3G Microcell - fun while it lasted

As you're aware, if you've been following along, when we went to the AT&T store in San Diego, we told them we were from out of town and the sales unit typed our home address here in Santa Clara into the computer and assured me that when I got the unit back home I could put in that address and it would be fine.

And, for a week, it was.

Apparently, that was a bug. A bug that AT&T fixed sometime before last night. As of now, I am no longer allowed to register our address here in Santa Clara with the microcell. And without it being located at the address to which it is registered, the GPS check fails and it emulates a doorstop.

I called up AT&T this morning and their plan of record is now to have me ship the unit back to San Diego for a refund.

Idiots.

There is a 30 day refund policy. If the unit is going to be made available in less than 3 weeks, I'd rather just keep it. But naturally, they refuse to "discuss future availability." I could understand not issuing advanced press releases, but these are extenuating circumstances. I cannot possibly believe that there is nobody up the food chain that has a clue as to even approximately when they'll roll these out.

Besides, they've been "testing" the microcells for, what, 9 months now? They work. Really well. Just flip the goddamn switch, AT&T!

3G Microcell, more day-to-day info

So it's been a few days since we got home and the microcell is still working perfectly. We haven't tried really abusing the upload bandwidth while a call was in progress. The same carve-out that was in place for Vonage on our server is still in place, so that should be sufficient.

I had an opportunity to test the call hand-off this morning. I called the USNO master clock (202-762-1401) in the driveway while the M-Cell logo was on the phone. I stayed on the call (via bluetooth, not with the phone up to my head) until I had driven about a half mile away. The call got a little spotty very briefly as I got a little bit away from the house (mostly because the coverage near our house is not so good), but the call was handed off. While the call was still going, the M-Cell logo stayed in place on the screen, despite the fact that the phone was well out of range and had handed off to the network proper. As soon as I hung up, and the call was ended, the logo reverted to the usual AT&T one. This likely is the phone's way of letting you know that if you have the unlimited calling plan that the call is still free.

When you fire up the Google maps app on the phone and then click the 'location' button, typically what happens is that the phone uses cell tower triangulation to get an approximate fix, then the assisted GPS functionality refines that down to a point. When the phone is in range of the microcell, it's clear that the microcell's GPS receiver has an impact on that initial triangulated fix - the dot starts up being in the right spot with a house-sized circle around it, then when the phone's GPS kicks in, the circle shrinks down to nothing.

One thing that's a little bit of a let-down for a techie guy like me is that the microcell is a bit of a black box. The status lights on the front don't ever change, and so far as I have found, the box has no open TCP listening ports, so it doesn't have an HTTP server or anything of that sort. It's the quintessential Internet appliance in that regard. It would have been slightly more fun had it had some sort of configuration or status UI. Apart from the blinking status indicators while it's booting up and activating, it doesn't do anything visible at all. Since there's no way to configure the box, it's likely that it requires a DHCP server in order to function (this is in contrast to lots of appliances that use DHCP by default, but allow manual overrides with a configuration GUI). This would have been an issue a while ago had I chosen to deploy the box to the outside network instead of behind the NAT, since it used to be that the outside LAN had no DHCP server. That has since changed, but if you have a completely manually configured network (likely a rarity nowadays), you may have some trouble.

It has had no trouble receiving a GPS signal, so far as I can tell, despite being pretty close to the center of the house. There is a window nearby, so it's conceivable that that helps, though probably the bigger factor is that above the unit is just a drywall ceiling, a layer of cellulose insulation, and a plywood and asphalt shingle roof. Wandering around inside with a Garmin tends to work most of the time. The box has a GPS antenna jack on the back, though at the moment there's no documentation about it, nor is there any mention on AT&T's site about an accessory antenna you can buy and situate closer to a window, if required. I assume that when the nationwide roll-out happens they might offer one.

Since the box plugs into AC power, it will die if there's a power failure. The network infrastructure here is on a UPS, but unfortunately it's also located out in the garage. I wanted the microcell a little bit more centrally located than that, since the path from the garage to the office goes through a lot of walls. One workaround I have used before is to cobble together a power-over-Ethernet solution - a pair of specially modified Ethernet cables that break out one pair to pass power. Then just set up one continuous Ethernet path from the back of the garage to the dining room (this is easier than it sounds), plug the power plug from one hack-cable into the microcell, plug the wall-wart into the socket hooked up to the other hack-cable, and then plug the wall-wart into the UPS.

3G Microcell

The AT&T 3G Microcells are for sale in San Diego, and coincidently, we're down there for the holidays. I went to one of the stores that was selling them, and despite being told we were from out of town, they were happy to sell us one. Registering it online was simple, although it did take about a half an hour for it to complete its setup and become active. During that time, you can't help but wonder if it is going to succeed. It needs to get good GPS reception before it will fire up. But once it did, the coverage at our parents' house was quite good upstairs. It was a bit more spotty downstairs, but once we get home I am fairly confident that the coverage should fill our smaller house. The web page at the AT&T site that manages the microcell allows you to put multiple physical addresses in and pick which one is the active one. You must put an address in, and the device's GPS will verify that it's accurate or it will refuse to work.

Normally you'd install the microcell someplace where you already have WiFi, so the fact that the microcell does 3G data is kind of moot. There are some points of interest, however, that make pondering the trade-off worthwhile. With WiFi turned off, I get a fairly consistent result from speed tests: a little less than 1 MB/sec down and about 60 Kbps up. WiFi here at my parents' house in San Diego gets easily 5 MB down and 2 or 3 up (their cable modem service is amazing - like, 15 MB down and sometimes 5 up). However, with WiFi, you're typically behind a NAT, and if you use 3G, you get a public address - even though you're using the microcell. Of course, apps that require local connectivity, like the iTunes remote app, won't work through the microcell because of that (you'll have to use WiFi).

I've made a number of calls through the microcell now, and they all sound as good as a cell phone can get. When we get back to Santa Clara, the uplink bandwidth will be a bit more constricted than it is here, but given that Vonage works ok, I would think the microcell will work just as well.

Update - 12/27

We got back tonight. Before leaving San Diego, I powered off the microcell and changed the address on the unit to our home address here in Santa Clara. When we got back home, the microcell took about a half an hour to come back up. When it did, we got the activation text message again, just like after the first activation. As I predicted, the coverage is just what I had hoped - 5 bars throughout the house and all around our property. Unfortunately, there is no unlimited minutes offer outside of the test markets, so we're not getting any breaks on our plan minutes. But hopefully that will change when the nationwide rollout starts. I suspect that the various test markets are really about testing the price points for those plans rather than any technical issues.

Really, NFL?

We're close to the playoffs. And of all of the games this weekend, there is one that is the most important.

The Chargers are hosting the Bengals at home. The Chargers are one game ahead of the Bengals in the drive for the 2nd AFC playoff seed - which means a first round bye.

In the latter part of the season, the NFL and the networks have the opportunity to "flex" the schedule - to make a more important game the Sunday night game, relegating what was scheduled in the pre-season back to Sunday afternoon (or morning).

So it didn't occur to anybody that this crucial game is worthy of being played on Sunday night? They think Vikings - Panthers is more important somehow? Really?! The Vikings are almost assured of either the #1 or #2 seed in the NFC and the Panthers are all but mathematically eliminated. Who gives a shit?

The only other game even close to how important that game is is the Eagles and 49ers. The 49ers need a win to even have a chance at a wildcard spot, while the Eagles are almost assured of the number 3 seed in the NFC. Of course, since the 49ers are the home team here, it'll be on Fox for sure.

The wrinkle is that the weather has screwed up the game schedules on the East coast. So potentially, the CBS morning game might have been the Chargers', but now that game will be played at the same time as the Raiders @ Broncos. Another "who cares?" game. But it's moot - Fox got the doubleheader today anyhow. Their rejiggered schedule today will give us the 49ers game in the afternoon, and of all of the remaining morning games to show us, they picked... Cardinals / Lions?! Does anybody honestly think the Lions have a shot at their third win this week?

Perhaps the single most important football game of the entire season. None for you!

Epic win

Look what I saw today:



The bumper sticker is available at I Park Like An Idiot.com. I can neither confirm nor deny any information concerning how the sticker got there.

Crossing the anti-spam Rubicon

I've resisted it for a while now, but it's time.

I have added the first DNS TLD to my spam filter.

I have yet to get a single e-mail that wasn't spam from any domain under .info. Not one. It's proven to be one of the best positive indicators of spamminess I've seen in a while now.

There are some .info websites out there that are, well, informative. Which is good. But if anyone can list any legitimate source of e-mail in the entire .info zone, then list it in the comments. Because I ain't seen one yet. And as things stand now, I never will.

There! I fixed the backyard gate!

We used to have a nice handle and lever style latch on our backyard gate. It was very attractive, and allowed operation from both sides. The problem was that it turns out it wasn't dog proof. Our new dog, Luna, was able to "bonk" into the gate and get it to flip open. She escaped into the world a couple of times, only to have neighbors return her. That's rather embarrassing, so that gate latch had to go.

Everyone has seen the standard kind of gate latch. It consists of a horizontal bar that attaches to the door (if it's inward-swing), and a metal bracket with a slot for the bar to fit into and a hinged metal piece that goes up and down to lock the bar in. It's serviceable, and (at least in our case) dog-proof. The difficulty is that it has no provision for operation from the outside.

Well, not quite. There is a small hole in a tab on the latch that you're supposed to tie to a piece of string run through a hole in the fence. We've all seen those, and the various failure modes.

If the string in any way restricts the free movement of the latch, the whole mechanism is bound to fail. And, of course, in this case, the failure mode is unlatched, dog-roaming mode. Even if the string works, it won't work for long - the wood fence material will no doubt shred the string before too long.

Well, you could attach a wire instead of the string. The problem there is that a solid wire is even more likely to impede the free movement of the latch.

No, what you need is a frictionless, low profile linkage between the cable and the latch.

What you need, is a fairly standard part from the RC model hobby industry: A clevis linkage. A clevis linkage is a small pair of long, thin metal plates that normally sit parallel to each other. At one end of each, there is a round grommet through which you run the cable and either crimp or solder to permanently attach the linkage to the cable. At the other end of one of the parallel plates is a pin, and at the end of the other plate is a hole inside which the pin sits.



Solder a foot long (or so) piece of 1/16" braided cable in the clevis, and use a crimp coupling to form a finger loop at the other end. Drill a 1/4" hole right behind the top of the latch. Run the clevis coupling end of the cable through the hole and attach the clevis pin to the latch.

I have to admit, I didn't come up with this idea myself. We used to have this same problem on our other gate - the string wore out almost immediately, and I was just used to reaching over the fence (I am tall enough) and flipping the latch by hand, but that didn't help Scarlet. So one day while I was in our local hardware store, they had a display for a gate latch improvement kit. It was virtually identical to what I've described above, except that their kit also included a very, very large spring intended to go between the gate and the latch to help insure closure, and a washer with a flange that was meant to sit in the hole and hold the base of the spring centered around the cable hole.

I'd heartily endorse that kit, except that the local hardware store doesn't carry that product anymore, and nobody else has ever heard of it.

The other problem is that while that kit worked flawlessly on the gate where I installed it, a quick test shows that it wouldn't have worked quite so well on this gate. The issue here is that the latch is installed on a 2x4, but the hole you drill for the cable is above that 2x4. That means that the washer / spring base would sit a good two inches behind where it's designed to. I'd either have had to build up the face of the fence at that spot so that the hole emerged "at grade" or modified the kit's spring somehow. But that implies I would have had the kit to start with - I didn't.

Instead, it appears that, at least in this case, the spring is unnecessary. Of course, YMMV.

That didn't take long

They kept telling me that if I voted for John McCain that we'd seen escalation in the US military presence in the middle east. I did vote for McCain, and sure enough!