|
From: David A. B. <dbu...@jc...> - 2009-09-28 17:49:31
|
Hi - On Sep 27, 2009, at 10:38 PM, Harald Welte wrote: > Hi Reynaldo, > > On Mon, Sep 21, 2009 at 10:53:33AM -0500, Reynaldo Baquerizo wrote: > >> I'm embarked in a project to measure the performance of openBTS. >> I have >> already build up openBTS and camped succesfully but now I look >> toward write >> a formal protocol procedure to test stability, fiability and its >> scope. > > This is really exciting news! Once you get to the actual > implementation, > please make sure that the RF layer testing is independent from the > layer 2 > protocol testing and layer 3 protocol testing. This way, other > projects like > OpenBSC can reuse the protocol 3 part, e.g. This might also be the basis for some kind of automated regression test, something we need more and more with every new feature. > > Also, for the actual RF layer conformance, you will have to > implement a full > MS-Side layer 1 of the Um interface - which is a really large task > in itself. > > The MS side interface is more difficult than the BTS side, as the > BTS is > clock master and on the MS side you have to do synchronization of > carrier > clock, bit clock and fram clock... Yes. We have given some thought to that as well. L1 is in two parts, the GMSK modem (sometimes called "L0") and the FEC part. The FEC part is nearly the same in an MS, but the GMSK modem part needs additional signal processing to sync to the BTS and track its clock. L2 is nearly identical in the MS and BTS. The only difference is in which side can initiate multiframe mode in SAP0. L3 is a kind of mirror image, with the MS decoding messages sent by the BTS and generating messages that the BTS expects to receive. > > I would be very interested in having that part as a standalone > project, > as it could serve as a building block towards an open source GSM MS > (something > many people have been dreaming about). Agreed. However, this is some risk in building the test fixture and the system-under-test from the same components, especially in L1 and L2 where you can simply have a symmetric error, where the two systems are compatible with each other but neither conforms to the spec. We've done that before, in our early standalone testing of L2. > > Regards, > Harald > -- > - Harald Welte <la...@gn...> http:// > laforge.gnumonks.org/ > ====================================================================== > ====== > "Privacy in residential applications is a desirable marketing option." > (ETSI EN 300 > 175-7 Ch. A6) -- David David A. Burgess Kestrel Signal Processing, Inc. (PS -- I do not endorse the spam that sourceforge appends to my e-mail.) |