Bug 2095

Summary: Wrong values in default-traces.h
Product: ns-3 Reporter: Esteban <muni_es>
Component: wimaxAssignee: ns-bugs <ns-bugs>
Status: RESOLVED FIXED    
Severity: normal CC: tomh, tommaso.pecorella
Priority: P3    
Version: ns-3.22   
Hardware: PC   
OS: Linux   
Attachments: Original default-traces.h file in ns-3.22
Modified default-traces.h file which gives normal values
Image with the BER-SNR relationship used in thw WiMAX module
Proposed patch for fix default-traces.h

Description Esteban 2015-04-12 12:29:56 UTC
Created attachment 2015 [details]
Original default-traces.h file in ns-3.22

I have observed wrong OFDM traces in default-traces.h, which have been made with an external OFDM simulator(IT++).

In the modulation 4 (16QAM 3/4) the BER values are abnormally high and the table lenght is not long enough.

The problem begins in the line: 257

static const double modulation4[6][115] = {  .... }

The file is src/wimax/model/default-traces.h 

Adtionally I am using an alternative default-traces.h with OFDM traces belonging to the 64QAM 1/2 traces (my simulations do not require accurate phy model), and the results of throughput are much more logical and precise with this modulation 4.

I attach both default-traces.h
Comment 1 Esteban 2015-04-12 12:30:57 UTC
Created attachment 2016 [details]
Modified default-traces.h file which gives normal values
Comment 2 Tom Henderson 2015-04-12 12:52:31 UTC
Thank you for providing patches.  Can you document any more how these new values were obtained, and can we leave a comment about this in the header file?

Would it be correct to summarize this as:

"Fix OFDM modulation values for 16QAM 3/4 coding in the WiMax module"

or if not, can you provide another summary statement that better captures what has changed in the model?
Comment 3 Esteban 2015-04-20 11:31:47 UTC
Originally, I only copy-pasted from modulation5 to modulation4, since for me was enough. However, in order to try obtain a less clumsy method, I have simulated new traces for modualtion4 with It++.

I attach the figure ber-snr-values.png which sumarize the problem. 

Modulation4 never reaches values close enough to zero, so the BER and BLER were always to high. I have contacted with the original authors, but none of them reminds or have anything of the simulator used originally. So instead of use the complex environment used originally (solomon-reed, interleaving and TDL channel) I have simulated a simple scenario with AWGN channel. Maybe is not completely alike to the real world, but at least it improve the older default-traces-h and make the module more usable.

I attach the new default-trace.h as patch whith a brief description.
Comment 4 Esteban 2015-04-20 11:33:09 UTC
Created attachment 2023 [details]
Image with the BER-SNR relationship used in thw WiMAX module

Image with the BER-SNR relationship used in thw WiMAX module
Comment 5 Esteban 2015-04-20 11:36:14 UTC
Created attachment 2024 [details]
Proposed patch for fix default-traces.h

Proposed patch for fix default-traces.h in ns-3.22. The only changes made are the header with the changes performed and the traces for modulation4 (16 QAM 3/4)
Comment 6 Esteban 2015-04-20 11:46:42 UTC
Comment on attachment 2024 [details]
Proposed patch for fix default-traces.h

It can be applied from the ns3.22 main working directory with patch -p2
Comment 7 Tommaso Pecorella 2015-05-02 15:53:40 UTC
changeset:   11344:c0f16f16643d