10-27-2011 6:55 PM
I have owned a AT&T/Cisco uCell for over 6 month and have now decided to find a solution for drop calls issue. After reaching the limits of ATT support I begun to seek information about that sick box.
As an introduction this device is shrouded with mysteries because there are hardly any available ways to troubleshoot it. Cisco has it locked down tight even to ATT technicians. What little mean of remote management available to ATT has been used by them already to solve common device registration issues (DSL networking, E911 geolocalization). It appears impossible for ATT support techs to address issues related to call disconnects beyond stating: "the uCell won't work at your location"
1- NETWORKING SIDE:
The uCell is connected at all time to the ATT network over a VPN connection on UDP port 4500. It updates its network time with ATT NTP servers on UDP123 couple times per hour and uses TCP port 443 for Https: another secured connection. The good thing here is that there is NO need to unblock incoming Firewall ports because the uCell initiates all connections from within our private network. Incoming data can always be assigned to existing opened connections.
All in all the networking design is rather solid and simple for users. The router MTU settings (1492/1500) may be worth exploring to prevent fragmentation of the uCell UDP.4500 packets if it is hardset at 1492 for aDSL networking...
The bandwidth requirements of the uCell have been kept small to work well within ATT aDSL limits. The up/down streams are below 500Kbps ie. connection bandwidth does NOT cause problem on cable Internet (Dwn: 30,000Kbps /Up: 6,000Kbps) especially if you run a QOS router to prioritize VOIP packet class ahead of any background traffic.
2- CELLULAR SIDE:
The firmware aspect that's half-baked is the cellular radio logic. There is a lot of patented material needed to emulate a cellular radio where Qualcomm is a key player with boat loads of patent attorneys so Cisco may not be playing with a deck of necessary licenses in that field.
Most of the uCell usability issues in my world fall within the cellular category: SMS, Apple 4S, dropped calls, lag time, noisy audio. There is a must-read 13 pages article on Anandtech ("A Comprehensive Exploration" by Brian Klug) about the uCell where one can learn how calls hand-offs should work.
3- TEST SOLUTION:
Armed with the above knowledge I began testing a few original ideas. Living on the outskirt of town ATT reception is a mixed blessing:
- my uCell is picking up to six 3G towers + five 2G towers
- my phone receiption has 1 bar indoors(-103dBm) , 3 bars outside with line of site.
At initialization the uCell does:
- builds a list of adjacent cellular towers
- sets its cellular frequency: 900MHz or 2.1GHz
- adjust its power output to match the reception level in its environment
- geolocalize either by 1/ cellular triangulation or 2/ by GPs signal
When the uCell is initialized without receiving 3G/2G macro-tower signals:
- has a clean/lean list of adjacent towers to process call hand-offs
- has to use GPS signal to geolocalize E911 address instead of triangulating 3G towers.
- it outputs a stronger signal inside the house for solid calls
Reset button while powered followed by GPS-only initialization with limited macrocell reception.
1- Made a shield with aluminum foil to cover everything but the GPS receptor antenna and LED's
2- Stuck the Reset button with a toothpick for 30 seconds without powering down the unit
3- Removed the antenna shield after all green lights (~10 minutes)
On "3G data" I now get 2,300 Kbps download and 300 Kbps uploads, stronger in-house coverage and so far no dropped calls but keeping my fingers crossed for more off peak testing.
10-28-2011 12:49 AM
10-28-2011 7:19 AM - edited 10-28-2011 7:40 AM
Thanks for the detailed article. My location sounds similar to yours - though our elevation is about 500' higher than the average within our 15-mile valley, I can probably "see" about 2 dozen cell sites. However none of the cell sites provide strong or consistent enough signal to provide reliable coverage at my house.
I've thought about shielding the MC while it starts up and builds it's list of nearby towers, but never had the time to experiment. I would consider resetting the device around 5pm when the cell towers are most likely loaded the most and their coverage area shrinks the most.
I have an app on my Blackberry which among other things shows me what frequency my MicroCell is operating at. I'm curious if it uses 800 MHz (vs. 1900 MHz) when it can't see any nearby cell towers?
I'll be watching for any updates over the coming days to see how your experiment pans out. I am now thinking of any metal boxes I have in my house that would be big enough to act as a signal attenuator while the MC initiates.
10-28-2011 11:10 AM
to build a faraday cage of some sort I've used what ever was easiest around me at that time: kitchen aluminium foil. I made sure the area around the GPS antenna was left open to pickup location and for cooling during the 10 minutes it took it initialize after a warm reboot (not from power-down state)
A metal box will render BOTH GPS and macroCell reception useless so the unit won't fly unless it can validate E911 location.
Results so far are excellent and meet my expectation: I have indoor coverage and make wireless calls that don't disconnect.
I do see that this uCell signal (0.9GHz / 2.1GHz) behaves much like WiFi (2.4GHz) where it is sensitive to echo-ing and wall attenuation. I translate that by install the uCell in a central point inside the house because the nature of the cell signal is a RF radio ie. circular as in round. Of course "line of site" with no obstruction will always works best.
At this time I have the uCell installed upstairs and I get my full bars downstairs at (-70dBm 22asu).
The perfect troubleshooting tool here would be an RF Spectrum Analyser to see:
- What carrier frequency the uCell is using
- Measure the carrier amplitude in dBm
- HOW CLEAN the carrier signal is couple rooms away du to signal-noise from building echoe
Our cell phones shows this information in a mixed format. The bar display (0 to 5) shows how strong is the useable part of the signal: cleaner is always better.
10-31-2011 5:32 PM
Can you share more info about what you have to cover and what you have to leave uncovered? I'm not sure where the GPS antenna is located. Can you give a more detailed description? Thanks.
10-31-2011 10:42 PM
If you read the Annantech article by Brian referenced in my post you will get a description and pictures of the uCell inards, specifically the GPS receiver and the cellular antennae.
- The GPS receptor is located at the top of the unit towards the rear - You can see it from the top once you have seen the article pictures. This antenna MUST receive GPS signal or else the unit won't go through its initialisation course. (GPS LED blinking for ever)
The two cellular antennae are printed coper traces on the main circuit board and located towards the front of the unit ie. towards the bank of LED's. This is what I purposely kept shielded in my initialisation procedure.
I kept in mind that the unit is aircooled and may overheat as the temporary shielding prevents any good airflow. Luckily my unit was ready with all solid LED's in less than 10 minutes after a warm boot (not powered out)
I believe there is some form of confusion when ATT tech support talks about Registration... I believe there are 2 types of "registrations":
1- there is the initial unit registration when unit is new fresh out the box the first time where you assign it to your account
2- there is the network sign-on/registration the uCell unit does everytime when it comes back alive after starting. That is the one I most interested about. This is when it updates is runtime information regarding its environment specifics. Hence the init procedure I came up with to make it operable.
Today I managed to send my first ever MMS with an attached 630KB picture. Of course I used the uCell and it worked but wayyy slow compared to normal 3G through put I measured with Ookla speedtest. As a reminder, just be sure to disable the WiFi Internet connection before sending SMS/MMS because they MUST go out on the 3G carrier network NOT the Internet
11-02-2011 7:55 AM - edited 11-02-2011 7:57 AM
Yesterday I followed your procedure and the results so far are positive.
First I wrapped the front 2/3 of the MicroCell in aluminum foil - paying attention to not cover what appears to be the GPS antenna - a 1" x 1" sqaure under the top rear of the device. I then got some aluminum and metal baking pans and layed them along the sides to help further attenuate the signal from nearby macro cells. Though after pressing the reset button with my first MicroCell I was told that I needed to get a new one becuase I had effectively bricked it, I trusted your procedure and held the reset button until all the lights briefly went out. After about 20 minutes I peaked through my aluminum foil sheild and all lights were solid. I removed the pans and foil.
Using the SignalLoc field test app on my BlackBerry I've noticed a few things. First, my MC is still using the same 1900Mhz control channel. I've also noticed that even while my BlackBerry is sitting still on a table one room away from the MC, at times the RSSI signal fluctuates wildly by up to 10dB. I don't recall this happening before.
I made about 20 minutes of calls last night from my office which is about 20 feet from the MicroCell and had no drops and very clear reception. One call I walked from my office, out the front door and down the street. About 2 houses away the call dropped. Trying that test before, about half the time the call would drop and the other half the call would hand-off to an EDGE or 3G macro cell. I will be testing more in the next few days and report back with any additional observations.
While doing some speed tests last night I was looking at my test history and found it interesting that sometime between July and September my ping time on the MicroCell went from ~600ms to ~250ms.
11-02-2011 3:02 PM
Right on "Wireless_Guru", now you are good! You are on your way to happiness again using AT&T... Can you test and report more data using your BlackBerry SignalLoc field test app. I am primarily on Android phone and don't have this sweet SignalLoc tracking App. Please do more uCell signal measurements inside and around your house for coverage levels.
Keep in mind that when the uCell looses power any shielded init benefits will be gone. In that case simply redo init.
It is normal for uCell Ping numbers to suck (>>100ms :...~500ms) because packets are essentially tunneled into ATT servers (secured UDP4500) then out on the Internet into the Ookla test servers and back. This adds a lot of overhead processing time when compared to using Wifi on your LAN: test packets go directly out of your ISP into the Ookla test servers (<50ms) and back to the SW test application.
My current approach to building a TEMPORARY shield is to WRAP the uCell in a BLANKET of kitchen aluminum foil leaving the GPS antenna uncovered. It has to be as good as possible so nearly no street cell signal reaches the tested uCell inside the shield. Unwrap as soon as GPS + 3G LEDs become solid green or in less than 15 mn in 65F homes to prevent any associated internal overheat damage. If these 2 lights don't go solid within that time frame: stop! ... something else is causing issue (likely bad GPS antenna reception - better shield opening needed)
I am glad to learn your unit is now handing over calls to existing cells the way it should when you walk too far as opposition to dumping your calls to remote ghost cells (>100dBm) that your phone can not reach thus dropping calls!
This procedure only serves as a proof of concept to show the uCell hardware is sound. It is only the current firmware logic that is sick!
In its current incarnation the uCell picks-up and holds-on to tower cells that are too far away, creating ghost cells our cellphone can never reach indoors. The uCell keeps trying to bump phones back to the street thinking they can reach these remote cells. Call hand-off to MacroCell is a desired function but at a different signal level than done currently in stock firmware.
So in essence the signal reception logic in the current uCell firmware makes it too sensitive. I believe this was done to help geolocate the unit by triangulation when indoor GPS coverage is poor but then afterwards the MacroCell signal level should be toned down to properly hang to cell phones listed under the subscriber registration preferences.
Call bumping when we walk too far from the uCell works properly after my init test procedure. Without shielded initialization the uCell prematurely keeps trying to bump phones. This firmware bug is in the logic dealing with signal levels. This is what is causing field issues for customers.
This serves as an example to show what can be done with time, a healthy dose of understanding and Internet forums. If ATT was true to its "Rethink Possible" marketing tag they would have uploaded a firmware fix - Proof: I was able to conceptualize and validate a creative solution without ATT countless resources.
Live and learn...
11-02-2011 7:59 PM
Well, the honeymoon is over. Shortly after this morning's post I was in my office approximately 20' from the MC, and my call got handed off to a macro cell after a few minutes. As soon as I ended the call, it immediately switched back to the MC. While driving slowly away from my house, the call handed off to a macro 3G cell.
Perhaps the MicroCell did a reset overnight, and built a list of nearby macro cells? Or perhaps the phone holds the list?
11-03-2011 1:39 PM
Here is my thought. It does the thing where it checks how many 2G and 3G AT&T signals it picks up every day, sometime between midnight and six a.m. It checks it against what it got the last time it did the search (the day before, or when you re-initialized it). It is looking for it to match. So if you do the thing and it is getting zero, because of the tin foil, and then it does its nightly check and finds different ones, it resets by looking for a GPS lock again. It is trying to find the same ones it found most recently, so zero before and zero again is a match, but if it had picked up multiple ones last time, it has to find at least half of the same ones again.
I have been having the worst time with mine. it is in a downstairs / sub-street level apartment with really thick walls, so it always has a hard time getting a GPS lock. Plus I have all the annoying problems that everyone else gets even when it is working. I haven't found any solution that works for me all the time.
11-03-2011 2:02 PM - edited 11-03-2011 2:35 PM
Georgiacc: you may want to ad the external GPS antenna using connector on the back. But honestly in your case I would concider throughing in the towel to use a better carrier that has coverage, support and sales new gear in working order.
Here is some hope:
If you manage to get GPS to lock then you may have to deal with the uCell flaky behaviors or perhaps not if there is No street receiption in you basement home. The uCell is reported to work okay in remote places where there is zero tower coverage thus my idea to simulate that condition.
Cheap fix solution for ya:
If you need to have the uCell outside to get sky view... plug it on a long extension power cord and use a longer ethernet data cable.
As for the uCell re-initializing itself daily, I can't say because I have not witnessed it.
Since your unit can not pass the activation stage (GPS sig) in the first place, I believe it never moved pass "initialisation", right?
You need to find a way to get those lights to go solid. Either using tower tringulation or better by GPS sky view only. Make sure you always have it connected to Internet during init.
11-03-2011 2:13 PM
Thanks! I am debating whether to get the GPS antenna or the electrical-outlet ethernet adaptor, because I actually can get it to do a GPS lock, but it involves stretching the power cable all the way in one direction, and the ethernet cable all the way in the other direction, creating basically a 15-foot-long unsightly stretch of cables that is also easy to trip over. If I relocate it and plug it in where the ethernet is, 10 feet away, it can no longer get the GPS lock consistently.
But it does get the GPS lock, maybe half the time, so what I'm saying isn't that I was never able to initialize it. I've had the thing for nine months or so, and it actually used to work pretty well, when I did get the GPS lock. But lately, the last four months or longer, it doesn't work well even when it has GPS. I can stand right next to it with my cell phone and it won't recognize the M-Cell. Or when it does, I get the "Call Failed" problem. Or it drops my calls, even though there is at most one bar of AT&T signal that is not Micro Cell. Or even when I turn my phone off and then back on, which sometimes works, it still won't pick up the M-Cell, which is ostensibly working.....OY. I have been reading these forums enough to get the sense that my frustrations with this thing are shared by others...
But my theory about the daily rechecking of the postioning? This help question:
was recently updated with totally new, weird extra info...take a look at it and see what I mean?
11-03-2011 2:34 PM
If you got handed over to a ghost cell in the street 20" from the uCell I guess somehow your uCell still had bad cells on its internal list.
Stick it good with a toothpick (30s) and run more tests. While the unit is powered off map the places in your house where you have some street coverage.
I am not sure I have all the relevant pieces on-hand, this thread is only experimental. We are here to try and learn something.
The uCell signals travels like wifi (similar frequency range): it easily creates bouncing waves (echo) that kills the signal (signal to noise ratio). In plain words: you can go quite far with direct line of site but not through many walls and stairs wells. That is why we have trouble receiving weak street towers indoors (attenuations + echoes).
So we don't need to hug the uCell but we need to stay on the good side with signal propagation - Thank God uCell output power is limited to miliWatts (0.001 W.) like cell phones not KiloWatts (1,000 W.) like FM/AM radios.
On my end so far things seem to be holding as solid as expected. No probs
11-03-2011 2:39 PM
What I mean is, on the link I posted, it looks like they put something up that was meant to be internal, for Customer Service. This is what it says:
All Data Support
Methods to Determine MicroCell Location
11-03-2011 2:48 PM
Georgiacc: very interesting - let me dig in some more!
I do love the PowerLine modules: my uCell is connected through one.
It seems the uCell should always be kept where it geolocalize even after Init is complete. I recall reading something like AT&T validates the actual uCell location at random intervals (just to be sure it has not moved to a different location)
Congratulations! You earned the Liz badge!