Russian adventures OpenBTS

I must say that despite the fact that all the GPL-projects are equal, some of them evenly. In other words, GPL-projects can be more open and less open. And OpenBTS is just less open-source project. How can this be, because the license says that all source code should be available? Yes, it is available, but only when you start to distribute the product based on these sources. Ie while you are working to develop, you can keep your change with you and no one to show. Or show, but limited. And it will be under the license. Such conditions are logical, but sometimes can lead to strange consequences.

If you look at the downloads page OpenBTS , you'll find there version 1.6 (New Iberia) - a version that Kestrel SP released in April this year, nearly six months ago. From the official blog of the project can learn that at Burning Man 2009, they used version 2.5, which they have worked well and they had during the festival well debugged. Moreover, the official wiki of the project is a compatibility list of phones for version 2.4.

Where are these versions 2.4 and 2.5? And they are only "active developers" from the community and commercial customers Kestrel SP. Why? David explains - the company needs funds to exist, and customers are willing to pay for something, to first gain access to some functionality. Active developers also can be equated to customers because it saves time and money the company. In addition, - David does not say, but it is obvious - thus somewhat difficult life for those who want to take the soft for free and sell under its own brand. Basically, everything looks quite logical, if not a few "buts". Suppose a person begins to develop in the project, making the first test and soon finds the first bug in the project. He's got options - (1) ignore the bug, if it is not a serious obstacle, (2) to report it to the list or in the tracker of the project, and (3) to fix it and send the patch to the project. To project the most useful option (3) - a project for free gets a correction, to which of the main developers did not get hands or that is rare. From the options (1) and (2) project in the long run neither cold nor hot. But remember that the person running the old version and knows it. And in this case he has a very weak motivation to fix a bug found - once developers write that they all works great, then they probably have it fixed and you can just wait until they lay out a general access the new version. Do not do the same, indeed, double the work, which will then nobody wants - the developers certainly prefer their bug fix than some party. As a result, many errors, which could be corrected by the community, developers have to fix themselves, wasting their time (and money).

There's another aspect of the problem. It turns out that the community give the release to which you obviously worse than working, and while they say that everything works fine (but the new version!). People have questioned this version, try to make it work, and they are not very good, they spend a lot of time and effort in trying to understand what is wrong - in fact wrote that everything works. Then they say: "Here's newer version, we will publish it soon in open access, it may be better to work. They are trying - and everything really starts to work better. How then feel these people? So, as if they threw. Not the most useful sense, when you work with open-source project.

PS Yes, we have received from David on a test version 2.4 and 2.5. Early results are encouraging - now all of our test Siemens'y calmly connected and working. More details about the new test results - in the following positions.

So, the day before, on Saturday, we held our first test we collected the system with different phone models. Before that we had to be limited only to those phones, that we have - a few old Siemens'mi, a pair of Nokia (N78 and 6021) and my Asus P750 (on WM6). In home tests:

• Asus p750 refused to connect (but tried a few times).
• Both Nokia connect and work. 6021, we connected via standard network search, and the N78 through «BTS test» - the channel is locked ( ARFCN ) to use our station.
• Once we managed to test Siemens, he happily worked for another three or four times disconnected-connected, normally called. We were delighted, and went to dinner and when we returned - he was no longer connected.

Saturday hosted the first autumn meeting of the Club ZPSH and we did not fail to take advantage of this to test our system with a more diverse set of phones. In the tests involved several Nokia phones varying degrees of antiquity, two phone Sony Ericsson, Motorola, one and two iPhone (2G and 3G). (Thanks to everyone who gave their phones in our hands:)) Result - all Nokia phones are successfully connected and worked without problems. All other phones have seen our network, but connect to it could not. An interesting exception was iPhone 2G, which, to our surprise, calmly connected to our network. But when you try to call him from another phone (Nokia 6021) was sound in only one direction - the iPhone could hear what was said in Nokia, but Nokia was complete silence. Unfortunately we did not have time to deal in detail with the bugs - maybe the problem is actually related to the SIP-part and has no relation to the iPhone. At the next opportunity to test the iPhone will try to understand the causes of such behavior.

The conclusion for those who want to raise in his OpenBTS - for the first test is better to use phones Nokia. They are not so choosy on the accuracy of hours of BS, and with them you soon get at least something works. You can try and use the iPhone 2G, but for them we do not have big statistics.

In general, it seems that we are faced with the famous problem of inaccurate clock , despite the fact that we used a stable low-noise oscillator own rations. We have not understood the reasons for the problems encountered - probably to blame for any interferences, which make the phase noise. Unfortunately, the equipment on which it can be verified, it is enough road to have it at home "just so" and we have to move almost to the touch. If you have an oscilloscope with a bandwidth of no less than 500MHz, a highly stable low noise frequency generator for 52Mgts or 64MGts (phase noise <<-60dbc on 1khz, stability of better than 1ppm) or a highly stable chastotometr (accuracy <<0.1ppm) and you do not mind to let These two men with a small base station,:) you will greatly simplify our debugging process.

Finally, the photo with one of our other exit test - in the basement at my cottage.

Test in the basement

The idea was that in the absence of other wireless LAN phone first tries to connect to our BS and, accordingly, will synchronize their watches in the first place with her, that should reduce the accuracy requirements of hours of BS. But whether isolation of the signal tower of MTS, which stands opposite, was too weak, or whether our assumption is not true, but half an hour of testing, no phone Siemens was not able to join us.

This is a blog where we write about our successes in working with OpenBTS , as well as all the interesting things that we find on the road. Welcome to this small Russian-speaking island open GSM communications. :)

PS If you also decided to plunge into the world of open GSM connection and you have something to tell - we are pleased to invite you to participate in this blog. Write!