Project: Melissa Hands - Perhaps a little easier..
by PaulL
Sat Mar 06, 2010 11:22 pm
It's stronger when pulling, weaker when pushing, so when on the inside of the finger opposing the knuckles, you get more strength when pulling, when the hand closes. Extending, from what I've done, is more weak. So, I doubt he'll be able to do pushups on his fingertips. Knuckles, definitely, though. 
Just for kicks, I did use a thin coated wire (like used for small pulleys, like I am thinking for head servo) on a finger through my guides on both sides, to do a "Pull / Pull" on opposite sides of the finger to test, and it was clunky, knuckles would randomly snap to position either way. Semi-flexible, like teflon or nylon is the only way to make the movement look right.
OR, some particularly wealthy soul could try something like this: http://www.newscaletech.com/ and control each knuckle independently... Heck, you could even use those little ones for facial expressions or some such (moveable face sections like transformers, anyone?). But, they're way too pricey for me at the moment.
Just for kicks, I did use a thin coated wire (like used for small pulleys, like I am thinking for head servo) on a finger through my guides on both sides, to do a "Pull / Pull" on opposite sides of the finger to test, and it was clunky, knuckles would randomly snap to position either way. Semi-flexible, like teflon or nylon is the only way to make the movement look right.
OR, some particularly wealthy soul could try something like this: http://www.newscaletech.com/ and control each knuckle independently... Heck, you could even use those little ones for facial expressions or some such (moveable face sections like transformers, anyone?). But, they're way too pricey for me at the moment.
by mog123
Sat Mar 06, 2010 11:48 pm
I still can't understand how this works 
If you're pulling the teflon inside the finger, you're shortening the effective part of it(the one in the finger "skeleton") so the hand opens. If you're pushing the teflon inside you make it squeeze in there, in order for room it needs to bend. Is that right?
If you're pulling the teflon inside the finger, you're shortening the effective part of it(the one in the finger "skeleton") so the hand opens. If you're pushing the teflon inside you make it squeeze in there, in order for room it needs to bend. Is that right?
by PaulL
Sat Mar 06, 2010 11:59 pm
If you look in the 4th picture I posted at the top of this thread, you will see a straight line of white down at the bottom of the knuckle on the closed hand. Pulling the teflon strip mounted at the finger tip relative to the finger draws the joints of the knuckle together as the teflon finds the shortest point between the mounted end and the pulling end, into a circle, shorter than the straight line of the strip with the finger extended, and consistently so, being held to the inside side of the knuckle in the finger by the gray retaining blocks. My english teacher would have a fit over that last sentence I wrote, I think.
Servo-Per-Finger Status Update...
by PaulL
Sat Mar 13, 2010 4:09 am
I received two Pololu Serial Servo Controllers today, one for each hand. Each of these controllers can control up to 8 servos, but 3 won't be getting used.
I could have gone with their USB servo controller, but went with this "kit" version of the serial controller that doesn't have the headers soldered on. This board is VERY flat. I will be putting this board underneath the aluminum on the back of the hand, with a couple holes drilled for the LED's on the board to show though (I think that will have a nice effect). In not using all of the servo connections, one of the LED's will illuminate when an "out of range" value is sent to it, so in short, I could send an invalid range to an unused servo channel to illuminate an LED on the back of his hand. Should be a neat feature, can flash a yellow LED at will on the back of the hand, and doesn't take anything more than a drilled hole and some code. There are also LED's for Data Receive and Error, those will get holes such that they also are visible from the back of the hand.
The USB version of the board is thicker, and I didn't want to use up any USB ports since I will be connecting this to an RB-100 that has plenty of serial ports, including TTL-level serial ports.
These boards are addressable, so I can tie both to the same TTL serial port and control them both through one COM port.
I am still waiting on my HobbyKing order for servo testing, and unless the cheaper of the two servos really impresses me, I'm already leaning towards the more expensive, metal gear servo that is slightly smaller with more than twice the torque.
I will have to give some thought to servo arrangement. The servos I'm looking at are 10mm thick, meaning stacked side-by-side, that's nearly 2 inches of width for 5, and I feel this might compromise the look a bit with my hand laid out as it is. I will need to experiment to determine what servo I will ultimately use, but a 2 inch wide block of servos at the hand doesn't seem very appealing right now. They're short, and not deep, but the width is more than I'd like.
The servos I want to test should arrive soon (status says they're in the states), and hopefully I can make a quick educated decision and get a pile of these in my hands to fiddle with configuration, and what will work out best.
I just can't give up on the idea of individual finger movement! Just seems like such a cool capability. One way or another, I will have separately controlled fingers, and I want to make sure the servos don't distract from the appearance.
Will post back when I have something to show.
Paul
I could have gone with their USB servo controller, but went with this "kit" version of the serial controller that doesn't have the headers soldered on. This board is VERY flat. I will be putting this board underneath the aluminum on the back of the hand, with a couple holes drilled for the LED's on the board to show though (I think that will have a nice effect). In not using all of the servo connections, one of the LED's will illuminate when an "out of range" value is sent to it, so in short, I could send an invalid range to an unused servo channel to illuminate an LED on the back of his hand. Should be a neat feature, can flash a yellow LED at will on the back of the hand, and doesn't take anything more than a drilled hole and some code. There are also LED's for Data Receive and Error, those will get holes such that they also are visible from the back of the hand.
The USB version of the board is thicker, and I didn't want to use up any USB ports since I will be connecting this to an RB-100 that has plenty of serial ports, including TTL-level serial ports.
These boards are addressable, so I can tie both to the same TTL serial port and control them both through one COM port.
I am still waiting on my HobbyKing order for servo testing, and unless the cheaper of the two servos really impresses me, I'm already leaning towards the more expensive, metal gear servo that is slightly smaller with more than twice the torque.
I will have to give some thought to servo arrangement. The servos I'm looking at are 10mm thick, meaning stacked side-by-side, that's nearly 2 inches of width for 5, and I feel this might compromise the look a bit with my hand laid out as it is. I will need to experiment to determine what servo I will ultimately use, but a 2 inch wide block of servos at the hand doesn't seem very appealing right now. They're short, and not deep, but the width is more than I'd like.
The servos I want to test should arrive soon (status says they're in the states), and hopefully I can make a quick educated decision and get a pile of these in my hands to fiddle with configuration, and what will work out best.
I just can't give up on the idea of individual finger movement! Just seems like such a cool capability. One way or another, I will have separately controlled fingers, and I want to make sure the servos don't distract from the appearance.
Will post back when I have something to show.
Paul
by PaulL
Sat Mar 13, 2010 4:45 am
When? Hmmm....
I am waiting on two servos to see which will work best, they have not yet arrived, should be in the next few days.
Once I make a decision on servos, I will place an order for 10 and maybe a spare or two.
Once the "big order" of servos arrives, I will experiment with how to arrange the servos to get the desired result and direct the creation of the final version of the hand / servo bracket assembly, keeping in mind an added Hitec Robotic Servo (HSR-5498SG or HSR-8498HB) for wrist rotation. If I can find a way to shave down the cases without compromising the servos, to make them thinner, I'll do that, too.
After I have a bracket assembly and servo arrangement figured out, along with the wrist rotation bracketry, I'll build a prototype or two to get the dimensions and look right, then I will fab the final parts, likely making 3 sets, one to put together ASAP to play with, and two to send out somewhere to have anodized in black (I'd say blue like the bracket set I have, but I don't know if I can get the color matched perfectly). I'll post some video of the hand once I have it together, moving.
While I'm waiting for things to arrive, I will likely hook up some servos I have lying around and get the code to run the Pololu boards setup and operational. It's serial stuff, it shouldn't take long.
I can't even guess right now how long this will all take. But, I'll post updates as I go. 
Hmmm....
I am waiting on two servos to see which will work best, they have not yet arrived, should be in the next few days.
Once I make a decision on servos, I will place an order for 10 and maybe a spare or two.
Once the "big order" of servos arrives, I will experiment with how to arrange the servos to get the desired result and direct the creation of the final version of the hand / servo bracket assembly, keeping in mind an added Hitec Robotic Servo (HSR-5498SG or HSR-8498HB) for wrist rotation. If I can find a way to shave down the cases without compromising the servos, to make them thinner, I'll do that, too.
After I have a bracket assembly and servo arrangement figured out, along with the wrist rotation bracketry, I'll build a prototype or two to get the dimensions and look right, then I will fab the final parts, likely making 3 sets, one to put together ASAP to play with, and two to send out somewhere to have anodized in black (I'd say blue like the bracket set I have, but I don't know if I can get the color matched perfectly). I'll post some video of the hand once I have it together, moving.
While I'm waiting for things to arrive, I will likely hook up some servos I have lying around and get the code to run the Pololu boards setup and operational. It's serial stuff, it shouldn't take long.
I can't even guess right now how long this will all take.
But, I'll post updates as I go.
by PaulL
Wed Mar 17, 2010 11:21 am
I made a decision on the servos, and am going with the MKS DS450's. The other one is, by the numbers, half as much torque. It's also a bit larger, and a little too "unusual" for easy mounting. The MKS servo seems like the best candidate, and I've already ordered 10, should arrive by Friday. I have the one from testing so I'll have 11 total, 1 as a backup for now.
I also have a preliminary design sketched on paper, I'll mock that up in poster board to see how it will work in aluminum. I've already started on the controller code for the Pololu serial servo controller, that won't take much to finish.
To connect the servo to the finger, I'm going to machine a piece of teflon such that it has a mounting hub for the servo, with the strip as before to run inside the finger. I've already tested the teflon twisted as it will go from the servo spline to the finger, and actually twisting it seems to alleviate the need to support it.
I also have a preliminary design sketched on paper, I'll mock that up in poster board to see how it will work in aluminum. I've already started on the controller code for the Pololu serial servo controller, that won't take much to finish.
To connect the servo to the finger, I'm going to machine a piece of teflon such that it has a mounting hub for the servo, with the strip as before to run inside the finger. I've already tested the teflon twisted as it will go from the servo spline to the finger, and actually twisting it seems to alleviate the need to support it.
by PaulL
Sat Apr 17, 2010 12:40 am
Short update. I received the servos, and had potentiometer issues with 4 out of 10 total. I have repaired one so far by adding thin flexible wire between the pot and the board to isolate pressure on the board from the servo leads (not easy work). I have moved all the rubber gromets away from the servo casings such that I can come out of the servo and head up along the side of a servo without the rubber "boot" causing problems. These micro servos are very strong, and should work more than OK, and likely have the power to snap the plastic fingers if I'm not careful.
I have started to try to mock up the design in Rhino3D, but it takes time to learn a CAD package, and with the demo having only 25 saves, I'm not sure if this is the best app to use to design these. I have made a number of decisions for how the pieces need to go together (and have a number of pieces of paper illustrating what I'm looking for), but I need to put it together in CAD so R&D will be faster for slight adjustments I'm expecting to have to make to the design with such tight tollerances to keep the micro servos from dominating the hand design.
I have the components, I have the materials, and I have a basic design, but now I need to do some CAD work and finish the design, then transform that into NC code so I can run my mill with it.
I have started to try to mock up the design in Rhino3D, but it takes time to learn a CAD package, and with the demo having only 25 saves, I'm not sure if this is the best app to use to design these. I have made a number of decisions for how the pieces need to go together (and have a number of pieces of paper illustrating what I'm looking for), but I need to put it together in CAD so R&D will be faster for slight adjustments I'm expecting to have to make to the design with such tight tollerances to keep the micro servos from dominating the hand design.
I have the components, I have the materials, and I have a basic design, but now I need to do some CAD work and finish the design, then transform that into NC code so I can run my mill with it.
Hands Update...
by PaulL
Sun Dec 19, 2010 2:39 pm
Wow, I hadn't realized it's been since APRIL that I've updated this thread!!!
The past several months sure went by quick!
I have finished up the hand bracket design, it's about 90% done. I have the Pololu serial board code ready to test. I have several sheets of .040 inch thick 6061-T6 aluminum. I have 10 DS450 servos to control the fingers. I have teflon sheet to do the tendons. I have plenty of chain for the fingers. I have all the materials I need for this project.
I did a couple test-bends in aluminum, but can't get consistent results to modify the CAD drawing, so I'll likely just oversize by about .015 inch and see how that goes. Without a bending brake that would work for something this small, I will have to cut some slots at the corners to make bending easier.
Also, when milling the sheet, I'll need to clamp the entire sheet down, as it's too thin on its own to mill cleanly (if I had a higher RPM spindle, it might be easier, but max is 4k rpm or so).
With all my effort in software, I'm tempted to farm out the hand brackets to a local fab shop just to get it done without having to fiddle. Same goes for my hip brackets and bearings. That may be cost-prohibitive, having a shop do a little R&D work, but I at least want to check it out. Besides, if I end up building sets for sale, it would be nice to have them professionally done without me having to fab the parts myself (and get them anodized as well).
Here's a picture of the brackets for one hand. On the right is the top plate, the tabs are for mounting the top plate to the bottom brackets. On the upper left is the base bracket, all 3 external panels fold upwards to make the base and sides of the hand.
The top panel doesn't have any wrist mounting holes yet, as I have not decided how best to rotate the wrist. I have 3 options: attach a servo directly to the wrist, but this will mess up the geometry of the Robonova; attach a pulley / bearing assembly like the one for my hip design, and rotate the servo with an offset HSR-8498HB or SG - this could look ugly, and could interfere with stock Robonova motion; using a pulley, attach a DS450 servo for wrist rotation, this will interfere less with the stock Robonova motion.
The bottom left bracket is for mounting the e-chain fingers. The area above the bottom two panels gets folded downward, and the left and right sides then fold backward to attach to the base side panels. The bottom left panel becomes the thumb mount, the bottom right panel is where the fingers get mounted.
I will likely need some sort of guide for the teflon tendon on the thumb from the back servo, there's a bit of distance there that is likely to have the teflon fold when extending the thumb flat.
The past several months sure went by quick!
I have finished up the hand bracket design, it's about 90% done. I have the Pololu serial board code ready to test. I have several sheets of .040 inch thick 6061-T6 aluminum. I have 10 DS450 servos to control the fingers. I have teflon sheet to do the tendons. I have plenty of chain for the fingers. I have all the materials I need for this project.
I did a couple test-bends in aluminum, but can't get consistent results to modify the CAD drawing, so I'll likely just oversize by about .015 inch and see how that goes. Without a bending brake that would work for something this small, I will have to cut some slots at the corners to make bending easier.
Also, when milling the sheet, I'll need to clamp the entire sheet down, as it's too thin on its own to mill cleanly (if I had a higher RPM spindle, it might be easier, but max is 4k rpm or so).
With all my effort in software, I'm tempted to farm out the hand brackets to a local fab shop just to get it done without having to fiddle. Same goes for my hip brackets and bearings. That may be cost-prohibitive, having a shop do a little R&D work, but I at least want to check it out. Besides, if I end up building sets for sale, it would be nice to have them professionally done without me having to fab the parts myself (and get them anodized as well).
Here's a picture of the brackets for one hand. On the right is the top plate, the tabs are for mounting the top plate to the bottom brackets. On the upper left is the base bracket, all 3 external panels fold upwards to make the base and sides of the hand.
The top panel doesn't have any wrist mounting holes yet, as I have not decided how best to rotate the wrist. I have 3 options: attach a servo directly to the wrist, but this will mess up the geometry of the Robonova; attach a pulley / bearing assembly like the one for my hip design, and rotate the servo with an offset HSR-8498HB or SG - this could look ugly, and could interfere with stock Robonova motion; using a pulley, attach a DS450 servo for wrist rotation, this will interfere less with the stock Robonova motion.
The bottom left bracket is for mounting the e-chain fingers. The area above the bottom two panels gets folded downward, and the left and right sides then fold backward to attach to the base side panels. The bottom left panel becomes the thumb mount, the bottom right panel is where the fingers get mounted.
I will likely need some sort of guide for the teflon tendon on the thumb from the back servo, there's a bit of distance there that is likely to have the teflon fold when extending the thumb flat.
by PaulL
Sat Nov 12, 2011 11:46 pm
It's been a long time since I started this.. But, all the same, I have been playing with Vectric's Cut2D, below is the result. I am heading to the workshop to dust off the mill and fire it up. I've been playing with the Demo (below is from Demo mode, the only "gotcha" is that you can't generate NC code, G-Code, until you license it, which I told myself I wouldn't do until I was ready to run the files). I really like this software, definitely worth the $150 USD to license it. It will also work with TurboCNC, which is what I use on my mill.
The image below is a "simulation" of what the end material should look like with the specified tooling and toolpaths - not just a 3D rendering from a CAD package, but a simulated CAM result. YAY!
I made some "designed for manufacturability" changes other than those in the image, such as going with .020 inch teflon sheet for the tendons, and cutting a tubular sleeve for the servo horn that will get tapped for the teflon strip. Next step is to prototype these brackets doing a test run in polycarb sheet (.040 inches, about 1mm) and see how the fit is and where I need to modify sizes to accomodate for bends, then one test run in alu sheet, then verify, then final run if needed.
The image below is a "simulation" of what the end material should look like with the specified tooling and toolpaths - not just a 3D rendering from a CAD package, but a simulated CAM result. YAY!
I made some "designed for manufacturability" changes other than those in the image, such as going with .020 inch teflon sheet for the tendons, and cutting a tubular sleeve for the servo horn that will get tapped for the teflon strip. Next step is to prototype these brackets doing a test run in polycarb sheet (.040 inches, about 1mm) and see how the fit is and where I need to modify sizes to accomodate for bends, then one test run in alu sheet, then verify, then final run if needed.
by PaulL
Thu Nov 17, 2011 11:48 am
Another update... Got the machine going (it's been sitting for a while), and cut some test pieces of 1 inch alu with v-bends. Carpet tape works well for holding down to particle board for sheet metal - clean cuts, no trouble, just a little tediousness to get the glue residue off parts. I've been trying to pin down some info on dimensional changes when bending metal (specifically alu), but such info is sparse on the 'net.
I CAN say, bending metal is generally not an exercise in precision. How you bend and what you bend can change the results greatly.
However, in the case of .038 inch 6061-T6 aluminum sheet (my .040 ordered sheets are actually .038 inch), manual bending with a .020 inch V-groove is a bust. The metal bends too easily and actually fractures in the process, but the bend is square and does hold and is repeatable within a range of .007 inches. Ironically, bending with the v-groove on the OUTSIDE of a 90 degree angle is stronger than INSIDE. Can't explain that one.
My next step is to lightly (.005 inches) score the aluminum to hopefully encourage a straight bend.
I do have some numbers, though: for a manual bend (not with a bending brake or machine) without scoring, one can expect with a tight bend of alu .2 inches (about 5mm) wide and one inch long with the bend in the middle to result in two sides of about .533 inches with about .033 inches of "lead in" on each side of the bend where the metal starts to curve in the bend.
I've looked for bending brakes / presses / etc, but few work with very small bends at .040 inches thick, none I've found are small enough to do what I need. I've considered just cutting some male / female v-cuts and making my own bender, but I don't have a vise with enough depth to do all the bending I'll need (mine are short with little clearance below the jaws - machinist vises).
Worst case scenario is the manual bend without any scoring, but now I have numbers to rework the CAD drawing (I know, another 12 months and I'll be done, right? No- it'll be sooner than that. ).
Next up, test with the very slight v-score and see if it helps with keeping the bend straight. If not, I'll just have to run several pieces and manually bend them until I get it right.
On a side note, I've found a suitable u-joint and bearings for my neck mod, will be like a heli swash with a u-joint in the center instead of a swivel ball to provide tilt with rotation (head will be able to lean in all directions and rotate). This will take 3 micro servos, probably DS-450's. On the RN-1, I'll cut two slots at the top edge of the front plate, putting the arms in the free space above the shoulder servos, and use piano wire to get motion to the neck assembly - that's 2 servos. The third will likely go middle-chest and drive one pushrod upwards (forward / backward head tilt).
The tricky part is height of the stack in the neck (bearing / rotate arm / u-joint / swash bearing). If not for the battery in the compartment being the full length of the compartment, I could extend the assembly into the body without raising his head. I have CAD drawings for the u-joint, but it doesn't quite look like the final product pictures, so I have to order a few and see what I can do. I don't need 45 degrees of tilt (thinking 30 or less), so I can ease into the joint's working area a bit with bearings. I'm trying not to extend too far into the head so I can put a cam inside it at some point (will make a transparent visor). I'm still looking for a small enough USB cam that will fit inside the head and work w/ OpenCV in XP on Roboard). The assembly should still be strong enough for head stands when done.
Neck is later, hands are now, before the hip joints or the wrist joints (will likely be the same components for the joint, but possibly different servos).
Take Care,
Paul
Btw - Yes, I did buy Cut2D, and actually used it to create the job for the test pieces. I've also run an "open air" test for the brackets, all went beautifully. Just need to modify the drawing for the bend geometry and then run it. Very close to having the hand brackets ready!!!
I CAN say, bending metal is generally not an exercise in precision. How you bend and what you bend can change the results greatly.
However, in the case of .038 inch 6061-T6 aluminum sheet (my .040 ordered sheets are actually .038 inch), manual bending with a .020 inch V-groove is a bust. The metal bends too easily and actually fractures in the process, but the bend is square and does hold and is repeatable within a range of .007 inches. Ironically, bending with the v-groove on the OUTSIDE of a 90 degree angle is stronger than INSIDE. Can't explain that one.
My next step is to lightly (.005 inches) score the aluminum to hopefully encourage a straight bend.
I do have some numbers, though: for a manual bend (not with a bending brake or machine) without scoring, one can expect with a tight bend of alu .2 inches (about 5mm) wide and one inch long with the bend in the middle to result in two sides of about .533 inches with about .033 inches of "lead in" on each side of the bend where the metal starts to curve in the bend.
I've looked for bending brakes / presses / etc, but few work with very small bends at .040 inches thick, none I've found are small enough to do what I need. I've considered just cutting some male / female v-cuts and making my own bender, but I don't have a vise with enough depth to do all the bending I'll need (mine are short with little clearance below the jaws - machinist vises).
Worst case scenario is the manual bend without any scoring, but now I have numbers to rework the CAD drawing (I know, another 12 months and I'll be done, right? No- it'll be sooner than that.
).
Next up, test with the very slight v-score and see if it helps with keeping the bend straight. If not, I'll just have to run several pieces and manually bend them until I get it right.
On a side note, I've found a suitable u-joint and bearings for my neck mod, will be like a heli swash with a u-joint in the center instead of a swivel ball to provide tilt with rotation (head will be able to lean in all directions and rotate). This will take 3 micro servos, probably DS-450's. On the RN-1, I'll cut two slots at the top edge of the front plate, putting the arms in the free space above the shoulder servos, and use piano wire to get motion to the neck assembly - that's 2 servos. The third will likely go middle-chest and drive one pushrod upwards (forward / backward head tilt).
The tricky part is height of the stack in the neck (bearing / rotate arm / u-joint / swash bearing). If not for the battery in the compartment being the full length of the compartment, I could extend the assembly into the body without raising his head. I have CAD drawings for the u-joint, but it doesn't quite look like the final product pictures, so I have to order a few and see what I can do. I don't need 45 degrees of tilt (thinking 30 or less), so I can ease into the joint's working area a bit with bearings. I'm trying not to extend too far into the head so I can put a cam inside it at some point (will make a transparent visor). I'm still looking for a small enough USB cam that will fit inside the head and work w/ OpenCV in XP on Roboard). The assembly should still be strong enough for head stands when done.
Neck is later, hands are now, before the hip joints or the wrist joints (will likely be the same components for the joint, but possibly different servos).
Take Care,
Paul
Btw - Yes, I did buy Cut2D, and actually used it to create the job for the test pieces. I've also run an "open air" test for the brackets, all went beautifully. Just need to modify the drawing for the bend geometry and then run it. Very close to having the hand brackets ready!!!