I am building this device.
I Have decided to use a name for it that I used for another test bed, but changed it a little.
Simple Circumferential Rotary Entrained Wave Device, SCREWD.
This is a pic of the setup that is described in this pdf.
I have been playing with three different AIs with this and this is what Clauede.ai wrote up after we talk about it.
Feel free to share this anywhere you wish.
Chat gpt can push back on things for a long time,, nitpicking over everything,,
Anyway Chat came to make the same observation and I asked Chat to write it up using its own words and this is one of the presentations that was created, I have been promised a much better one tomorrow when I have more time.
Claude wrote this up before invoking the Law of Conservation.
shouldn't it work? ;)
you know what is not so easy,,, trying to get the AIs to be able to use a single protocol with device and not get derailed. I thought that I would just pop in to Chatgpt with the list from Claude, nope, chat started injecting the lab frame and all sorts of other considerations, why you should ask,, because it thought that was the way to go even tho I told it no other considerations, well when Chat got on board we discussed why this had to happen and what must I provide to ensure that it does not happen again,, well a long list of stuff came from Chat.
Hours spent trying to find the correct words, so Claude thought the Chat thing was good, but then broke that protocol any way,, or they would loose the geometry and say I'm sorry, or good for pushing back on that. I will try again today to get some tight protocol in place with a matching description so that any AI or analyst will follow the simple requirements, do the geometry just on the shapes and bodies within the system, do the kinematics for only the system as it has been presented, do the force flows, no other considerations should be used before these have been done. no other reference frames should be used and no changing of actual dimensions or points.
This is what Chat said to use
Analysing geometry first followed by forces is a good approach to solving these.
I am missing force diagrams and torque diagrams in these analyses, though.
I will work on that for you,,
In the meantime I have got this write up from Chatgtp,
I also will post these
Quote from: Verpies on 2026.05.01, 11:42:44
Analysing geometry first followed by forces is a good approach to solving these.
I am missing force diagrams and torque diagrams in these analyses, though.
This just a little fun, but change the file extension to html and it will run in your browser.
this is weird watching it, the blue arrow I think is following a 5/6 ratio, so many of these showing different things but wrong,
You may not notice that the blue inner gear is actually moving, place your mouse pointer on a spot that the blue ring passes, then you can see that it is orbiting.
If this actually works in the real world I am stuck asking why we are not already using this relationship, it is so simple, why was I not taught this in school?
So the 3 AIs agree on the geometry, they agree on the kinematics, but the energy created big hurdles so the write up from Chatgtp did not go into energy. Gemini was excited about the force multiplication to the pivot, that is no big deal because the pivot does not move, but it showed that the full path from the ring/sun mesh to pivot is a real path and that the frame is absorbing a large amount of force.
With the sim that is the .txt file that you change to html, (read the file first if you don't feel safe with doing that), It shows all the motions very well, the input from the cross slide assembly orbits the blue gear and provides some force to that orbit that is then transferred into the ring gear, there can be no argument there from any person or AI, That ring gear then must push back on the force that the sun is trying to apply, Newton. Kinematics then tell the path and velocities and all that that the geometry creates by virtue of the hard bodies existing and in intimate contact, the system is a whole closed system. That means that the ring gear must rotate and that the arm must follow,, and then you get the full force flow, input from sun to ring, ring through the pin that is fixed to the arm and that the ring must rotate around that pin, but for the ring to spin and supply a counter force to the sun mesh it must apply that into the pin, Newton, and then the arm has a reactionary force that then pushes on the pivot. All of this passes the scrutiny of the 3 AIs WHEN they are held to the testing protocol as outlined by Chatgtp and me, and Claude, and Gemini,, my idea there feedback and push push push back.
The energy for the 1N at the sun/ring mesh is 0.0628J per cycle, again this part has no push back, it is when then you look at the ring to pin exchange, that energy is 0.628J, no typo that is 10x more out. Now the thing is they all agree on the force from the ring gear, and they all agree that the ring must rotate and they all can identify all of the other forces, but that one creates problems.
The stepper motors I chose showed the same current whether they were loaded or not, so I ordered some little brushed motors and I will have to redo the whole shebang.
for those that like to see things here is the test bed. I was in the process of trying to get the steppers working nicely when I realized that they used all of what they were going to use.
stepper 28byj-48
board uln2003
atmega mini controller
this is the build I am now working on.
It will use 10 N20 motors mounted in the frame with 10 10mm arms with bearings to drive the Sun gear.
I am setting it up so that I can drive 6 TT motors at the ring gear and am using a gear for the pin connection to drive the TT's at a 3:1 step up.
These two pics show 2 solutions from a change of one variable. Now what this means is that the system I am building could be set up so that it looses some of the mechanical gain to shift away from a singularity and into a more stable run condition.
The mechanical Engineers or the Physicists should understand that,, no arrows needed.
big design error on my part,, trying to "improve" things I got rid of an important relationship,,,,
I will change that up and see what happens.
I have the 10 N20's all installed and wired up to a power rail and they all spin along together, so that part is done.
changes/issues/life
Here is a pic of the mesh I am now running with, this is an offset of 180 degrees
These are some data points I have collected during some shake down runs with my modified system.
I am not claiming anything other than I find the results interesting, they do not prove anything.
The no load runs were with the system complete and all parts moving.
The holding hands runs were with only the Sun gear installed.
The loaded runs were with an open TT motor using the 3:1 gear but the system has an initial step down of 5/6 then the 3:1 step up, or 2.5:1
Holding Hands
2V @ 0.41A to 0.47A
6V @ 0.58A to 0.65A
No load system run costs
2V @ 0.44A to 0.54A
5V @ 0.53A to 0.65A
6V @ 0.61A to 0.71A
Loaded system
2V @ 0.46A to 0.56A
6V @ 0.75A to 0.91A
Direct out
2V @ 0.58A to 0.68A
5V @ 0.59A to 0.72A
6V @ 0.59A to 0.72A
What I find interesting is that the 2V geared up RPM delivered to the TT motor is about the same RPM as the 5V direct out.
so I did a few runs loaded with a resistor across the motor and a little volt meter and diode bridge reading the voltage. I have 2 100 ohm resistors in parallel for 50 ohms resistance.
using the 3:1 aka 2.5:1 gear
Input 3.5V @ 0.84A to 0.94A
output across 2 100 ohm resistors in parallel
0.52V to 0.63V
using the direct drive
input 7.5V @ 0.72A to 0.93A
output across 2 100 ohm resistors in parallel
0.54V to 0.84V
I turned up the input voltage to get close output voltages, the same motor with the electronics on it was used on both test points.
Quote from: chief kolbacict on 2026.04.30, 16:03:46
shouldn't it work? ;)
It won't work. The behaviour of lossless mechanical systems can be described by a Lagrangian; therefore, regardless of the complexity of the mechanical system – and this one is clever but tricky – since the system is closed and evolves over a single cycle between identical initial and final states, it is invariant under time translation, and so energy is conserved.
Archimedes' force is equal in magnitude to the weight of the displaced fluid and is supposed to be vertical, but not in the system you describe, where centrifugal acceleration must be taken into account. The 'weight' of the displaced water, and therefore Archimedes' buoyancy, is no longer vertical but always directed towards the surface.
Quote from: F6FLT on 2026.05.19, 14:55:51
It won't work. The behaviour of lossless mechanical systems can be described by a Lagrangian; therefore, regardless of the complexity of the mechanical system – and this one is clever but tricky – since the system is closed and evolves over a single cycle between identical initial and final states, it is invariant under time translation, and so energy is conserved.
Archimedes' force is equal in magnitude to the weight of the displaced fluid and is supposed to be vertical, but not in the system you describe, where centrifugal acceleration must be taken into account. The 'weight' of the displaced water, and therefore Archimedes' buoyancy, is no longer vertical but always directed towards the surface.
Wrong thread??
The freecad drawing is my compenstor device, it sits in the middle of the sun gear and pivots on the shaft, in the test bed I only used 2 gears and 4 bearings, the bearings are located under the gear locations and run against a shoulder on the inside of the sun gear. This was in an attempt to figure out why the apparent gain was there and what the actual force paths and flows are.
The one part of the runs I did with this setup that is most interesting is where the draw drops when I use the ring and its 3 gears,, adding more frictional stuff lowered the draw, that is interesting.
Quote from: webby1 on 2026.06.04, 04:11:11
The freecad drawing is my compenstor device, it sits in the middle of the sun gear and pivots on the shaft, in the test bed I only used 2 gears and 4 bearings, the bearings are located under the gear locations and run against a shoulder on the inside of the sun gear. This was in an attempt to figure out why the apparent gain was there and what the actual force paths and flows are.
The one part of the runs I did with this setup that is most interesting is where the draw drops when I use the ring and its 3 gears,, adding more frictional stuff lowered the draw, that is interesting.
Hey webby1
I also use freecad, I suspect your using the gears workbench. What other workbenches do you recommend.?. I have a project similar to this I want to build.
That's an impressive build, nice job.
I use part and part design all the time,, Sketcher for more than just creating a new shape, I use it as a chalk board kind of thing.
I am using an older version because I find it easier for what I do, version 0.21.1
this is my mechanical build as it is evolving.
I needed to get it to lock every time without any other parts, and this one does just that.
Very nice build but does it manifest any anomalous behavior ?
Well it can not be anomalous if it is predicted, right? So no it does not because all of the interaction have been calculated, or predicted, and I am just building to get as close to those predictions as I can.
Consistency without the electric motor generator guess work I think is the best way to move forward and so I have shifted to mechanical.
So this system will lock the mesh to arm, or sync the 2 together, with any offset values that total greater than 20mm, and as you should know, at 20mm exactly the toggle event is very unstable but as the system is moved away from that point of infinite force between the sun/ring/arm and pivot that force emerges into the real world and is what forces the arm into rotation and locking it to the mesh. That force has a cost in the real world and it boils down to strength, the stronger the parts against the massive force the closer the offset distance can be and the less force is used by that part, so more for the output.
I have 2 bearing adapters for the ring gear and a built in 10mm offset for the pivot bearing, so that gives me the choice of using either a 0, 5, 10, 15, or 20mm ring gear offset, I am going to re-do my sun arm-gears so they have a 5, 10, 15 and 20mm offset,, so then any of those can be used for testing.
Any person with a large format 3D printer could then take my files and adjust them for there printer and print out there own test bed to experiment with,, just add bearings and screws and filament.
Here is the system opened up so you can see how simple the actual system is.
the green is the input.
The gear-set is a 6/5 step down from input to output so the output arm should be 60mm for the input arm of 50mm for the system to balance in force.
The pic I have posted has the output rubber band at 70mm to have the same amount of stretch, between 40 and 42mm as the input,, resetting and retesting many times has given me the little window on the measurements but they are almost always within about 0.1mm of each other.
This then is a picture of a device in a static test state that is showing a mechanical gain for the operator.
I forgot to mention that that little gain is after all of the input costs have been paid for by the input,, so it cold be more,, it is a little sad that it is not yet at the 10:1 that the model predicts.
here is a link to a youtube video of Mr. Finger moving the input lever
https://youtube.com/shorts/2y2TRE8bvuo?feature=share
this new build is a deviation and I am calling it Maxwell's Engine,, the Demon sits in the gap.
This is an exploded view of the parts so far, more to design and print.
this is the inside of the TT motor reduction unit
the build so far
printing off parts that should be the final parts,, should be because so far I have messed up a few print changes,,,,,
An interesting observation is that the "pin" when only between 2 pivot bearings can twist out of alignment, meaning that the pin is not 90 degrees vertical to the pivot bearing face,, good thin I built in a gear sync system for that part.
With all of the test fits so far the system actually can spin very smoothly which, with all of these gears,, is a little surprising but very good.
I still need to print out some more spacers, cut 2 shafts shorter, build the electronics package, build the motor holder and the motor coupler.
It is coming along and so far it all spins fairly well.
Interesting - and clearly a lot of work!
Best wishes to your progress.
a look inside while I have the 80T gear and the TT motors removed,, a spin up test using a yellow body TT motor was interesting,, it took less to spin using the 2:1 step up and spinning the TT motors than it did to just spin the system 1:1,, using that 2:1 step up does reduce the the upper and lower gears rate of rotation,, but the TT motors are a little drag forward you know,,
these are only preliminary tests trying to set a base line. On this run the Blue TT motors had a 4:3 step down ratio to the input drive gear, I am printing new gears to bring that up to a 1:1
Configuration / Setup,Input Voltage,Total Input Current (Iin),Net Current over Motor Tare,Generated Output / Status,Bench Notes
Yellow Motor Alone (Tare),6 V,0.11 A,0.00 A,N/A,Baseline internal motor + gearbox friction.
Yellow + Bare Frame Stack,6 V,0.22 A,0.11 A,N/A,Frame bearings (8 mm shafts) & structure add just 0.11 A drag (no 80T gear).
1 Blue TT Motor (Direct Drive),6 V,0.35 A,0.24 A,5.0 V Open Circuit,Smooth single-branch baseline; yellow motor stays cool.
2 Blue TT Motors (Direct Drive),6 V,0.54 A,0.43 A,—,Linear load scaling up to 2 direct-driven gearheads.
3 Blue TT Motors (Direct Drive),6 V,0.70 A→0.95 A,>0.59 A,Stall / Thermal Limit,"Torque limit exceeded; armature back-EMF collapses, cooking motor windings."
3 Blue TT Motors (In Planetary Loop),6 V,0.31 A−0.34 A,0.20 A−0.23 A,Feedback Loop Active,Key Result: Loop force cancellation drives all 3 motors for less current than 2 direct-driven motors!
1 Blue TT Motor (Under Peak Short-Circuit Load),6 V,0.78 A,0.67 A,0.110 A−0.120 A Peak,Measured across DMM 200 mA range (∼0.5Ω shunt load). Heavy magnetic back-torque.
the yellow motor was not going to cut it so I changed it out to a Blue all metal gear TT motor,, then I got carried away running some tests.
The pdf explains all of that fairly well.
It is nice when the math matches the bench,, lower draw.
there seems to be an error on the calculated TT motor turn ratio,, so those conclusions are wrong. Sorry for the confusion.
The printer I used for the frame sections had an issue that I have now resolved. I reprinted those parts and also made the dual ring gear, or hybrid ring gear, a little stronger so it is less prone to deforming under load. I also found two bad pivot bearings and replaced them.
As tested:
28 bearings
17 gears
Blue motor alone: 0.07 A
1:1 direct head-to-head, 1 Blue driving 3 Yellow:
0.48–0.50 A
2:1 system, no load motors:
0.16–0.23 A
1:1 system, no load motors:
0.18–0.24 A
2:1 head-to-head through the system, 1 Blue driving 3 Yellow:
0.45–0.54 A
All tests were run at the same 6 V power-supply setting and allowed to run until reasonably stable.
These are the measured numbers from this build. They are not being presented as evidence of gain; they are simply the results from the current configuration.
something that has been bother me with this is that the math only shows 1/2 of the input coming back to assist and I did not see why that would be, well I finally figured that out and it is so silly that it is embarrassing that I missed it.
There are 3 systems lets say,, there is the upper input section, the slip and then the lower return section. The upper and lower share a physical connection on one side and a force connection on the other and it is that force only connection where the slip is.
Now if the input goes in on the upper and it spins the upper and then the upper goes across the slip and spins the lower and then the lower comes back to assist the upper,, the simple thing is that the slip boundary has the upper traveling at say 100 RPM and with the 2:1 return gearing the lower is traveling at 50 RPM, so the same force but at half the rate means half the return work,,, and that is why there is only 1/2 of the input returned to assist.
a small side track
Concept test bed, not an optimized machine. Same hardware, same 12 V Yellow TT drive, same NEMA23, and two 50 Ω coil loads. Only the return/sun gear ratio was changed between runs. The table shows measured current/RPM data plus calculated carrier/slip values. Estimated coil voltage/power values are based on separate NEMA bench calibration and are not direct in-system measurements.
90/30 open RPM values in the table are the corrected rerun, 263/792, not the original bad tach reading.
no one will get this but Chat Gpt came up with a funny,, Not-Links,, if you only new what that was about you wold find it funny,,,, :)
https://magmovenergy.com/magmov3000/
https://www.accio.com/business/magmov-generator-review
https://www.youtube.com/watch?v=Z9aY6wsVfDU
Quote from: panyuming on 2026.09.18, 22:09:14[Link1]
[Link2]
[Link3]
Do not make posts with links only.
Quote from: panyuming on 2026.09.18, 22:09:14https://magmovenergy.com/magmov3000/
https://www.accio.com/business/magmov-generator-review
https://www.youtube.com/watch?v=Z9aY6wsVfDU
If you can provide this device for us to test, we will review it.
If thats not possible then we will need to delete it from this thread, if you can provide the device for review we will move to a separate thread.