Problem with OFDM system

[thavamaran]

Hi guys, I am currently working with my OFDM system. My back-to-back system works in perfect condition.

My OFDM was done with 16-QAM, 72 subcarrier carries data and 56 carrier zero (zero padding) total 128 subcarries, then IFFT and DAC, then upconverting it to 1 GHz. Perfect back-to-back system.

But when I include a laser model in my program to make optical OFDM, there is a phase shift at the constellation point after FFT at the receiver, the laser model is a nonlinear ODE model. Can anyone please advice what I should!

I am really lost, I even try to compensate the phase shift which is about 45 degree, but at every loop the phase shift varies, like 45, and then 135, it just jumps with a difference of 90 degrees.

Please advice me on this. If you dont understand my explanation, please let me know, I will explain in more detail.

 

[jmpac]

As you say, a non-linear element is modifying amplitud and phase of your transmitted symbols. So first, ensure the device is not in the saturation region, and second: I think you will need to make an algorithm in order to dynamically pre-distort base band signal before the digital-to-analog converter.

The info you need to pre-distort base band signal is the amplitud and phase at the output of your non-linear device: so RF signal must be downconverted / demodulated and analog-to-digital converted again in order to construct tables with the changing values.

Have you already built the hardware of the tranmitter and receiver? or you have only made simulations with your program.

 

[thavamaran]

Thanks very much for replying!! You are the first person replying to my thread!!

How do you define the saturation region, cause as I understand, the response of the device increases linearly and then become constant by all the way till the end, I believe this is the saturation part, and my signal is only modulated at this constant part.

Please explain what is pre-distort base band signal, I have never heard of this term before. And when you describe pre-distort after ADC, it sounds more like synchronization.

Its pure simulation. I can only go to practical after completing this. Thank you very very much!

 

[jmpac]

If your constellation is good just before the input into the LASER, and your LASER is working at the maximum constant output optical power, maybe it's the photodiode, a non-linear element, which is eventually distorting the electrical signal.
First I woud try to reduce optical power in order to see some improvement on the constellation. But I am afraid that whether it isn't work, you'll finally have to demodulate and decode the electrical signal just after the photodiode, and compare it to the original I and Q base band signals: if you see any imbalance in both amplitud and phase you have to apply dinamically the digital inverse correction (pre-distort) at the digital-to-analog converter at the transmitter.

Why don't you work with a higer order modulation, like 64-QAM, and much more subcarriers, for example 2K or 8K for a commercial standard? Too many operations for a MATLAB program? Have you already identified the FPGA or DSP?

 

[thavamaran]

Hi jmpac, I have identified the problem, looks like i need pilot tone to estimate the channel and with that i need to equalize the system. So Im looking into it now, I just need a small help from you if you dont mind:

I am using 128 subcarriers, 72 of it carriers my data, and 56 are zero padded, 27 subcarriers of zeros at the beginning which are at DC, then 36 subcarries of data, then 2 subcarries of zeros in the middle, then again with 36 subcarriers of data, at the end another 27 subcarries with zeros.

So 27 beginning, 2 in the middle and 27 at the end (Nyquist frequency). Correct me if I padded it wrongly.

I went through a few books and papers, but I still dont get the idea of Pilot symbol.

Can you rigorously explain to me why pilot tone, how to include them in subcarrier and a bit of explanation on that, please, thank you very much.

 

[jmpac]

Hello, this is my reply:

The edge carriers at the upper and lower channel edge are set to zero, they are inactive and carry no modulation at all, their amplitudes are zero. The continual pilots are located on the real axis, the I (in-phase) axis, either at 0 degrees or at 180 degrees and have a defined amplitude. The continual pilots are boosted by 3 dB compared with the average signal power and are used in the receiver as phase reference and for automatic frequency control (AFC), for locking the receive frequency to the transmit frequency.

There are two basic reasons for the existence of unused zero-information carriers:
• Preventing adjacent channel crosstalk by facilitating the filtering of the shoulders of the COFDM spectrum.
• Adapting the bit capacity per symbol to the input data structure.

Because IFFT is used, however, it is necessary to select a power of two as the number of carriers which, after subtracting all data and pilot carriers, still leaves carriers, namely the zero-information carriers.

I dont know if your number of zero-carriers is correct. But take account of this: The various types of carriers used in DVB-T are briefly summarized as follows. Of the 2048 carriers in the 2K mode, only 1705 carriers are used and all others are set to zero. Within these 1705 carriers there are 1512 payload carriers which can be QPSK, 16QAM or 64QAM modulated, 142 scattered pilots, 45 continual pilots and 17 TPS carriers.

Perhaps scattered pilots and TPS carriers aren't necessary for your project. Only fixed continual pilots.

I also attach a figure of the physical layer frame for the DVB-S2, in which you'll see that the pilot block is periodically put into the FEC (Forward Error Correction) frame.



Regards.

 

[thavamaran]

Wow!! that really helps!! thanks very much!!! So I should add this pilot before IFFT right? But how can I just add pilot carries to I phase? Cause after serial to parallel conversion, say I have 128 sub-carries with 6 symbols in each, all of them will be in complex form according to 16-qam amplitude level. So I should just add continuous pilot with BPSK modulated data after every 16 subcarries is it?

Im just confused a bit, just to clarify, so if its too much of question. Thanks a lot, your explanation really really helps.

 

[jmpac]

After serial to parallel conversion, the QAM Mapper outputs I/Q QAM symbols (amplitud-phase of the vector; or real-imaginary parts) to the IFFT block, but before doing IFFT the Framer block must insert pilots (BPSK) at various carriers within each symbol. IFFT block will then add the guard interval and even an optional non-linear predistortion can also be performed.

Regards.

 

[thavamaran]

Right, ok jmpac, thanks a lot!! I will try this and update you. Thank you very much!