Clear my doubts about OFDM

[Antonio_Magma]

OFDM, my misconception?

Went to see my project supervisor today, he cast some doubt on my understanding of OFDM...

Lets say i have a string of bits [1 0 1 0 1 1 0 0]. To transmit it with OFDM with 4 subcarriers, i have to convert it from serial to parallel,
1 1 1 0
0 0 1 0
with rows representing OFDM symbols, columns representing subcarriers. Each subcarrier will carry part of the OFDM symbol.

My supervisor says that it is wrong. Instead with data [1 0 1 0 1 1 0 0], the result should be
1111
0000
1111
0000
1111
1111
0000
0000
with the data bits converted into parallel and copied 4 times (based on number of subcarriers) meaning that there will be 8 OFDM symbols with each subcarrier carrying the same copy of part of the symbol. If that is the case, then if i have lets say a string of data with 1000 bits, wouldn't i have 1000 OFDM symbols?

I am quite sure i am correct with my theory, my supervisor is quite sure of his too.
So which is which?

 

[flatulent]

Re: OFDM, my misconception?

There is no redundancy. Each subcarrier has N modulation states (amplitude and phase) For 16 QAM this represents 4 bits of information per subcarrier. There are M subcarriers that carry information. There are several that act as phase and amplitude references. (they have the same amplitude and phase independent of the information sent.) You can set up your own proprietary relationship between the incoming data bits and the subcarrier modulation states. You can also comply with the standard 802.11 or 802.16 arrangement.

 

[Antonio_Magma]

Re: OFDM, my misconception?

flatulent, so meaning that there isn't any problem with my theory right? Because most if not all books just place a serial to parallel block before the IFFT block, and i could not find any examples.

And i do not plan to modulate my baseband into QAM,BPSK etc before converting into OFDM. I just creating a prototype. It should be ok right?

And what is the proper convention for serial to parallel conversion? Horizontal conversion or vertical conversion? Meaning lets say a string of [ 1 0 1 0 1 1 0 0], would the serial to parallel conversion of it be,
1) 1 0 1 0
1 1 0 0

or
2) 1 1 1 0
0 0 1 0

1 and 2, which is correct?

 

[bpu]

OFDM, my misconception?

the overal convention of the serial to parallel conversion isn't exist. you can be translate input data with the 1) or 2).

 

[Antonio_Magma]

Re: OFDM, my misconception?

If so,
Assuming i'm transmitting 2 characters with binary A=1010 and B=1100...wouldnt that mean if i were to use

1)
If frequency selective fading distorts subcarrier 1 (column 1), i would have both A and B distorted since 1 bit of it is distorted.

2)
If frequency selective fading distorts subcarrier 1 (column 1), i would have A distorted but yet i could still recover B..

so which type does exactly OFDM use? and when ppl use serial/parallel blocks in books, papers or anywhere, which convention do i assume? since both 1) and 2) produces very difference results...

 

[flatulent]

Re: OFDM, my misconception?

There are many ways to do the FFT. This is why there are different ways to feed the raw data into the FFT to get the OFDM outputs.

 

[Antonio_Magma]

Re: OFDM, my misconception?

Is there a standard convention of FFT? Because in my case, i'll generating C codes from Matlab/Simulink to my TI TMS320C6713 DSP..would different type of fft produce different results or is there a specific type of fft i should use for my DSP?

 

[Gandharva]

Re: OFDM, my misconception?

hi antonio_magma,

your understanding of OFDM is quite right.

as for seriall to parallel convention is concerned 1) is proper.

as for your next question
"Assuming i'm transmitting 2 characters with binary A=1010 and B=1100"
this is taken care by error correction codes like Convolutional codes or turbo codes.

 

[Antonio_Magma]

Re: OFDM, my misconception?

as for seriall to parallel convention is concerned 1) is proper.

But wouldn't 2) b better? since i would b able to recover B correctly?

as for your next question
"Assuming i'm transmitting 2 characters with binary A=1010 and B=1100"
this is taken care by error correction codes like Convolutional codes or turbo codes.

For my project, i only have to design and develop a basic OFDM transciever with DSP which will b able to communicate through power lines. Thus, according to my supervisor i need not include channel coding..

Question is that assuming i am not expecting impressive data rates (maybe <100kbps), can i for d sake of the project ignore channel coding and yet b able to transmit?

-------------------------------------------------------------------------------------------
If lets say i DO have to do coding, since i plan to build my design using Simulink and convert its C code into my DSP..

would just adding a FEC block or some other predefined coding block in my Simulink model and downloading it to d DSP work? or do i have to go through d whole process programming manually with C things like that?

 

[joinfaisal]

OFDM, my misconception?

Hi Antonio I think u r right... ur view of ofdm is right....actualy first bits are converted into symbols and then there symbols are converted into serial to parallel form.. by symbol here I dont mean actual ofdm symbol...... I mean simple mapping from bits to symbol and then you can convert serial to parallel and then can assign sub carriers so in short ur view is right I far as I know....

 

[Antonio_Magma]

Re: OFDM, my misconception?

thx faisal..

1 more thing, when it comes to mapping the symbols..is it equivalent to modulating the baseband with other methods such as BPSK? before IFFT?

 

[azaz104]

Re: OFDM, my misconception?

Hi there,
ur prof is NOT right, if what he says is true then it means that the data rate wil not be the same, conversting ur bits or symbols from one domain to the other doesn;t mean an increase in information, it's just a mathematical manipulation. but try asking him nicely if he meant, converting from binary to another bandwidth efficient type of modulation,

 

[Antonio_Magma]

Re: OFDM, my misconception?

His reasoning is that with redundancy, if 1 subcarrier gets cancelled or distorted by frequency selective fading, the others carrying the same symbol will survive and carry their load to the receiver.

I wasn't able to explain clearly to him how OFDM is robust against frequency-selective fading.

 

[sph3r3]

Re: OFDM, my misconception?

His reasoning is that with redundancy, if 1 subcarrier gets cancelled or distorted by frequency selective fading, the others carrying the same symbol will survive and carry their load to the receiver.

It is completely up to your implementation. If you choose to transmit n symbols on n channels (i.e. no redundancy) then you will still lose data if one of the channels fades... or you will correct the data with some error correction scheme. However, you will lose less data than if you had one wideband channel as the symbol rate in the OFDM system will be lower (a fade lasting 1ms may only destroy 1 symbol instead of 10). You can also choose to transmit less than n symbols on n channels, but this isn't normally done (IIRC?) as errors are normally corrected by LDPC, turbocodes, etc. Errors aren't all that frequent.

I wasn't able to explain clearly to him how OFDM is robust against frequency-selective fading.

Simply answer: It isn't, in general, although most textbooks say it is.
More thorough answer: In the frequency domain an OFDM/QAM signal is a whole bunch of overlapping sinc functions. If a narrowband interferer/fade miraculously happens to coincide with a peak of one of the sinc functions it will knock out that channel. Since the sinc functions are orthogonal, this won't effect the other channels.
However, what usually happens is that the interferer will not coincide with a peak => it's energy leaks into all of the channels (at least to an extent). There are ways around this (using shaped pulses in time domain instead of rectangular windows => results in a frequency localised spectrum), but to my knowledge they are only used for research currently.

 

[Antonio_Magma]

Re: OFDM, my misconception?

Errors aren't all that frequent.

But with power line channels, they are..

More thorough answer: In the frequency domain an OFDM/QAM signal is a whole bunch of overlapping sinc functions.

Referring to 1 of my questions before, when i see mapping in OFDM system block diagrams is it more or less equivalent to QAM modulation, BPSK etc of the baseband before performing IFFT?

Block diagrams from some books perform such modulation after the data has been separated into parallel streams, some convert into parallel streams after the whole baseband is modulated...

Is there a difference? Or is it again, up to my own implementation? What are the differences?

 

[sph3r3]

Re: OFDM, my misconception?

Errors aren't all that frequent.

But with power line channels, they are..

Ah. I don't have any experience with that sort of channel. I guess spikes caused by changing loads may cause problems? Error correction for that type of channel may be interesting to research.

More thorough answer: In the frequency domain an OFDM/QAM signal is a whole bunch of overlapping sinc functions.

Referring to 1 of my questions before, when i see mapping in OFDM system block diagrams is it more or less equivalent to QAM modulation, BPSK etc of the baseband before performing IFFT?

Yes. OFDM can be layered on top of QAM, BPSK, etc. The "whole bunch of overlapping sinc functions" arises because (for example) in OFDM/QAM we use the IFFT to modulate a vector of QAM symbols onto different carrier frequencies. Each QAM symbol is a sinc function in the frequency domain (due to it's rectangular time envelope) with different amplitude and phase (due to different symbols carrying different data).

Block diagrams from some books perform such modulation after the data has been separated into parallel streams, some convert into parallel streams after the whole baseband is modulated...

Is there a difference? Or is it again, up to my own implementation? What are the differences?

Not sure what you mean? Is it that some books show the following:

• 
  serial data stream -> [S/P converter] => // data stream, then
  // data stream => [QAM encoder] => vector of QAM symbols, then
  vector of QAM symbols (baseband) => [IFFT] => OFDM/QAM (RF)

... and other books show:

• 
  serial data stream -> [QAM encoder] -> stream of QAM symbols, then
  stream of QAM symbols -> [S/P converter] => vector of QAM symbols, then,
  vector of QAM symbols (baseband) => [IFFT] => OFDM/QAM (RF)

These are functionally equivalent. Which one you choose may depend on implementation (hint: the first option has more parallelism if you can exploit it) and other functions in your radio chain.

 

[Antonio_Magma]

Re: OFDM, my misconception?

Ah. I don't have any experience with that sort of channel. I guess spikes caused by changing loads may cause problems? Error correction for that type of channel may be interesting to research.

Yeah, impulse noise, echo (impedance mismatches) and other unpredictable, time-varying factors makes it quite a crappy channel...

Not sure what you mean? Is it that some books show the following:

• 
  serial data stream -> [S/P converter] => // data stream, then
  // data stream => [QAM encoder] => vector of QAM symbols, then
  vector of QAM symbols (baseband) => [IFFT] => OFDM/QAM (RF)

... and other books show:

• 
  serial data stream -> [QAM encoder] -> stream of QAM symbols, then
  stream of QAM symbols -> [S/P converter] => vector of QAM symbols, then,
  vector of QAM symbols (baseband) => [IFFT] => OFDM/QAM (RF)

These are functionally equivalent. Which one you choose may depend on implementation (hint: the first option has more parallelism if you can exploit it) and other functions in your radio chain.

Yep, thats wat i meant. Although i don't know what parallelism (and how can i exploit it) is all about :?

 

[azaz104]

Re: OFDM, my misconception?

1-actually it doesn't make a difference if the modulation is done before or after.
2- the thing that makes ofdm really good is the insertion of a guard band, and to make it a cyclic prefix, in this way if some multipath happens it will not effect the signal on the subcarrier, or will the energy in the subcarrier leak into other subchannels.
3- the risk in ofdm is of timing and frequency mismatch at the receiver, and THAT could destroy the orthogonality between the subcarriers.
as for impulse noise :
reffering to a good and easy to read book try reading chapters 3,4 and 5.
the diagrams there are pretty helpful

MODERATOR ACTION: Attachement deleted. Already uploaded in ebok section: .
Warning

 

[joinfaisal]

Re: OFDM, my misconception?

thx faisal..

1 more thing, when it comes to mapping the symbols..is it equivalent to modulating the baseband with other methods such as BPSK? before IFFT?

yes... mapping to symbols is same as mapping with any digital modulation scheme before IFFT.