Hi,
Very thanks!
I think one possible reason is that I narrow the frequency offset searching
range from about -200:5:200kHz to -50:5:50 kHz. Because the scanner now use
.1MHz step, why not get rid of extra searching?
You may try "--freq-start 1814.9e6 --freq-end 1815.1e6" to verify that.
But the sample is definitely useful. Would you please capture the bin file
for me? (You may find the method in README).
And,
Would you please also try to use "-t" option for the new Scanner (without
--freq-start and --freq-end) and try again? ("-t" force the scanner into
original mode).
BR
Jiao Xianjun(Ryan)
On Sat, Mar 29, 2014 at 8:08 PM, Dimitri Stolnikov <[email protected]> wrote:
> Hi Jiao,
>
> first of all, thanks a lot for sharing your work!
>
> I've tried it, but for some reason it fails to find the local cell here:
>
> LTE-Cell-Scanner/build/src$ ./CellSearch -s 1815e6 --opencl-device=1
> OpenCL LTE CellSearch v1.1.0 (release) beginning. 1.0 to 1.1:
> TDD/OpenCL/ext-LNB/faster added by Jiao Xianjun([email protected])
> Search frequency: 1815 MHz
> Found Elonics E4000 tuner
> OpenCL: number of platforms 1
> Platform 0 NAME: AMD Accelerated Parallel Processing
> Platform 0 PROFILE: FULL_PROFILE
> OpenCL: number of devices 2
> Platform 0 Device 1 NAME: Intel(R) Core(TM) i5-3570K CPU @ 3.40GHz
> Platform 0 Device 1 PROFILE: FULL_PROFILE
> Platform 0 Device 1 LOCAL_MEM_SIZE: 32768
> Platform 0 Device 1 MAX_CLOCK_FREQUENCY: 3399
> Platform 0 Device 1 MAX_COMPUTE_UNITS: 4
> Platform 0 Device 1 PREFERRED_VECTOR_WIDTH_FLOAT: 8
> Examining center frequency 1815 MHz ... try 0
> 6RB filter cost 0.00318098s
> PSS XCORR cost 0.833934s
> No LTE cells were found...
>
> LTE-Cell-Scanner/build/src$ ./CellSearch -s 1815e6
> OpenCL LTE CellSearch v1.1.0 (release) beginning. 1.0 to 1.1:
> TDD/OpenCL/ext-LNB/faster added by Jiao Xianjun([email protected])
> Search frequency: 1815 MHz
> Found Elonics E4000 tuner
> OpenCL: number of platforms 1
> Platform 0 NAME: AMD Accelerated Parallel Processing
> Platform 0 PROFILE: FULL_PROFILE
> OpenCL: number of devices 2
> Platform 0 Device 0 NAME: Tahiti
> Platform 0 Device 0 PROFILE: FULL_PROFILE
> Platform 0 Device 0 LOCAL_MEM_SIZE: 32768
> Platform 0 Device 0 MAX_CLOCK_FREQUENCY: 975
> Platform 0 Device 0 MAX_COMPUTE_UNITS: 24
> Platform 0 Device 0 PREFERRED_VECTOR_WIDTH_FLOAT: 1
> Examining center frequency 1815 MHz ... try 0
> 6RB filter cost 0.0368459s
> PSS XCORR cost 3.36666s
> No LTE cells were found...
>
> When i run the original program eit does receive the cell:
>
> ./CellSearch -s 1815e6
> LTE CellSearch v1.0.0 (release) beginning
> Search frequency: 1815 MHz
> PPM: 120
> correction: 1
> Found Elonics E4000 tuner
> Examining center frequency 1815 MHz ...
> Detected a cell!
> cell ID: 263
> RX power level: -15.8443 dB
> residual frequency offset: -68142.8 Hz
> Detected a cell!
> cell ID: 261
> RX power level: -17.8496 dB
> residual frequency offset: -68142.3 Hz
> Detected the following cells:
> A: #antenna ports C: CP type ; P: PHICH duration ; PR: PHICH resource type
> CID A fc foff RXPWR C nRB P PR CrystalCorrectionFactor
> 263 2 1815M -68.1k -15.8 N 100 N one 0.99996245719114640327
> 261 2 1815M -68.1k -17.8 N 100 N one 0.99996245745901013446
>
>
>
> How can i help you to debug this? Do you require a capture of this cell?
>
> Best regards,
> Dimitri
>
>
>
> On Sat, 29 Mar 2014 12:24:21 +0100, Jiao Xianjun <[email protected]>
> wrote:
>
> Hello there,
>>
>>
>> New features coming:
>>
>> 1. Now MATLAB isn't needed. All in C/C++. All you need are rtl-sdr dongle
>> and Linux computer. TDD and FDD are both supported. External LNB/Mixer
>> (allow mixer and A/D use non-coherent clock sources) is also supported.
>>
>> 2. OpenCL parallel computation is supported! It searches per frequency
>> point in
>>
>> 0.5s (desktop i7-3930K 6 [email protected]) 22x faster!
>> or 2~3s (ThinkPad T410 i5 CPU [email protected]).
>>
>> Intel, AMD and Nvidia OpenCL SDK are tested. Acceleration ratio is gotten
>> by Intel SDK.
>>
>> Nvidia GPU hasn't reached so amazing acceleration ratio. It maybe because
>> of slow PCI-E and memory. I haven't found time to optimize on that.
>>
>> Another open source OpenCL lib pocl - Portable Computing Language (
>> https://github.com/pocl/pocl) is also tested, and it seems as fast as
>> Intel's! (See video I attached in the end.)
>>
>> 3. Some improvements on algorithm. Higher sensitivity and more robust.
>>
>> video: http://www.youtube.com/watch?v=SxZzEVEKuRs
>> (in china) http://pan.baidu.com/s/1o6qbLGY
>>
>> Have fun with LTE. Have fun with OpenCL!
>>
>> BR
>>
>> Jiao Xianjun (Ryan)
>>
>>
>>
>> On Wed, Feb 19, 2014 at 11:21 PM, Jiao Xianjun <[email protected]>
>> wrote:
>>
>> Hi there,
>>>
>>> As LTE-Cell-Scanner doesn't support TDD mode:
>>> https://github.com/Evrytania/LTE-Cell-Scanner
>>>
>>> I fork LTE-Cell-Scanner and add TDD support to it:
>>> https://github.com/JiaoXianjun/LTE-Cell-Scanner
>>> (Not reviewed by James Peroulas, so it is just experimental currently.)
>>>
>>> It works fine with rtl-sdr E4k tuner dongle below 2.2GHz, but doesn't
>>> work
>>> in 2.5~2.7GHz even with external MMDS-LNB.
>>>
>>> Because the algorithm assumes analytic relationship between carrier and
>>> sampling frequency error.
>>>
>>> I write some matlab scripts (https://github.com/JiaoXianjun/rtl-sdr-LTE)
>>> to separate carrier and sampling processing in algorithm. Aided by
>>> MMDS-LNB, the scripts can detect TDD&FDD LTE cell in 2.5~2.7GHz now!
>>>
>>> 12 LTE Cells information are decoded. 2 are FDD LTE, the rest are TD-LTE.
>>> Because here is China! (Partly because that TD-LTE is announed earier
>>> than
>>> FDD by government)
>>>
>>> Hope that in the future these features can be merged to original
>>> LTE-Cell-Scanner.
>>>
>>> Scanning results and video are attached.
>>> http://www.youtube.com/watch?v=4zRLgxzn4Pc
>>>
>>
