Wednesday, January 26, 2011

Serial Wiring for NanoStationM2-Loco


Long Day at the Office

Some days you win, some days you lose, some days you get 3hrs of sleep, work for twelve hours and learn all about what to do differently next time.  Today was showtime for Nick and John. While Tom and I have been hacking network configuration for the last week or so, Nick and John have had the responsibility of designing and building a permanent installation solution and working out all of the logistics for meeting node hosts.   Today we were scheduled to deploy three nodes, beginning at 7am. As Tom and I were glued to computers, John and Nick were left to make sure that when the tech was ready, the world would have waterproof housing, power and rigid mounting poles to accept it.

Having done this a couple of times already, I tried to be adamant about the importance of having EVERYTHING ready (and tested) in advance and thinking through contingencies in case something went wrong in the field, but I don't (yet) know the answer to every management problem and got a lot of yes, yes, yes, without the associated substance.  Not surprisingly, the team got sent straight to the headmaster of experience at 6:45am.

When I walked in at ten-of-seven (hoping that my own residual debugging wouldn't be critical path) I found Nick feverishly hammering away at a mounting box and John was nowhere to be found -- great for my moral high-ground but not for today's productivity.  By 9:30 I had sorted the network, but the hardware train didn't roll until after a shopping trip an hour or two of packing and a hasty lunch. We had to cancel our first appointment, got lucky that the second cancelled for us, and had to bail on the third for lack of time by day-end.

 As I look at what worked and what didn't, the importance of design and planning stand out strikingly.  The team did a lot of things right, but in neglecting the details undermined a lot of their own hard work.  Nick spent the better part of two weeks designing, sourcing, and finishing a mounting solution for the new Fabfi

In an unexpected move, John made an on-the-fly design mod to put the pole behind the bracket and the box on the front, cutting down the materials list and solving a tension problem with the pole attachments.  John also pounded the pavement to cobble together an effective power backup out of computer parts, solar charge controllers and UPS batteries.  Between the two of them, they created a beautifuly integrated, reasonably priced installation:


  • Mounting box: 250KES
  • Mounting bracket + HW:  600KES
  • Charge controller: 2900KES
  • Power supply: 800KES
  • Battery: 1000KES
  • TOTAL: 5650KES (=$70USD)
With all the components mounted in the box installation involves only drilling three holes and tightening a few bolts.  Power-in, ethernet/POE out, done (it's very clever) While the mounting was sturdy and simple to attach to the wall, by deciding to assemble the solution in the field, the team lost most of the benefits of a well-thought-out design.  During the four hours in the field today, more than three were spent mating the box to its bracket and mounting components inside.  The other four hours that could have been spent in the field were occupied by packing the field kit that could have been ready to roll the night before.  In all, what could have been an 8-hour workday was strung out to a 12-hour marathon where less than half the time translated to progress.

That said, the results for the one node that the team did complete were impressive.  As a mentor, I can only hope that wisdom is a better motivator of behavior than I am.

Note: the rig on the left in the picture is KDN butterfly.  As far as I can tell, nobody uses it, ever....

More nodes tomorrow.  

~Wrench

Friday, January 21, 2011

Power Strugglez

Continuing on the "ends in Z" theme...

One of the big questions for this trip has been how to provide backup power for the nodes. It's been in the back of my mind to DIY a power solution in the lab, and with the help of the interwebs (Google: LVD circuit) I don't think it would be that big a deal, however there's always too much to do and too little time. As a result, we did some shopping around, and here's what we found / tested.

Sparing you the long iterative process, we came upon two solutions.  The first, coming in at about $55USD was a computer power supply ($9), wired with +12,-3.3 to an Apple 5 charge controller ($22 in USA, $36 here!) and a 7Ah battery ($10).   We originally tried to save on the power supply by using the Ubiquiti PoE bricks to drive the charge controller, but they were nowhere near up to the task. Here's Nick hacking the power supply together (or apart, as the case may be):



The Second, coming in at $38 was a good ol' Chinese-brand UPS with the same battery capacity.  

I figured the OTS UPS would perform markedly less well given the extra DC-AC-DC conversion, but the results were more striking than I imagined:



With the same load (roughly 6W), the OTS UPS lasted less than three hours (partially because the UPS doesn't overcharge the battery quite as much) while the Apple+some crap lasted longer into the night than we cared to stay.   (X axis on graph is fraction of a 24-hour day) Needless to say we're going with the former solution.

A side benefit to the above experiments is some hands-on education with LVD circuits (because I'm certainly no electrical engineer...). A previous worry that had kept us from building the LVD circuit is that if the battery drained too far and became too strong of a load when the power returned, it would hang the routers.  It turns out, that with LVD and a reasonably strong power supply, the  internal resistance of the charging battery is sufficient to bring the line voltage right back up to 14ish volts and turn the electronics right  back on when the mains kick in.

With a little fabbing, we could probably bring the cost  of the UPS down to $25. 

Sneak Previewz

Sooo, Clooooooose....



Thursday, January 20, 2011

Google Business

One of the most frustrating things about being in the field these days and writing this blog is I feel like I never have enough interesting pictures.  Wireless networking may be cool, but it's definitely not glamorous.  To boot, a lot of what we've been working on in Fabfi-land these days isn't tech, it's business.  There's a huge leap between having somthing that works once in a lab, and a reliable product with the business processes in the background that make it viable, especially in the chaotic world of a startup in a developing country. 

A big challenge We've been working on lately is process development, in other words - what do you do when you buy something and have a receipt?  How do you account for when users pay you?  How do you file things so they don't get lost? 

Most small businesses here, with the possible exception of tech businesses, still run on full-manual: Cash transactions and paper accounting.  Personally, I've never trusted paper.  You can't automatically search it.  It's a pain to sort it.  It gets lost or damaged easily, and you have to be physically co-located with it if you want to use it. 

With the JoinAfrica team spread out across multiple continents, and the team's local workspace being temporary at-best, we've gotten a lot of mileage out of running JA through cloud-based services like Google Apps, Google Code and Dropbox:


Above is a screenshot of some of the things we're automating:  We have web forms for accounting, spreadsheets for timelines, and shared folders for photos or receipts and technical documents.  I was a little skeptical, at first, that the team wouldn't latch onto the idea of running a business from a computer screen, but they seem to really like the forms entry, and have slowly been getting used to the idea of uploading photos to dropbox, which is great for the local team because keeping everyone up to date online allows the rest of the team to be more helpful from abroad.

With a pair or robot arms, you could just about run our business from anywhere in the world, which nice when "anywhere in the world" is often your address.

Monday, January 17, 2011

When all dimensions matter....

It never occurred to me that there existed standards for just about everything.....drill bits, bolt sizes and drilling standard clearance. Imperial vs. Metric doest help either. Bottom line: no shortcuts



Wrench Edit:  The above drawings, an updated set from the ones posted earlier, were developed entirely by Nick, from scratch.    Nick is responsible for designing, sourcing and coordinating the fabrication of mounting hardware for the new network, and so far is doing a bang-up job.  Keep up the good work!

Chinese Brand = Fail

Anybody who knows the core fab crew knows we're snobs about good tools.  Here's an example of why you should spend a few extra bucks on good tools and take care of them:


The photo above is of a Chinese-Brand masonry bit.  It's about half the price of the "Not Chinese" brand.   In Kenya, "Chinese Brand" seems to speak more to to the quality than the place of origin.  Here, Chinese Brand is synonymous with "Work one week.  Then buy new you." (the instructions often also read like that)  Consistent with local expectations, our Chinese Brand bit lasted approximately one minute before total destruction.  I suppose it's not surprising, given the obvious lack of carbide in the tip...  We will be buying the other ones next time.

Thursday, January 13, 2011


Shopping for a new box to house our electronics-headnode, POE adapters(see earlier blog by the wrench)

and later this month, the frame will hold a 2-metre mounting pole.

Mounting design for a box to hold a switch and power adapters......fabfi mounting brackets.

Network upgrade: The Health centre


Layout for the AMURT rooftop layout.

Monday, January 10, 2011

Workin', Day 1: Site visit

In case my teaser didn't give it away already, I'm back on the ground in Kenya for some follow-up to the project we started in August.  The official start of network hardening and scaling festivities is tomorrow, but I got a chance (at the cost of my pocketknife on the bus, Grr) to tag along with the team on a site visit to the pilot site and one of the early adopters' houses.  In this particular case the user we visited was three hops from the gateway, and was reporting great satisfaction with the speed and reliability (sweeeeet!), but more importantly, he had some very positive feelings about the prospects for demand (something that, honestly, I was pretty skeptical about).  He says:

"My family loves the service, it's much better (faster) than the 3G modem" and "everbody [in the neighborhood] wants wifi now.  In the last year or two everyone has gotten a laptop, and all the phones have wifi, and having the usb dongle all the time [for 3g] is too cumbersome.  If this service is stable and priced like 3G, people will pay for it"

Right on!

The team has been doing a very good job building relationships at the pilot site, and the work shows.  The hardware, however, needs some work.  The team had expected to upgrade all of the mounting you'll see below months ago, but an administrative obstruction kept the funding for new mounting materials out of reach until... wait for it...  Later this week.  Hopefully I won't have any more to say about that. EVER.

At any rate, what follows is yet another testament to the resilience of solid state electronics and the fact that well-trained installers can make even the most cobbled-together system function well.  A professional installation is an important part of a permanent system and making some serious upgrades is top on our list of January todos.  In the meantime we're all about transparency...

Here's the downlink from the gateway.  It shows some weathering, but otherwise it's totally unchanged from August.  In the coming weeks it will finally get moved across the platform to a more permanent mount.


 I'm more than a little surprised that the alpha-version Nano-loco feed mount hasn't warped or broken. 


This jauntily mounted little pico serves the users right around the Health Center, note the angle to direct the coverage circle down.  Gotta love zip-ties, btw.  



Finally, my personal favorite: the headnode.  This little linksys is the gateway to all of the pilot network. A close look will reveal some chunks of concrete in the bottom of the box.  Purpose?  Keep the router from being submerged when it rains.   {shudder}.  It does, in fact, work when it rains.  


Not shown are two other distribution hubs that consist of nano-locos zip-tied to poles and don't make for particularly interesting photos.  

Stay tuned for some before and after shots coming soon.  

Thursday, January 6, 2011