Archive for May, 2012

 

D.P.R. Chibikart: The 80/20 Principle

May 28, 2012 in D.P.R. Chibikart, Project Build Reports

There’s lots of variations on the phrase “80/20″ – usually it’s some corruption or adaptation of the Pareto Principle, bent and shaped to your specific industry. Legend has it that 80/20 framing got its name from having 80% of the strength of a solid bar of aluminum of the same outer dimensions, but at 20% of the weight. I think it actually weighs a little more than that, but even according to 80/20 themselves the name is just derived from the classical “80% of the results come from 20% of the effort”.

Now, in terms of hours spent building this version of Chibikart, I don’t think I’ve even come close to that. I’ve probably spent around 4 or 5 hours staring intently at a waterjet head versus maybe the same amount actually putting parts together and fabricating. And if I factor in the many hours spent stewing over little details in the design, like how to shove brakes onto this thing, I’m even worse off, probably closer to 20/80…

But what I’m trying to say is, 80/20 is awesome and being able to pitch so many parts of DPRChibikart together in a single day sure as hell makes it feel like I’ve gotten 80% of the way there…. if only. I’ve been taking some notes and modifying the design to make the assembly steps easier. I’ve also been finding that some of the parts I specified in the design are not as appropriate as I had imagined, and while I would totally just modify or custom-machine a part to fit the bill better, having to find a workaround that is a repeatable process by others is challenging as well.

First, a brief recap showing more Pretty Instruction Pictures and some more completed subassemblies.

I put together the “uprights” for the front wheels, and I must say that I’m really proud of these. I had expected some fit issues with the head of the bolt and generally keeping everything together, but once I tightened down the 4-40 screws which lock the layers together, and tightened the wheel spindle bolt, the whole thing was rock solid. I have no doubt this will be able to carry any reasonable load DPRChibikart will experience.

My intent with the eventual Instructable isn’t so much presenting a step-by-step of how to build Chibikart as putting together a sneaky way to present resources and techniques while applying them directly to an example build. I’m convinced that said resources and techniques will be way more useful to people than building the whole kart will be – it is going to cost a ton of money and you can readily extend the concepts to your own weird vehicle anyway. Making the uprights for a kart is generally one of the harder things to do – right angle part mates are almost always more difficult to make rigid (short of welding, I suppose), and this will be one more page in the book of tricks.

The rear motor mount/corner modules also came out as expected. A small amount (<10min) of file-fitting was required, but I think the tolerances on these slots is more than reasonable given my experience with different waterjets.

When the wheel is clamped by the spindle bolt, the whole thing becomes rock solid and all the parts are constrained by screw pressure – nothing can come loose and slide off, for instance.

One of the fine tuning details I was playing with was raw-waterjet bearing fits. Normally you would precision bore the hole on a milling machine to get a proper bearing press fit, but if you know the qualities of the machine you use well, then you can make a bore which is natively the right size. However, this is a very risky procedure because then your parts are not gauranteed to be duplicable on anyone elses’ machines.

Hence why I will elect to keep these steering kingpin bearing bores oversize than risk having them be too tight of a press fit, even though they were just right for me. Because these are flanged bearings and the kingpin will keep them tightly bound together, there is almost no need for a tight press fit. It could even be sort of jiggly to start with, held in place later by loading.

Cool assembly methods will not be the only thing in the Instructable. I intend to put a healthy amount of manual fabrication tips in as well – the first of which is DON’T TRY TO SUPERMAN-GRIP YOUR PARTS AS YOU DRILL THEM.

I was reaming out the kingpin holes in the uprights with a 1/2″ drill bit when it suddenly bound and whipped the whole thing out of my grip. Not bad damage at all (and I jut kept working), but I stopped to think a second about what I’m having people do with tools, and how tools can eat you.

The remainder of the holes were done with the upright securely locked in a vise.

After finishing as many subassemblies as I could, I started cutting 80/20. Normally this would be a horizontal bandsaw (“drop saw”) operation, but did I have a horizontal bandsaw in high school? Not really.

So I hit these with the Sawzall. Very brutal and surprisingly clean and straight. The “S” indicated that the section was cut starting from a clean square end of 80/20, which I deemed critical for some alignments. “NS”, of course, then stands for “Not Square” – a piece which won’t be involved in any end-fastening.

Alright, with all of the freestanding subassemblies done, it was time to to assemble the frame! Here’s a shot of everything so far, along with the ‘critical hardware’ needed for the procedure. This will probably be a theme in the Instructable: necessary materials/parts, tools, and fasteners will be listed, with allowable deviations also listed. If there’s more than one way to make a part, then I’ll try to discuss that too.

Magic happens, and here’s the whole frame. 80/20, anyone?

If I didn’t stop every 15 seconds to take a picture, this would have been an under-1-hour job.

For kicks, I rushed ahead a little and put the wheels on. This is not a step, but I just wanted to see how it looks compared to Chibikart proper (would it be the Republic of Chibikart?). I was missing some critical nylon and bronze washers at this point, so I called it a night here.

If you’re paying attention, you can clearly tell by the ambient lighting what work was done during the day and what was at night.

Another day! No, my left arm is not sublimating as I work. This is an “action shot” from when I was filing out the internal bore of a 0.75″ bronze bushing, cleaning up the result with some sandpaper. I should not have found it surprising that raw aluminum tubing is not made to shaft-fit tolerances. I had specified a 0.75″ OD 6061 tube for the steering column, but really it was more like 0.76″ or so.

While I wouldn’t have had second thoughts boring out the bushing on a lathe (and in fact did for Chibikart, just forgot), I realized I wasn’t allowed to do that. So, elliptical filing action it was. This was how they USED to make bearings, dammit.

I will most likely make a note in this step that intsead of using aluminum/steel tubing, McMaster sells precision aluminum shafting in 0.75″diameter which would natively fit in such a bushing. It would weigh significantly more, but aluminum is light anyway.

After the steering parts are assembled…

The brake pedal came out nicely. I only had one concern, and it was that the vertical pieces are way too close together. I specified their distance based on a scooter brake-holding-cross-drilled-screw-thing that I already had, which was pretty short. I then lost it, so had to source new ones from Cambridge Bicycle. It turns out that bicycle caliper BHCDSTs are usually alot longer, and I wanted most of these brake parts to be sourceable from bike shops for repeatability. I may make an untested design change for the Instructable in which the pedal itself is made wider to accomodate different length screws. It would also cut down on the sheer number of washers I had to stack between the plates…

And the front end is off the ground! Putting together the steering involves cutting some allthread, which I did using a hacksaw. Man, it’s been a while since I hacksawed the shit out of something.

Mounting the sprocket on the motor was a chance for me to finally use a mostly bullshit tactic that I had been saying to people to get them to stop asking questions. Yes, that’s a chunk of a soda can hanging out of the sprocket there. I’d jokingly suggested to people before that soda cans in fact are sources of precision shims from 0.004″-0.006″ and that they can be used as crude metric-to-imperial bore converters if needed.

Well, I’m not joking any more. I measured the can wall of a Sprite Zero at 0.0045″ +/- fuck, this isn’t 2.671 near the 1/3 mark. 6mm is 0.236″, and 0.25″ is… uhhh, 0.250″. So there’s about 0.007-0.008″ per side to make up, or 2 full turns of soda can. Procedure: Cut small strip of soda can, wind into coil, stuff into sprocket.

Note that the thickness increases near the endcaps due to the drawing process, so if you need a thicker shim, you can cut closer to the ends. The can walls are thinnest in the middle, which according to my old 2.671 Soda Can lab paper is about 0.0041″.

To add to the list of things I’m really proud of on this build, these half-caliper half-railcar brakes came out pretty awesome. The circular notch holds the little cable-end-ball-thing of a road bike brake cable. A general cheap road bike brake pad is the friction element, and the cable-tension-adjusting-hollow-bolt-thingie is also sourced from a road bike caliper brake.

I spent an incredible amount of time just learning what these specialized bike brake parts were called. I’ve gathered that:

  1. “cable-end-ball-thing” is  a nipple
  2. “cable-tension-adjusting-hollow-bolt-thingie” is a barrel adjuster
  3. “brake-cable-holding-cross-drilled-screw-thing” is an anchor bolt or pinch bolt.

So yeah. If you want to pre-buffer parts, go get two standard road bike brake cables with at least one nippled end, two smallish brake pads, two barrel adjusters, and two anchor bolts. Use the proper terminology so your bike shop guys don’t give you the “wtf?” face when you literally use those Buffy-speak names.

Clamping the cable this time is much, much easier compared to the fiddling I had to do with Chibikart’s brakes, which didn’t use brake-cable-holding-cro ANCHOR BOLTS. The waterjetted cable sleeve holders also worked out very well, requiring no drilling or messing with.

BAM! Rolling frame.

The brakes on this are obnoxiously good. Probably because I overcompensated since Chibikart proper had no brakes to speak of until 2 weeks ago. This will lock up and skid with 65A durometer Colsons, which means if I fit hard wheels on the back it will drift readily. This is exciting.

What’s left but to have your friends push eachother around while looking silly?

Here’s what’s left.

  • Assemble the electronics plate. This involves waiting for a new shipment of Jasontrollers, which should arrive this week. If they don’t come by Wednesday, I might actually knock two Jasontrollers off Chibikart for now just to get it over with. Because I want to ride it. Badly.
  • Put all the pictures into rough assembly order and make smaller sizes  of them, because there are currently 235 of them. Not all will be used, but most will be!
  • Write the damn thing. I need to start now if I have any hope of finishing before the Make It Real Challenge is over!

D. P. R. Chibikart: Well, I did the hard part.

May 24, 2012 in D.P.R. Chibikart, Project Build Reports

The first step of the Great Proletariat Revolution of Chibikart will actually be sort of bourgeoisie. I’m running a few waterjet experiments to determine the best tolerance for parts – since while I can totally just throw the stock files out there, it’s not very useful directly. With laser cutters, which cut “on the line”, slots are automatically made bigger and tabs made smaller. Waterjets, though, make an attempt to compensate for the kerf of the stream – which is usually 0.030″ or more for larger machines, and changes with distance of the nozzle from the surface, wear of the nozzle, and other factors.

For instance, to get 1/4″ thick aluminum tabs and slots in parts to fit without sanding and filing, I have to change the nozzle offset from 0.014″, the default setting, to 0.012″. The machine I most frequently patronize does not have taper-free cutting accessories, so the extra 0.002″ per cut is just due to the stream changing shape. Using a commercial service would mean the offset is not under my control, so I’d have to design extra loose – just how loose is something I’m trying to work out.

Additionally, piercing and traversing (non-cutting movement) all factor into machine runtime and hence expense. One thing I’m trying is combining multiple parts into a single “metapart”:

Those are some of the brake parts combined into one profile. For small parts, this also prevents them from falling into the tank – while I’m sure shops will take care of this on their own, it’s one of those things which will make their lives easier.

This operation is actually not easy – I have to make the little tabs as a separate part (all of them at once using the other parts as adaptive references), then put them into a 2D drawing which I then edit in AutoCAD to get rid of unnecessary lines. It’s kind of a pain to change things, since AutoCAD imported drawings aren’t really parametric.

Here’s an example panel which Big Blue Saw‘s rendering engine made of the panel of 1/8″ parts:

The quoted cost of this panel was $138. This is actually an incredibly good deal. Let’s say that I buy a panel of 24 x 24″ 1/8″ 6061 aluminum from a reasonably priced metal dealer – that’s about $40 including shipping, typically (not counting surplus deals and random sellers as possible sources because it’s too nondeterministic). The only really “public” machine on campus is the MIT Hobby Shop, which charges $2 per minute of run time if you’re a student and $3/min for funded/departmental project (this, too, is an incredibly good deal at $120-$180/hr). This panel would take at least half an hour to cut

So if I actually had to buy everything, even cutting on campus is already exceeding what Simon can provide me without the hassle of dealing with a big panel of metal and working in between machine maintenance – since being a public machine, it is prone to n00b damage.

Let’s face it – waterjetting is expensive as fuck if you actually have to pay for it and do it in small quantities.

I wanted to get the riskiest part out of the way first, and that was the set of sprockets and hubs (and the whole wheel assembly in general). These parts were all cut using standard offset – no magic was applied.

The sprockets in particular are actually not raw generated profiles. I “profile shifted” them inwards by 0.005″ during modeling in order to make sure they come out either on-size or even slightly loose. Historically my waterjet sprockets need a significant amount of ‘running in’ because the extra material from the taper would cause the rollers to not seat completely. That is something I can not assume others will be able to (or have the patience) to do, so it is a good example of the kind of pre-compensation that is going into these parts.

After breaking the sprues apart, I wrapped a #25 chain around it and… hey, it works. The fit is pretty snug, indicating there can be even more shift if needed. I’m having a hard time believing that the thing is cutting nearly 5 thousandths out of spec – this alone might make the experiment invalid as no commercial machine being revenue-employed will be allowed to run that far out of whack.

This is one example “nice instruction picture” which might make it into the final Instructable. Yes, there is tapping involved. Start practicing now.

And another! Chamfering your sprockets by spinning them haphazardly in a drill press while attached with a 1/2″-bolt to the chuck. If there is one Instructable which could get thrown out for encouraging dangerous behavior…

Here are the completed test wheels. This took less than an hour, and is actually the most complex manual fabrication job on the whole thing – save for wiring. Everything else ought to be plug-and-play or assembly only, as far as I can tell.

The tools involved were a drill press, cordless drill, tap (and handle), 1/2″ (to clear out the sprocket bore), #29, and #42 drill bits, metal file, and some 4-40 x 1/2″ cap screws. Oh, and a 1/2″-13 bolt.

At this moment, I’m out of 1/8″ aluminum to finish the rest of the 1/8″ frame bits, so I’m going to experiment next on the critical 1/4″ parts such as the steering knuckle blocks. I may be able to assemble the brake lever soon, too, but it depends on what 1/8″ aluminum I can scrap together. I should have more metal coming later this week or very early next week.

 

We Interrupt This Regular Scheduled Update BECAUSE REPLICATOR

May 23, 2012 in Beyond Unboxing

About 2 months ago exactly, I commissioned a Replicator for our research group because I at that point was clearly never going to get anywhere with my plastic-pooping EZ-Bake oven. Several weeks (about 8, actually) passed, a few seedlings did some burnouts in a parking garage, I allegedly invented Mario Kart (again), and a bunch of other stuff happened and I kind of forgot about it.

Then this showed up.

Oh dear. Well, I’d gotten cryptic emails a while back regarding awaiting some kind of special shipment…

Alright, it’s time for another episode of BEYOND UNBOXI…well, I guess i’m only unboxing it right now, so nevermind.

With the top layer of packing material removed, I SEE THE THING. Glossy printed setup manual and cut-to-shape cardboard packing structure? Makerbot is getting so legit that it’s still funny because it’s awesome.

THE THING. GUYS. IT’S THE THING.

The whole process of unpacking it speaks to how much effort was put into packing it in the first place – there’s alot of well fitting custom packing material. I guess if you’re shipping a completed and tested machine there’s no other choice. The platform is all pre-coated with Kapton and there’s a free roll of it included.

I presume the little calibration dongle was made using the machine itself before shipment.

The level of engineering in this thing is leaps and bounds over the Cupcake and even the Thing-o-Matic, as far as I can tell by eye (and by jiggling axes). Injection molded brackets and bearing holders! It looks like the focus has shifted way from ‘kitting’ to integrated, tested machine – which, IMO, is probably better. As my past adventures in designing for easy lasering and waterjetting, and DPRChibikart’s build process is showing again, there’s alot of optimization potential and performance you sacrifice by restricting your build to a certain process or making it so generalizable that anyone can put it together. For silly vehicles, my view is that alot of these compromises are acceptable because there’s invariably many solutions to the same problem; I don’t think the same is true for machine design, especially machines relying on precision and repeatability.

Anyways, this new gantry seems rock-solid and ripe for serious overclocking.

OH, DID I MENTION IT’S NOW A GANTRY HEAD

One of the things I didn’t like about MaB from the start was the fact that I hurried through the design and just kind of copied and pasted what everyone else had going on at the time. The moving bed design really sucks because the axes have significant inertia and you accelerate the workpiece itself, which is bad if it’s remotely heavy.

The little details are great, like these integrated spool holders and all the cable snaps.

It even comes with allen wrenches. Guys, this is just like IKEA.

(The little black things are rubber edge bumper stock cut up to make convenient legs for the machine)

With two screws, the dual head extruder pops on. We went ahead and sprung for the twin head (can you say profit margin?) in case fully integrated support material becomes a thing in the near future, which I am positive it will. Right now, you still have to “merge” two separate STL geometry files within the software (ReplicatorG) to use the dual extrusion feature. More legwork on your end, but legitimate dual material is possible right now, such as ABS + PLA or similar.

Because I think we’re mostly going to be making machine parts and robot dong(les) with it, I’m more interested in integrated streamlined support material deposition than the ability to print a blue and green world. That would make this thing roughly 90% of a Stratasys uPrint at 10% of the cost.

If Makerbot were steampunk, this would be a cast-iron or forged brass badge with THE REPLICATOR MAKERBOT INDUSTRIES BROOKLYN N.Y. U.S.A. in fat script or squared off block letters arranged in circular outline. Or if it was really hardcore, just straight across with no stylization whatsoever because your machine is too badass for cute logos.

Because old machine badges are awesome.

After setting it up and powering on, the machine has a first-startup script that tests the extruder and helps you level the build plate (which is fully retracted for shipping). It also changes colors as it heats up – the ‘underglow’ is blue for cold plate or extruder, gradually fading to red when it is fully heated.

Well, technically that’s the opposite of what a black-body radiator (“heat source”) would do, but just like conventional current vs. electrons, who’s gonna argue…

IT’S POOPING

IT’S POOPING

I tried one of the dual extruder files that were included on the SD card (whose slot I searched on all 5 sides of the machine which did not have the SD card slot) just to see if I did the levelling thing right.

While I guess I did, the actual level that this sets seems to be too high. This file was the fish looking thing, and it didn’t end very well. Any other Replicator owners notice this? When printing the bottom most raft layer, I’m used to seeing the head mashing a thick track of plastic into the build plate – with the ‘one paper thickness’ first-run platform height, it seems to be nearly .5mm too high, and even the base layer sort of just squirts onto the surface but doesn’t get spread out at all.

I kept adjusting all of the leveling screws until the ABS trace was more like what I was expecting. This involved moving the plate up at least another full turn and a half, or something like .75mm if I estimated the screw pitch correctly – I was afraid that the nozzle was going to plant into the build surface, but it seems to have a preset altitude.

The RepG software is infinitely easier to use this time around. This is the first time I’ve used the “Print-o-Matic” feature in person, and I must say it makes the user experience more intuitive.

It looks like the default infill is scribbly hexagons.

I swore I would make no setting or hardware changes to the machine to appraise it for its out-of-box funcationality, but I couldn’t resist. The default full-fill setting extrudes using only one axis – either X or Y. I’ve historically favored the ’45 degree’ method which ensures both axes move simultaneously to draw a fill line, because it forces any axis inconsistencies to average out through vector addition – keeps things symmetric. Therefore, I changed the “infill direction” to 45 degrees instead of 90. It’s like printing X’s instead of +’s.

So that’s two things I’ve deviated from so far – platform trim height 0.75mm closer to nozzle after the levelling script has run, and infill direction to 45 degrees. Not bad yet…

Alright, so let’s make that three.

I’m still not sold on the whole “hot kapton tape” surface, unfortunately. It’s included with the machine, it looks great, but I just couldn’t get it to stick well at all. I tried cranking the platform temperature to 110, then 115 celsius, but the results were not much different (two 6%-bunnies died for this cause). Maybe it’s just extra cold up in Boston or something  – MaB had never fared well during winter either, and the shop was about 55 degrees that day due to leaky windows and a friendly cold front.

So I did the THIS IS HOW I DID IT IN MY DAY thing and dumped some blue tape onto it. Sorry Makerbot D:

Quantiatively speaking, I could get about 6mm of bunny out of the blue tape before signs of raft warping showed up, whereas the raw Kapton gave up after what seemed like only 4-5mm. I let two half-bunnies print to see at what point the prints fell off – the blue tape bunny detached around 60% whereas the raw Kapton bunny only made it to 45%. This was all during the chilly evening, so I’m sure ambient temperature played a role as well, but an equal role for both.

Yes, I watched this thing print bunny asses for about 4 hours. In the name of Science, I swear.

While I could have tried roughing up the Kapton with sandpaper, seemingly a common tactic, I’d have to do that every time the surface needed to be replaced. I’d much rather just rip off more blue tape. To me, anyway, it makes more sense that molten plastic would tend to stick better to a porous and micro-rough surface such as the painters’ tape. A Stratasys build tray is actaully very finely textured if you look up close.
In other words, I’m going to need alot more convincing before accepting hot Kapton – even though it seemingly works for alot of people, I suspect there’s things which go unreported to maintain the validity of the ‘community solution’.

Alright, after the aforementioned minor finangling, the FIRST CALIBRATION BUNNY! The effects of the less than warm room (which is also pretty drafty) can be seen in some of the split layers. Extrusion settings for this run were bone stock, though, and the results are impressive. There was absolutely no tuning of extruder settings or making 10 calibration cubes to reach this stage.

To alleviate the breeze problem, I might consider walling off the sides and front (with a door in the front) like some people have. Breezes and incidental wind gusts are what do the most damage to these unshielded prints, and just the turbulence of you coming up to stare at it will generate a puff of cold air. I’m not inclined to try and turn it into an oven because some of the internal major structure is made of ABS plastic, and to stress-relieve ABS being held up by ABS will not end well.

It’s not my machine to hack, technically, but maybe if I plan it well and make it reversible I can make aluminum replacements for the ABS moldings and then heat the interior up to 50-60C.

Also, it’s been christened Fatbot, because it’s pretty wide.

I’m going to harp a little more on the development of the direct-drive stepper motor controlled extruder and its associated control hardware/software from even a year and a half ago to now. This amount of detail on the tip of the ears is phenomenal, and done without support structures. This just about beats the uPrint machines we have already in terms of finish quality. That’s a ~2.5mm long single line at the very tip.

So to recap… I think this contraption is awesome. But I’m not so sure on some of the very fine details that I took note of immediately which novice buyers and users might miss on – such as the seemingly extra low build platform and the inconsistency of hot Kapton. Any machine can be made to work much better once nudged enough, and while I didn’t have to nudge it as much as MaB, I’m wondering how these tests would have ended up if anyone else in this lab space – who don’t necessarily follow the hobby 3D  printer scene – were the one unpacking and testing.

There is a Calibration Naked Lady printing right now, overnight, without supervision (just like the manual told me not to do). We’ll see how it is in the morning!

Chibikart’s New Unobtainium-Free Sibling

May 21, 2012 in D.P.R. Chibikart, Project Build Reports

Hi. I’m building another Chibikart.

But this one will be just a little different. Historically speaking, most of my projects have involved a little bit of “unobtainium” – or perhaps to put it more accurately, an Unobtaining Machine. As well documented as they tend to be, they are admittedly difficult to reproduce by anyone outside of a university or similarly well-equipped makerspace setting, the reason being that they tend to involve much manual or CNC machine work; waterjetting, laser cutting, precision boring, etc.. I’ve generally made a attempt to explain the processes involved in addition to investigating ways that things can be built by more people with different resources, but it isn’t usually the focus of a project.

I support the methodology of getting up and building something as an effective way to learn engineering. Part of my research here (if it could be qualified as that yet) is to engage more  students in learning engineering and design concepts using electric vehicles – it was the whole motivation for 2.00EV this past semester, an experiment that I consider to have been successful (and useful as a data point to work from for future iterations of the course). Working with a more life-sized system, in my view, makes you understand the mechanical engineering concepts commonly taught during the 2nd year such as bending/flexing, bearings and constraints, power and work, etc. more than building a small robot would. The numbers are in a range which make more sense to think about – you understand more what 10 pounds-force is like than 0.1 lbf, or why your vehicle can’t have 1/4″ aluminum axles. You can’t bend 1/2″ aluminum vehicle frame plates to suit your misaligned bearings, but 1/32″ sheet metal robot frames just don’t illustrate that well.

And that talking in the customary units system is fun and all, but Charles will make you write everything in SI because things make more sense that way.

While “2.00EV” engaged the students in learning the machining, designing, and analytical skills  that MIT typically expects out of sophomores, I’m wondering if a slightly detuned version would be accessible to more people – younger, older, different university, or just out to learn & build.

Through the activities of the Collegiate Silly Vehicle League, we have both discovered and created a sizeable array of resources and links for collecting knowledge about buildables. Many of the vehicles that we have built (and zip around on) have their own web pages and long, detailed build reports – so many that listing them here is going to be impractical, so feel free to click through the links in the left sidebar. Some of the principles and sources of parts have been amassed together – for instance, the ‘scooter instructable‘ and my very own ‘hub motor‘ instructable. These have all been guides more so than telling people how to build one specific thing in the spirit of Instructables, but both their 5 star ratings and the emails I get related to them speak alot. Some times, knowing the resources is all the push that someone needs to get started.

Despite pushing several billion dollars a year in revenue, very few people know what McMaster-Carr is, or that they deal directly with end users and customers. When they find out for the first time, it’s quite daunting because there’s everything there and how are you going to find the thing you need? Shopping on McMaster (and related enormous catalogs like DigiKey, for example) is a skill which is most definitely not taught here at MIT, at any level, and I’m definitely not okay with that. It’s something you have to pick up by going above and beyond the call of handing in problem sets. That’s why I made it a point for students this year to order things from McMaster that they found by themselves, but I digress.

Very few people anywhere also know that you can in fact have parts custom made for you by places like Big Blue Saw and ShapeWays, just to name two. There is an interesting “uncanny valley” of personal fabrication here – the general nontechnical public, of course, does not know about the existence of it and online rapid prototyping services. Yet the university and academia crowd also tends to not know about it, from my surveys, probably because if you needed an Unobtaining Machine it is usually right there. In the middle of that is the realm of makers and amateur engineers for whom these services are a boon. (The other big reason is that personal fabrication is still quite new, so it’s unfair to equate it to established systems of multi-shop quoting and contracting machining services.) Oddly enough, even as some people complain that infinite waterjet is the hard part of my projects to replicate, it is actually the milling and turning which will stump most people – there is as of yet no equivalent fast service for custom 3d-machined parts.

I also believe it’s important that you not immediately drop the textbook on people from the very outset. I sometimes betray a disapproval of ‘academic engineering’ when I talk to people, a fact that I used to be ashamed to admit in a place like the MIT Department of Mechanical Engineering. About 10 years ago (…. whoa…) I started out building things from the opposite end of the spectrum that academia is used to – I didn’t learn mechanical physics, structures, beam bending, electric motor dynamics, etc. first and then go out and build a robot – I started throwing junk together and then wondering why it didn’t work, then went to local robotics clubs and tournaments and asked people where they got parts from. Alot of reading online (and real books! Even in 2002 the Internet wasn’t as everywhere as it was today). While that sounds like alot of “Well it worked for me…”, I understand from personally knowing many people who also became interested in engineering early and then subsequently were driven to excel that relevant experience is of serious value even if you cannot explain fully to a professor why something works. You incrementally… build… both experience and knowledge, which are two important qualifiers of competence people conflate alot.

Building for your own satisfaction and learning is also discredited alot – both academia and the curiously onlooking public contribute to this effect. I’ve been asked countless times why I built something when it’s already on the market, or already been invented or done. I’ve been asked “how is it useful?” both facetiously and seriously. Other students I have known in the past years have told me they don’t really build projects because they don’t think they can invent anything new. It’s a hard mentality to change, especially in academia where cutting-edge research is going on all around you and you feel pressured to take as many advanced classes as possible to get to that level so you can contribute. With the growth of MITERS and awareness of projects on campus in general, I hope this is something that will improve in the future. No matter how much we joke about the MITERS “build fads” – robots, scooters, motor controllers, quadrotors, tesla coils, go-karts… (roughly in that order actually, from my recent memory) the fact is that one person starts out building something and their peers look on, go “hmm, I want one too”, and many people benefit from the learning experience in the end. Even though we’ve made literally 5 of the same thing like 4 times over. So what? There’s now 4 more people who can, for example, more confidently design a power converter or mixed-signal closed-loop controller than who could before.

Is it just me or do I insert a long rant or philosophical essay before every build post now?

So, finally getting to the point. What does “detuning” mean for this project? It will mean fully documenting online the build process of a small electric vehicle using commonly available but modern parts and a reasonable set of garage tools and hardware store purchases, while utilizing available digital/online/personal fabrication resources. It will not feature cutting-edge science nor attempt to teach principles of engineering simultaneously. The goal is a fun buildable project of more difficulty and exploration of the hidden underbelly of engineering than a trip to Home Depot will net someone. Again, some times all you need is the resources and the technique – I’m guessing individual sections of this documentation will be way more useful to more people than the whole project is.

Now, Chibikart is probably the worst possible project to do something like this with. First of, it’s not going to be cheap at all – it’s not a weld-together-a-steel-frame-insert-scrap-lawn-mower-engine-here kind of project. It’s also substantially more complicated than the aforementioned WTSFISLMEH go-kart. I decided to pursue it, though, both because it has novelty factor (while it may be impractical it’s fun as hell to ride around), is made of a pretty adjustable/extensible platform (80/20 extrusion), and because it is harder. It’s abundantly clear to me that not everyone will be able to pick up and do it, but in my opinion, to throw it out there is better than just relegating it to a build category in my sidebar that I will occasionally linkdump to.

Let’s set the ground rules:

  1. fully documenting online the build process“. I will use Instructables and create a step by step document that shows the vehicle’s construction from start to finish. This is a departure from my usual style, but I will try to keep things as open ended and customizable as possible.
  2. a small electric vehicle using commonly available but modern parts.” No old car batteries, galvanized frame tubing, and hacked wheelchair DC motors. This will be brushless, lithium ion, aluminum framed, and powerful & lightweight using parts that can be purchased new. Much McMaster, Hobbyking, and Surplus Center (with other minor vendors) will be involved.
  3. “a reasonable set of garage tools and hardware store purchases.” This will be the biggest constraint on technique. I will limit the use of tools to those which could reasonably be found in someone’s garage, or the average high school builder had access to. For me, it was a drill press, hand drill, Dremel tool, a fairly common 29-piece drill set (couldn’t afford the 115 number-letter-fraction set!), and a set of hand tools including some wrenches (not a whole set), pliers, cutters, and limited taps and dies.
  4. Finally, the kicker: “utilizing available digital/online/personal fabrication resources“. The limitations set forth above may make it sound difficult, but this is the part where I get to essentially pick and choose the fabrication exercises. The advent of on-demand waterjetting means that nothing substantial has to really change from Chibikart’s design. Like its conceptual predecessor tinyKart, Chibikart is a pile of 80/20 and waterjet-cut aluminum plates that I really didn’t have to do that much to in order to put together. I’ll publish the flat layout files publicly such that anyone could conceivably Big Blue Saw a copy.

the interesting part

I should be clear that this Chibikart is not like the current Chibikart. In fact, I took Chibikart’s Inventor assembly file and pretty much ripped everything out of it, and the first things to go were the hub motors. That’s right – no more hub motors, since as much as I would love to corner the market on small hub motors, I don’t have the time or patience yet.

We start with a redesign of the back end, now featuring indirect drive. The horror.

The motor is a Turnigy Sk3 5065. It’s pretty weird for an outrunner, since it’s long and skinny, but it’s compact. I chose this largely because it was left over from 2.00EV – I had purchased a “sample plate” of motors at the start of term for students to fondle. It should be more than powerful enough for this purpose.

The Jasontrollers are staying. Even though I’ve seen that they are not very good at starting low inductance and resistance loads like R/C outrunners, I’m interested to see how well they perform at this scale. My experience is with Melon-sized motors and 8″ tires, and the very small 4″ wheels (those are staying, because lol) will give a comparatively larger force at the ground. The reduction is also pretty high – 2.5:1 as designed, resulting in a max speed of 20mph as limited by the Jasontroller.

Note the “live axle” suspended in bearings – this is actually an outdated design that I scrapped in the interest of making people build as few live spinny things as possible. I couldn’t figure out a good way to couple sprocket to wheel except through risky bore-shimming and drilling of steel sprockets in precise alignment with the wheel bore.

The updated method takes a page from Jamison and uses a sprocket cut from 1/8″ aluminum. I hope to avoid needing as much “drill chamfering” since the material is closer to the native width of the #25 chains. The wheel is now a 1.25″ wide, 4″ diameter Colson wheel, of robot fame, because of their softness and pliability in the face of cutting tools. I picked the ball bearing version, and the bore of the sprocket hub is designed to be two soda can wall thicknesses larger such that a shim can be inserted for drilling the bolt holes, then removed for smooth rotation. This is risky, and I have to build one soon to make sure it can work.

Not being able to do this whole “precise right angle hole” thing means that I need to think of creative ways to retain a right angle dead axle. I used methods commonly seen in the DIY 3d printer universe here – trapping the hex head of a bolt against an internal pocket. Once the front wheels are cranked down with a nut, then the bolt head will have to strip the entire aluminum thickness around it to fail.

Because I only have control of material in Z resolution of 1/8″, I decided to make ‘frame thickness spacers’ such that I have a full 1.25″ between the uprights. Enough to put in a few bronze washers to space the wheel block away from the bearings so there can never be a case of my aluminum parts wiping off on eachother like I experienced with Chibikart’s first steering arrangement.

Up front, I’ve changed the big mill handle to a waterjet-cut yoke, since it’s actually quite hard to interface with a 3/4″ hex shaft. The lovely flange-bolt shaft collars make a return, and they will be the adjustable hubs which hold the steering components in place.

Working out the brakes took the whole weekend. Mostly because I wanted to avoid an excessively complex solution like Chibikart’s eggy cam thing. I was also kind of stuck on the mechanism to use, until I started staring at bike parts at Cambridge Bicycle across the street from MITERS. The result after some more thinking is this half-of-a-single-pivot-brake dealie that uses a stock road bike brake pad and stock brake cable mount. In this case, the cable sleeve actually moves too – usually I’m used to seeing fixed cable sleeves. The modeled spring thing is commonly found with scooter/bike brake part kits (I plucked a random one off McMaster to make it complete the look).

And the outrunner can is the brake drum. Isn’t this a wonderful idea? I decided to not do rubber (pad)-on-rubber (colson) because the engagement would have been less predictable, it would have eaten the colson pretty fast, and the motor can was just right there. This is not possible unless the motor has a “can bearing” – the brake pad is basically pressing right on top of it, so there’s no forces which can cause the can to become misaligned.

Oh, and during this process I found out what all of those random little brake parts are called. For future reference, the cross-drilled screw doohickey is a “Brake Cable Anchor Bolt” and this tubey thing is a “6mm Barrel Adjuster”. I went through so many iterations of “brake cable adjusting tubey thing with the nut on it” or similar search terms before finding the correct looking object.

On the other end is a custom brake pedal. The Jasontrollers’ don’t have regenerative braking anyway, and so having the left Hall sensor pedal on Chibikart was kind of silly. It’s also a bit fragile, and the pedal plate actually bends significantly since it’s made from some seriously bullshit steel. Having it as a mechanical brake pedal was therefore suboptimal, so I got rid of it in favor of a stronger aluminum one.

This design will use two of those cross-drilled-anchor-bolt-thingies and pull on both brake cables simultaneously. It does not have a ‘wiffletree’ or force splitting hinge, so some adjustment of the tension on either side (using the tubey-things-with-nuts) will be needed. For rear wheel braking, though, it’s not as bad because unbalanced brakes aren’t as likely to jerk the vehicle around as much as having unbalanced front brakes, which might cause it to pull suddenly in one direction.

One minor change to the steering – I’m reconfiguring the location of the collars and bushings a little because of the flange-bolt collars being annoying.

They use clearance holes for a #10-32 screw which are actually tapped holes for 1/4″-28. Well, nothing else on this design so far uses #10 screws, and it would have required a 3rd tap size (#10-32 or #10-24) since the collar’s holes are clearance (necessitating threads elsewhere). To avoid the need to do this, I elected to expand the holes in the linkages to be 1/4″-28 clearance so the tapped holes in the shaft collar can be used. McMaster conveniently sells 1/4″-28 screws for cheap in packs of 10.

The protruding screw heads necessitated a bit of rearrangement. I actually like this better, because currently the lower shaft collar is the lowest point on Chibikart and it is the first to hit anything.

So here’s where it’s at so far. It still looks vaguely like Chibikart, is still the same size (34 x 26 at the outer ends of the wheels), and the estimated weight is 28 pounds without hardware (so maybe 30-32 with hardware, wiring, duct tape, etc.)

Through a bit of late night Wikipedia jiggling, I’ve decided to christen this the Democratic People’s Republic of Chibikart. Because like North Korea, I swear this is for everyone to build and learn from but the Instructable will just mean it’s me telling everyone what to do anyway.

(Occupy Chibikart was briefly considered as a name before I realized that DC motors and lead-acid batteries are more 99%-ish than brushless and lithium)

More details to come. Parts are due to arrive in massive amounts this week, and I want it rolling by next week.

Beyond Unboxing: K2 Energy 12v Lithium Iron Phosphate Lead-Acid Replacement Bricks

May 17, 2012 in Beyond Unboxing

Whoa, it’s back! Another episode of Beyond Unboxing, the series of random curiosity-driven posts by me which has so far shown light (literally speaking) on the inner workings of a few shady motor controllers. This time, I explore what is going on inside a commercially available generic lithium ion based “lead-acid replacement” – a battery made of a bunch of lithium iron phosphate cells hiding inside the shell of a standard-size lead-acid battery, with (or….without?) a battery management system to make them act like a plain lead acid battery.

Why would you want such a thing? Reduction of weight, longer lifecycle and shelf life (1500-2000+ charge cycles compared to 500 or less), no lead or acid, and higher power density and efficiency under high discharge loads, among other things. First, a little bit of philosophizing and rationale, though:

I’m looking at them as a possible prepackaged lithium ion battery solution for hobbyist & amateur electric vehicle construction – it’s a solution that is less white-hair inducing to suggest to people than “Oh, you buy this shady hobby lithium ion battery with no hard shell or output protection and use this rather complex multi-purpose charger to charge it, occasionally making sure to use the ‘balance’ function so your battery doesn’t explode”. While that solution is perfectly workable for someone with some technical ability or experience in the field, a plug-and-play solution can reach a wider audience.  I can reasonably assume that the people I see day to day in the engineering departments can handle such instructions as “use this battery charger on this setting and don’t plug it in backwards”, for instance, but from experience with the average curious audience member at the Maker Faires who wants to build a vehicle, it is not something I would suggest immediately without gauging their technical experience more.

Batteries are one of the last finicky items on the list of cheap & repurposed EV parts that we have compiled here over the years – I consider the problem of motors and controllers to be well-solved, especially with things which are coming down the pipe locally, but the answer to “What battery do I use?” has always been sort of difficult. The first response is the one stated above – hobby equipment and soft, plushy lithium pouch cells. Workable? Absolutely. Unforgiving if you are an idiot or just don’t pay attention for a second? Definitely. The second class of answers usually centers on chopping power tool batteries, e.g. DeWalt 18v XRP or 36v lithium ion batteries, which are nice because they already come with the charger for the hapless drill or saw you are about to dissect.

For example, the DeWalt DC9360 36v pack, popular with electric bicycle hackers, is only 2.3Ah and $150 (about $2/Wh), not including charger, but you either have to find an interface for the proprietary DeWalt drill connection or modify the battery/solder discharge leads yourself. They are by far the closest to plug-and-play generic battery solutions I’ve seen yet, though. In terms of modern Li battery solutions for small vehicles, there’s a clear trend between cheapest but highest user experience required (hobby and R/C batteries, chargers), and most expensive but plug and play.  There’s always the classic fallback of nickel cadmium or nickel metal hydride batteries – easy to charge, easy to use. But as it turns out, a good NiMH pack is actually as expensive, if not more expensive, than a lithium battery of the same watt hours these days. It would seem that the magic Chinese manufacturing cloud has largely moved towards producing cheap lithium.

Now, the other big question is, do I really want everyone to be scurrying around on Chibikarts? Probably not, but one step in pushing the construction of electric vehicles as a…… vehicle…. for engineering design education (I’m going to use that pun so much it’s somehow going to end up in my thesis, just watch…) is making sure that more people can do it, and there exist many different starting points and “upgrade paths” as you go. It’s similar to what happened to the robot fighting/combat robots community that I’ve been part of since the beginning of it all for me over 10 years ago: Starting out as a niche sport for people with tons of money and expertise with access to expensive tools, and gradually having commercial solutions and clever hacks for common problems emerge and letting more builders participate.

Anyways, rant over. Onto the relevant subject matter.

A product exactly like this, the A123Systems ALM12V7 module, is what powered this year’s 2.007 EV section to victory:

They are pretty awesome, I’ll admit. Externally accessible ATO fuse, automatic charge cutoff, automatic discharge cutoff too. These were pretty much foolproof, and fainted like Pokemon when the fuse blew or they were overdischarged. Really, that was the word we settled on describing it – when the battery shut off its outputs due to undervoltage or fuse tripping, a brief application of charging voltage would wake them right back up.

But what I found disappointing about them is that nobody else can get them. We were lucky to get them as a sponsorship / donation, but just typing “ALM 12v7″ into Google shows alot of press releases, datasheets, and articles, but no product. This is consistent with A123′s (and most American companies’) position of not dealing with the public directly, something which kind of irks me a little – no matter how much I love the fact that they are an MIT-affiliated battery company that regularly throws prototypes and QC line rejects at us for our own consumption, it’s not reproducible in a setting which is not us. Kind of hard to test your hypotheses then, huh?

They are not the only producers of these lead acid replacements, though. In fact, here’s a whole page of them.  And another.  And that’s just two of my favorite shady Chinese back-of-the-truck battery dealers, out of many! It’s my firm opinion that A123 has already lost the game here – when the 12V7 finally hits the market, they will be seen as just another player in the game, not the player or the leader.

As with all Chinese products, I viewed the generic SLA replacements with a healthy skepticism. I’ve already observed how the A123 modules behave under “out-of-spec” charging and discharging conditions, but I also understand some of the engineering and testing that went into them. The Chinese modules? Not so much. Thus, even though I knew of their existence for a long time, I hesitated recommending them to people. For all I know, they could be just cells in a box.

I was lucky enough to come upon some “class droppings” recently, in the form of these guys:

As far as I know, a 2.009 team purchased 2 of these units for a project last fall, but they were unused. They were found sitting in the Course II teaching laboratory and conveniently hijacked to run the 2.007 tables this year for a little while. Afterwards, they were moved back to the graduate student nest. My interest in them fell to a low point after finding out they were pretty well sealed – the A123 brick prototypes we got last year still had removable shells for testing purposes – and I just figured these were the same thing inside, or somehow knocked off of A123.

Renewed interest in their internals came during a battery search for a new variant of Chibikart (to warrant its own discussion later). So, last night, I decided to crack one open just to see what was going on.

The module above is sold by K2 Energy, which I can’t quite tell from their website if they’re a front for a shady Chinese operation or not, but who cares?

update!

A representative from K2 Battery actually wrote to me in response to the above:

“K2 is a Nevada corporation founded by myself and 5 other partners in 2006.  Most of the technical group, myself included, came to K2 from other battery companies, where we had been developing and manufacturing phosphate-based cathode materials.” -K2 Battery dude

Well there you have it!

end update!

The funny thing about it is, if I search ‘K2 lead acid replacement” or “K2B12V7EB” , the only things that show up are places selling them! It’s kind of a negative A123 problem – I can’t find anyone who’s used them in depth, dissected them, dissed them in general, or anything. They exist. In product form. That’s definitely a sign that something needs poking internally.

It’s interesting to note that these come in two forms – there’s an “EB” version which, according to the short shopping site blurbs, has an internal BMS, and the “E” version which does not. Presumably the latter is literally the box of cells of legend, but here I had an EB version, so it was a chance to see just how BMS-y it was.

I had Shane randomly select one (of two…) to be sacrificed to the scooter gods. Next was figuring out how to crack them open – I didn’t want to do that literally, since I did want to use them as batteries. I settled upon a more dramatic but significantly cleaner solution:

If there is one image which I should not post on the Internet, it is this one. Please, everybody, do not mill your batteries.

I positioned a thin slitting saw right on the green-black plastic seam which is cemented together very thoroughly. I made sure to do very light and shallow passes, poking with a screwdriver each time to see if I broke through the welds – if the top moved, then it was a sign to stop increasing the cut depth and move on to another side.

Whatever they make the casings out of, it smells positively disgusting. Also, the case is thick. Like well over 2mm thick!

After all four sides, the lid is popped off… and we have…

……

Well that was anti-climactic.

It’s cells. In a box.

With what appears to be a generic “PCM” board glued to the underside of the lid. These boards are sold throughout the Inexpensive Chinese Battery Markets (YES, ANOTHER USE FOR “ICBM”!) to append to your own battery packs. In my adventures in EV design at the MIT Media Lab, I’ve had the joy? of using them for a few different custom packs.

They function primarily as output limiters – the 3 bars of metal soldered to the board are current sense shunts – and will turn off the output FETs if the voltage becomes too low or the discharge current becomes excessive. These are nominal 25 amps, peak 40 amps (according to manufacturer datasheet). They also usually perform low-current charge balancing if any of the cells become out of line by a certain amount – purely voltage-differential triggered. As far as I can tell, they also do not limit input current – presumably because the shunt is only set up as a single-sided measurement so the simple logic cannot handle negative voltages. As black-box devices to append to your battery for some semblance of failsafe behavior, they are more than adequate.

These boards, though, seem to have the FETs back-to-back – the Source leads of the row of devices goes to both the negative output terminal and the shunts. So it very well could be more sophisticated – I didn’t take the opportunity to disassemble the board, since it was plastic-welded in place and I accidentally stripped two of the screws trying to get them out.

So I guess it’s not just cells in a box…but that still doesn’t say anything about the “E” version. Now I kind of want to get one, but also don’t feel like dumping $100 on cells in a box.

(By the way, this is my favorite PCM board by far – the 100 amp model. It really shows the Chinese design paradigm of “CTRL-C CTRL-V” well.)

Well there it is. I do like this arrangement of cells, though – it’s almost like there’s enough space in the case, with the cells laying down and the BMS board in the otherwise wasted space of the lid, to add another row of paralleled cells for more capacity…

….oh.

i see what you did there.

There is a plastic insert that is conveniently one-cell in height. I wonder why?

There is a “10Ah” model which is oddly enough the exact same dimensions despite it having 50% more cells. Could it be that it’s just rows of 3 paralleled cells instead of 2, and the insert is not used?

Nah. Couldn’t be.

A better view of the board, for anyone who wants to do some armchair trace-guessing. In keeping with the tradition of posting FET datasheets, here’s the AOD4184, which is pretty reasonable I suppose since 6 devices are used in parallel to share the output current.

My assessment of these modules:

I’m going to guess that anything which looks superficially similar will use a similar setup inside (The Law of Chinese Packaging Inertia), just with different PCM boards. The rudimentary charge-balance and output protection these devices offer is reasonable if the battery is similar in voltage to the lead-acid systems they are designed to replace. Most common non-automotive SLA chargers do a rough constant-current constant-voltage charge profile anyway – the former is called  “fast charge” and the later “float charge”. With 4 cells being 12.8v nominal and 14.4v charging, that’s pretty damn close to a 12v SLA setup. 24v systems are similar, with the nominal voltage being 25.6 and peak voltage being 28.8v – most 24v SLAs charge at 28 volts.

Some of them come with a “2 maximum modules in series” specification and it’s clear why that’s the case. A “36v” system made of these would actually be 38.4v nominal and 43.6v charging – the deviation from SLA nominal voltages gets more severe with each added cell. There is also a potential for inter-module imbalance. As some 2.007 students have already found out, the modules themselves can take on different levels of charge and during heavy use, one will fall under its minimum voltage and shut off. It is exactly like the inter-cell balance problem, just more meta. The way around this would be to make sure the packs are at the same level of charge (e.g. full) before connecting them in a string. Real SLAs and Ni cells will all suffer from this problem too, so it’s not necessarily any worse. With the output protection circuitry, I think it is manageable, but over charging is still a concern. I haven’t tested to see if these generic modules shut off on overcharging yet.

In terms of Wh / $, it’s expensive but not terrible – on par with the DeWalt DC9360, but more general purpose. The K2 “EB” module sells for $140 and at 82 Wh that works out to be about $1.70 per Wh. The “10Ah” version is in fact a better deal overall at $1.40/Wh. For a 24v 7Ah battery system with decent plug-in-wall charger (6+ amps), it would run about $350 and charge in an hour. If you don’t mind waiting all day, then generic 24v scooter chargers are like $20. As a reference, two of these 8.4Ah Hobbyking LiFe packs would run $160 (That’s only $0.75 per Wh… damn you Hobbking), and a charger which can handle 8 cells starts at $45 or so. But you will need a 12v power supply which can handle the wattage (example – $40). So the hobby solution is certainly still the cheapest for those who can accept its lesser level of integration and can properly secure the battery from damage.

Perhaps most importantly, the modules do have a (very) tough shell and rectangular brick form factor. It’s way better than watching someone zip tie a soft-pouch pack to a bicycle frame. I did a little bit of further shopping on BatterySpace, and found that I could buy the parts for my own 8-cell, 12.8v 6.6Ah pack with a 30 amp PCM board for about $100. However, then I would spend several hours putting it together and then still not have a cute box in the end to put it in. So certainly I could buy the parts for much less, but the balance of the $140 price is made up in convenience.