Atkinson Differential Engine - Making it work?

Home Model Engine Machinist Forum

Help Support Home Model Engine Machinist Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.


Well-Known Member
HMEM Supporting Member
Dec 31, 2010
Reaction score
Even if I get plans from reputable sources I always re-draw them in my 3D modelling program. By the time I get finished modelling the engine I have a full understanding of how it goes together, and I almost always find errors or omissions in the drawings which I am able to correct. Back in 2015 I acquired the book by Vincent R. Gingery " Building the Atkinson Differential Engine" pictured below.
I had already built the "Atkinson Cycle Engine" also by Mr. Gingery with success. So I thought it would be a good addition.
Right from the beginning I was not impressed with the book. It starts off with how to make patterns to make your own casting and the dimensions required there and follows on with machining the castings. There are several places in the book where dimensions are confusing to say the least. To top it all off in the assembly instructions it states that you need to fiddle a bit to make it all work and even suggests that some parts might have to be re-made in order to get the mechanism to operate properly. Following that it is suggested that a very long run-in period of turning the engine over with an electric motor might be required.
To me none of that should be necessary. In my opinion if the plans are worth anything and you build it according to the plans the engine should be capable of running with not much more than typical fuel and ignition adjustments.
One of the advantages of a 3D model is that if the parts are constrained properly you can actually manually animate the engine by say turning the flywheel which will make all the other pieces move. Being that this engine has a very strange mechanism it was interesting to "operate" the model and see how it achieves the typical 4 cycles. You can turn the flywheel a few degrees and take measurements and figure things like compression etc.
What I noticed right off is that the engine was going to have VERY low compression. I figured about 2:1. I checked my drawings with what was in the book (factoring in the confusing drawings) but I still could not realize better compression.
I concluded that something had to be wrong with the drawings so I put it on hold and decided I'd ask one or two builders of the engine I'd seen running at shows what they had found to be wrong and what they did to fix it. I asked one fellow and he just said simply that he had to change something but he couldn't remember what. I contacted another fellow that had one running on YouTube. He said he made a wooden model first before he built it but had no details on what might have been altered.
SO I decided I wasn't going to build it and the files have been sitting on my computer for 5 years.
Then about a year ago on this forum I saw a post by Gordon entitled "Atkinson Frustrations'
Gordon had built the engine and was asking for help to make it run. One of his many comments was that the engine had poor compression. There was also another contributor (Ramoye) who was having the same problems.
So it appeared my drawings might have been correct. It was agreed that the plans were sketchy at best and unfortunately both of the Gentlemen had been through making a lot of new parts with limited success.
I decided that maybe I should analyze the model more closely to see what changes could be made to the mechanism to correct at least the compression issue.
This was no easy feat because almost any change of only a few thou in positioning of parts can make a HUGE difference in the operation in the differential motion of the two opposed pistons. I went through a couple of weeks of change something / observe the motion / see if it looks reasonable / make measurements to see if compression improved and mostly to see if there were collisions of parts and if the engine was still able to be built.
In the end I think I've (more or less stumbled) come up with a configuration that MIGHT work. Unfortunately a lot of the changes are not suited to being made to an existing engine built from the original plans. Too many parts would have to be re-made.
BUT I can certainly build a new engine from my plans to see what happens.

***** WARNING - This project may be a complete failure - but I think I can trust my drawings to build it. Although my analysis of the operation of 3D model may end in disaster when it tries to run. *****

SO follow along on what might turn out to be yet another unsuccessful Atkinson Differential Engine build.
I'll get started in a couple of days.


Last edited:
I am really looking forward to seeing your results. It certainly is one of the most frustrating engines I have ever built. I have played around with a lot of changes and have not found anything which makes much difference. I have played around with my 2 D cad moving stuff around to simulate actual intake, compression, ignition but have not found anything which makes much difference. Occasionally I can get the engine to run for a short time but then it will not even try the next time even when nothing has changed in parts or adjustment.

Well..... We'll see what happens. It might be a complete failure. But I think I've gained enough knowledge of what affect various part positions have on the operation that maybe I can "adjust it" (with minimal part changes) to get it to be happy.
My fear right now is that on the power stroke the right piston moves down the cylinder quickly as it is supposed to but the left piston begins to follow too soon after. I guess that effectively makes one piston fight the other with the expanding pressure. It's not too bad but it's one of those things that remains to be tested. I think I can make "adjustments" to reduce that effect when it fails.
Another concern is that it may not have much "Suction" on the intake stroke. i.e no incentive to bring in any fuel. You need a good "suction" to make a carb vapourize the fuel. This is not unique to my changes. I don't think the original design had much either.
Obviously this is all a crap shoot. This may end up being a shelf queen.
I am thinking of you (and others) in that it may be possible to modify an existing engine with my changes. But at this point I need to test some of my assumptions.
Suction on intake has not been a problem in that when it is turning over there is fairly good suction at the carburetor throat. I also wondered about atomazation and tried a vapor carburetor but that does not make any difference. Obviously this engine was never going to power anything useful. It barely keeps itself running. I would certainly like to know what other YouTube posters did to make the engine run.
It's good to know that suction won't be a problem. In that respect I also wondered how much time it had to take in a charge with the small hole into the cylinder they specified. That's why my plan is to make the hole into a slot to take advantage of what my simulation shows is the whole portion of the intake stroke where it appears the pistons are moving apart from each other (creating vacuum) and cut the port off before they start to move in unison to the left.
I tried a vapour carb on my other Gingery Atkinson engine when it stopped running reliably for some reason. In the end it was an ignition problem (you know the saying "most fuel problems are ignition" and vise versa).
Anyway I found the vapour carb to be just as much trouble to get the fuel air mix correct. Although they don't have the problem with raw fuel being ingested.
I use propane on my Parcell and Weed engine but even with it's special mixer intake valve it's a chore to get it just right to fire. There is a VERY narrow range of mixture where propane is combustible.
So I'm glad I'll be able to stick with a carb which is pretty adjustable.
Time will tell.
Being able to power anything requires horsepower and that's mostly dependent on compression. I'm hoping my changes will boost the compression without adverse effects.
The other Atkinson engine was a pretty sketchy design as well. Mine has very low compression. It runs and will run fast but it doesn't have much power.
Last edited:
I decided to start with the flywheel. When I made the Atkinson Cycle engine I made a pattern and my friend and I cast a flywheel for it from aluminum. The first attempt had a lot of casting flash so we made a second one that turned out much better. Luckily I kept the first one and was able to use it for this engine (they are the same - close enough anyway). I had experimented with powder coating this flywheel back when it was first made because I'd heard that powder coating cast material can be an issue. It was true. If you just follow the normal procedure and put it in the oven cold, when it heats up to the typical 400deg or so the gasses in the casting come out an bubble up the curing coating. It leaves a rice crispy like finish of bubbles. The trick is to heat the casting first above the normal baking temperature and coat the part while it's hot and put it back in the oven.
Apparently this works - I have yet to try it - but if you've ever tried removing power coating after it has cured it's very tough. So I figured it best to just use conventional paint this time.
I first turned the rough (and powder coated) casting to true it up and to remove the powder coat. A lot of hand work was required to remove the casting flash that was mostly around the spokes. I blasted it with ground glass and applied a bit of body work to make it pretty, primed it with etching primer and painted it.
In the picture the wheel is hanging inside of my 3x3x3ft paint booth from a rotisserie that turns the part slowly. This not so much a proper paint booth it's more of an exhaust booth with a reclaimed furnace fan ducted outside. It does a really good job at removing paint fumes, anodizing fumes and powder coat dust (or any other smelly stuff) from the shop but it also sucks a lot of air out of the house along with dust from the shop. So a painted part can sometimes end up with a bit of dust in the finish. I just add an extra coat of paint and if it's a problem I can buff the finish after.
The flywheel is held into the crank shaft with a collet. Although a lot more work than a key and set screw they work well. The flywheel needs to have a matching taper turned into the hub.




Last edited:
Most of the engine framework and moving components can be made from 1/4" extruded aluminum so I bought standard 6061 extruded material that comes in a range of widths. In order to simplify the order I purchased a 6ft long piece of 1/4" thick x 6" wide material. What I didn't expect was that a plate that width would be cupped. Not a lot - maybe 15thou across the width (not as bad as it looks in the picture) but it was not acceptable when a couple of pieces had to be stacked such as the two tier base plate. So I had to cut each stock piece and put it in my hydraulic press to take the cup out of it. Once each piece was "flattened" to an acceptable amount it was machined to size. In the original engine the base plate is a single casting 1/2" thick. It was easy enough to get the tiered shape by stacking two 1/4" pieces rather than machining the step into a 1/2" piece.
The other issue was that the material was only nominally 1/4" thick but measured .260" . Not a big deal except perhaps when things start to stack up like in the front of the engine where the oscillating arms are stacked on top of the links etc. etc. The spacers will need to be adjusted else it will affect the position of the cylinder. And so it goes. I'll have to adjust my drawings.
I started making pieces from the bottom of the engine up. The first picture is of milling the four feet to go under the engine.
Then I proceeded up to the base plate. This is where I noticed the finished plate was about 10thou larger in both X and Y than it was programmed to be. I'm using a mill-drill converted to CNC. I guess I beat it up quite a bit making the George Britnell V-twin (actually two of them). I found the gibs on both axis loose allowing the table to shift around. I also found the quill had quite a bit of side play. I have an adjustment to fix that. So I spent a half day playing around getting the machine back in shape. I had to trim the previously made base plate to the correct dimensions.
After the "tune up" some test pieces were coming out right on size. In reality there isn't much on the engine where the outside dimension really matter since the original design is made from castings which can be notoriously off dimension. Hole placement is the only critical aspect of the build and they are now right on.
Sorry nothing of great interest to show so far. I'll be making plates for a bit.



Last edited:
Nothing too exciting today either. I just turned out some more plates. The side panels, like the rest were originally castings. They would have to be made from 5/8" thick material. Instead I used my 1/4" plate and some 3/8 x 3/8 (which I had to machine down from 3/8 x 3/4) for the corner supports. I purposely made them a bit long so they could be fastened to the side panels with screws and then trimmed to the height of the side panels.




Things are starting to come together. I assembled the base plate and side panels. You might note the one upright has drilled and tapped holes ready to accept the front plate. I forgot to do the others but that turned out to be fortunate because it was a bit presumptuous of me to think that the panels were all going to fit nice and flush at the corners given material and machining tolerances. After making the front and back plates I clamped them in place and used a transfer punch to mark the proper place for the holes in the uprights.
The front and back plates is where I'm committed to my design. One of the major changes was to move the left oscillating arm pivot point down from the standard location. I found this changed the geometry of the movement to increase the compression significantly.
One of the tricky parts of CNC machining is holding the material. You have to come up with a secure fastening method that also (hopefully) allows you to do all the machining in one setup. It's especially tricky when you have a crappy mill drill machine like me because without the ability to raise the knee you have to get the height of the head correct to accept all of the tools used to complete the job. Raising the head during the job is not accurate. I have an assortment of different length collets, drill chucks and cut off reamers so I can drill and mill without losing my reference point.
The easiest way to mill on-the-flat is to screw the stock down to a piece of MDF (which is surprisingly flat).
The job required reaming the holes for the shafts so in the picture you'll see I've pre-milled pockets into the board so the reamers can pass fully through the stock.
The original casting had bosses for the shafts to pass through. I turned some bosses separately and glued them with JB weld to the surface of the plate. I also machined the water jacket mounting boss plate and glued it in place. There is no stress on these parts. They are essentially just spacers so the JB weld should be sufficient.
I extracted the locations for these pieces from the machining code and wrote a program to cycle around to the locations and stop at each. The bosses started with only small holes in the center so I could use a fine scribe pointer to get the boss centered over the proper location and press it down into the JB weld.
You'll see in the picture that both the front and back plates are stacked so the holes in both plates are perfectly aligned. The first step was to drill all the holes in the front panel and to ream the shaft holes. Some of those holes are used to hold the panel itself to the board because milling the profile will release it from the stock plate. This is another tricky part of CNC because you have to hold the stock and also hold the finished piece without having clamps in the way of the profiling process. (I avoid tabs for various reasons).
Yes I forgot the slot in the front of the water jacket plate before the picture. But that has to be done only to the front plate so I fastened it down to the board separately and finished that step. I didn't take a picture. I also made a couple of the shafts to stick through in order to get the front and back aligned before fastening them to the side panels.
SO it's finally coming together.






Last edited:
I pressed in some bushings for the crankshaft and added the shaft and crank disk.
The bushings are 660 bearing bronze that even with a brand new reamer is resistant to reaming on size. So a bit of honing with an expanding lap took the last tenth or so out of the bore and surprisingly the shaft went right through the holes in the front and back plates. I was expecting that because the two plates were not perfectly flat (why mentioned previously) that when they were separated after being clamped for machining that the plate would spring back and the holes may not be aligned, perhaps tilted forward or back or side to side. Luck was with me.

Thanks. A few more pieces to make and then I get to the water jacket placement, cylinder, pistons etc and work will slow down. I'll have to make and check placement of almost everything to be sure it's turning out according to my drawings. Small changes in the placement of those items appears to throw the whole mechanism out of whack.
Maybe like the original Hubble telescope mirror it will be "perfectly wrong".
The oscillating arms are a prime example of why I always re-draw plans that I plan to build. On page 36 where you are creating the castings for the arms they give dimensions for the patterns. On page 81 they give dimensions for where the holes should be drilled in the castings. Unfortunately the hole in the tip of the oscillating arm turns out not to be concentric with the radius of the end. I just doesn't look "nice" to have the hole off center. Assuming that the placement of the holes is important to the geometry of the movement I kept those dimensions and adjusted the end radius to be concentric. Of course this throws off the general shape of the arm slightly because the curves joining the various bosses are no longer arcs as laid out on page 36. On paper one could (I suppose) just join the features with a French curve but that would make it more difficult to machine if you're using a rotary table. I kept the curves close to the original shape so as not to introduce interference between the arm and the cylinder (for instance). Since I'm CNC milling it, it makes no difference the shape. If producing the parts manually from my drawings, simple arcs and a bit of hand work could reproduce the shape I arrived at.
As with the front plate machining the machining process was to locate the position of the bosses on the wooden plate and mill pockets into the board so the reamer could pass through the plate.. The bosses were glued to the plate and the holes were drilled and reamed. Then those holes could were used to hold the arm to the board when it is release from the plate upon machining. Since the bosses were glued to the plate they needed to be faced off after to be the correct height above the plate since they are effectively spacers standing the arms off the front plate.

There are only a couple of more simple pieces to produce and then the work is going to slow down as I make the water jacket, cylinder, pistons, connecting rods etc. and position them accurately according to plan. Most of my changes in the geometry were made in these remaining pieces and their positions are critical.









  • upload_2020-3-2_7-14-28.jpeg
    105 KB · Views: 273
Last edited:
You are making amazing progress in a very short time. FWIW I made the bosses with a shoulder to fit into the main part and attached them with four 4-40 socket head screws Worked well without depending on Loctite. Example: 1.25 dia boss w/ 1/2" bore and 5/8 dia shoulder to fit a 5/8" hole in the main part with 1" bolt circle.

The problem with the hole location in the oscillating arm was one of the first problem I ran into also. I am amazed that of the many folks who have made this engine I could not find any mention of this or several other discrepancies I found.
I guess maybe I should / could have taken that approach with the bosses. My thinking was that there should be no stress on them. They are more or less just spacers and the torque transfers through the arm itself. Then again I'm not a structural engineer. I guess if they break loose I can screw them in place (even from the front)
I'm somewhat in a "let's just build it and see what happens mode" because I'm thinking I shouldn't spend a lot of time on what might become a shelf queen.
You might see that my plates are all held in place with flat head / counter sink screws into non- countersunk holes. This helps locate the plates exactly where they need to be because the plates can't shift around under the screw head. I hate it when there are so many pieces that can be adjusted and you are trying to get reamed holes to line up.
I plan to replace those with socket head cap screws one at a time when everything is running smoothly.

I have a question for you: I didn't read the book on this. I've given up on the book and it's sketchy suggestions.
What did you do to seal the water hopper? I'm pretty sure I don't want to take it all apart again to put sealer between the plates. I 'm also concerned about painting it for the same reason. I'm paranoid that taking it apart and adding sealer and paint will shift stuff enough that it will be a nightmare to get things aligned again.
I'm thinking I'll just put clear silicon on the seams from the inside, not paint the inside (can't paint silicon) and then just paint the outside. OR maybe I'll just bead blast the whole thing and leave it plain??
Usually I anodize my aluminum pieces but it's VERY time consuming, my tank is too small for the larger pieces and I'd have to take it apart.
Suggestions welcome.

I didn't think about (and I didn't analyze) the position of the hole in the tip of the arm and it's affect on the operation. I don't think it would matter much. I was more concerned with the looks of it. The books whole avoidance of the issue and not mentioning it is what annoyed me. It brings doubt to the drawings at a time when you are trying to make things right.
I presume it was because it was easier to make the patterns using basic round curves.

Thanks for your interest.
I did not seal it at all. I used a different set of drawings and the cylinder is made from a solid piece of cast iron which dissipates the heat better. Ray Moye made his using the original Gingery and I don't think that he used any water either. He had his running for several runs of 30 minutes. I can send you my 2d drawings if you want to see what the differences are. Pivot points are different and the arms are quite a bit different.