Forum: GNU Radio Fwd: Proposal for GSoC on gr-gsm

A07755847ec8d05052221d351a3ae20f?d=identicon&s=25 zhenhua han (Guest)
on 2014-03-11 11:16
(Received via mailing list)
---------- Forwarded message ----------
From: zhenhua han <hzhua201@gmail.com>
Date: 2014-03-11 16:00 GMT+08:00
Subject: Re: [Discuss-gnuradio] Proposal for GSoC on gr-gsm
To: Bogdan Diaconescu <b_diaconescu@yahoo.com>


Thank you, Bogdan. Your work is a great help in developing the channel
hopping part.
As there are only 14 weeks in GSoC, the schedule is a little tight.
However, I will continue working on this project after GSoC (if I am
selected). And Channel hopping will be the first task after I finish all
the tasks planned in GSoC.

Best wishes,
Zhenhua




2014-03-11 15:34 GMT+08:00 Bogdan Diaconescu <b_diaconescu@yahoo.com>:

 Hi Zhenhua,
B4ffbc711addde4c649b1ed526df6157?d=identicon&s=25 Martin Braun (Guest)
on 2014-03-11 13:32
(Received via mailing list)
On 03/11/2014 11:14 AM, zhenhua han wrote:
> As there are only 14 weeks in GSoC, the schedule is a little tight.
> However, I will continue working on this project after GSoC (if I am
> selected). And Channel hopping will be the first task after I finish all
> the tasks planned in GSoC.

You should definitely check out the PFB channelization, though. For
multi-ARFCN applications, this will always be a requirement.

M
616d6d8c8b18b9bbd5a998ff7ae69066?d=identicon&s=25 Bogdan Diaconescu (Guest)
on 2014-03-11 14:42
(Received via mailing list)
I totally agree with Martin regarding PFB channelizer. The PFB for gsm
will be also a good challenge for gnuradio in general as obtaining only
a moderate number of channels(say 50) takes a lot of processing power
and achieving realtime processing is not possible currently. Split per
thereads and VOLK should be taken in consideration.

Bogdan







On Tuesday, March 11, 2014 2:30 PM, Martin Braun
<martin.braun@ettus.com> wrote:

On 03/11/2014 11:14 AM, zhenhua han wrote:
> As there are only 14 weeks in GSoC,
the schedule is a little tight.
> However, I will continue working on this project after GSoC (if I am
> selected). And Channel hopping will be the first task after I finish all
> the tasks planned in GSoC.

You should definitely check out the PFB channelization, though. For
multi-ARFCN applications, this will always be a requirement.


M
A07755847ec8d05052221d351a3ae20f?d=identicon&s=25 zhenhua han (Guest)
on 2014-03-13 04:26
(Received via mailing list)
Hi,

I have implemented the algorithm in the paper for test and uploaded my
code
to github.

https://github.com/hzhua/gr-fchdetection

The result given by the algorithm seems great.

Here is a figure generated by my program which is the ratio of average
error power to input power. It is very low at frequency burst and high
at
other bursts.
[image: ǶͼƬ 1]
I have also added this part in my proposal.

Best,
Zhenhua



2014-03-11 21:41 GMT+08:00 Bogdan Diaconescu <b_diaconescu@yahoo.com>:
A07755847ec8d05052221d351a3ae20f?d=identicon&s=25 zhenhua han (Guest)
on 2014-03-13 04:28
(Received via mailing list)
Oh, I forgot to say. The data is sampled by a RTL-SDR.

Zhenhua


2014-03-13 11:25 GMT+08:00 zhenhua han <hzhua201@gmail.com>:
A07755847ec8d05052221d351a3ae20f?d=identicon&s=25 zhenhua han (Guest)
on 2014-03-13 14:13
(Received via mailing list)
The figure in the previous mail is incorrect.
It is too weird that the ratio is so low and with narrow transition. And
the sample count is not match the FB.

After some debugs, I found the reason. I got the sample data with GNU
Radio
companion. When I execute the flow graph, it starts to write data into
the
file before my RTL-SDR started. So there are lots of invalid samples at
the
start.

I have fixed this bug. Here is a correct (maybe :) ) figure. The sample
rate is 1.25M sps. The FCCH burst lasts about 700 samples which equals
0.56
ms. The duration of a standard FB is 0.57 ms. It seems correct.

I have updated this part in my proposal. Sorry for my mistakes.[image:
ǶͼƬ
1]


Best,
Zhenhua


2014-03-13 11:27 GMT+08:00 zhenhua han <hzhua201@gmail.com>:
A07755847ec8d05052221d351a3ae20f?d=identicon&s=25 zhenhua han (Guest)
on 2014-03-16 08:28
(Received via mailing list)
Hi,

My proposal has been updated on github.
The proposal has been changed a lot according to the comments on
Melange.

1.The flow graph has been changed. A new architecture is designed.
2.A state machine is added which explains more details on GSM decoder.
3.The deliverable list has been changed.
4.The schedule has been changed according to the new architecture.

Looking forward to further advises :)

Best wishes,
Zhenhua

2014-03-13 21:12 GMT+08:00 zhenhua han <hzhua201@gmail.com>:
A07755847ec8d05052221d351a3ae20f?d=identicon&s=25 zhenhua han (Guest)
on 2014-03-21 16:18
(Received via mailing list)
Great thanks to all of you here and in melange! The content below is
updated according to the advises above: 1.More details in time
synchronization and burst extraction; 2.More details in using streamed
tag
to get more stable frequency resetting; 3.More details in frequency
correlation; 4.A state machine is added;
5.The structure of flow graph is changed; 6.New deliverable list and
schedule; 7.More references.
8.Deleted the words of promising some tasks after GSoC.




2014-03-16 15:27 GMT+08:00 zhenhua han <hzhua201@gmail.com>:
Please log in before posting. Registration is free and takes only a minute.
Existing account

NEW: Do you have a Google/GoogleMail, Yahoo or Facebook account? No registration required!
Log in with Google account | Log in with Yahoo account | Log in with Facebook account
No account? Register here.