Manufacturing guy-at-large.
Share

EAGLE Libraries on Github

Added on by Spencer Wright.

This has been *such* a PITA, but I think I just got The Public Radio's EAGLE libraries onto Github in a way that makes sense.

The goal: To have everyone be able to access all of the parts that we use on The Public Radio, no matter where they are or who created the package/symbol/part, by just syncing the Github repo.

Our Github page now has one active repo (Embedded_Hardware) - and two for firmware & firmware libraries that are mostly inactive for now. Within Embedded_Hardware are two big subdirectories (EAGLE_CAD and EAGLE_lbr), plus an ARCHIVE and a place for us to put production files (Gerbers, etc).

Within EAGLE_lbr, there is one EAGLE CAD library file - PR_Parts_Library.lbr. From now on, that will be the default library for all packages, symbols, and parts that we create or modify for use in The Public Radio.

In order to make this all work, you need to add the local directory that you sync to Github to your EAGLE Libraries search path. I keep our Github repositories at ~/Documents/The-Public-Radio, so I added :$HOME/Documents/The-Public-Radio/Embedded_Hardware/EAGLE_lbr to the end of my Libraries search path:

And now the EAGLE libraries show up just fine!

Lastly, I opened up our .brd and .sch and went through all the parts I had created. I drew the speaker, potentiometer, antenna hole and batteries. The .brd and .sch included copies of those parts that came from a library on my local disc, so first I had to find those parts and copy them to our new shared library. Then I went into the .brd and .sch and swapped the old versions of those parts for the new ones. 

Anyone else who's working on the project right now (that'd be Zach and Andy) can do the same with any of the footprints that they created. And anyone who's hoping to make their own (not sure why they would, but who am I to ask) can grab the entire Embedded_Hardware repository and modify packages to their heart's delight. If you do so and find a mistake we've made, please let me know - we're happy to take any pull requests that improve on our design!

Share

Bunnie Huang

Added on by Spencer Wright.

From a post called "Soylent Supply Chain:"

This point is often lost upon hardware startups. Very often I’m asked if it’s really necessary to go to Asia–why not just operate out of the US? Aren’t emails and conference calls good enough, or worst case, “can we hire an agent” who manages everything for us? I guess this is possible, but would you hire an agent to shop for dinner or buy clothes for you? The acquisition of material goods from markets is more than a matter of picking items from the shelf and putting them in a basket, even in developed countries with orderly markets and consumer protection laws. Judgment is required at all stages — when buying milk, perhaps you would sort through the bottles to pick the one with greatest shelf life, whereas an agent would simply grab the first bottle in sight. When buying clothes, you’ll check for fit, loose strings, and also observe other styles, trends, and discounted merchandise available on the shelf to optimize the value of your purchase. An agent operating on specific instructions will at best get you exactly what you want, but you’ll miss out better deals simply because you don’t know about them. At the end of the day, the freshness of milk or the fashion and fit of your clothes are minor details, but when producing at scale even the smallest detail is multiplied thousands, if not millions of times over.

More significant than the loss of operational intelligence, is the loss of a personal relationship with your supply chain when you surrender management to an agent or manage via emails and conference calls alone. To some extent, working with a factory is like being a houseguest. If you clean up after yourself, offer to help with the dishes, and fix things that are broken, you’ll always be welcome and receive better service the next time you stay. If you can get beyond the superficial rituals of politeness and create a deep and mutually beneficial relationship with your factory, the value to your business goes beyond money–intangibles such as punctuality, quality, and service are priceless.

This is really smart.

Share

Directions in automation adoption

Added on by Spencer Wright.

I've been spending a lot of time looking at industrial automation the past few days, and had an idle thought:

I've touched on this before, if only obliquely, when writing about MFG.com's role in manufacturing logistics. Much attention is being paid to companies who want to simplify (or circumvent) some part of the product development value chain. Many of these are companies I admire, and think are doing really valuable things. Take Within, whose 3D design software generates structures that are driven directly from functional constraints (but can't, as far as I can tell, deal well with thin-walled structures). Or Willow Garage's PR2, the really slick research robot (that takes charmingly long - 20 minutes per bath towel - to fold laundry).

Each of these is an incredibly impressive feat, and one that follows an ambitious (and I would argue honorable) line of thinking:

If we can encode all of the information needed to complete a routine yet complex task, then we can use machines to automate the process, freeing up our minds to do other (presumably more important) things.

But consider an alternate proposal:

If we can get machines to mimic a series of behaviors that humans can plan and execute with relative ease, then we can decrease the amount of rote mechanical work that humans need to do.

This is the tact taken seriously by Baxter, the admittedly not-too-serious (but cool nonetheless) humanoid task robot built by Rethink Robotics. Baxter learns by physically training his movements, presumably by the technician who he's "collaborating" with:

Even the traditional robotics companies, like Kuka, are moving in the direction of using robots simply to execute the complex tasks that humans calculate and perform with ease. Here a Kuka robot is trained how to clean a permanent mold by a BMW employee:

Both of these robots' use cases share a key feature: There's still a human doing the "hard" planning and calculation about how the task will be completed. In each case the robot doesn't understand the physical constraints or goals per se. Baxter has some awareness of his surroundings for sure, but all he knows is that his arms hit something; he doesn't have the vision or awareness of why that happened or how to correct for it.

Similarly, the Kuka bot doesn't understand that he's cleaning a mold, or have the facilities to learn how to do better work. He's just repeating a toolpath that he knows a human told him to do. Which, in this case, is good enough - and a hell of a lot faster than waiting for a computer vision expert to give him the intelligence required to do better.

I'm not sure what the implications of this are for the companies working to automate the design and supply chain. But the philosophical difference is striking, and I must say that the more hands-on model is very compelling - and I expect it to be so for the foreseeable future.

 

Parenthetically: All of the industrial 3D printing market is currently driven off of this same model: An intelligent, experienced technician makes manual edits to 3D CAD data in order to get a part to print within its design constraints. Anyone who suggests that build optimization is "right around the corner" is, in my opinion, *not* to be trusted. We're in a world of basic research still, and an automated design-print-post process chain is many years away.

Share

Three interviews

Added on by Spencer Wright.

I've done three interviews recently. The first, with Brian Barela, was on the future of work. It's in two parts:

The second was with New Hampshire Public Radio's "Word of Mouth," and was about The Public Radio's Kickstarter campaign:

Spencer Wright/The Public Radio Word of Mouth interview

The third was with Matthew Lesko, who you'd know as the question mark suit guy from late night infomercials - and who ends up being a really intelligent, thoughtful person. This one was also about The Public Radio:

Being interviewed is, IMO, *okay.* Holler if you want to chat :)

Share

DMLS in process

Added on by Spencer Wright.

This is my seatmast topper being printed from titanium 6/4 powder on an EOS M280 at DRT Medical - Morris.

This video clip is about 1-2mm into the build, so this is the very close to the bottom edge of the part (which, to be clear, is lying on its side). A lot of what's being printed during this clip is support structures & ribs that will help hold the part to the build plate, but you can clearly see the general shape of the part already.

This is 6th iteration on this build. In other words, we printed 5 parts before this one, and each of them failed for one reason or another. We've (and by we I mean mostly Dave Bartosik, the head Additive Technologist at DRT, with me trying to look over his shoulder) made a bunch of modifications to the build to help the part come out within spec, and I'm hoping that today we end up with something that has consistent inner diameters and is more or less useable.

Anyway, what you're seeing here is a 400 watt ytterbium fiber laser in the process of melting 30 micron layers of titanium powder. The recoating arm spreads a thin layer of powder, and then the laser scans a cross sectional slice of the part, and then the process repeats. 

When the video goes into slow motion, notice the smoke that's coming off of the weld pool. The machine has a laminar flow of argon gas that's blown across the build platform (from top to bottom in this video) that pulls the soot away and filters it outside the machine.

More soon.

Share

Initial topper builds

Added on by Spencer Wright.

This.

Today I stopped by DRT Medical - Morris to check out the first builds of my titanium DMLS seatmast topper. I'll be writing up a longer report in the next few days, but the short story is that it's moving forward - but that the process of printing a part is *not* at all straightforward, even when you have some of the most experienced people in the business working with you.

More soon. And - fucking cool, right? This thing started out as titanium powder, and was literally made by laser sintering. Nuts.

Share

Updated drawings

Added on by Spencer Wright.

Changes:

  • Oriented all drawings the same way
  • Much improved BOM with screw & nut specs & names, more details on custom parts
  • New spacer setup
  • Section & detail drawings show interface between spacer, lid, and antenna/potentiometer

Full PDF here, in case you're really curious.

Share

An idea I'm not so sure about

Added on by Spencer Wright.

We need a better way to sandwich The Public Radio together. It needs to be a spacer between the speaker and the lid, and needs to insulate the antenna from the lid, and probably needs to interface with the potentiometer too.

This is my first real crack at it. Not sure how I feel about it. Will probably be expensive.

Needs more work + some supplier DFM feedback.

Share

Charmed

Added on by Spencer Wright.

I have about 50 of these in my inbox on MFG.com right now:

Not all Chinese manufacturers are as charming as this one, but in general they're *much* more outwardly communicative than American suppliers. 

Share

Island scanning, and the effects of SLM scanning strategies

Added on by Spencer Wright.

From a (highly academic) paper titled "The influence of the laser scan strategy on grain structure and cracking behaviour in SLM powder-bed fabricated nickel superalloy" (emphasis mine):

Changing the scan strategy however, or indeed investigating the effect of a different scan strategy, should be approached with caution as the geometry of the component or sample being produced may have a serious influence; the data presented in this study relates to a simple cuboid. Consider a sample with a large cross-section; the time between each laser pass will be much greater than that of the sample seen here conversely for a thin section the time between each laser pass may be significantly reduced and a band heating effect (similar to that seen with the ‘island’ scan-strategy) may be observed. More complex still would be a geometry transitioning between a thick and thin section or visa-versa. Therefore by applying a single laser scanning strategy for the entire geometry, the actual structure of the material may not be the same as that predicted from simple cuboidal trials.
Ideally the optimum scan strategy would be calculated on a layer by layer approach taking into account the part geometry and even adjusting the fundamental parameters of laser power and scan speed allowing the microstructure to be tailored to the final application of the component; this would require significant research into the thermal modelling of the SLM process.
Currently however this is not possible and so the ‘retro-fix’ solution of utilising a post-fabrication HIP treatment has been shown to be effective at closing the internal cracks within the material.

Some background: the DMLS and SLM processes use a focused laser beam to selectively melt (or sinter; the distinction can be hazy) powdered metal into a solid part, cross-sectional layer by cross-sectional layer. Since the beam can't melt an entire cross-section all at the same moment, it instead traces back and forth across the part, melting a single line at a time until the entire cross section is done. This gif, from the Solid Concepts blog, gives a good sense of the process:

Once a layer is complete, another 20-40 micron layer of metal powder is laid out, and the laser scanning process starts again.

The SLM/DMLS (and SLS, for that matter) build process.

The SLM/DMLS (and SLS, for that matter) build process.

You might not stop to consider it, but the pattern that the laser takes as it traverses the part can have a huge effect on its structural properties. As the paper at hand discusses at length, one strategy is to break thick cross-sectional areas into a grid pattern, and scan adjacent grid zones in opposing directions. This method is called "island scanning."

Much like a brickwork bond pattern, the 5mm x 5mm grid is shifted 1mm for each successive layer in both the X and Y dimensions. Interestingly, the crystalline structure of the resulting part varies as you look across the island boundaries. As a result, if you look at a SLM part under a scanning electron microscope, you observe distinct 1mm x 1mm crystalline grids:

In addition, micro-cracking and porosity (which is generally quite low; these parts are commonly around 99.7% dense) follow the same pattern. These effects can be mitigated - indeed, nearly eliminated - by hot isostatic pressing (HIP), and most SLM and DMLS parts are HIP treated by default. 

The interesting question for me, though, is how these factors apply to thin-walled parts. My seatmast topper's wall thicknesses are in the neighborhood of 1mm, and thinner in areas. As the authors of this paper note, cracks that are open to the surface of the material aren't always closed by HIP. My suspicion is that the feature sizes at issue here are orders of magnitude smaller than they'd need to be to have a noticeable effect during the lifetime of the part, but only destructive testing (or a better trained material scientist than myself) will bear that out.

I'm also not sure, to be clear, how these crystalline structures compare to those of samples produced via conventional (i.e. non additive) methods. If anyone knows of a good study on the micro-structure of drawn titanium tubing, I'd love to see it!

Share

Pro supplier moves

Added on by Spencer Wright.
  1. Calling me out of the blue to say "hey." Before that, you're just some guy on the internet who sent me a quote - and I've got like 50 of those to deal with right now.
  2. Following up with an email within the hour.
  3. Following up with another email a week later with a picture of some samples (minus the knurling, but you acknowledge that) that you made.

Now you've got my attention. And then I respond and say something about how important the feel of the knurling is to us, because the knob is really the only UI element in The Public Radio. And now we have a relationship.

So: Tru Precision Manufacturing LLC. Good stuff.

Share

What's with Hoffman Estates, IL?

Added on by Spencer Wright.

I've been searching through machine tool manufacturers today, and stumbled across this town of 51,000 residents twice. What's going on here?

DMG Mori, which is headquartered in Hoffman, doesn't even rank big enough to show up. Ditto Renishaw USA.

It could be that this is just a manufacturing-heavy suburb of Chicago; I guess the name is just throwing me. Anyway, I'd be interested in visiting.

Share

NYIO Session Two Notes: Sims Metal Management's Sunset Park facility

Added on by Spencer Wright.

Last week I visited Sims Metal Management's Sunset Park facility with the New York Infrastructure Observatory. The whole group posted notes in a public Google Doc, and we shared photos & videos on Dropbox - both of which I'd encourage you to check out. My overview is here.

Location: Sims Metal Management’s Sunset Park facility

472 2nd Ave, Brooklyn, NY 11232

Present:

  • Spencer Wright
  • Jason Spinell
  • Daniel Suo
  • Athena Diaconis
  • Jiayi Ying
  • Benny Zhu
  • Mark Breneman

Arrived at: 1400

Tour began at: ~1430

Tour completed at: ~1600

Tour guide was named Eadaoin Quinn

First: When we arrived at Sims, we immediately saw that they were installing a wind turbine in the front of the building. It will eventually generate about 100kW, which is in the neighborhood of 2% of the facility's energy usage.

This facility processes streetside recycling for the Bronx, Queens, Manhattan and Brooklyn (I believe Staten Island goes elsewhere). Brooklyn and Queens deliver directly by truck; the Bronx delivers by barge from a transfer station in Hunts Point; Manhattan will soon be delivering by barge from a station on the West Side. Sims does *not* handle commercial or construction waste. Our tour guide said that there are about 250 commercial carriers, each of which has their own sorting facilities and policies. At some point in the future, it's possible that Sims would contract to process their product as well, but it's not clear when.

It costs $65-70 per ton in "tipping fees" to process waste at Sims. For comparison, it costs the city $100+ per ton to put waste in a landfill. 

All told, recycling costs the city money. In the early 2000s, Mayor Bloomberg cancelled most of NYC's streetside recycling, saying that it just wasn't economically feasible. He later reversed his stance, but the cost-benefit analysis (at least in the short term) still isn't great.

Sims processes 600+ tons of product (i.e. recyclables: bottles, cans, paper, etc) per day. They have two daytime shifts, running from 0800 to 2400. They run maintenance daily from 000 to 800, plus another eight hour shift on the weekend.

They try not to shut down for maintenance during the workday, but it does happen. They *hate* plastic bags here (our tour guide had testified in favor of a plastic bag tax the day before we arrived). They're difficult to process, clog up the machines, and have essentially zero resale value. When unplanned maintenance does happen, it seems to usually be the result of plastic bags. Sims processes about 30 tons per day of bags alone.

Another surprise was the way they process glass. Because glass isn't sorted and handled carefully, it essentially lands at their door partially crushed. It's then crushed further (by their disc screens), and then allowed to filter through their process chain, ending up literally at the bottom of their machines. From there it's sent by barge to Sims' Jersey City facility, which expects everything to be roughly 3/8" in any dimension. There, it's scanned and sorted optically between clear and non-clear. Clear glass can be resold and recycled, but it's *really* difficult to sort brown from green - and the two cannot be recycled together. As a result, they're sold as underlayment for public works projects - roadbeds, foundations, etc. Oh - and Sims has some in a planter in the front of their building, too.

The facility's big bragging point, though, is its 16 Titech optical sorting machines. These track the frequency response as items run down conveyors and have infrared light shown at them; different plastics respond differently, and are sorted by pneumatic sprayers as a result. The whole process is totally automated and pretty cool:

Once all the different products are sorted, they're baled into blocks weighing between 800 and 1200 pounds. Sims has relationships with 3-5 customers for each product, and sells it to those customers depending on how their bids (which are updated monthly) compare to the national prices for the commodities they're selling. The contract that Sims has with the city requires a portion of the national price to be paid back to the city each month, so presumably Sims is incentivized to sell the product for more than those national prices.

Share

Dog collar

Added on by Spencer Wright.

After 3 or 4 years, Libo's old collar was falling apart. So I took it apart and made him a new one.

The buckle and D-ring I made by hand back in 2010. They both started out as solid copper rod; 1/4" for the buckle frame and all of the D-ring, and 3/16" and 1/8" for the center bar and prong. The D-ring is just bent, cut, and brazed together with (IIRC) 56% silver, but the buckle was a bit more complicated. There I bent the frame first (on my old DiAcro #2 bender, which was *so* great). I turned the center portion of the center bar down on my lathe and brazed it to the frame. Then I bent the prong, mostly by hand, forming it in-situ over the center bar. The whole thing took some finishing time with a disc sander, a hand file, and aluminum-oxide abrasive cloth.

The nametag I cut from stainless steel stock and engraved on an Engravograph/New Hermes manual pantograph engraver. Then I bent it over the biggest mandrel I had in the DiAcro.

The leather is by far the easy part. The belts come precut from Tandy; I dyed them with Fiebings dye and then bevel the edges by hand. All of the holes are punched with a hole punch and mallet. The rivets are solid copper with washers from McMaster-Carr; you just put the parts together, secure the washer in place (usually driving it softly with a mallet and a tubular punch) and then give the end of the rivet a swift smack with a hammer. When you do that, the whole thing bulges out a little and squeezes on the inner diameter of the leather belt and the washer. Then you take a ball peen hammer to the rivet, smushing it down so that it's flush and soft to the touch.

Rebuilding the collar didn't take much time; figure 45 minutes or so. Building the buckles from scratch was kind of a pain, though; although it was fun and satisfying, I'd like to find a way to do that via Shapeways in the future.

Anyway. A good Sunday morning project, and one that will get a lot of use.

Share

Speaker mods

Added on by Spencer Wright.

One of the big pain points to date on The Public Radio is that our speaker - a Dayton Audio CE32A-4 - was *not* designed to be PCB mounted. Like most good quality full range speakers (it's one of the most expensive items on our BOM), it's got solder lugs, which are designed to have wire soldered to them.

So, why did we choose this speaker? Partly because of sound quality - it's really the best available given our size constraints. It was also relatively easy to attach to on the cone side; the plastic flange is sized nicely and has four good screw holes in it. And anyway, we figured we could work something out if we ever had to build a whole lot of radios.

Fortunately, at 2000 units we start to be in the range where getting custom PCB pins could make sense. It'll cost us some time (13-15 weeks) but that's okay if we decide quickly - and we'd probably wait that long anyway just getting the full quantity from our supplier. So I drew up a basic version of the pins we'd want and sent it to Dayton:

I'm not sure what the net cost will be on the parts, but the assembly will be significantly less expensive and *much* more reliable. 

Share

Public Radio updates

Added on by Spencer Wright.

My last week has been spent mostly trying to squeeze money out of our potentiometer knob. The one we had been using - a Kilo product - has terrible pricing, even at quantity ($2.38 @ 2000 units). So I remodeled the part and started poking around.

I got a bunch of quotes from some US machine shops I know that are closer to the $2 range. I also got a quote from Taiwan for about $.50 apiece (though I'll expect additional shipping/customs costs). 

Then, I put the part on MFG.com. At quantities of 2000, this should hit a good spot for a lot of US mill-turn shops - and so far the response has been pretty good. I'll let the quoting period run for a while still, but I expect to pay no more than $1 for these parts, and that's including clear anodizing and possibly shipping to NY.

These are all for quantities of 2000 units.

Meanwhile, Andy has been researching FM IC pricing; Zach has been looking into batteries and tactile switches; and Eden has been poking around at PCBA shops. 

My next focus will be on the lid and the speaker. The former is in a pretty good place right now, but the speaker wants some real love... updates soon.

Share

Fulton Center

Added on by Spencer Wright.

The Fulton Center opened today, after 12 years and $1.4B in spending. I stopped by after work and wandered around - which I'd recommend that you, dear reader, do as well. It's the closest thing we have to an actual modern subway station, and that's pretty cool.

If you're looking for an actual description of the space & its launch, check out Second Avenue Sagas - it's good.

Share

NYIO Session One Notes: Amazon's Middletown, DE Fulfillment center

Added on by Spencer Wright.

Last week I visited Amazon's Delaware fulfillment center with the New York Infrastructure Observatory. The whole group posted notes in a public Google Doc, which I'd encourage you to check out; my overview is here.

Location: Amazon’s Middletown, Delaware fulfillment center

560 Merrimac Ave, Middletown DE.

Present:

  • Spencer Wright
  • Rob Snowden
  • Dan Suo
  • Jason Spinell

Arrived at 0945

Tour began at 1000

Tour completed at ~1115, and was quick almost to the point of being rushed.

Tour guide was named Evan

Notes taken retrospectively by everyone above.

This facility is 1.2 million square feet. The parking lot had somewhere around 2-3000 parking spots, and was mostly full. There was a sign saying “First Day Starts Here” in the front, and a number of what looked like brand new employees doing orientations. This is a “sort” facility, as opposed to a “non-sort” facility. The difference is that all of the items in a sort facility fit in a yellow bin (about 14”x14”x24”).

There are two sides to the facility: Inbound and Outbound. Inbound receives stuff from suppliers and also from other Amazon facilities (“transfer”). They’re doing a lot of transfer right now in prep for “peak season” (i.e. the holidays). Outbound ships orders to customers, and also presumably to other distribution facilities.

Most of the time, employees work 4x10hr days. There are two daily shifts, leaving the facility closed for a few hours a day. There are about 3000 people in normal employment at the facility. During peak season, they ramp up to 6x10hr days. I believe they also shift the schedules a bit, such that the facility is humming 24x7. They also increase their staff, up to about 6000 people.

At inbound, pallets come in and the bill of lading is checked. Then the boxes are loaded onto a conveyor. One team of associates *just* open the boxes. Then another team picks a box out, empties the items out, and checks quantities. Then (I think) they add an ASIN if the item doesn’t have its own UPC.

They use random storage at the facility. There are cubbies everywhere, and each has a bar code. Inside each cubby is up to 6 unique SKUs. If a SKU is in one cubby, it CANNOT be in any adjacent cubby.

There are about 300 Pickers at this facility. They are managed by two managers. Pickers find the cubby they’re assigned to look for, scan its barcode, and then find the item within that cubby that they’re supposed to pick. They scan the item’s barcode, and drop the item into their bin. If a picker finds a broken item, they put it in a red bin at the end of the aisle, where QC can find and fix/dispose of it. If a picker finds an item that’s not in its proper cubby (e.g. on the ground), they put it in a blue “amnesty” bin, where QC can pick it up and return it to its proper location.

Once they get all the items on their current list (which does *not* correspond to a customer order), they put the bin on a conveyor, and it goes up to sort.

At sort, an associate has a bin (from the pickers) and an 8020 rack on casters that has a few dozen cubbies on it. They pick an item at random out of the bin, scan it, and then a monitor tells them which cubby to put it in. Here, the cubbies *do* correspond to customer orders. Once the rack’s cubbies are all full, the rack gets wheeled over to the pack stations. Each cubby has a packing list in it.

There are about 100 Packers at this facility. They have one manager, and two or three assistant manager types. The packers take a packing list and its items out of the rack. They scan the packing list and it tells them which box to use. Different packers have boxes of different sizes; so one packer might have small/medium boxes, and another might have larger boxes.

They pack all the items into the box and fold it up, putting the packing list inside. They have a machine that wets water-activated shipping tape and cuts it to length automatically. All they have to do is press a green button and the machine already knows what kind of box they’re using and spits out a piece of tape that’s just the right length.

Each packer has a set of Andon lights above their station, which indicate their ability to perform their work. During normal operation, the Andon light is green. If the associate is running low on supplies, they turn it to blue; if they have a serious blockage or shortage problem, they turn it red. Andons are constantly monitored by support staff, who will come to assist if a packing associate is unable to do their job.

They seal the box up. Then they take a barcode sticker off of a reel, scan it, and put it on the box. At this point that barcode is assigned uniquely to the customer’s order. The box goes on a conveyor. It is automatically scanned, and has a shipping label printed and slapped onto it by a machine (this was *cool*). A spot on the conveyor weighs the package, and if the weight is off then it goes into QC. 

The packages are sorted roughly by size and go onto a long oval conveyor. At one point on the conveyor, a 360* scanner checks the barcode and figures out which truck (by shipping zone) the package needs to go on. Then, as it’s going around the conveyor, little pushers kick the package off the conveyor at the right chute, and it’s sent down to the truck to be loaded. 

Humans load the trucks, creating big walls layer by layer. They try not to stack the same box size very high, because it’ll become unstable unless the stack is interlocked at a lot of points.

Other notes:

  • The whole place was rather spartan. 
  • Our tour guide used an iPhone, but claimed he’d use a Fire if he wasn’t locked into his carrier.
  • Our tour guide is normally an Ops guy; tours are *not* his full time job.
  • Their “___ Days since last workplace injury” sign was *not* filled out.
  • They had Grainger vending machines with “free” PPE supplies inside - but they track employees use of those supplies.
  • They were very excited about FBA (fulfillment by Amazon) and their MOS (make on site i.e. print books on site) programs. FBA packages get plain brown packing tape (as opposed to the Amazon branded tape), but otherwise the whole process seemed to be very similar to Amazon owned items.
  • Our tour guide was very enthusiastic about a few TPS things. He mentioned Kaizen (literally “good change”), which to Amazon is a program where people from throughout the organization get together to propose and implement significant improvements in their processes. The Andon lights in the packaging department are also a derivative of a Toyota production feature.
  • Amazon has a number of “etiquette” policies, which refer to the “nice” way of doing things. For example, heavy items shouldn’t be stored on shelves above head height). It was unclear whether these guidelines were substantially different from rules, but with friendly names.
  • The whole place was highly automated. The box & tape length selection systems were really interesting, and have the effect of making the packers into just highly adaptable robots.
  • When the tour was over, Evan (with some enthusiasm) gave us all Amazon branded cigarette lighter USB chargers.