Posts tagged UMTS

AT&T Bands in Las Vegas – 850 GSM/EDGE, 1900 UMTS/3G

Last time I was in Las Vegas it was for MIX 10 and Windows Phone 7 (back when it included ‘series’ at the end). This time, the reason is CES 2011 with AnandTech and a whole bunch more mobile devices.

I thought it was interesting last time I came that most casino floors in Las Vegas had shockingly poor or non-existant UMTS (3G) coverage on AT&T. I guess I didn’t find it too shocking, since coverage inside buildings in a dense urban environment is probably the most challenging for mobile networks, but it seemed to be a consistent problem. After getting frustrated about 6 hours into my stay, I decided to switch entirely to EDGE for the duration just because of how annoying being constantly handed between GSM/EDGE and UMTS is when you’re trying to do things. For whatever reason, back then I didn’t think to pull up field test on the iPhone 3GS I was currently carrying to see what bands were assigned to which network technology.

Now that I’m back, I decided to check. Thankfully, Apple has restored most if not all of the Field Test data products in iOS 4.2.1, a huge step forward from 4.1 just allowing signal strength in dBm at top left, and a far cry from 4.0 which shipped with no field test whatsoever. To save potential readers some googling, to get here, enter *3001#12345#* from the dialer and hit call – if it hasn’t been removed yet, you’ll get dumped into Field Test on iOS.

In EDGE and tapping on GSM RR Info, it’s immediately obvious why I saw that behavior last time I was here:

ARFCN dictates what channel inside what band we’re on, and 142 just happens to lie inside the GSM 850 band. It’s a number basically used to refer to the FDD pair of frequencies the phone is currently using. You can calculate exactly what frequency downlink and uplink are on with a little math and some reference guide (there’s a good table here), but basically with an ARFCN of 142 we know immediately that GSM/EDGE is on AT&T’s 850 MHz spectrum. Between 128 and 251 is that GSM850 spectrum.

Now, what about UMTS/3G? Enabling 3G (look at how weak that signal is…) and going into UMTS RR info, I saw the following:

Looking at the fields “Downlink Frequency” and “Uplink Frequency” we can see the device’s UARFCN channel numbers. It’s the same thing, but U for UMTS. Again, with a reference aide (read: wikipedia) we can see that UMTS/3G is working in the PCS 1900 MHz band.

Remember that higher frequencies are less effective at propagating through buildings. It’s pretty obvious now why getting good 3G coverage on AT&T is a challenge deep inside a casino in Las Vegas. There’s nothing inherently wrong with putting GSM/EDGE on 850 and UMTS on 1900, it’s just interesting in practice how immediately obvious the difference is walking around. Propagation is a challenge in dense urban environments with lots of people moving around to begin with, I’m sure this doesn’t help in Las Vegas. AT&T promised to put all of its 3G (UMTS) network on the 850 MHz band (wherever it’s licensed to use it) by the end of 2010, but sadly that hasn’t happened quite yet, at least in this market. I’ll keep checking, but thus far it’s been solidly in 1900 PCS. Oh well.

AT&T 3G in Las Vegas

While I was in Las Vegas for MIX10, I couldn’t suppress my inexplicable urge to run as many speedtests as I could muster. Of course, I was packing the usual iPhone 3GS with AT&T. Sadly, nearly the entire visit speeds were barely 250 kilobits/s down, 220 kilobits/s up, if I could even get the application to run. Take a look at the following:

(kilobits/s) Average Min Max
Downstream 251.8 14.0 552.0
Upstream 220.8 0.0 357.0

This data is from 13 tests taken during my 3 day stay. They’re from over 3G UMTS when it did work, and GSM EDGE when it didn’t, and that was virtually the entire time. 3G was either slow, or didn’t work at all; switching to EDGE was the only way to do anything.

How is this possible?

Now, it’s fair to say that some of this is sampling bias and the fact that I was at a conference, but even then, there’s no excuse. This is a city used to a huge flux of visitors in a short time for trade conferences. Frankly, I can only begin to imagine how overloaded networks are during major conferences like E3.

Take a look at the following plot of the average speeds for each day:

Average Downstream Speed

Can you spot which three days are the ones I’m talking about? Note that on the 16th, I couldn’t even get a test to run to completion; it just didn’t work. There’s nothing more to really say about the issue than simply how bad this is. If this is the kind of performance AT&T users see and complain so vocally about in the San Fransisco Bay Area and Manhattan, I can completely understand. Frankly, I can see no other reason for that kind of performance degradation other than congestion.