Dyne's LD-F1 (D-O concept) build

[Dyne]

Current Status: Rolling & Repair, Debut complete

• Intro and Table of Contents - This post
  
• Initial Track Design - Post #2
  
• Sizing and Lift Geometry - Post #6
  
• Revisiting the build, Belt drive details - Post #9 (and 10)
  
• Lift system (neck design and geometry) - Post #11
  
• Build plan for Dragon Con 2022, Initial head mechanism - Post #13
  
• Upper Stalk structure - Post #14
  
• Actual Physical Evidence - Post #17
  
• Tank Tracks - Post #18
  
• On the topic of levers and trying to position a linear actuator - Post #22
  
• Front chassis design - Post #23
  
• Rear Chassis design - Post #30
  
• Idler structure - Post #32
  
• Larger nozzle and suddenly, a chassis! Details on prints, list of brass inserts and bearings - Post #36
  
• Printing the last parts, notes on wiring, assembly, starting painting - Post #37
  
• Painting, details on radio config, more wiring, the assembly progress - Post #41 (and 42)
  
• Mid-Dragon Con 2022 update - Post #47
  
• Post-Dragon Con 2022 Photos & To-do list - Post #48
  
• Side Plate Design - Post #50
  
• Weathering and Side Plate Printing - Post #54
  
• Side Plate Painting & Weathering - Post #59
  
• Prepping a future file release - Post #63
  
• Working antenna mechanism & updated wiring diagram - Post #68
  
• Musings on the top Bay Door - Post #73
  
• New Neck Design - Post #75 (and #76)
  
• THE GREAT REWIRING - Post #77
  
• Gear Drive Thoughts - Post #87
  
• Gear Drive Implementation - Post #88
  
• Gear Drive Test #1 (benchtop test) - Post #89 (and 90)
  
• Gear Drive Test #2 (failed the Turning Test) - Post #91
  
• Gear Drive Test #3 and thoughts on idler - Post #92
  

Intro

Since my droiditis has been flaring up and I'm never content with just three or four to focus on (Current list: BB-5M, R2, Treadwell, plus revamping Artie Deco), I thought I would also start musing about/planning one of the other droids high on my build list before more droids from Visions start to crowd in.  (Technically I already did some preliminary CAD for this last year but nothing very significant, so I'm starting from scratch.)

It's that D-O concept that I posted about in the D-O forum, the one by Luke D. Fisher (thus my chosen droid name LD-F1, though I waffled over whether to move the hyphen and call it LDF-1).

It's kind of similar to Treadwell, both being tracked droids with spindly necks and a head on top.  The head also resembles BD-1.


Decision Gate

There are a couple of decisions I know I'll need to make early on for this.

Size - The art suggests a smallish Droid, though to me it feels somewhat larger than D-O's final design.  I'll likely decide on a rough size for the radar eye and tracks that seems practical, then scale everything to that.  That coiled cable could also be a guide, assuming a phone handset cord size.

Treadwell's base without tracks is roughly 20 inches wide by 36 long iirc, and I think this droid's base is a fair bit smaller than that with tracks included (which are each about one third of the total width).  Maybe around 12-15 inches wide and 18 inches long?


Tracks - The art suggests that the segments making up the tracks are rigid on a hidden belt (see where they curve around the front), but design wise it might be easier, quieter, and more practical to make them just a flexible outer part of the belt.  Either way, I doubt I can get away with an off the shelf belt part like Treadwell uses.  Maybe something printable in TPU if it's small enough to fit on a print bed.

For simplicity, I'm treating the base and tracks as having a constant height.  They can also be read as having a slight taper front to back that isn't entirely perspective.  That could be more visually interesting for the droid, but I'm inclined to doubt that was the intent.  As an artist, I wouldn't have made those panels on the sides of the tracks look the way they do if it were meant to taper.  Tapering would also complicate the horizontal cover over the concealed portion of the neck.


Neck - The folding neck design is going to be tricky (I need to look more at the design Matt Hobbs and the Wall-E builders are using to deal with Wall-E's neck to see if it'll be applicable, as it does have a sort of 90 degree bend and lifts in a similar fashion).  Look around 6:17 here:





But I also have to resolve some minor artistic ambiguity.  To me, in the right-hand (lifted) illustration, the two arms of the upper neck read as being side-by-side due to how the main joint is drawn, but the upper connection in both images and the base of the neck in the left (non-lifted) image can all be read as one arm being in front of the other.

[Dyne]

I've been thinking about the tracks.

After playing with numbers and sketches, I decided that the dimensions I proposed before felt off.  The 18 inch length with the 3 inch wheels I was considering using felt too thin (6:1 length to height ratio), and the 12 inch width was too narrow (2:3 ratio of width to length).  Three inch wheels would also not leave much room for motors or other internals, because the bottom of the center section would need ground clearance and it also has those recessed greeblies on top.  

Treadwell uses 4 inch wheels, so I decided to go up to that size, then boost the body length to 20 inches and the width to 15 (5:1 length to height, 3:4 width to length).  I'm using a belt thickness of 1/4 inch for now.  Half of that is for the belt itself, the rest for the segmented outer portion.  Might reduce the thickness later if that isn't flexible enough.

As each belt occupies roughly 1/3 of the 15" width of the droid, I set them to 5 inches wide (the side plates appear to be entirely surrounded by the belts, so little or no additional width from those. I'll have to get clever to mount them).



Though only the front and rear wheels are shown here, it will have at least one additional set hidden in the middle.


Assuming that the front and rear wheels are the same size, and that the belt runs straight between them -- i.e. the middle wheels are not larger like Treadwell's are -- then the inner circumference of the belt (blue line on the sketch above) is:

• Twice the distance (once for the top span and once for the bottom) from the center of the front axle to the center of the rear axle
  
• plus the circumference (PI * diameter) of one of the wheels. (It's actually half the circumference of the front wheel and half of the rear, but those are the same in our case, so they add up to one whole circumference)
  

Then you can divide the result by PI to get the diameter of the belt as laid out in a circle.


With the new numbers, the calculations for the belt look like this:

Axle_Spacing = Body_Len - Wheel_OD = 20 - 4 = 16

Spans = 2 * Axle_Spacing = 32 (the top and bottom spans of belt between the wheels)

Wheel_Circum = Wheel_OD * PI = 4 * PI = 12.566 (the length of belt that's in contact with the front and rear wheels).

Belt_Inner_Circum = Spans + Wheel_Circum = 32 + 12.566 = 44.566 inches.  That is how much space you have to allocate to the outer tread segments and the gaps between them.

Belt_ID = Belt_Inner_Circum / PI = 14.186 inches.

Belt_OD = Belt_ID + (2 * Belt_Thickness) = 14.186 + (2 * 0.25) = 14.686 inches.  This belt would be printable on my Anycubic Chiron (bed size of 400 x 400 mm or 15.74 x 15.74 inch).  

Belt_Outer_Circum = Belt_OD * PI = 46.137


I decided to set the arcs that define each segment (the Segment_Length) at 2.2 inches, which looks like a decently rectangular aspect ratio given the 5 inch width of the belt.  Could reduce the segment length to improve that further, but I also want to preserve the relatively small number of segments.  It appears that the illustration is showing the droid with 18 segments per belt (I count 9 on the half of the nearer belt that's visible).  I'd prefer to stay close to this number.

Once you know your Segment_Length and the Belt_OD, you can draw that arc length in Fusion using this formula to calculate the angle between the edges of the segment, in degrees:

Segment_Angle = (Segment_Length * 360) / Belt_Outer_Circum = ~17.2 degrees




The maximum number of segments that you can fit on the belt is the integer portion of Belt_Outer_Circum / Segment_Length.  In this case, that's 46.137 / 2.2 = 20.971 segments, or a max of 20.  The gaps all come from the 0.971 that's left over (times Segment_Length) -- the lower this remainder, the tighter the segments will be spaced.

If you want to reserve a minimum gap between segments, then just add that to the Segment_Length before dividing.  So for 1.4 mm, convert to ~0.055 inches, then add it, to get 2.255 inches.  46.137 / 2.255 = 20.46 segments ... still a max of 20 segments either way in this case.  The nonzero remainder here means our gaps will end up wider than 1.4 mm (in fact they end up more like 2.7 mm).


From the sketch above, I first extruded the inner belt (the wider blue curve), and then the calculated segment.  I reproduced the latter feature with a circular pattern around the belt in order to get the model.

[savagecreature]

Very cool!
I'm sure this'll be an awesome one when completed.
I totally understand the having five droids going at once. That was me for *years*. However, I have turned over a new leaf and am only working on them one at a time.

[Dyne]

I totally understand the having five droids going at once. That was me for *years*. However, I have turned over a new leaf and am only working on them one at a time.

I'll likely get to that point (or at least down to two at a time) once I have droids recharging in my bed, sitting at my computer (I've already been there a few times), at my workstation at work (if only), etc. Sadly, it's pretty obvious already that my desire to build will not be limited by practical matters such as "finite storage space".

In any event, I posted two in-progress body pics to Twitter last night...

[r_saggio]

I've been pondering this droid a while (since the concept pictures were posted). For the size, I was thinking more in line with BB-8 scale, maybe because of the color scheme. The neck, lift, and head motions are going to be the trickiest part and what have been holding me back.

[Dyne]

I've been pondering this droid a while (since the concept pictures were posted).  For the size, I was thinking more in line with BB-8 scale, maybe because of the color scheme.  The neck, lift, and head motions are going to be the trickiest part and what have been holding me back.

We aren't too far off BB-8's bounding box.  This just occupies a lot less volume.

IIRC, BB-8's body is nearly 20 inches in diameter, and about 26 inches tall counting the dome but not the antennae.

With a base at 15 by 20 inches, LD-F1 has a similar footprint.  Then the height is the sum of:

• the head itself (I'm eyeballing at roughly 4.5 to 5 inches without antenna)
  
• the 4 inch height of the tank base
  
• the double piston neck length (8-10 inches judging relative to the base's height)
  

So the top of the head is circa 16.5 - 19 inches high. Here's a size comparison with a roughly BB-8 sized sphere:



On top of that we can add whatever height trigonometry says we can get from the bit of neck that's covered by the panel when at its maximum angle.  I'm figuring on something like a 45-60 degree upper limit.  Too steep an angle will mean the droid risks tipping over (or requires adding weight in the base to prevent that)

I've placed the mid neck hinge 5.5 inches forward of the midline, so the maximum possible length of the lower neck section is 13.5 inches (15.5 for the body aft of that point, minus 2 inches because the cover starts curving around a 4 inch circle directly above the rear axle, so the neck can't protrude further back than that.)

That lower end drops as it rotates backwards around the circle, so you'll lose a bit of height, but the most you'd lose is another two inches if the neck was completely vertical (circle rotated 90 degrees).

This lower neck sizing gives me a maximum head height of 25.46 to 27.96 at 45 degrees (depending on the height of the other parts) and 27.191 to 29.691 at 60 degrees.




As for the mechanism, yeah, it'll be tricky.  I'm still thinking Wall-E might be a good model to use for the lift ... if the mechanism can be adapted.

As shown in Matt's video that I linked in post 1, Wall-E has a linear actuator that pushes a slide at the base of his neck to raise his head.  You can see the movement around 1:40 here to see what I mean about it possibly being a good starting point:





I really need to touch base with Dave Ferreira (who designed that), or look around more on the builders club forum to see if there are any details beyond the blueprints for the external parts.  I suspect he has a parallelogram in the lower neck section and that'll be hard to fit, but maybe not impossible.

[kresty]

Just saw the thread, looks like a fun build

I was wondering for scale maybe looking at the antenna compared to other antennae? Or the wires, like the coil compared to wires like in D-O? I wonder if about D-O's height in the first position before the section of the base lifts the head higher? So maybe fully raised about BB-8's size - but far less massive?

[Dyne]

Just saw the thread, looks like a fun build  

I was wondering for scale maybe looking at the antenna compared to other antennae?  Or the wires, like the coil compared to wires like in D-O?

So far it is fun, though the neck is giving me fits. I have a post about that being drafted right now.

I'd have to go measure to be sure, but from memory I think the coiled wires that I have lying around (mostly from 12V car cigarette adapters) looked like they have coils roughly a half inch in diameter. I'm not going to go pixel counting the concept art, but to my eye it looks like those coils could plausibly be 1/16th of the height of the neck.

I wonder if about D-O's height in the first position before the section of the base lifts the head higher?  So maybe fully raised about BB-8's size - but far less massive?

That sounds about right.  D-O is about 18 inches tall according to Wookieepedia.  If you ignore the slanted section of the diagram above, you get 4 + 8 + 4.5, or 16.5 inches tall. I'm still playing around so this could vary, but probably not by a huge amount.



In vaguely build-related news, I FINALLY have a solution (?) for my printing woes.  

One reason I haven't been very active for awhile is because at some point last summer, I ended up abandoning use of my new FDM printer's heat bed (which meant I could only use PLA, and even that tended to warp if the part was big enough) because it made all the lights flicker and wasn't great for the amount or type of stuff running on the same circuit.  At best, it was probably asking for the breaker to trip.  At worst, it could've been risking a fire.

I've since cleared some space in the garage and moved it there, where there's almost never anything else active on the circuit (including the lights) and it's unlikely to annoy anybody even if they were flickering.  There are drawbacks to this, but it's better than not printing at all.

With the reactivation of my printer, I've also resumed work on my R2 build.  This weekend was back-to-back dome printing.  I wrapped up the last main sections of the Baddeley dome yesterday and I'm working on the second ring and pie sections.  This is consuming most of my remaining PETG, so I ordered some more filament (including white PETG for builds like LD-F1, and some grey TPU for the belts.

I plonked the LD-F1 belt model in my slicer with the profile for my current flexible filament ... each one would apparently take over 40 hours to print and consume 284 meters of filament. Woof.

[Dyne]

Well, after printing the remaining major sections of my R2 dome, I got very distracted by assembling, sanding, and priming them.  Then it got too cold to do outdoor work, and I got busy with the holidays.  I've poked at things here and there on my active projects in the meantime, but only recently started making real progress again.


Lately, I've been thinking about LD-F1's drive system.

Since I intend to use the 380kv iFlyRC outrunner motors I picked up last year, I need to work out some sort of drivetrain for them.  Using a 12V power system (well, 11.1V, the nominal voltage of a 3S LIPO), 380 kv motors can produce up to 11.1 * 380 = 4218 RPM.  That's 4218 * 60 = 253080 revolutions per hour.

For the moment, I'm going to ignore the droid's treads and pretend it runs directly on the wheels.  Adding the treads will slightly increase it's speed, but the difference is small enough that I'm happy to ignore it for simplicity.

The linear distance that these motors would move the droid per revolution is equal to the wheel's circumference, 4 * pi = ~12.57 inches  That makes it 253080 revs/hour * 12.57 inches/rev = 3181215.6 inches per hour.

1 mile is 63360 inches, so if we assumed the motors had enough torque to move it at all with no gearing, this would move the droid at a top speed around 3181215.6 / 63360, or roughly 50 mph.

Briefly...

I want the droid to be limited to something far more practical, like a fast walking pace (2-4 mph).  This suggests a gear reduction between 25:1 and 12.5:1 (50/2 or 50/4).

As we've discussed elsewhere, it's hard to backdrive much above 16:1.  This is probably less of an issue for chain drives and belt drives than it is for gears, since you can just disconnect the chain or belt, but that's a pain so it'd be best to stay below 16:1 nevertheless.

Commercial gearboxes are expensive as a rule, and if they sell any for these 3508 outrunners that don't need a lot of faffing about with shaft replacements and such, I haven't seen them.  So I was leaning toward a DIY transmission from the start.  I wouldn't trust a printed gearbox with a droid of this scale

(Edit: On the other hand...)





(End Edit.)

, and I don't want to fool with chains, so that leaves me with a timing belt drive (HTD-5mm feels about right as the profile).

I eventually decided that, due to the space constraints, I would at least need a two-stage belt drive.  Otherwise it'd require the pulley on the driven wheel to have a very large diameter to accommodate the ridiculous number of teeth necessary to get a useful ratio. (To get within my desired range with one belt, I'd need at least 12.5 times as many teeth on the driven pulley as are on the motor pulley, and there's a practical minimum on the latter's tooth count).


I could probably get away with using a smaller gear reduction if I had to.  Maxstang's Maxdrive (belt version) features a 12 tooth to 42 tooth belt -- only 3.5:1 ratio -- in order to roll his 150 lb R2 unit around.  (However, I note that he is also using Neo motors, which are beefier than these little outrunners and have nearly 100 kv higher rating.  The low ratio means he's also limiting the droid to a small fraction of the what the motors are capable of.)

I'd be surprised if LD-F1 ends up weighing even a fifth of what that R2 weighs, so my power requirements are a lot lower.  TPU tracks also aren't going to be nearly as grippy as proper rubbery materials like Treadwell's drive uses, so there'll be a bit less friction during turning.  Unfortunately this still leaves me guessing which gear ratio would be best.  While I potentially could go as low as 3.5:1 or even less, I probably wouldn't go below, say, 10:1.  The extra speed isn't of much use.




My initial design had a 10 tooth pulley on the motor (green), connected to the larger 30T input of the yellow pulley.  The smaller 10T output of the yellow pulley was connected to a 50T pulley on the wheel (red).  That's a 30:10 reduction for stage 1 and 50:10 for stage 2, or in other words, 3:1 and 5:1.  

Multiply 3 * 5, and you get a 15:1 total reduction, which fits within my 12.5:1 to 25:1 range, and leaves me room to move it further below 16:1 by altering the design.  

Going down to 42 teeth on the red pulley would lower the ratio to nearly the bottom of my range (30/10 x 42/10 = 3 * 4.2 = 12.6:1).  And that means I can use the same 42T Vex versapulley that the maxdrive uses, if I want.

I can fiddle with the pulleys to get down to 10:1 if needed.  Like 10-50,21-42 (50/10 * 42/21 = 5 * 2 = 10:1)

Modeling the pulleys in fusion for printing isn't too hard.  I kind of have to do that for the green pulley because the mounting needs are specific to the motor, and for the yellow pulley because it's a custom compound part.

Anyway, I'll continue to noodle at it (especially the neck mechanism, which is the other major holdup).

Edit: There's a fair amount of space in the "bays" forward and aft of each middle roller, and I will probably have to make good use of it.  At a minimum, the drive system will probably be tucked up under the tracks, as shown below.  Some of the remaining space will have to be occupied by supports that connect the outboard panels (the ones that cover the ends of the rollers) to the rest of the droid

The central chassis of the droid contains decorative elements as well as the neck lift mechanism and whatever portions of the drive system extend beyond the confines of the tread bays.  I also have to fit the microcontroller, ESCs, amp, speaker, battery, fuses, and switches somewhere.

[Dyne]

I've continued working on the drive system.  I think I settled on the 12.6:1 given by the 42 tooth pulley mentioned last time.  I've also been planning mounts, axles for the rollers, and the supports that I mentioned last time.

Speaking of the rollers, I knew all along that Treadwell's ... well, treads ... use a centering mechanism, and I also saw that Paul recently converted his to effectively become synchronous timing belts.  I recently started getting to the point where I needed to look at this for LD-F1.

That's when I realized I might have a problem finding a tooth pitch that would work with both the roller diameter and wheelbase I chose.

The crux of the matter is that the roller circumference and the tread circumference are both functions of Pi (an irrational number), so neither are rational numbers, much less integers.  Finding any tooth count that would divide evenly into both numbers seemed unlikely (it has to divide more or less evenly, because you can't have partial teeth).  Either the roller diameter or the wheelbase would have to be adjusted until I found something with minimal difference between the pitches.

Eventually, I ended up making a spreadsheet so I could play with the values to find something that worked.  I'm going to limit the values below to 6 digits of precision.

If I increased the wheelbase by 2.158 mm to 408.558 mm (so 16.085 inches), then I get a Tread Circumference of 1136.301814 mm.  (Roller Circumference wasn't changed, it remains 319.185813 mm.)  

If I put 25 teeth on the roller and 89 on the tread, then that gives me a roller pitch of 12.767433 mm and a tread pitch of 12.767436 (both around half an inch) -- a difference of 0.000004 mm per tooth.  If I multiply that by the number of teeth, then the roller has an overall error of 0.000089 mm, and the tread's total error is 0.000317 mm.

Those numbers are all well below the accuracy of the printer, which for my printer is 0.0125 mm in X & Y (and 0.0020 in Z).  So I don't see much point in trying to optimize any further.

There's another concern, which is whether the pitch chosen will affect the flexibility of the tread.  I can reduce that by doubling the number of teeth on both roller (to 50) and tread (178), which halves the pitch to a bit over 6 mm.  The error is also halved, but overall that doesn't change anything, since it's applied to twice as many teeth.


In the physical world, I printed a couple of test pulleys for the drive and also got a new set of ESCs for the iFlyRC motors.  I hooked these up to one of my ESP32 boards, with Penumbra installed for testing.  It worked pretty well. With the battery fully charged, I'd be getting more RPM out of it.  I only show one motor/esc here, mostly because I don't currently have an XT60 splitter to supply battery power to both ESCs at once and didn't feel like making one.





The timing pulleys shown on the motors here are the equivalent of the green one in the screenshots from the previous post, just with different attachment mechanisms.

(More in a moment; the forum will only allow one video per post.)