is the area in contact times 30 Kpsi. The total area in contact depends on the geometry shown in the diagram above, specifically, the roller diameter, the gap between the rollers, and the initial thickness of the penny.

Choosing a roller diameter was a complex design problem, that I'll talk about later. For now, it's enough to know that I chose 1.5". Given those parameters I was able to calculate that when the widest part of the penny was being pressed, the area in contact would be at most 0.11 square inches.

That gives a total force per roller of 5,070 lbs and an estimated torque per roller of 380 inch-lbs.

Step 2: Initial Design Choices

I considered three different approaches designing a penny crusher that could produce a significantly larger number of images than the standard commercial design. My primary concerns during this design phase were to ensure that the mechanism was simple enough to build efficiently, that I could use bearings and gearing similar to those I'd used successfully in my prototype, and to ensure that the gap between the rollers could be adjusted easily.

The mechanically simplest solution I could think of was to increase the diameter of each roller to make room for more images. Two large rollers with nine images on each roller would have a diameter of about 3.5 inches. These larger rollers would have doubled the required force per roller and tripled the required torque. Since I was already operating near the maximum load for the bearings and gears I planned to use, I decided not to pursue this design. It was also unclear how to create arbitrary pairings of images without introducing a clutch mechanism to allow one wheel to rotate without the other rotating simultaneously. Preserving alignment between the wheels (so that the images on either side engaged simultaneously) once a clutch was introduced seemed like it would result in additional complexity. In retrospect, it may have been a better idea to switch to larger bearings and gears with a higher load rating rather than pursue a more mechanically complex design to avoid them.

The second solution I considered was to gang three smaller rollers side-by-side into a single wide roller. By sliding past each other and rotating, the rollers could bring any pair of images into opposition. I abandoned this design because of the complexity of allowing the drive gears to remain engaged during lateral motion, and the need to index linear motion.

The third solution was to mount smaller rollers on a revolving carriage that could bring each pair of wheels into opposition. This is the design I eventually built, and that I'll discuss it in detail in the next step.

Step 3: The Indexable Die Carriage Design

Of the three designs I considered, the index-able die carriage is the only one that allows any pair of die images to be used without the need for a clutch that temporarily disconnects the drive gears to allow the Die Rollers to move independently.

This is achieved by the use of an epicyclic (planetary) drive gear in the carriage that runs at a 2:3 ratio. If the sun gear in the middle of the carriage is fixed in place, and the carriage is spun around it one full revolution, the dies will have completed two thirds of a full revolution. Since there are three die images on each face, this essentially advances the die image one step without moving the drive gear.

The procedure to create an arbitrary pairing of images is as follows:

One die carriage is rotated until the die with the desired image is selected.
The drive shaft is rotated until the desired die image is on that die is in starting position.
The second die carriage is rotated until the die with the desired image is selected and the desired image is in the starting position. Note that this may require upto three full rotations of the die carriage.

When assembled, the die rollers are placed so that if the first die has its image facing radially outwards from the center of the carriage, the second roller is advanced 2/9 of a rotation from that position, and the third die 2/9 of a rotation from the orientation of the second die. This ensures that each die roller will present its image in the same relative position as the carriage rotates.

Step 4: Load Frame Design

The Load Frame transmits the forces necessary to deform the penny between the die carriages. It is also designed to be adjustable so that the gap between dies can be finely tuned.

The die carriages rest in bearing blocks made from 3/4" thick steel plate. These bearing blocks press against the end plates. The end plates then transfer load to the tie plates that run above and below the die carriages. Because the connection between the end plates and the tie plates is in tension, I used load rated bolt for that connection. I also used the end plate to tie plate connection as the location to insert shims that are used to adjust the die roller gap.

I also used steel shims in the prototype, placing them between the die plates and the roller wheels, and had run into trouble with the shims deforming over time and becoming thinner. My load frame design avoids this problem since the shims only experience the preload tension of the tie plate bolts, but are not under additional force during operation.

Step 5: Die Carriage Indexing Design

When the carriages rotate the need to move exactly one third of a revolution and then lock in place. An indexing mechanism is what creates this type of motion.

The video above shows my first design for this mechanism. I wanted to ensure that the mechanism would provide a strong positive locking engagement, and require only single rotational motion to operate. I started with the idea of using a strong pin passing through both the bearing block and a flange plate connected to the die carriage to lock the die carriage rotation relative to the bearing block.

Because I wanted the mechanism to lock in any of three positions I created three pins and three holes, although it may also have been possible to use a single pin, three holes in the flange plate, and one hole in the bearing block. I didn't like that option because I wanted the mechanism to be symmetric around the die carriage's rotational axis, so that the locking forces wouldn't load the die carriage bearings.

Based on the initial idea of three pins that repeatedly insert and withdraw from indexed holes, I designed a mechanism to automatically withdraw the pins under a spring load, then use the pins to transmit a rotational force from the input shaft to the die carriage, and then automatically re-lock the die carriage after it had completed its rotation.

Ultimately, however, I decided this mechanism was overly complicated, and built a pair of 3-position Geneva mechanisms to perform the indexing. Unfortunately, because of the small torque arm during the middle of the index-advance movement and the unexpectedly high friction in the die carriage bearings the Genevas, as built, require manual assistance to advance the carriages, although they do an excellent job of locking them at their indexed positions. Eventually I plan to separate the drive-power for advancing the die carriages from the indexing mechanisms, using reduction spur gears to drive the carriages and an indexed ratchet to hold them in position.

Step 6: Wooden Prototype

Before beginning to fabricate the penny crusher in steel, I made a quick prototype using laminated plywood and a laser cutter. I built two die carriages, including the gearing the die wheels, the load frame, and a mockup of the indexing geneva.

This prototype identified two issues that would have been difficult to correct if I hadn't identified them before beginning work on the steel assembly. First, I'd left zero clearance between the drive gears and the middle plates on the opposite drive carriage. This interfered with rotation of the die carriages, so I added a 1/16" clearance by making the die carriage plate spacing asymmetric.

Second, the geneva mechanism tended to bind when the drive pin support passed over the driven cam. This again was due to insufficient clearance, and was corrected before beginning the final assembly.

Step 7: Die Carriage Fabrication

I began by fabricating both die carriages. The three-lobed carriage plates and the rectangular tie-plates that join them together were cut from plate steel on the waterjet, as shown in the video. The end plates are designed to have a short shaft protruding from them that rests on the load frame's bearing blocks. I created this shaft by first finishing the rough hole cut on the waterjet in the vertical mill, enlarging it to a precise diameter and adding a chamfer to both sides of it. Then I turned mating pieces on the lathe that were about .001" larger than the holes. These pieces also had matching chamfers.

I used a hydraulic arbor press to insert the shaft "spuds" into the end plates using about 5000 lbs of force. Then I TIG welded the plates and spuds together, using the matching chamfers to ensure good penetration of the plate steel. I cleaned up the back-side welds with a face mill, then turned down the spuds to the final diameter on the lathe.

Using the turned spuds as my reference datum I was able to create the bores through the spuds that the drive shafts pass through, and bore out the three smaller recessed areas for the die roller shaft bearings. Because it's essential that the die rollers be a consistent radius from the center of the die carriage rotation, using the turned shaft as my reference was important.

I finished the tie plates in the vertical mill, facing and chamfering all sides, and drilling/countersinking the six holes for the bolts that fix them to the carriage plates.

I used the water jet to cut a small jig that let me hold the carriage plates on end in the vice so that I could use the vertical mill to drill and roll-tap the holes for the tie-plate bolts. This was a slow process at three setups per plates and six plates, if I'd had access to the TR-160 5-axis trunnion at this time the work could have been done in one setup per plate.

I also made the mistake of tapping the holes for the tie plates before welding the spuds. The heat from the welding warmed the threads enough to make the thread fit noticeably tighter. I chased all 36 holes by hand to fix this.

The final step in assembling the die carriages was to use the arbor to press the die shaft and drive shaft bearings into each plate.

Step 8: Load Frame Fabrication

The load frame was a simple but time consuming machining process. The bearing block's primary features are the bores for the bearings and the tapped holes used to attach them to the end plates. In addition to these features each block required three setups to clean up the edges after the rough cut on the waterjet.

The end plates have no bores, and were relatively quick to make because their edges could be left with the rough waterjet finish, since no other pieces bear on them.

The tie plates proved somewhat difficult to clamp in the mill due to their size and awkward shape, and required four setups per plate, two to tap both ends and two clean up the edges.

Step 9: Gear Fabrication

The smaller epicyclic gears are stock gears available from manufacturers like Martin or Boston Gear. The larger ones, however, were custom cut on the waterjet. Because the waterjet cuts a beveled profile and is unable to cut with better than .030" accuracy (YMMV, but that's my experience) I designed the gears in Autodesk Inventor to include about .020 of backlash. I was also able to design gears with non-standard diametral pitch, which let me achieve specific shaft-to-shaft distances at particular gear ratios with theoretically perfect tooth engagement, rather than design the ratios or the shaft distances around a stock diametral pitch.

I found that water jet cut gears worked best with diametral pitches of eight and below. After cutting the teeth and a rough undersized bore on the waterjet, I used a vertical mill to bring the bore up to size, then used an arbor press and broach to cut a keyway to lock the gear to the shaft.

The exception to this process is shown in the final picture of gears that mount on the die carriage flanges with bolts, rather than on a shaft.

Step 10: Machining the Geneva Mechanisms

I used a two setup flip-milling process for both parts of the Geneva mechanisms using soft jaws to hold the work after flipping it. The simple setups and soft metal made this one of the fastest parts of the project, despite the relatively complex geometry.

The photos above shown one of the driven elements during machining, and the full process for machining one of the driver elements. The process begins with a block of aluminum, from which the basic shape is roughed in one pass. Then the drive pin feature is cut down to the height of the drive pin. Next the pin is cut away from the surrounding material. After this, several finishing passes countour and chamfer the edges of the piece.

After this the soft jaws are machined to match the profile of the cut part, and the part is held in them while the remaining stock is removed from the back of the piece with a roughing mill and then finished with a combination of a face mill and chamfer mill.

Step 11: Dies and Shafting

Before test assembly of the penny crusher I needed to make the die shafts and dies.

The dies are cut from solid rod stock of A2 tool steel. I begin by drilling and boring the inner diameter on the manual lathe. Since I'm working with ground shaft stock I aim for an ID of 1.000 +.0005/-0. I made a custom mandrel that fits in the jaws of the 4th and 5th axis trunnion. This mandrel has a 1" OD that fits the die blank, and positions the blank inside the very small working envelope of the mill. Once bed travel limits and tool holder clearances are taken into account, there's only about 1.2" of Y-axis travel, so the placement of the die blank relative to the trunnion chuck needs to be carefully planned.

Before I can engrave the dies I need to reduce their OD to the final size. It is essential to the engraving process that the surface of the die blank be concentric with the trunnion's B axis to within 0.001" or less, since any variation in that distance will change the depth of the engraving. Since the engraving is only .0075" deep at mosta variance of even .001" is significant. Rather that turning the OD on the lathe I rotate the B axis while keeping an end mill rotating in a fixed position near it. By varying the distance from the B axis to the end mill axis I can get very consistent diameters and excellent concentricity. The final photo shows a dial test indicator measuring the runout in the die blank after this operation. There was no observable runout.

I used carbide engraving cutters at 10,000 RPM spindle speed to cut the images into the dies. The deep boundary dots in the pattern are cut with a 90 degree drill/mill. Then, the larger areas of the images are removed using a 0.010" 45 degree flat-tip engraver. More detail is added using a 0.005" 45 degree engraver, and final touch-up is done, depending on the image, using a pointed 45 degree engraver.

The toolpaths for the engraving started as black and white raster images provided by the artists. I used ArtCAM to convert those images to vector outlines, then did my layout on a rectangular work area that would 'wrap' onto the surface of the die blank. I used V-cutting and smart-engraving tool pathing to generate the toolpaths, adjusting the vectors slightly as needed. I customized the default ArtCAM post processor to use the Haas mill's G107 mode, which substitutes B axis motion for X axis motion, effectively wrapping the design onto the cylindrical die.

After machining the dies were sent off for heat treat, with a target hardness of RC 56.

The original illustrations were drawn specifically for this machine by a number of artists. The images for the front sides, and the artists were:

Ally Reeves - Passenger Pigeon
Bec Young - Red Wolf

Pete Railand - Eastern Elk

Katie Kaplan - Caribbean Monk Seal

Josh MacPhee - Kemp's Ridley Sea Turtle

Santiago "Mazatl" Armengod - Jaguar

Mary Tremonte - Eastern Cougar

Roger Peet - Great Auk

Shaun Slifer - Mexican Grizzly Bear

Step 12: A Work in Progress

I assembled the components I've built to date and produced and was able to produce several pressed pennies. However, much work remains to be done. I still need to fabricate the image selection mechanism, that lets the user pick what pictures they want on their penny and control the rotation of the die carriages and timing of the coin drop. After that I'll make a case to keep the moving parts safely away from users, and create a more ornate handle.