Project: Melissa Hands - Perhaps a little easier..

Discussions regarding building a walking robot at home. Most of the robots participating at Robo-One competitions are custom fabricated.
135 postsPage 5 of 91, 2, 3, 4, 5, 6, 7, 8, 9
135 postsPage 5 of 91, 2, 3, 4, 5, 6, 7, 8, 9

Post by limor » Wed Jul 24, 2013 10:40 pm

Post by limor
Wed Jul 24, 2013 10:40 pm

sorry for being thick again, but can you please explain how this break works. We've been struggling with aluminum bending for a while.
btw: do you carve into the aluminum sheet to get the perfect bend ?
sorry for being thick again, but can you please explain how this break works. We've been struggling with aluminum bending for a while.
btw: do you carve into the aluminum sheet to get the perfect bend ?
limor offline
Savvy Roboteer
Savvy Roboteer
User avatar
Posts: 1845
Joined: Mon Oct 11, 2004 1:00 am
Location: London, UK

Post by PaulL » Wed Jul 24, 2013 11:48 pm

Post by PaulL
Wed Jul 24, 2013 11:48 pm

Hi Limor,

No problem, I'll tell you everything I know / have learned about bending aluminum. :)

The brake I'm using is this one:

Image

This brake doesn't come with any kind of backstop.

The modification I posted above is to add an adjustable back stop, a reference surface to butt the aluminum against to position the material such that the bend will occur where it's supposed to be. For really accurate bends, you just can't do them by hand or by eye. At least, I can't. :)

I tried bending 6061-T6 by hand. That doesn't work, at least not for me. The metal would break instead of bend.

Even if I managed to get a decent bend, I would test it only to find that it fractured on the outside edge, and would break with only a little pressure.

I tried scoring at the bend, this just gives more leverage to split the metal on the outside of the bend.

I got frustrated with hand bending and bought the brake above. This brake will do great bends in 6061-T6, and the bend is very strong. I haven't given much thought as to why this brake works so much better than bending by hand, but it does. I feel as though if I question why it works, the sheet metal gods will ruin every part I make from now on, so I let it be. :)

I've read up on bending physics and calculations and techniques. Now, let me tell you what I've found in practice:

There are brakes out there that use a wedge into a V, they're pretty common. I considered one of these, but I thought the design of the Grizzly G0556 was likely to be better for my needs (precise bends) since one side of the bend is actually clamped down prior to bending instead of positioning a shaper over a part over a die. It's also straightforward to do something other than a 90 degree bend. And, it can do pans, as it is a pan and box brake. I think I made the right choice. If you look at higher end pan-and-box brakes, they're similar in design to this one.

As for bending calculations: nothing beats a true test. Brake setup (you can adjust this one such that the fingers are closer in for a tighter bend or farther away for a more soft bend), material, maybe even the wind, all affect how a bend turns out. In reading, they say not to make a bend tighter in radius than half the thickness of the material. My experience doesn't agree with this. I found that you can make the bend tighter than that, almost square using this brake, and it can still be a good bend. Practicing and playing with the fwd / back adjustment for the fingers helps get a feel for what you can do.

I took a look at some online calculators for bend radii, but they really aren't 100% accurate. My measurement of .007" (0.1778 mm) works great on .040" (~1mm) 6061-T6. I simply add .007" to each inside length on either side of a bend in CAD, making up a total of .014" added right where the bend goes. This works, and is close enough for me. When I say close enough, I mean I can't get any closer. :)

In testing the length change when bending, I bent some CNC'd strips that were very precise in length. I then measured the result with a caliper, and repeated some 10 times or so until I was satisfied with the consistency of .007".

I tried 5052, but I hate the way it machines - gummy finish. The original RN-1 brackets seem to be soft like 5052, but they look like they were pressed with a custom die.

As for the angle of the bend, there is some springback in 6061-T6, but not much - a degree or two. I mounted an angle finder to my brake, and this is OK, but I'd like to make an adjustable hard stop for setting the angle to result in a 90 degree bend.

This is the angle finder I mounted to the brake:

Image

About the modification above: I mounted an aluminum plate to my brake a while back and clamped down an alu bar to the plate to act as a fixed backstop. With shims to space the sheet the required distance from the fixed stop, I could get OK results. This modification is to provide an adjustable backstop - a very accurate one. :) No more shims, and more precise than messing with shims. I say shims, you could also call these shims "spacers", that might make more sense.

In looking around on the net, you can get a bending brake that has a computer controlled backstop, but they're huge and extremely expensive, way beyond my means and space. I looked for anyone putting a Digital Readout (DRO) on an adjustable backstop, but only found adjustable backstops by crank on larger brakes, and no DRO. So, I don't know how novel it is, this thing I built above, but it seems pretty useful to me. :)

If you have any questions, by all means, just ask. :)
Hi Limor,

No problem, I'll tell you everything I know / have learned about bending aluminum. :)

The brake I'm using is this one:

Image

This brake doesn't come with any kind of backstop.

The modification I posted above is to add an adjustable back stop, a reference surface to butt the aluminum against to position the material such that the bend will occur where it's supposed to be. For really accurate bends, you just can't do them by hand or by eye. At least, I can't. :)

I tried bending 6061-T6 by hand. That doesn't work, at least not for me. The metal would break instead of bend.

Even if I managed to get a decent bend, I would test it only to find that it fractured on the outside edge, and would break with only a little pressure.

I tried scoring at the bend, this just gives more leverage to split the metal on the outside of the bend.

I got frustrated with hand bending and bought the brake above. This brake will do great bends in 6061-T6, and the bend is very strong. I haven't given much thought as to why this brake works so much better than bending by hand, but it does. I feel as though if I question why it works, the sheet metal gods will ruin every part I make from now on, so I let it be. :)

I've read up on bending physics and calculations and techniques. Now, let me tell you what I've found in practice:

There are brakes out there that use a wedge into a V, they're pretty common. I considered one of these, but I thought the design of the Grizzly G0556 was likely to be better for my needs (precise bends) since one side of the bend is actually clamped down prior to bending instead of positioning a shaper over a part over a die. It's also straightforward to do something other than a 90 degree bend. And, it can do pans, as it is a pan and box brake. I think I made the right choice. If you look at higher end pan-and-box brakes, they're similar in design to this one.

As for bending calculations: nothing beats a true test. Brake setup (you can adjust this one such that the fingers are closer in for a tighter bend or farther away for a more soft bend), material, maybe even the wind, all affect how a bend turns out. In reading, they say not to make a bend tighter in radius than half the thickness of the material. My experience doesn't agree with this. I found that you can make the bend tighter than that, almost square using this brake, and it can still be a good bend. Practicing and playing with the fwd / back adjustment for the fingers helps get a feel for what you can do.

I took a look at some online calculators for bend radii, but they really aren't 100% accurate. My measurement of .007" (0.1778 mm) works great on .040" (~1mm) 6061-T6. I simply add .007" to each inside length on either side of a bend in CAD, making up a total of .014" added right where the bend goes. This works, and is close enough for me. When I say close enough, I mean I can't get any closer. :)

In testing the length change when bending, I bent some CNC'd strips that were very precise in length. I then measured the result with a caliper, and repeated some 10 times or so until I was satisfied with the consistency of .007".

I tried 5052, but I hate the way it machines - gummy finish. The original RN-1 brackets seem to be soft like 5052, but they look like they were pressed with a custom die.

As for the angle of the bend, there is some springback in 6061-T6, but not much - a degree or two. I mounted an angle finder to my brake, and this is OK, but I'd like to make an adjustable hard stop for setting the angle to result in a 90 degree bend.

This is the angle finder I mounted to the brake:

Image

About the modification above: I mounted an aluminum plate to my brake a while back and clamped down an alu bar to the plate to act as a fixed backstop. With shims to space the sheet the required distance from the fixed stop, I could get OK results. This modification is to provide an adjustable backstop - a very accurate one. :) No more shims, and more precise than messing with shims. I say shims, you could also call these shims "spacers", that might make more sense.

In looking around on the net, you can get a bending brake that has a computer controlled backstop, but they're huge and extremely expensive, way beyond my means and space. I looked for anyone putting a Digital Readout (DRO) on an adjustable backstop, but only found adjustable backstops by crank on larger brakes, and no DRO. So, I don't know how novel it is, this thing I built above, but it seems pretty useful to me. :)

If you have any questions, by all means, just ask. :)
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Post by PaulL » Mon Aug 12, 2013 11:53 am

Post by PaulL
Mon Aug 12, 2013 11:53 am

An update...

It's been a few weeks! Well, I found a problem with one of the ballscrews on my mill - debris in my X axis ballnut. I ended up tearing down and rebuilding my X and Y on the mill, including cleaning out and rebuilding the ballnuts / ballscrews. I also found a problem with the kit I bought years ago on the Y axis - the block that mounts the ballscrew and motor to the mill wasn't machined correctly. There are two thrust bearings that clamp the ballscrew to the block, and one side was way off from the other (about half a mm). I turned the ID of the pocket that holds the thrust bearings using the lathe to enlarge the pocket - the results are much better! There was a good bit of binding before that I compensated for by keeping a few pieces loose - this caused a fair amount of backlash. Now, I have things tightened up with no binding, and backlash is down to .002 inches on Y as is on X. With things loose, it was up to 2 or 3 times that!

The mill is back together, but I am going to see about a better way to keep chips off the ballscrews. I added hot glue seals to both X and Y (Y had one before, but one side failed!), but it seems the oil is attacking the hot glue (which is why the one on the Y axis failed). There's hardly any room to add a different kind of seal to the ballnuts.

I also mounted the adjustable backstop to the brake - there's some more tweaking to do. I may turn some spacers on the lathe to get the height adjustment right.

For now, I'm going to give some thought to how to keep the chips off the screws - I'll come up with something. Worst case, I do nothing, put it back together, run a while, then rebuild again - not my preferred option.
An update...

It's been a few weeks! Well, I found a problem with one of the ballscrews on my mill - debris in my X axis ballnut. I ended up tearing down and rebuilding my X and Y on the mill, including cleaning out and rebuilding the ballnuts / ballscrews. I also found a problem with the kit I bought years ago on the Y axis - the block that mounts the ballscrew and motor to the mill wasn't machined correctly. There are two thrust bearings that clamp the ballscrew to the block, and one side was way off from the other (about half a mm). I turned the ID of the pocket that holds the thrust bearings using the lathe to enlarge the pocket - the results are much better! There was a good bit of binding before that I compensated for by keeping a few pieces loose - this caused a fair amount of backlash. Now, I have things tightened up with no binding, and backlash is down to .002 inches on Y as is on X. With things loose, it was up to 2 or 3 times that!

The mill is back together, but I am going to see about a better way to keep chips off the ballscrews. I added hot glue seals to both X and Y (Y had one before, but one side failed!), but it seems the oil is attacking the hot glue (which is why the one on the Y axis failed). There's hardly any room to add a different kind of seal to the ballnuts.

I also mounted the adjustable backstop to the brake - there's some more tweaking to do. I may turn some spacers on the lathe to get the height adjustment right.

For now, I'm going to give some thought to how to keep the chips off the screws - I'll come up with something. Worst case, I do nothing, put it back together, run a while, then rebuild again - not my preferred option.
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Post by PaulL » Sun Sep 08, 2013 12:43 pm

Post by PaulL
Sun Sep 08, 2013 12:43 pm

I've got a chip tray I've been working on, this should keep screws on the mill much more clean. For now, I'm building it from plywood - it just seems wrong to do that on a cast iron mill!!! Maybe I'll contract a shop to make me one from sheet metal once I'm set on the dimensions - I'd make one myself, but my machines don't have enough capacity (I'd need about 30 inch of 18 gauge steel sheet).

I machined a new recessed hub on the lathe to work with the custom thrust bearing / pulley / joint rotation assembly. The dimensions are good, and I'm going with this design. Recessing the nut side of the screw that clamps the two brackets together (and applies force to the thrust bearing) means the length added to arms with the wrist rotation setup (with hands!) will be shortened, keeping closer to the stock RN-1 dimensions.

I'm going to use the same thrust bearing design for the hips, but I won't need to recess the hub that goes in the hip joint - there's plenty of room on either side of the joint.

For the wrists, anything I put between last arm servo and hand adds to the length.

From here:

* Finish up the chip tray, get the mill going again.
* Machine hip and wrist rotation brackets (design is ready to test).
* Machine the pulleys from 2mm alu (design is done).
* Cut a few more hubs on the lathe (design is done).
* Cut a few more ball cages from teflon (design is done).

Then, I'll test-assemble the wrist and a hand to one of my RN-1's and see how it all works. If all is good, I'll machine a few hand brackets and some teflon hubs and strips, and build it all up to test - test configuration will be an RN-1 with a Roboard RB-100, hip servos and brackets, wrist rotation servos and brackets, and 10 finger servos and hands. I'll probably try on my gold one (meaning stock servos), as he still has his servo wires laced up and Roboard mounted.

Beyond that, I have some CAD work and testing remaining for the torso and the neck brackets.

I'm slowly getting closer to getting the hardware done!! I'd like to take a few weeks off of work to finish up the hardware, but that's not going to happen until the holidays.
I've got a chip tray I've been working on, this should keep screws on the mill much more clean. For now, I'm building it from plywood - it just seems wrong to do that on a cast iron mill!!! Maybe I'll contract a shop to make me one from sheet metal once I'm set on the dimensions - I'd make one myself, but my machines don't have enough capacity (I'd need about 30 inch of 18 gauge steel sheet).

I machined a new recessed hub on the lathe to work with the custom thrust bearing / pulley / joint rotation assembly. The dimensions are good, and I'm going with this design. Recessing the nut side of the screw that clamps the two brackets together (and applies force to the thrust bearing) means the length added to arms with the wrist rotation setup (with hands!) will be shortened, keeping closer to the stock RN-1 dimensions.

I'm going to use the same thrust bearing design for the hips, but I won't need to recess the hub that goes in the hip joint - there's plenty of room on either side of the joint.

For the wrists, anything I put between last arm servo and hand adds to the length.

From here:

* Finish up the chip tray, get the mill going again.
* Machine hip and wrist rotation brackets (design is ready to test).
* Machine the pulleys from 2mm alu (design is done).
* Cut a few more hubs on the lathe (design is done).
* Cut a few more ball cages from teflon (design is done).

Then, I'll test-assemble the wrist and a hand to one of my RN-1's and see how it all works. If all is good, I'll machine a few hand brackets and some teflon hubs and strips, and build it all up to test - test configuration will be an RN-1 with a Roboard RB-100, hip servos and brackets, wrist rotation servos and brackets, and 10 finger servos and hands. I'll probably try on my gold one (meaning stock servos), as he still has his servo wires laced up and Roboard mounted.

Beyond that, I have some CAD work and testing remaining for the torso and the neck brackets.

I'm slowly getting closer to getting the hardware done!! I'd like to take a few weeks off of work to finish up the hardware, but that's not going to happen until the holidays.
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Re: Project: Melissa Hands - Perhaps a little easier..

Post by PaulL » Fri Sep 20, 2013 10:15 am

Post by PaulL
Fri Sep 20, 2013 10:15 am

Update!

I abandoned the chip tray for now, and made some parts this past weekend:

* 6 more teflon ball cages (more than enough).
* Hip rotation brackets, ready to bend and test.
* Wrist rotation brackets, ready to bend and test.
* One more recessed hub for wrist rotation, done with these for a while.

I ended up using some dull tooling I had close at hand - one drill bit and the end mill I used for the hip and wrist rotation brackets are both on their last legs. The end mill won't center-cut anymore - that made for some "not good" sounds when running the rotation brackets, but these brackets came out well enough. I also made a mistake in CAM, had the brackets cut in one pass, so the mill put quite a bit of effort in pushing a worn center-cutting end mill through 1mm alu sheet. I like to do finer passes for a better finish, but these are prototypes, probably not the last version to get cut.

I have the CAM job for the 2mm thick pulleys set and ready to go, but I ran out of time last weekend. This weekend, I should be able to finish up the pulleys, and get the brackets bent and tested. If things go well enough, I'll have a hand mounted on my RN1 to see what it all will look like. I may have enough time to finish up the pulley mandrel (for cutting the groove in the pulleys) and test out the whole pulley method.

Take Care,
Paul
Update!

I abandoned the chip tray for now, and made some parts this past weekend:

* 6 more teflon ball cages (more than enough).
* Hip rotation brackets, ready to bend and test.
* Wrist rotation brackets, ready to bend and test.
* One more recessed hub for wrist rotation, done with these for a while.

I ended up using some dull tooling I had close at hand - one drill bit and the end mill I used for the hip and wrist rotation brackets are both on their last legs. The end mill won't center-cut anymore - that made for some "not good" sounds when running the rotation brackets, but these brackets came out well enough. I also made a mistake in CAM, had the brackets cut in one pass, so the mill put quite a bit of effort in pushing a worn center-cutting end mill through 1mm alu sheet. I like to do finer passes for a better finish, but these are prototypes, probably not the last version to get cut.

I have the CAM job for the 2mm thick pulleys set and ready to go, but I ran out of time last weekend. This weekend, I should be able to finish up the pulleys, and get the brackets bent and tested. If things go well enough, I'll have a hand mounted on my RN1 to see what it all will look like. I may have enough time to finish up the pulley mandrel (for cutting the groove in the pulleys) and test out the whole pulley method.

Take Care,
Paul
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Re: Project: Melissa Hands - Perhaps a little easier..

Post by PaulL » Mon Sep 23, 2013 9:51 am

Post by PaulL
Mon Sep 23, 2013 9:51 am

More progress!!! It's funny how things don't always go as planned (didn't spend as much time on my project this weekend as I had hoped), but that's OK... :)

I've had the bending brake backstop mod mounted to the brake for a while now, but I hadn't tried using it on actual parts until this weekend. Let me just say, this (the precision adjustable backstop) is an EXCELLENT way to get precise bends. This thing works EXACTLY as I had hoped!! If you make your own brackets and need precision, get a pan-and-box brake that clamps the work, and make yourself an adjustable backstop like the one I made. I didn't do much fine tuning, as the first precision bend I did after a quick calibration (eyeball, test bend, apply offset, zero the scale) was within .002 inches (a little less than 1/20 of a mm) of what I had calculated!! That's NUTS for a $200 USD hand-operated brake!! :)

The adjustable backstop is the perfect way to do bends at varying distances from the end of parts. Once I wrote down the distances from the edges of my parts from CAD and applied the offset, the rest was simply dialing in the number, placing the part, and bending. That's about as good as it can get!!!

There is only one aspect that could be addressed for even better precision: a stop for the clamping lever. The part moves forward away from the backstop when clamping, the fingers push the part away from the backstop as they bend (taken into account when calibrating the backstop somewhat). If the clamping force is relatively consistent (by feel), the repeatability is acceptable. I have to say, I doubt I'll bother messing with making a clamping lever stop. :)

One problem I have with this brake setup is in reading the scale - when doing the really close (6 mm or less) bends, it's very hard to see the scale, as it ends up far under the clamping mechanism. A few small mirrors might fix that problem - or, I can interface the scale (it has data out) and add a display elsewhere (a useful weekend Arduino project).

Another issue - not a big one: With this bending brake, the fingers are fixed width, which means that if you want to bend a pan shape (two or more bends at 90 degrees to each other), the bend is best if the finger edges run the width of the bend. However, the fingers that came with the brake are only in 1 inch, 2 inch, and 3 inch widths - which means that bending something like a pan-shaped bracket along the length of a Hitec robotic servo is a compromise. I used a 1 inch finger for the bends, which resulted in 1" of the bend being very tight and the rest was more "soft". It's not terribly bad, but it doesn't look like a factory bend (I suspect the factory bends were done on a press brake with a specially made arbor plates). To do better on the ~1.7 inch wide bends, I need to have a finger or two cut down (hardened tool steel). Alternatively, a 1 inch finger could be cut down for the width of the servo and then bend the long sides on a 2 inch finger first. Structurally, the bends are fine, special width fingers would be purely for aesthetics.

And to mention once more - perfect 90 degree bends are subject to some variation with only an angle finder as an indicator. A calibrated hard stop would resolve this, but the results I'm getting are good enough. :) Hmmm - an accelerometer might be of use here - if I bother with a remote display, I could add an accelerometer to find the "stop" angle.

***

With the brake working nicely, I bent all of the odd bends for the hip and wrist rotation brackets after cleaning up the brackets a bit (dull drill bit left some raised material behind).

In test-assembling the brackets and 4 servos, I found two problems - one, I forgot to make a hole for the business end of a servo, and two, I didn't notice where I needed to cut a relief for the business end of a servo. The relief was easy enough to do with a Dremel and sanding drum. The missing hole wasn't really an issue for testing, as I just mounted the servo by it's opposite end.

The whole point of testing was to check clearance of various points and strength of the bracket configuration. SUCCESS! :) The only changes I'll make in CAD is to soften a few radii on some fillet'd corners, and add the relief and missing hole.

I wasn't sure how well the arm servo end bracket was going to turn out, but it's fine, with enough clearance for the recessed wrist rotation hub. The hip rotation brackets result in a pretty stiff mechanism, it should work fine. The wrist rotation brackets are a bit less stiff (only one side is mounted by design), but it's fine for wrist rotation.

I can say, I'm starting to have some concern for how this guy is going to walk after all I'm doing to him - the weight of adding servos and hands will definitely be enough to require changing some of his moves / positions.

In testing these brackets, it again became clear that the HSR-5498SG servos are far better than the HSR-8498HB stock servos. Metal servo hubs are MUCH stronger than plastic ones. And, there is NO play in the 5498 gear train. In the end, I think his wobbliness won't be as much of a problem as it is when stock (all of the servos on my final incarnation will be HSR-5498SG's - in total, I have 40 of them!!!).

Bit by bit, it's coming along. :)
More progress!!! It's funny how things don't always go as planned (didn't spend as much time on my project this weekend as I had hoped), but that's OK... :)

I've had the bending brake backstop mod mounted to the brake for a while now, but I hadn't tried using it on actual parts until this weekend. Let me just say, this (the precision adjustable backstop) is an EXCELLENT way to get precise bends. This thing works EXACTLY as I had hoped!! If you make your own brackets and need precision, get a pan-and-box brake that clamps the work, and make yourself an adjustable backstop like the one I made. I didn't do much fine tuning, as the first precision bend I did after a quick calibration (eyeball, test bend, apply offset, zero the scale) was within .002 inches (a little less than 1/20 of a mm) of what I had calculated!! That's NUTS for a $200 USD hand-operated brake!! :)

The adjustable backstop is the perfect way to do bends at varying distances from the end of parts. Once I wrote down the distances from the edges of my parts from CAD and applied the offset, the rest was simply dialing in the number, placing the part, and bending. That's about as good as it can get!!!

There is only one aspect that could be addressed for even better precision: a stop for the clamping lever. The part moves forward away from the backstop when clamping, the fingers push the part away from the backstop as they bend (taken into account when calibrating the backstop somewhat). If the clamping force is relatively consistent (by feel), the repeatability is acceptable. I have to say, I doubt I'll bother messing with making a clamping lever stop. :)

One problem I have with this brake setup is in reading the scale - when doing the really close (6 mm or less) bends, it's very hard to see the scale, as it ends up far under the clamping mechanism. A few small mirrors might fix that problem - or, I can interface the scale (it has data out) and add a display elsewhere (a useful weekend Arduino project).

Another issue - not a big one: With this bending brake, the fingers are fixed width, which means that if you want to bend a pan shape (two or more bends at 90 degrees to each other), the bend is best if the finger edges run the width of the bend. However, the fingers that came with the brake are only in 1 inch, 2 inch, and 3 inch widths - which means that bending something like a pan-shaped bracket along the length of a Hitec robotic servo is a compromise. I used a 1 inch finger for the bends, which resulted in 1" of the bend being very tight and the rest was more "soft". It's not terribly bad, but it doesn't look like a factory bend (I suspect the factory bends were done on a press brake with a specially made arbor plates). To do better on the ~1.7 inch wide bends, I need to have a finger or two cut down (hardened tool steel). Alternatively, a 1 inch finger could be cut down for the width of the servo and then bend the long sides on a 2 inch finger first. Structurally, the bends are fine, special width fingers would be purely for aesthetics.

And to mention once more - perfect 90 degree bends are subject to some variation with only an angle finder as an indicator. A calibrated hard stop would resolve this, but the results I'm getting are good enough. :) Hmmm - an accelerometer might be of use here - if I bother with a remote display, I could add an accelerometer to find the "stop" angle.

***

With the brake working nicely, I bent all of the odd bends for the hip and wrist rotation brackets after cleaning up the brackets a bit (dull drill bit left some raised material behind).

In test-assembling the brackets and 4 servos, I found two problems - one, I forgot to make a hole for the business end of a servo, and two, I didn't notice where I needed to cut a relief for the business end of a servo. The relief was easy enough to do with a Dremel and sanding drum. The missing hole wasn't really an issue for testing, as I just mounted the servo by it's opposite end.

The whole point of testing was to check clearance of various points and strength of the bracket configuration. SUCCESS! :) The only changes I'll make in CAD is to soften a few radii on some fillet'd corners, and add the relief and missing hole.

I wasn't sure how well the arm servo end bracket was going to turn out, but it's fine, with enough clearance for the recessed wrist rotation hub. The hip rotation brackets result in a pretty stiff mechanism, it should work fine. The wrist rotation brackets are a bit less stiff (only one side is mounted by design), but it's fine for wrist rotation.

I can say, I'm starting to have some concern for how this guy is going to walk after all I'm doing to him - the weight of adding servos and hands will definitely be enough to require changing some of his moves / positions.

In testing these brackets, it again became clear that the HSR-5498SG servos are far better than the HSR-8498HB stock servos. Metal servo hubs are MUCH stronger than plastic ones. And, there is NO play in the 5498 gear train. In the end, I think his wobbliness won't be as much of a problem as it is when stock (all of the servos on my final incarnation will be HSR-5498SG's - in total, I have 40 of them!!!).

Bit by bit, it's coming along. :)
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Re: Project: Melissa Hands - Perhaps a little easier..

Post by PaulL » Tue Sep 24, 2013 11:46 am

Post by PaulL
Tue Sep 24, 2013 11:46 am

Pictures!

HipAndWrist1.jpg
HipAndWrist1.jpg (63.55 KiB) Viewed 9797 times


HipAndWrist2.jpg
HipAndWrist2.jpg (75.64 KiB) Viewed 9797 times


HipAndWrist1Illustrated.JPG
HipAndWrist1Illustrated.JPG (51.37 KiB) Viewed 9797 times


HipAndWrist2Illustrated.JPG
HipAndWrist2Illustrated.JPG (69.71 KiB) Viewed 9797 times
Pictures!

HipAndWrist1.jpg
HipAndWrist1.jpg (63.55 KiB) Viewed 9797 times


HipAndWrist2.jpg
HipAndWrist2.jpg (75.64 KiB) Viewed 9797 times


HipAndWrist1Illustrated.JPG
HipAndWrist1Illustrated.JPG (51.37 KiB) Viewed 9797 times


HipAndWrist2Illustrated.JPG
HipAndWrist2Illustrated.JPG (69.71 KiB) Viewed 9797 times
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Re: Project: Melissa Hands - Perhaps a little easier..

Post by PaulL » Wed Sep 25, 2013 11:55 am

Post by PaulL
Wed Sep 25, 2013 11:55 am

Hubs and Teflon ball cages...

SmallParts.JPG
SmallParts.JPG (116.1 KiB) Viewed 9774 times
Hubs and Teflon ball cages...

SmallParts.JPG
SmallParts.JPG (116.1 KiB) Viewed 9774 times
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Re: Project: Melissa Hands - Perhaps a little easier..

Post by PaulL » Tue Oct 08, 2013 11:18 am

Post by PaulL
Tue Oct 08, 2013 11:18 am

Not hand joint, but hip joint assembled. The hand joint will be similar in construction. It seems to be very strong, very little play, very square. It works better than I expected!!! Wire is nylon coated, .019" (about half a mm in dia).

Hip 1.JPG
Hip 1.JPG (83.16 KiB) Viewed 9577 times


Hip 2.JPG
Hip 2.JPG (82.57 KiB) Viewed 9577 times


Hip 3.JPG
Hip 3.JPG (95.24 KiB) Viewed 9577 times


Hip 4.JPG
Hip 4.JPG (74.67 KiB) Viewed 9577 times


Hip 5.JPG
Hip 5.JPG (93.84 KiB) Viewed 9577 times
Not hand joint, but hip joint assembled. The hand joint will be similar in construction. It seems to be very strong, very little play, very square. It works better than I expected!!! Wire is nylon coated, .019" (about half a mm in dia).

Hip 1.JPG
Hip 1.JPG (83.16 KiB) Viewed 9577 times


Hip 2.JPG
Hip 2.JPG (82.57 KiB) Viewed 9577 times


Hip 3.JPG
Hip 3.JPG (95.24 KiB) Viewed 9577 times


Hip 4.JPG
Hip 4.JPG (74.67 KiB) Viewed 9577 times


Hip 5.JPG
Hip 5.JPG (93.84 KiB) Viewed 9577 times
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Re: Project: Melissa Hands - Perhaps a little easier..

Post by PaulL » Sun Dec 01, 2013 11:10 am

Post by PaulL
Sun Dec 01, 2013 11:10 am

Long time since I posted an update, nearly 2 months!!

The project hasn't stalled, I just haven't been posting.

I've made a bunch of progress.

More brackets cut, done making parts for the hips and wrists. Only bending and finishing left (thinking paint, maybe anodize some time in the future).

The real progress has been on the torso. I've got the first rev of torso brackets CAD'd, CAM'd, and ready to run on the mill today. I have to say, I underestimated the amount of work involved with the torso - lots of tedious details. Hopefully there will be few if any revisions.

I've settled on a bunch of stuff - speakers, neck design, fitting the Kontron Z530 board in, cooling, laser selection and mount, wire routing, etc, etc. Tons of progress.

Parts left to order: 3M VHB tape and some screws and standoffs. Ordered more MKS DS-450's for the neck plus some spares. Laser module is on its way, too (line laser, it's part of a 3D scanner system). Mirrors showed up Sat (I only need 1, bought 2).

I started looking up parts for the control / power board, found that GE makes some DC to DC converters w/ PMBus, might shelf the GE Naos Raptors I bought. The PMBus converters let you read source volts, out volts, and out amps - handy, and keeps me from adding more components to do that externally. The PMBus interface for these also lets you do things like voltage margining and such, more control over power. I'm thinking I'll need 3: one for the PC board, one for 13 x 5v micro servos, and one for the main servos (these will get a special bit of circuitry to switch to / from full batt voltage, I'm planning on 2c LiPo, 7.4v for the HSR-5498SG's).

I wasn't planning on doing any electronics work until later this month, but I still managed to finish up the first rev torso design this past week.

I'm hoping to use a 2 layer board for the electronics, but it may end up being a 4 layer - I lost some space due to speakers and fan.

In final form, he'll be about 4mm taller, and his hands will be 3 or 4 mm longer.

Sorry, no pics this time, haven't bothered to take any.
Long time since I posted an update, nearly 2 months!!

The project hasn't stalled, I just haven't been posting.

I've made a bunch of progress.

More brackets cut, done making parts for the hips and wrists. Only bending and finishing left (thinking paint, maybe anodize some time in the future).

The real progress has been on the torso. I've got the first rev of torso brackets CAD'd, CAM'd, and ready to run on the mill today. I have to say, I underestimated the amount of work involved with the torso - lots of tedious details. Hopefully there will be few if any revisions.

I've settled on a bunch of stuff - speakers, neck design, fitting the Kontron Z530 board in, cooling, laser selection and mount, wire routing, etc, etc. Tons of progress.

Parts left to order: 3M VHB tape and some screws and standoffs. Ordered more MKS DS-450's for the neck plus some spares. Laser module is on its way, too (line laser, it's part of a 3D scanner system). Mirrors showed up Sat (I only need 1, bought 2).

I started looking up parts for the control / power board, found that GE makes some DC to DC converters w/ PMBus, might shelf the GE Naos Raptors I bought. The PMBus converters let you read source volts, out volts, and out amps - handy, and keeps me from adding more components to do that externally. The PMBus interface for these also lets you do things like voltage margining and such, more control over power. I'm thinking I'll need 3: one for the PC board, one for 13 x 5v micro servos, and one for the main servos (these will get a special bit of circuitry to switch to / from full batt voltage, I'm planning on 2c LiPo, 7.4v for the HSR-5498SG's).

I wasn't planning on doing any electronics work until later this month, but I still managed to finish up the first rev torso design this past week.

I'm hoping to use a 2 layer board for the electronics, but it may end up being a 4 layer - I lost some space due to speakers and fan.

In final form, he'll be about 4mm taller, and his hands will be 3 or 4 mm longer.

Sorry, no pics this time, haven't bothered to take any.
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Re: Project: Melissa Hands - Perhaps a little easier..

Post by PaulL » Wed Dec 18, 2013 12:42 pm

Post by PaulL
Wed Dec 18, 2013 12:42 pm

Progress Pics...

Hand, Mounted - haven't posted this one here yet:
Image

Neck Joint, 3 DOF (Like a heli swash plate, top arms and guide plates in the back needs to be bent, but you get the idea):
Image

Pile o' Parts. It's coming along. Z530 board looks good mounted to the back plate. Some parts I've posted before, some are as recent as the other day. Gold pointy thing with round base in upper left corner is a tap in a coupler I cut this past Sunday to tap parts using the shaft on the end of my Y axis (double shaft motors come in handy!). Tested yesterday, tapped 4 holes pretty quick - I've got about 48 holes to tap for the torso parts, and more for the other brackets - I needed a faster method. The arbor in the upper right corner is for cutting grooves in pulleys, cut on my lathe. Note the speakers on the side plates - 2W, 8 Ohm, with enclosures on the back. Fan is mounted to back plate, but screws are to be replaced w/ M3 cap head screws. There's even a laser mount in there. You can see how the hip servos are rotated inward on the plate a bit to the right on the bottom.
Image

Next up is tapping, bending, and sanding (the metal I get often comes with free scratches!! Yay!!).

I still need to revise the hands one last time, I keep putting it off. One last rev, and they're done.

Obviously, parts are piling up. :)

Meanwhile, I've started a parts library for some DC to DC converters from GE (PMBus devices, lots of features). I'm waiting on a reply from their tech support about clamping one of the FET's in a UDT020A0X by shorting RTrim (max power to servos), haven't heard back yet.

Board work will begin in earnest starting at the end of this week. I'll probably fit in some part finishing and the hand bracket rev. I have a few tweaks to do (like the plates on top of the Z530 board, they need a slight change). It's going to be tight in there.

Also waiting on some Arctic Silver Adhesive for the copper heat sinks on the Z530 board (Barely visible in the pic). The 3M 8815 tape just isn't very tacky, not enough to secure parts in a bot that will take a fall.

Coming along.
Progress Pics...

Hand, Mounted - haven't posted this one here yet:
Image

Neck Joint, 3 DOF (Like a heli swash plate, top arms and guide plates in the back needs to be bent, but you get the idea):
Image

Pile o' Parts. It's coming along. Z530 board looks good mounted to the back plate. Some parts I've posted before, some are as recent as the other day. Gold pointy thing with round base in upper left corner is a tap in a coupler I cut this past Sunday to tap parts using the shaft on the end of my Y axis (double shaft motors come in handy!). Tested yesterday, tapped 4 holes pretty quick - I've got about 48 holes to tap for the torso parts, and more for the other brackets - I needed a faster method. The arbor in the upper right corner is for cutting grooves in pulleys, cut on my lathe. Note the speakers on the side plates - 2W, 8 Ohm, with enclosures on the back. Fan is mounted to back plate, but screws are to be replaced w/ M3 cap head screws. There's even a laser mount in there. You can see how the hip servos are rotated inward on the plate a bit to the right on the bottom.
Image

Next up is tapping, bending, and sanding (the metal I get often comes with free scratches!! Yay!!).

I still need to revise the hands one last time, I keep putting it off. One last rev, and they're done.

Obviously, parts are piling up. :)

Meanwhile, I've started a parts library for some DC to DC converters from GE (PMBus devices, lots of features). I'm waiting on a reply from their tech support about clamping one of the FET's in a UDT020A0X by shorting RTrim (max power to servos), haven't heard back yet.

Board work will begin in earnest starting at the end of this week. I'll probably fit in some part finishing and the hand bracket rev. I have a few tweaks to do (like the plates on top of the Z530 board, they need a slight change). It's going to be tight in there.

Also waiting on some Arctic Silver Adhesive for the copper heat sinks on the Z530 board (Barely visible in the pic). The 3M 8815 tape just isn't very tacky, not enough to secure parts in a bot that will take a fall.

Coming along.
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Re: Project: Melissa Hands - Perhaps a little easier..

Post by PaulL » Tue Dec 24, 2013 9:20 am

Post by PaulL
Tue Dec 24, 2013 9:20 am

Managed to get a decent amount of board work done by Dec 22nd morning (all day and then some on the 21st!). Also did some work reorganizing the "bot shop" (my home office) to make it a bit more productive as I come closer to completion of the hardware and electronics. Busy with tasks of the holidays, but all the same, this is coming along better than expected.

"Shoehorning" is what I'm doing at this point - trying to figure out where I can fit all that needs to fit inside. One bit I still haven't got down is the main power switch - I want it to be extremely low current draw when off, but capable of digitally controlled "off", meaning it has to be some kind of digital switch. Pololu makes a 10A one, and they can be paralleled (so I've read), but I don't know where I could stuff 2 or 3 of them yet. He'll have various power level modes, but "all the way off" is a bit trickier without using a mechanical switch. Of course, I don't want to foul the LiPo's!! I could home brew a design, perhaps. We'll see.

It quickly became obvious that there was no way I was going to fit everything into such a small space of one 2 layer board (not anything I could physically build myself) - SO, I opted for a 2nd "power board" (top board in pic). The board outline in the bottom pic is from DXF, based on dimensions in CAD for what I have left after speakers (left and right notches) and fan (topwards notch). A separate board for power is better for a few reasons, not the least of which is that other choices are less optimal (ex, re-sizing the torso enclosure) - otherwise, it puts the high amp stuff further away from the PC board (not sure how the Z530 would feel with a 500khz or so high ampere signal into a coil a few mm right above it! Who knows, may need copper shielding even after this). There are terminals missing here and there, and some values need to be assigned, but it's coming along pretty well. Looking good for the goal of this end-of-year ready-to-fab controller board design.

Chips on this board are: STM32F103CBT6 Micro, CP2103 USB to UART, TS2012 Audio Amp, and 2 x PCA9685 for PWM. DC to DC modules are GE (formerly Lineage Power) DLynx 6A (logic, to reg for Audio power, etc), 12A (5v hand servos), and 20A (adjustable V for main body servos) - PMBus versions (I won't pull full amps from any of these, but over-rating them means longer life while having less air flow - 20A converter will get the most airflow). Outputs for 23 servos (spaced .105 inches to make room for the plugs). Audio is lower right, as far away from the digital stuff as I could put it - Ok, almost, I have a few more mm I could go. Hands will have their own servo controllers by serial connection (maybe Pololu's boards, maybe home brewed and built - STM32's on the backs of the hands, maybe? LOL!). Still, this board will behave like a Maple Mini - I actually started with a copy of their schematic (Github). I may just go with breakout boards (Sparkfun) for some I2C things (BlinkM Mini in the head, 6 DOF IMU, etc) since I have little room left on the main board. There are a few more things to add, so this will change a bit more before it's done.

One bit of fun is placement of the NTSC to USB converter board (to use the tiny camera for the head). It looks like it's going to get stuffed under the PC board next to the RAM. Might have to shield it.

Image
Managed to get a decent amount of board work done by Dec 22nd morning (all day and then some on the 21st!). Also did some work reorganizing the "bot shop" (my home office) to make it a bit more productive as I come closer to completion of the hardware and electronics. Busy with tasks of the holidays, but all the same, this is coming along better than expected.

"Shoehorning" is what I'm doing at this point - trying to figure out where I can fit all that needs to fit inside. One bit I still haven't got down is the main power switch - I want it to be extremely low current draw when off, but capable of digitally controlled "off", meaning it has to be some kind of digital switch. Pololu makes a 10A one, and they can be paralleled (so I've read), but I don't know where I could stuff 2 or 3 of them yet. He'll have various power level modes, but "all the way off" is a bit trickier without using a mechanical switch. Of course, I don't want to foul the LiPo's!! I could home brew a design, perhaps. We'll see.

It quickly became obvious that there was no way I was going to fit everything into such a small space of one 2 layer board (not anything I could physically build myself) - SO, I opted for a 2nd "power board" (top board in pic). The board outline in the bottom pic is from DXF, based on dimensions in CAD for what I have left after speakers (left and right notches) and fan (topwards notch). A separate board for power is better for a few reasons, not the least of which is that other choices are less optimal (ex, re-sizing the torso enclosure) - otherwise, it puts the high amp stuff further away from the PC board (not sure how the Z530 would feel with a 500khz or so high ampere signal into a coil a few mm right above it! Who knows, may need copper shielding even after this). There are terminals missing here and there, and some values need to be assigned, but it's coming along pretty well. Looking good for the goal of this end-of-year ready-to-fab controller board design.

Chips on this board are: STM32F103CBT6 Micro, CP2103 USB to UART, TS2012 Audio Amp, and 2 x PCA9685 for PWM. DC to DC modules are GE (formerly Lineage Power) DLynx 6A (logic, to reg for Audio power, etc), 12A (5v hand servos), and 20A (adjustable V for main body servos) - PMBus versions (I won't pull full amps from any of these, but over-rating them means longer life while having less air flow - 20A converter will get the most airflow). Outputs for 23 servos (spaced .105 inches to make room for the plugs). Audio is lower right, as far away from the digital stuff as I could put it - Ok, almost, I have a few more mm I could go. Hands will have their own servo controllers by serial connection (maybe Pololu's boards, maybe home brewed and built - STM32's on the backs of the hands, maybe? LOL!). Still, this board will behave like a Maple Mini - I actually started with a copy of their schematic (Github). I may just go with breakout boards (Sparkfun) for some I2C things (BlinkM Mini in the head, 6 DOF IMU, etc) since I have little room left on the main board. There are a few more things to add, so this will change a bit more before it's done.

One bit of fun is placement of the NTSC to USB converter board (to use the tiny camera for the head). It looks like it's going to get stuffed under the PC board next to the RAM. Might have to shield it.

Image
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Re: Project: Melissa Hands - Perhaps a little easier..

Post by PaulL » Sun Jan 12, 2014 3:25 pm

Post by PaulL
Sun Jan 12, 2014 3:25 pm

An update... I got sick with the flu a few days after my last post, lost 5 days to it - tried to work on it, but didn't get very far - hazy from Nyquil and Sudafed. Didn't make my deadline of submitting the board for fab by end of vacation. So, I'm back to work, trying to fit in what I can when I can. Worked on the boards all day yesterday.

Up to now: I found a MAX1614 - perfect for controlling power (LiPo cutoff) with a large FET. And speaking of FET's, I found one that's 40A at ambient, up to 100A. I added two 3.3v regulators to the controller board, then widened the power board and moved them there. I've changed packages on many of the devices, meaning more parts added to libraries. Seems the parts I'm finding are either too new or uncommon. There are NO parts on the bottom of the controller board, and some optional capacitors under the power board that I may or may not need.

I've added a bunch of features, and it's getting close to fab-ready. If I'm very lucky, some time in the next 12 hours, I may be ready to send to fab.

It's coming along nicely. So much so, I have decided to do STM32 (Maple Mini - based design, like the controller board) for the hands. I'm thinking PWM for LED's on the fingertips - would be an interesting effect, and the STM32 certainly has enough hardware PWM for that and the hand servos.

I finally tested the TS2012 audio amp board from Adafruit with the speakers and the Z530 board - it sounds pretty good for not having the covers on the back of the speakers (it will when done, I just took them off and wired right to the speakers to test). The speakers and amp are a good match. I did this test to see if I could omit a few components for selecting gain - turns out, full gain is needed - easy enough to wire up, and I saved space by getting rid of a few more components. I might be able to get away with using less gain on the TS2012, but the volume controls in Windows allow plenty of control, and there is no significant amount of noise using this gain in the TS2012. I also made the decision to control "shutdown" of the audio amp. I noticed there was some noise when I shut the Z530 board off, so I'll control the audio amp in conjunction with the Z530 board from the STM32. This is probably going to be the best sounding bot for his size. Also pinned down the wiring from the harness for audio (pins seem to be backwards, but maybe I was looking at the connector wrong - wire colors are meaningless).

I also added a couple FET's for controlling laser power and mirror deflection. Also added controls to the MAX1614 to the STM32. Today, I need to add Z530 Reset and Power On connections to the board.

I still need to beef up connections / vias, but that's straightforward.

I got power routing on the control board improved quite a bit over what I had, so that's good (notice the fat traces on the control board).

I plan to put a ground trace between the I2C for the PWM controllers, I want to make sure I'm good at 400khz. The other I2C Bus (DC to DC converters, MPU-9150, BlinkM Mini) should be fine at 100khz (but you know I'll try it at 400khz!). I also moved the I2C traces on the power board, maybe I can keep them safer where I've moved them to.

I need to re-route the USB to UART chip (bottom left on controller board), but that's easy, I've got plenty of room.

The challenge right now is moving the 3A regulators (right in top board) up so that I can give them a better ground path - what they have at the moment isn't enough.

I also centered the MPU-9150 (bottom middle), makes sense, I think.

Also did many improvements to the audio section (bottom right), added ferrite beads and capacitors for longer wire runs (around 3 inches to the far side), improved access to power.

Overall, some things are very very tight in there. It's a constant fight to keep things spaced enough to keep the Design Rule Check happy.

Parts are going to be expensive, over $100 USD just in parts - much of that is in the DC to DC converters, but they're worth it - if they work as well as I hope they will!!

Image
An update... I got sick with the flu a few days after my last post, lost 5 days to it - tried to work on it, but didn't get very far - hazy from Nyquil and Sudafed. Didn't make my deadline of submitting the board for fab by end of vacation. So, I'm back to work, trying to fit in what I can when I can. Worked on the boards all day yesterday.

Up to now: I found a MAX1614 - perfect for controlling power (LiPo cutoff) with a large FET. And speaking of FET's, I found one that's 40A at ambient, up to 100A. I added two 3.3v regulators to the controller board, then widened the power board and moved them there. I've changed packages on many of the devices, meaning more parts added to libraries. Seems the parts I'm finding are either too new or uncommon. There are NO parts on the bottom of the controller board, and some optional capacitors under the power board that I may or may not need.

I've added a bunch of features, and it's getting close to fab-ready. If I'm very lucky, some time in the next 12 hours, I may be ready to send to fab.

It's coming along nicely. So much so, I have decided to do STM32 (Maple Mini - based design, like the controller board) for the hands. I'm thinking PWM for LED's on the fingertips - would be an interesting effect, and the STM32 certainly has enough hardware PWM for that and the hand servos.

I finally tested the TS2012 audio amp board from Adafruit with the speakers and the Z530 board - it sounds pretty good for not having the covers on the back of the speakers (it will when done, I just took them off and wired right to the speakers to test). The speakers and amp are a good match. I did this test to see if I could omit a few components for selecting gain - turns out, full gain is needed - easy enough to wire up, and I saved space by getting rid of a few more components. I might be able to get away with using less gain on the TS2012, but the volume controls in Windows allow plenty of control, and there is no significant amount of noise using this gain in the TS2012. I also made the decision to control "shutdown" of the audio amp. I noticed there was some noise when I shut the Z530 board off, so I'll control the audio amp in conjunction with the Z530 board from the STM32. This is probably going to be the best sounding bot for his size. Also pinned down the wiring from the harness for audio (pins seem to be backwards, but maybe I was looking at the connector wrong - wire colors are meaningless).

I also added a couple FET's for controlling laser power and mirror deflection. Also added controls to the MAX1614 to the STM32. Today, I need to add Z530 Reset and Power On connections to the board.

I still need to beef up connections / vias, but that's straightforward.

I got power routing on the control board improved quite a bit over what I had, so that's good (notice the fat traces on the control board).

I plan to put a ground trace between the I2C for the PWM controllers, I want to make sure I'm good at 400khz. The other I2C Bus (DC to DC converters, MPU-9150, BlinkM Mini) should be fine at 100khz (but you know I'll try it at 400khz!). I also moved the I2C traces on the power board, maybe I can keep them safer where I've moved them to.

I need to re-route the USB to UART chip (bottom left on controller board), but that's easy, I've got plenty of room.

The challenge right now is moving the 3A regulators (right in top board) up so that I can give them a better ground path - what they have at the moment isn't enough.

I also centered the MPU-9150 (bottom middle), makes sense, I think.

Also did many improvements to the audio section (bottom right), added ferrite beads and capacitors for longer wire runs (around 3 inches to the far side), improved access to power.

Overall, some things are very very tight in there. It's a constant fight to keep things spaced enough to keep the Design Rule Check happy.

Parts are going to be expensive, over $100 USD just in parts - much of that is in the DC to DC converters, but they're worth it - if they work as well as I hope they will!!

Image
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Re: Project: Melissa Hands - Perhaps a little easier..

Post by PaulL » Wed Jan 22, 2014 12:45 pm

Post by PaulL
Wed Jan 22, 2014 12:45 pm

I managed to finish up the board that Sunday, and sent it off to OSHPark for fab. Since then, I've got everything ordered, I'm just waiting on deliveries before I can continue. Parts and the stencil are here, I'm waiting on solder paste and the boards. If deliveries go well (bad weather here), I may be able to build V1 this coming weekend. I have enough parts to build two boards. We'll see how they go - there's a lot that has to be perfect for everything to work.

I've made a lot of changes, including adding an optocoupler to tie in to the board's Reset and Power buttons from the microcontroller, as well as adding the ability to read the board's HDD LED output. According to the documentation, this same header doesn't have a "Power" LED - it'd have been nice to have one of those. In a second version (not sure if I'll need one yet!), I'd set up the 4th optocoupler (it's a quad package) for power LED, then tap off the LED on the board if I had to, or experiment and see if any of the undocumented pins are useable.

Waiting on the boards and paste, then I can fab these boards. I'm hoping things go smoothly, but there are a lot of variables - who knows. Perfect first time out? Well, no - at the least, I forgot two resistors - I can add them, there's a spot on the board I can use if needed. I hope I can make v1 work, but there are quite a few "what if's". We'll see.

In total, there are more than 100 parts to populate for the two boards. The BOM export says 182, but some of those "parts" are just holes and headers.

I should add, I did get a miss-ship, two rather large surface mount capacitors instead of 2 diodes - makes me wonder if anything else is off - but that's difficult to check, some parts don't even have markings. Fingers crossed. First time I had a Digikey order that was wrong in any way. The diode miss-ship won't keep me from building them, though. I'll manually solder the diode when it arrives, one per board - it's for a non-critical function, so no worries.
I managed to finish up the board that Sunday, and sent it off to OSHPark for fab. Since then, I've got everything ordered, I'm just waiting on deliveries before I can continue. Parts and the stencil are here, I'm waiting on solder paste and the boards. If deliveries go well (bad weather here), I may be able to build V1 this coming weekend. I have enough parts to build two boards. We'll see how they go - there's a lot that has to be perfect for everything to work.

I've made a lot of changes, including adding an optocoupler to tie in to the board's Reset and Power buttons from the microcontroller, as well as adding the ability to read the board's HDD LED output. According to the documentation, this same header doesn't have a "Power" LED - it'd have been nice to have one of those. In a second version (not sure if I'll need one yet!), I'd set up the 4th optocoupler (it's a quad package) for power LED, then tap off the LED on the board if I had to, or experiment and see if any of the undocumented pins are useable.

Waiting on the boards and paste, then I can fab these boards. I'm hoping things go smoothly, but there are a lot of variables - who knows. Perfect first time out? Well, no - at the least, I forgot two resistors - I can add them, there's a spot on the board I can use if needed. I hope I can make v1 work, but there are quite a few "what if's". We'll see.

In total, there are more than 100 parts to populate for the two boards. The BOM export says 182, but some of those "parts" are just holes and headers.

I should add, I did get a miss-ship, two rather large surface mount capacitors instead of 2 diodes - makes me wonder if anything else is off - but that's difficult to check, some parts don't even have markings. Fingers crossed. First time I had a Digikey order that was wrong in any way. The diode miss-ship won't keep me from building them, though. I'll manually solder the diode when it arrives, one per board - it's for a non-critical function, so no worries.
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

Re: Project: Melissa Hands - Perhaps a little easier..

Post by PaulL » Sun Jan 26, 2014 7:27 am

Post by PaulL
Sun Jan 26, 2014 7:27 am

SUCCESS!! So far, so good - Power Board Assembled!!

I got the boards from OSHPark today, they look beautiful, just awesome.

I couldn't wait but so long before I put one together, so I did just that, today - I built the power board, the one that was sure to be more problematic (via-in-pad, atypical devices, etc), the one I had less faith in.

I've never reflowed a board before, but I have to say, it was much easier than I thought it would be. I had 3 locations that I touched up after reflow, and I'm not even sure they needed it, but they didn't quite look right (I think I saw bridges, not 100% certain).

As Digikey miss-shipped the diodes, I didn't bother trying to find a replacement to install - I can add that later. As for the input capacitors (3), they were rated for 10v (0805, TDK, 47 uF), so I just omitted them for now. I wanted to get these things up and running on my bench 12v supply, I'll deal with input capacitors later.

SO, I have 5v for Micro Servos, 6v (and they said the UDT020 was only good to 5.5!) for Main Servos, 5v for Logic and LDO Regulators, 3.3v Digital, and 3.3v Analog, all pretty tight on actual values. The on / off works as expected, the MAX1614 is doing what it should as far as the power board dictates. IT WORKS!!!

First rev, first run, first reflow, first success. Amazing.

As it turns out, I had some resistor packs on the PCA9685 PWM chip outputs, and I ordered the wrong ones - about twice as large as the board footprint, no realistic way to make them work. I have a few same size resistor packs on some boards, I might "borrow" them tomorrow and build up the controller board. I really don't expect any problems with that board, the components and schematics are based on known good designs. Meanwhile, I'll order the right RP's so I can swap them later.

As OSHPark ships 3 copies of every design, I have 6 boards. I tried a Kapton tape via-in-pad thing, ended up with more solder to wick off. This was on the same board I used for my first reflow, I only attempted this on one DC to DC converter, and that one works, so I did no harm. As these DC to DC converters have a fair amount of solder on them to start, I don't think I need to do anything for the via-in-pad's - two sets of pads for the other two converters were left alone, and they're working.

The only casualty of the day was a 26.7k resistor - not sure where it went. I pulled two out of a strip, but only one of those made it to the board. So, when I order the RP's, I'll order another 26.7k resistor. It'll go in a backup board, had enough to finish up the first power board since I ordered enough parts for two sets of boards (controller, power).

At this point, I haven't tried PMBus interfacing, so that could still be fouled - but I doubt it, my meter says I didn't short anything, and that part of the board is pretty clean.

There are a few things I learned from this:

* Reflow on a skillet WORKS. No doubt about it.
* Stencils are awesome, even the cheap ones from Pololu.
* You can even smear a bit with the stencil and paste and still have your design work.
* With caution and care, you can have a first design that works well enough to be the last.
* If I had to do another run of the power board, I'd space components out a bit more.

When I did the reflow, I used an Oster 12" Electric Skillet, with a piece of 5052 alu sheet, with the board taped to the alu with Kapton tape (high temp). I loaded the board with it taped to the alu sheet, made it easier to handle. I smeared some paste at the corners of the alu to provide a "melt" gauge (no thermometer here). I set the skillet to 420 (max), then waited for solder to flow. After all had melted, I let it sit for a minute or two before powering the skillet off and removing the cover.

Angled tweezers are a necessity. I bought some from Adafruit, and they work just fine. Also made for a nice jumper to bridge the yet-to-be-added external power switch.

A hint for those with shaky hands - use both hands, close in to the part on the tweezers, and try to drop the part in place, then nudge it to final position, using both hands steadied against the table to move the tweezers.

I used 0805 size parts as the smallest size on these boards. Trace width is 10 mils minimum, but the board manufacturer can go to 6. I didn't need 6, and figured I'd have a better board using thicker traces. The tweezers come in handy for peeling the cover off of carrier strips.

The 3 DC to DC Converters plus dual LDO regulators pulled around 300 mA at 12v, most of this is due to the efficiency drop relative to output voltage from the 3 DC to DC converters at 12v. When running off 7.4v from 2C LiPo, efficiency will be over 95% or so, and the regulators won't use as much current dissipated as heat.

Hopefully controller board tomorrow!!!

I'll definitely post some pics tomorrow as well.
SUCCESS!! So far, so good - Power Board Assembled!!

I got the boards from OSHPark today, they look beautiful, just awesome.

I couldn't wait but so long before I put one together, so I did just that, today - I built the power board, the one that was sure to be more problematic (via-in-pad, atypical devices, etc), the one I had less faith in.

I've never reflowed a board before, but I have to say, it was much easier than I thought it would be. I had 3 locations that I touched up after reflow, and I'm not even sure they needed it, but they didn't quite look right (I think I saw bridges, not 100% certain).

As Digikey miss-shipped the diodes, I didn't bother trying to find a replacement to install - I can add that later. As for the input capacitors (3), they were rated for 10v (0805, TDK, 47 uF), so I just omitted them for now. I wanted to get these things up and running on my bench 12v supply, I'll deal with input capacitors later.

SO, I have 5v for Micro Servos, 6v (and they said the UDT020 was only good to 5.5!) for Main Servos, 5v for Logic and LDO Regulators, 3.3v Digital, and 3.3v Analog, all pretty tight on actual values. The on / off works as expected, the MAX1614 is doing what it should as far as the power board dictates. IT WORKS!!!

First rev, first run, first reflow, first success. Amazing.

As it turns out, I had some resistor packs on the PCA9685 PWM chip outputs, and I ordered the wrong ones - about twice as large as the board footprint, no realistic way to make them work. I have a few same size resistor packs on some boards, I might "borrow" them tomorrow and build up the controller board. I really don't expect any problems with that board, the components and schematics are based on known good designs. Meanwhile, I'll order the right RP's so I can swap them later.

As OSHPark ships 3 copies of every design, I have 6 boards. I tried a Kapton tape via-in-pad thing, ended up with more solder to wick off. This was on the same board I used for my first reflow, I only attempted this on one DC to DC converter, and that one works, so I did no harm. As these DC to DC converters have a fair amount of solder on them to start, I don't think I need to do anything for the via-in-pad's - two sets of pads for the other two converters were left alone, and they're working.

The only casualty of the day was a 26.7k resistor - not sure where it went. I pulled two out of a strip, but only one of those made it to the board. So, when I order the RP's, I'll order another 26.7k resistor. It'll go in a backup board, had enough to finish up the first power board since I ordered enough parts for two sets of boards (controller, power).

At this point, I haven't tried PMBus interfacing, so that could still be fouled - but I doubt it, my meter says I didn't short anything, and that part of the board is pretty clean.

There are a few things I learned from this:

* Reflow on a skillet WORKS. No doubt about it.
* Stencils are awesome, even the cheap ones from Pololu.
* You can even smear a bit with the stencil and paste and still have your design work.
* With caution and care, you can have a first design that works well enough to be the last.
* If I had to do another run of the power board, I'd space components out a bit more.

When I did the reflow, I used an Oster 12" Electric Skillet, with a piece of 5052 alu sheet, with the board taped to the alu with Kapton tape (high temp). I loaded the board with it taped to the alu sheet, made it easier to handle. I smeared some paste at the corners of the alu to provide a "melt" gauge (no thermometer here). I set the skillet to 420 (max), then waited for solder to flow. After all had melted, I let it sit for a minute or two before powering the skillet off and removing the cover.

Angled tweezers are a necessity. I bought some from Adafruit, and they work just fine. Also made for a nice jumper to bridge the yet-to-be-added external power switch.

A hint for those with shaky hands - use both hands, close in to the part on the tweezers, and try to drop the part in place, then nudge it to final position, using both hands steadied against the table to move the tweezers.

I used 0805 size parts as the smallest size on these boards. Trace width is 10 mils minimum, but the board manufacturer can go to 6. I didn't need 6, and figured I'd have a better board using thicker traces. The tweezers come in handy for peeling the cover off of carrier strips.

The 3 DC to DC Converters plus dual LDO regulators pulled around 300 mA at 12v, most of this is due to the efficiency drop relative to output voltage from the 3 DC to DC converters at 12v. When running off 7.4v from 2C LiPo, efficiency will be over 95% or so, and the regulators won't use as much current dissipated as heat.

Hopefully controller board tomorrow!!!

I'll definitely post some pics tomorrow as well.
PaulL offline
Savvy Roboteer
Savvy Roboteer
Posts: 423
Joined: Sat Sep 15, 2007 12:52 am

PreviousNext
PreviousNext
135 postsPage 5 of 91, 2, 3, 4, 5, 6, 7, 8, 9
135 postsPage 5 of 91, 2, 3, 4, 5, 6, 7, 8, 9
cron