Dyne's Treadwell

[Dyne]

For the record, this was printed with Anycubic Clear resin, though I suspect similar clear resins would work just as well.

Treadwell's lenses are not colored, but you never know what people may try to build. I've seen ordinary food coloring being used to tint the clear resin before printing BB-8's holoprojector lens.  I know food coloring is water-based, but it seems to work, so that's something to experiment with if you don't want to buy transparent colored resin just for this sort of thing, or you want colors not available (like highlighter dye).

[savagecreature]

That's really cool!
Ever since the source of the lens frames was discovered I've been wanting to make a treadwell. Watching yours come together has been a good substitute!

[Dyne]

When last we left our intrepid droid builder, he was printing lenses and still waffling over what sort of drive system to use for Treadwell.  For a variety of reasons, absolutely no other progress has been made in the interim.



However, another Dragon Con has come and gone (two if you count the virtual con in 2020), and once again the urge to build has seized hold of his soul.  To be specific, the urge to build a droid that people might actually recognize.

So, yeah.  At Dragon Con this year, I decided that I would make every effort to get Treadwell built in time for next year, or at least a lot closer than he is now.  It may be a severely optimistic goal, considering the nearly 0 percent progress I've made on actually building him since I started back in 2016, but that's how I roll.

As mentioned, I was waffling on the drive system ... err ... three years ago.   I still don't really have much more of an idea on what the track drive might need than I did back then, so at some point, I'm going to have to just commit to something and hope it'll work.

I'm going with a brushless system using Oliver Steeples' brushless Jaycar replacement for R2 as a guide, as previously discussed. However, some things have changed since Oliver wrote the Astromech thread (and even since the part links were updated in 2019). Here is my logic and the parts I am leaning toward:

Unlike Paul's wooden base, I suspect that a styrene base Treadwell is likely to be lighter than a styrene R2, as there's less material to its body.  Given that Oliver's R2 drive was capable of propelling Oliver himself, I think (hope) the stronger brushless motors and the weight difference will be more than sufficient to counter the friction difference between the wheeled R2 and Tracked Treadwell drives.


* 2 ESCs @ $31.75 each:
These are the HK-60A speed controllers that Oliver listed for those who wanted to run higher voltages.  More importantly for my purposes, they are 60A, which may be necessary for Treadwell.  Unfortunately for me, the programming card for the ESC is currently backordered, with no ETA.


* 2 Motors @ $19.99 each (and 50% off the second motor):
The Multistar 268kv motors that Oliver recommended have been discontinued.  I found a variety of motors that might work on Amazon and Hobbyking (like the Turnigy Aerodrive 4250 350kv), but there were open questions with all of them.  Another option is the Rev Robotics Neo motor that Maxstang uses in his Maxdrive R2 drive system:






However, for now I found these iFlyRC 3508 380kv motors on ebay.  They are similar enough to the original multistar motor and at a low enough price that I'm willing to roll the dice to see if they are sufficient.  If not, I can likely use them elsewhere (such as my D-O concept droid).  And if they are junk, I won't be out all that much.  I just need to make sure the rest of my parts can support upgraded motors if I need something stronger.  Especially the most expensive components, the...


* 2 Gearboxes @ $67.25 each:
Because I chose a higher kv motor than Oliver recommended, I want to gear it down more than he did, in order to trade some of that excess speed for power.  However, as Oliver cautions, higher gear ratios also make the geartrain harder to backdrive and therefore the droid is more difficult to push around when unpowered.  I'm electing to go with a 20:1 ratio gearbox instead of Oliver's 16:1 or my original plan of 26:1.  I'm also choosing the "RS-5nn or 7nn" variant of the gearboxes, in case I need to upgrade to a motor with a 5mm shaft later (like the Aerodrive motor above).

If that's a good compromise between power and pushability, fantastic.  If not, I'm not too concerned with pushing him, since I could remove the treads and slightly tilt Treadwell to simply roll him around on his non-driven wheels.  I could also build in removable casters.  But Vagabond Elf on the Astromech forum pointed out that real life track drives have the drive sprocket off the ground.  If I built in a liftable drive wheel or a removable spacer, that'd leave the droid with his other 8 sets of wheels to be pushed on, once the belts were removed.

Then there are the sundries, like the shaft kits to go down from the motor's 4mm to the 3.17mm of the pinion, the programming card, etc.


I'm planning to do either a full RC or a dual RC/Bluetooth system.  To that end, I've ordered a Taranis Q X7 16 channel ACCST radio and a 2100mah LiFePo battery to power it (~$150 total).  Still deciding on what receiver to get.  (By the way, FrSky is currently controversial because of their new ACCESS and ACCST v2 protocols breaking compatibility with legacy and -- by no coincidence -- third party receivers. I went with this radio primarily because it was affordable and I knew it was compatible with the Kyber system. But I have no qualms about ditching companies that are too anti-consumer for my tastes, so if it becomes a problem I'll be happy to jump ship to their competitors. The RadioMaster TX16 was what I was originally looking at and seemed to be a great option, esp. if it also proves to be Kyber-compatible.

On the bluetooth side, I'm leaning toward Skelmir's Penumbra system, which uses ESC32 boards and their built-in Bluetooth.  These are a bit easier and cheaper to source than a Mega ADK + BT dongle setup.



If you are curious, Oliver's original 268kv motor/16:1 gearbox combo running at 3S LIPO nominal voltage of 11.1 would give an output of 185.925 RPM.  With Treadwell's 4 inch drive wheels, that equates to a speed of 3.56 kph or 2.21 mph.  The iFlyRC motors are 380kv and at 20:1 would have an output of 210.9 RPM, for a speed of 4.04 kph or 2.51 mph.  The Neo motors are 473 kv and at 20:1 would have an output of 262.5 RPM, for a speed of 5.03 kph or 3.12 mph.


Edit: OK, I've written this edit at least twice and haven't managed to post it either time.

The above parts list were in fact my intent.  As of right now, though, the only drive parts I've bought are the iFlyRC motors, and as mentioned before, I was willing to accept it if I ended up not using them.  Everything else is still in flux until the dollars fly out of my wallet and off to their new homes.  

Around the time of writing it was suggested to consider the Neo motors used by the Maxdrive, and while doing that, I remembered that brushless speed controllers need to know where the motor is in its rotation to time things correctly.   At high RPM, even a sensorless ESC can determine that info from the energy fed back from the motor.   But that doesn't help when the motor first starts up, because nothing can be coming back.  I'd totally forgotten that the result for a sensorless motor/ESC is poor startup behavior and acceleration at low rpm, esp. in high torque situations.

As a sensored motor with a sensorless ESC doesn't help that situation, if I switch to the Neo, I'd also have to go with some other speed controller (the Spark Max that is designed for it is Out of Stock, however), and it's an 8mm shaft motor so I'd need to rethink the gearbox too (or find a shaft kit).  Fortunately, BaneBots and AndyMark both sell compatible gearboxes.

[savagecreature]

Dude.
Awesome post! thank you for sharing all of that. I'll have to read it a couple more times.
I've been a FIRST mentor for ten years this season. It's so strange (and cool) to finally see all of our competition suppliers being listed in people's droid builds.
the Sparks will be back in stock soon, as competition season fast approaches.
The Neos have had mixed reviews among team competitors, so I find it particularly interesting how happy so many droid builders seem to be with them.
Anyway, thanks again

[Dyne]

Thanks!

I've never actually used the Neo motors (or any brushless setup, really). Do you know what problems people have with the Neos?


I received the iFlyRC motors, but since I'm lacking any other parts, I can't yet test them. But they do seem legit enough insofar as I can tell by looking at them.

I'm preparing a new post regarding my work on the Treadwell base design, but I need to flesh it out more.

[Dyne]

While designing the base for my treadwell, I'm simplifying the dimensions and angles a bit, just like I previously did for the head.  Since I'm building it out of styrene, I'm using a tab and slot design and designing for 3/16" styrene, though one could omit the tabs and slots and just weld the edges together.  If it works out, I plan to release printable plans much like folks do for R2.  



The intent is to skin the side panels and the sloped parts (not shown) with thinner styrene to cover most of the seams.  The sloped parts will be removable panels (held in with magnets).

One side effect of Treadwell's design is that the slopes mean beveled edges.  The only thing I've scratch built out of styrene sheet in the past is a shell for my MSE-4 out of very thin plastic.  I'm still trying to decide how much of a pain making the bevels will be.



For the stick portion of the Treadwell, I'm hoping to be able to incorporate a linear actuator in the base (if I can get one to fit) to get the extendible neck.  The actual throw distance needed is 4.25" according to the plans but that's unlikely without modifying the throw somehow.  More likely options are 4", 6", or 8".

The weight capacity of most of these actuators is generally far greater than what is needed to lift Treadwell's head, so the other important rating is speed.  The ones I looked at in the $30 to $40 price point range from 5 to 14 mm per second.


Regarding the drive, I doubt that I can get away with driving wheels directly from a gearbox like the one previously mentioned, as the output shaft wouldn't be long enough for the pair of wheels on each side.  Much like Paul, I'll probably have to transfer the power to the drive axles with a chain and sprocket setup.

This would, however, make it unnecessary to lift the drive wheels off the ground or even remove treads to enable the droid to be pushed if the gearing were too much; instead I could remove the chains, perhaps by building in a locking slider for the motor mounts.

Yes, I could also use a chain to do the gear reduction from the motor (saving the need to buy gearboxes), but 20:1 feels excessive for a single stage, and I don't want to make things too complicated.

[Dyne]

I've also been experimenting with the tank drive mixing on my radio, since I know I'll need it later.  The radio supports up to 60 different models (i.e. vehicles, droids, etc.) ... more with an SD Card IIRC.  

I know this will be old hat to some, but I haven't done it in the R/C world before (though I did my own implementation in Shadow for Artie Deco), and thought the explanation might be helpful.  So here's the inputs page of my Treadwell model:



Don't worry about the names of the inputs too much, I simply used an R/C plane model as a starting point.  For our purposes, just know that in this setup that Input 4 (Aileron) is the right stick's horizontal axis, and Input 2 (Elevator) is the right stick's vertical axis.  I will call these RX and RY respectively.

Input 1 (Rudder) and Input 3 (Throttle) are the equivalents for the left stick, which I will be using to control Treadwell's head.  They are LX and LY.


As you can see, I have three lines each for the RX and RY inputs, which correspond to the three positions (Up, Down, and Center) of Switch B, one of the toggle switches on the radio.  I'm using this switch to alter the maximum speed of the drive system.

Currently, all three RX lines are identical; I've been experimenting and haven't decided how I want to handle them at different speeds yet (see the end of the post for thoughts), so the switch position doesn't actually matter for this input.  There's just a consistent 50% scale on the RX axis for now.  The switch is only really affecting the three RY lines, and what those lines mean is this:

• If the switch is flipped in one direction, then give the mixer 25% of the right stick's true vertical value (for a slower than normal speed).
  
• If the switch is centered, then feed the mixer 50% of the true value of the right stick's vertical position.  This is what you'd typically use for a tank mix, so it's the "Normal" speed.
  
• If the switch is flipped in the opposite direction, then give the mixer the true value of the right stick's vertical axis (for a faster than normal speed).
  

I'm setting the numbers on the input page because I have three speeds.  If you weren't using a speed control switch like this, then you could just as easily leave all inputs at 100 and use 50 in the mixer settings below.  But because there are 3 "switch" lines per input like you see above, if I put the scaling in the mixer then I'd need 6 lines per input to do the same job (each input has to affect the left track and the right track, so 3 switch values * 2 tracks).

You can see the effect of my toggle switch by the number being returned for channel 1 (left track) and channel 2 (right track) in the images below.  In all three images, the right stick is at full forward (Center Top), and the droid would be moving forward at one of its three maximum speeds.



Moving the stick to the Center Bottom would simply make both numbers negative, and the droid would go backward instead of forward.


For the steering, we'll need to look more at the mixer settings:



The lines for Channel 3 and Channel 4 above simply connect the lefthand stick to the outputs that control the head.  (I'll have to set limits on the tilt servo - LY - later down the road.)

If you'll notice, both lines have the text "Slow: (U2:D2)".  This is not a complaint about some weird U2-D2 knockoff droid being slow.  U2 means that when the value goes up (U), then instead of instantly changing to the new value, the output will take 2 seconds to reach it from wherever it was.  D2 means the same when the value goes down (D).  The idea is to make the head smoothly accelerate and decelerate rather than instantly starting and stopping, similar to what James Bruton is doing in this video.

I also have similar slowing factor on the drive channels because I don't want Treadwell popping wheelies or charging to full speed instantly.  Both are problems I've had with Artie Deco when I accidentally hit the turbo button while driving.


The pairs of lines shown for Channel 1 and Channel 2 are what actually translate a single stick control input into a tank drive output.


To summarize how the mixer calculates the output, remember that the physical position of the sticks feed the named inputs (Elevator, Aileron, etc.) with values between plus or minus 100%, with the center stick being 0, and left or down being negative.

On the input page, those numbers get scaled.  RY will be Elevator (vertical axis), but scaled to a percentage based on Switch B.  RX will be half the Aileron (horizontal axis).

For the left track (Ch. 1), the output value from the mixer is RY + RX
For the right track (Ch. 2), the output value from the mixer is RY - RX

Note: Any output result beyond the range of plus or minus 100% will be clamped at that limit (So 100% + 100% = 100% output, not 200%).  This is why tank mixing typically starts with the inputs scaled down to 50%.


For example, assume the right stick is neither up nor down.  The RY input is therefore 0% regardless of the speed switch, because 0 times anything is 0.  If you then move the stick all the way to the left (-100%), then you are telling the droid to rotate on the spot by turning to its left.  Since the stick is at -100% horizontal, and this is being halved on the input page, RX equals -50%.

If RY is 0 and RX is -50, then channel 1 returns their sum, -50.  The left track goes backwards at half speed.  Channel 2, meanwhile, is SUBTRACTING the RX value.  0 - (-50) equals 0 + 50, which equals +50.  The right track goes FORWARDS at half speed.  The droid is now rotating on the spot.  

And when the stick is Left Center position, that is exactly what happens, as the image below shows.



The signs of the numbers would be swapped if the stick was instead in the Right Center position, and Treadwell would rotate in the opposite direction.


So what if the stick were left top instead?  Stick returns -100 horizontal, 100 vertical.  With switch B centered, both numbers get halved.  RX = -50, RY = 50.  So channel 1 is 50 + (-50) = 0.  Ch. 2 is 50 - (-50) = 100.  Droid turns left but in a wider circle.

With the other Switch B settings, RY becomes either 25 or 100, giving outputs of either -25/75 or 50/150 (clamped to 100).

If you look closely, you'll notice that the difference in channel 1 & 2 when the horizontal axis is at full throw is always 100 (more accurately, always twice the maximum RX).  This makes sense.  The same RX is being added to one and subtracted from the other.

This suggests that if I alter the RX scale factors, I might want it below 50 in both slow mode (so it turns slower overall) AND in fast mode (since Treadwell should turn more gradually)

With RX scaled to 25% in either mode, and stick in top left as before, we get outputs of 0/50 in Slow and 75/125 (clamped to 100) in Fast.

That's not without quirks, however.  Full left gives us -25/25 in either mode.  It feels odd that spinning in place is slower in fast mode than in normal mode (-50/50).  This is why I'm still debating the RX scales.

Maybe I should just have a fast mode line in mixer that exaggerates the turn rate as RY decreases.  Hm.

[savagecreature]

I think it's awesome you're planning on releasing plans. Can't wait to seem 'em

[Dyne]

I think it's awesome you're planning on releasing plans. Can't wait to seem 'em

It might be awhile before I finalize anything enough to actually release any plans, but yeah.  One of my goals is to document stuff enough to take out some of the unnecessary guesswork and hopefully encourage more Treadwell builds.


Pondering the telescoping neck mechanism.  There are several ways to go about it, but here's what I'm thinking...

The physical neck is mostly decorative (and hollow).  It might be made of short overlapping sections, or a few longer subsections that connect (if I went with aluminum pipe, I would probably do the former for cost, as shown here).




Either way, inside the physical neck there is a long hollow 1/2" shaft attached to the pistoning section of the neck.  If it were made of PVC, then the 0.8" OD of 1/2 inch PVC would just barely fit within the lowest and thinnest section of the neck.  If it's aluminum tube, which is still fairly cheap, then a couple of 1/2 ID bushings would be used to center it, and that's what I've modeled for now.  These would allow the shaft to both slide vertically (to raise and lower the head) and rotate around its axis (to turn the head).






The shaft has a slip ring at the bottom end (for the head tilt servo -- the arm wires don't need it). I know the original uses a slip ring at the top, but it makes more sense to me to put it nearer the point of rotation. The slip ring rests on a sandwich of two washers (I didn't bother to model these) with a thrust bearing in-between, something like McMaster-Carr 5909K24, which sit on the lift platform.  The washers are because the lift platform and slip ring flange wouldn't be precision surfaces.

This entire assembly can then be raised and lowered by some mechanism, be it a linear actuator like I was thinking upthread, a crank shaft powered by a wiper motor, or -- as I have it here -- a lead screw.  There should also be something to keep it reasonably level (which I haven't modeled yet, but could be something simple like a pair of drawer slides or even 3d printed rails).




The thrust bearing is there so the shaft can also be rotated by a friction wheel (similar to how R2's rockler is turned by a small wheel, at least on those that aren't using dome gears).  Thus, the head is both lifted and rotated via the same shaft.  This wheel and motor would probably be mounted in the region shown below, under the top plate of the logic housing between the two bushings.




I'm thinking I would probably need some form of omni wheel there to allow the shaft to slide vertically when the friction wheel isn't rotating.


Yeah, it's possible to raise and lower the head with a coaxial lead screw, but then rotating the head and routing the wires becomes a bit harder.

[Dyne]

Well, a few things have changed since my last post, none of it strictly Treadwell building, but it does have some relevance.

As mentioned in my LD-F1 thread, I finally got my printer set up in a location in the garage where the obnoxious flickering the heat bed's power draw causes A) isn't risking tripping the breaker or overloading the circuit, B) won't bother anyone, and C) is greatly reduced anyway because it's on a different circuit from the lights (which means I can also turn those off).

With the printer working, I can eventually get back to printing Treadwell parts (such as the Eye boxes I never got around to doing on my old printer, or the wheels). I have done some more Treadwell CAD, but it is relatively uninteresting stuff like remodeling the head mount in the new project and adding mounting holes for a pair of 4 inch speakers in the logic unit on top.


Also as mentioned in the other thread, now that I can print parts again, I've resumed my R2 build.  Once I started making more progress on R2, I ... went a bit crazy.  Among other things, I was seduced by the Aluminum Side, and now have a set of CS:R ANH Aluminum skins.

The reason I mention going crazy is that when I saw that the Spark Max finally came back in stock the other day, I ordered a pair, as well as a pair of Neos.  Should be here early next week.  Meanwhile, my wallet is crying in the corner.

Whether these ESCs and motors end up in R2 or Treadwell is up in the air at this point.  It's kind of a race to see which one gets to the point of needing motors first.  (While Treadwell is far less complex, it's also a less established build, so we'll see).

If these end up in Treadwell, then I'm still likely to be loosely basing the internal drive system on Maxstang's Maxdrives (I'm thinking shafts, pulleys, and so forth ... not so much the housing, wheels, etc.).  Mostly because I know it works for a heavier droid than Treadwell is likely to be.  I'm inclined toward the belt-driven version for R2.  Treadwell may or may not demand some tweaks or even the chain version due to the need for more torque.