Monday, October 31, 2005

WiMax OFDM

I started implementing Wi-Max OFDM.
So far, from 11a no much difference found.
The big difference is subchannels and CTC FEC.
And I don't think they are so difficult. I'll go forward.

Thursday, October 27, 2005

wrap-up WiMax-SC

Today I released my Wi-Max-SC cmodels in sourceforge.net.
I reduced the multipath range from 100 to 50ns and reduced RS block size to get the multipath tests to pass. As the result, 200 random generated tests are passed successfully.
But 64-QAM modes are not working very well, so I guess it may need to tune the cmodels more.
Anyway , I'll move to Wimax-OFDM tomorrow with skipping uplink PHY and WiMaxSCA.

Wednesday, October 26, 2005

multipaths & Freq detection.

When multipaths are exist, Freq detection is tend to screwed up.
So I may need to take Freq detection circuit after DFE in Framecontrol section.
Today's results are very good with some multiple paths.
But still many errors are happening with multiple paths.

Tuesday, October 25, 2005

DFE

I have found my DFE was wrong.
Because multipaths did not work at all,
I tried to look for the reason and I found the existing design of DFE
was wrong from WiFi 11b.
I fixed the problem and tests with a single path worked well.
Let's see what will be happen in the new DFE with multipaths
tomorrow.

Monday, October 24, 2005

Multipaths

The integration of QAM64 and TurboProductCode has been completed today.
Including these 2 new functions , 200 tests are passed without any problems.

But TurboProductCode is not as good as type2(RS+viterbi).
And I tried to add multipaths, but None of them are worked correctly.
I need to consider its solution.

Saturday, October 22, 2005

getting better

Today I fixed a lot of bugs.
So far, 200 random tests besides 64QAM and Turbo Product FEC
are working correctly.
In Next week, I'll try 64QAM, TurboProduct FEC & multipaths.

Friday, October 21, 2005

Bug Fixs

The problem of Clock Freq Offset got better than yesterday.
But still 64-QAM does not work well.
And I randomized many parameters of Wi-Max SC.
And I'm seeing a lot of failures. Some may be a real problem like 64-QAM,
some are because of my simple mistakes...
I'll contine to fix these bug fixs.

Thursday, October 20, 2005

clock freq offset

The clock freq offset problem is taken place.
This time, the offset is less than one ppm,
but it's still affect to payload decoding.
And I should solve the problem with QAM16/QAM64.
This may be difficult..

Wednesday, October 19, 2005

NO Noise. it worked.

Today I made a lot of progress. A frame can be encoded and decoded
without Noise.
Tomorrow I 'll put some noises and see what's happening.

Tuesday, October 18, 2005

Preamble detection

Today I modeled Preamble detection, freq offset detection
and DFE for multiple paths. They worked with a very simple test.
And I'm also writing decoding frame controler.
Tomorrow I 'll test frame control decoder and probably
payloads decoder.

Monday, October 17, 2005

SC Decoder

Today I satrted to model SC- Decoder.
I found one problem that In TDD mode only a single Frame
preamble defined in a frame which can be as long as 2ms.
I can detect the preamble and I can lock the SS's clock to BS's.
But the locked clock must have an error.
If we have 1 ppm error , the clock offset will be accumulated 2pi*20
during 2 ms.
I don't think it's easy to have such accuracy.
It seems that I have to track the clokc offset even in 64-QAM mode.

Friday, October 14, 2005

Wi-Max Encoder

I'm writing the encoder cmodel.
It's much more complicated than the one in Wi-Fi.
It support both half/full duplex, FDD & TDD,
more network types, broadcast, PnP, etc..

So an frame structure and its configulations are very messy.
Its taking some time...
The problem I have mentioned yesterday was solved,
but the burst length calculations are not easy one...

Thursday, October 13, 2005

burst length ??

Today I code Wi-MAX-SC downlink encoder.
But I cannot understand how I can get burst length.
I understood burst start point and FEC configulations,
so I may be able to encode the first completed FEC codeword.
But I don't know how I can know the length of the last
shortened FEC Codeword.
It's strange to me..

Wednesday, October 12, 2005

Back to Wi-max

Today I worked on WiFi 11b to remove some bugs.
It's completed.
I also read Wi-Max-SC PHY spec and got some idea to make
its encoder.
From tomorrow I'll start to implement the encoder.

Wednesday, October 05, 2005

FEC wrap up

Today I worked on Turbo Product Decoder.
It's improved a lot and worked on all modes.
The performance is still poor, but there are some ways to improve it.
Also I wrap up all of FEC including Code type3.

Though These FEC blocks are still some problems( see below),
I'll move forward to modulation block and go back to 11b's modulation
because sc modulation should be almost same as 11b and it's till incompleted.

The problems on FEC:
1) Reed solomon Encoder result did not match with the spec.
2) Turbo Product Decoder has still poor performance.
3) Code Type3 Decoder also has poor performance.

Tuesday, October 04, 2005

x^6 +1 ???

I misread the standard for generator polynomials since the last monday.
In my hard copy, it looks like "x^6 +1", but in the real world,
it should be "x^6 x 1" . The charactor "+" could not be read in my hard copy..

After the quick fix, both n=31 and n=63 worked!!!
But its performance is still poor.
I need more tunes.

Monday, October 03, 2005

SISO

Today I tried to use SISO method on Turbo Product Decoder.
It worked well and it improves Error collections a lot.
But still it's not enough. Tomorrow, I'll try to implement "parities"
and see how it affects the Error Correction.