Talk:Orthogonal frequency-division multiplexing

From Wikipedia, the free encyclopedia

You have new messages (last change).
 Orthogonal frequency-division multiplexing was a good article candidate, but did not meet the good article criteria at the time. Once the objections listed below are addressed, the article can be renominated. You may also seek a review of the decision if you feel there was a mistake.

Date of review: 2 December 2006
Orthogonal frequency-division multiplexing was a good article candidate, but did not meet the good article criteria at the time. Once the objections listed below are addressed, the article can be renominated. You may also seek a review of the decision if you feel there was a mistake.

Date of review: 14 April 2006
This article is within the scope of WikiProject Telecommunications, an attempt to build a comprehensive and detailed guide to telecommunications on Wikipedia. If you would like to participate, you can edit the article attached to this page, or visit the project page, where you can join the project as a "full time member" and/or contribute to the discussion.


Contents

[edit] OFDMA

OFDMA links here. Should they be merged?--Gbleem 21:36, 14 December 2005 (UTC)


[edit] Doppler: "(...) sender or the receiver is moving at a high speed (...)"

I removed this sentence from the text. I am not 100% sure, but is seems very unlikely to me that normal speeds (car, train, plane, etc) would affect this, especially when we talk about frequencies around 5 GHz. I wrote a small thesis about OFDM in uni and I cannot recall any major issued with moving receivers/senders. Please correct me if anyone has more accurate information. --83.109.152.151 22:12, 20 December 2005 (UTC)

Well, at least according to recent tests by Digita in Finland, mobile use at speeds of "over 200 km/h" is no problem for Flash-OFDM operating at 450 MHz. --80.222.254.10 16:01, 22 December 2005 (UTC)
Okay, I'm removing that statement for now. R6144 01:52, 23 December 2005 (UTC)
As the Doppler frequency offset is proporcional to the carrier frequency, the offset would be different for every subcarrier thus causing loss of orthogonality Danielcohn 03:20, 1 May 2006 (UTC)
I put the sentence back since it is a problem in for example DVB-T, but I added that "Several techniques for ICI suppression are suggested, but they may increase the receiver complexity substantially." I have seen several papers on very complex equalizers for this purpose. Are there a simpler method? Can the sample rate be adopted, based on the pilot carriers?Someone said that modern DVB-T receivers are less sensitive to doppler. Howcome? Mange01 10:17, 13 October 2006 (UTC)
The problem comes from the combination of Multipath and Doppler. Consider the simple case of travelling into the direction of a steady transmitter. This slightly increases the frequency observed by the receiver. If at the same time there is a reflection of the tranmitted signal coming from the back of the driver, than the frequency shifts to a slightly lower frequency. So one tone results into two simultanious tones received. This corrupts the orthogonality and so the carrier breaks through to all the other carriers and thus introduces distortions. This is THE failure mechanism. The closer the intercarrier spacing the higher the distortions. —The preceding unsigned comment was added by 161.85.127.152 (talk) 14:47, 6 December 2006 (UTC).
Thnx for a helpful comment. User:Oli Filth has now incorporated it into the text. HOwever, in modern OFDM based systems such as DVB-H, the dopper shift does not seem to be a problem. It would be interesting to know which ICI cancellation algorithm that is used in practice. Mange01 23:00, 6 December 2006 (UTC)
The robustness against multipath can be improved by extending the guard interval duration. Usually the length of OFDM symbol will be scaled too to keep the guard overhead relatively the same. The carrier-spacing is then reciprocal with the symbol duration (1/T) and thus becomes smaller by better multi-path robustness. However the Doppler robustness will decrease at the same time, so there is a trade-off between multi-path and doppler robustness. In DVB-T they just forgot to define one intermediate intercarrier spacing. This gap appeared to be a good compromise between doppler and multi-path robustness. So they included that carrier spacing in the DVB-H standard. (PS: I'm the same writer as talk:161.85.127.152)

[edit] Reorganization - and GA nomination

This is a very understandable article and I'd love to see it promoted to good article. I did not pass it this time because:

  • The lead is too short.
  • The article has structural/formatting problems.

Some suggestions on how to improve the article's structure/formatting:

  • The "Digital radio and television" heading has no content beneath it and should be removed.
  • The table dominates the "Wireless LAN" section.
  • The "OFDM feature abstract" is a list not a section consider making it an inset for the "Characteristics" section.
  • The "Ideal encoder" and "Mathematical Description" (should be "Mathematical description") sections should probably come before "Usage".
  • The "History" section seems tacked on, maybe it could also be made an inset.

Examples of how to do insets are available in the Columbine high school massacre article. Please renominate once the above problems are fixed.

Cedars 00:09, 14 April 2006 (UTC)

I have restructured the text by removing overlapping text; moving the list of key features (which I divided into advantages and disadvantages) to the top; moving the list of applications to the top and adding a numerical example in blockquote; and dividing the section "Characteristics and principles of operation" into sub-sections.
I have clearified why OFDM is considered a modulation scheme as opposed to a MA scheme, and added references to OFDMA and MC-CDMA. I have also corrected som incorrect statements, for example that OFDM would be sensitive to time synchronization errors.
(Someone else has addressed most of Cedars suggestions, except using insets.) Mange01 12:29, 13 October 2006 (UTC)

[edit] Multipath resistance only when coded?

Why does the article claim that multipath resistance exists only when combining OFDM with coding schemes? Multipath resistance is added by the fact that OFDM allows using longer symbols and therefore decreasing inefficiency caused by GI.Danielcohn 01:35, 22 June 2006 (UTC)

OFDM uses a cyclic prefix guard interval to convert a frequency-selective channel to a set of frequency-nonselective fading channels. As a consequence, intersymbol interference is avoided. The thing is, however, OFDM does not have frequency diversity. With coding, OFDM achieves diversity and performs well in multipath. ---sct

[edit] "Ideal encoder" section

I believe that the Ideal encoder section is in severe need of revision. I'll make these changes at some point in the near future; just wondering if anyone had any thoughts before I do.

[edit] Scrambling

Firstly, I don't think that it's necessary to include scrambling (shown as multiplication by ( − 1)k in the diagram) in a hypothetical "ideal" encoder. Secondly, it is shown as occurring in the time-domain, which is completely incorrect (see section 17.3.2.1 in the 802.11a spec, for instance). Thirdly, the reason given, "in order to have a null mean value", is also incorrect.

[edit] Inter-symbol interference

In the second paragraph, orthogonality of the sub-carriers results in zero inter-carrier interference, not zero inter-symbol interference, and only in the case when a cyclic prefix is used, which is not mentioned or illustrated.

[edit] Diagram

The blocks marked "ROM" are clearly meant to represent constellation mapping, but what does "ROM" stand for?

In my opinion, it would be better to replace the "impulse generator" and Ht(f) blocks with "DAC".

It's also debatable whether illustrating frequency-domain zero-padding is necessary for an "ideal" encoder.

Oli Filth 17:47, 20 August 2006 (UTC)

All fixed. Oli Filth 13:50, 2 September 2006 (UTC)

[edit] Flash-OFDM 450MHZ Data Network

In 9th October 2006, Finland has licenced the 450MHz band to an operator for building a nation-wide Flash-ODSM data network. Press release in Finnish: [1]. Could this be added to the main article?

[edit] Suggestion: OFDM standard comparison table

I suggest a table that summarizes the most crucial features of common OFDM systems. I have made a similar table for two broadcasting systems in the publication http://ieeexplore.ieee.org/iel5/49/20698/00957306.pdf?arnumber=957306.

Examples of features are:

  • Standard name.
  • Ratified year.
  • Frequency band [GHz].
  • Channel bandwidth [MHz].
  • Number of subcarriers.
  • Subcarrier spacing [kHz].
  • Net bit rate [Mbps].
  • Link spectrum efficiency [bit/s/Hz].
  • Symbol length [s].
  • Guard interval [s].
  • Subcarriermodulation scheme.
  • Inner FEC.
  • Outer FEC (if any).
  • Sub-carrier adaptive transmission (if any). Yes/no.
  • Multiple access scheme (if any). Example: OFDMA uplink.
  • Maximum travelling speed.
  • Time interleaving depth [ms] (if larger than the symbol).
  • Requied carrier-to-interference ratio (for AWGN without fading). Example: 5dB for BER 10^-5.

Mange01 13:11, 13 October 2006 (UTC)

[edit] Failed GA

This is a very promising article, but it is let down by an extreme scarcity of citations, and by a number of lists that would read better if converted to prose. Once these improvements have been made, please feel free to renominate the article. MLilburne 20:20, 2 December 2006 (UTC)

[edit] Merge with DAB really a good idea?

It has been suggested for a while that DAB COFDM section should be merged into the OFDM article, but none of us commented on the suggestion. Now it is accomplished, but I'm not sure that it was such a good idea.

Anyway, some of the merged text is overlapping with the old OFDM text, or is generic, not specific for DAB, and should therefor be removed or moved up in the article.

Secondly, Wikipedia now warns that the article has become longer than 30kB. Is that a recommended maximum length?

Should every application of OFDM be described as detailed as DAB in this article? In case the article should be extended with something, I would prefer more illustrations, and a comparison table summarizing the features of several systems.

Whats your vote? Should the merge be reverted?

Mange01 23:37, 2 December 2006 (UTC)

I completely agree; the verbatim insertion of the text from the DAB article completely disrupted this article, introduced a lot of repetition and redundancy, and also brought over a lot of the errors that were present in the DAB article. I have removed this addition, and have pasted the inserted text here; I'm not going to attempt to re-edit the DAB article; it's a complete mess; I'll leave that for someone else to sort out! Oli Filth


[edit] Removed text

[edit] Modulation

The modulation used in DAB is Coded Orthogonal Frequency Division Multiplexing (COFDM). According to this acronym the three properties of COFDM are: 'C' for coding; 'O' for orthogonal modulation and 'FDM' for frequency division multiplexing. These are described here.

[edit] Coding

Coding refers to convolutional coding and means that the original data carried over the multiplex is deliberately manipulated by splitting it into small blocks and adding some intelligently designed redundant information to each, thus generating a data 'overhead'. The overhead bits added to each block are determined according to rules applied to the true data content of the block. After demodulation at the receiver the digital signal processor examines both the actually received data and overhead bits and regenerates what it believes to be the original data based on a set of statistical rules known as an algorithm. The regenerated data may include a number of data bit corrections. The algorithm used in DAB is known as a Viterbi algorithm, and is an example of a maximum likelihood algorithm. This works by maintaining a history of demodulated bit sequences, building up a view of their probabilities and then using these to finally select either a 0 or 1 for the bit under consideration. This type of coding is also known as an example of forward error correction (FEC).

To some extent the types of errors most likely to be present with DAB can be mathematically predicted and therefore corrected for. The addition of FEC requires extra information to be transmitted at the same time as the original traffic data and therefore requires an increased channel capacity, needing extra bandwidth, compared to if it had been uncoded. DAB carries different 'strengths' of FEC, a stronger one being used for the control of critical features in the receiver.

[edit] Orthogonal

Orthogonal is the mathematical term applied to two RF sinusoidal signals when their phase relationship is precisely 90 degrees. Alternatively they may be said to be in 'quadrature'. In DAB the sub-carrier frequency spacing is chosen to be the reciprocal of the active symbol period. Under this condition the DAB modulation results in successive sub-carriers having a quadrature relationship with each other. The frequency spectra components of one modulated sub-carrier will therefore integrate to zero at the corresponding components from both of the adjacent sub-carriers. This has two advantages: (a) the modulated sub-carrier spectra will efficiently occupy the allocated bandwidth with a degree of controlled overlapping and (b) simple I-Q demodulation to zero intermediate frequency (zero-IF) can be used in the receiver without needing the costly hardware overhead of many bandpass filters to extract the sub-carriers.

[edit] Frequency division multiplexing

Frequency division multiplexing (FDM) is the process where two or more basic information channel bandwidths or basebands are shifted in frequency and added to others to form an aggregate wider bandwidth containing the information from all of the constituent basebands. To avoid mutual interference, their bandwidths would normally require shifting (translating) in frequency and no two translated basebands would occupy any part of the same frequency spectrum. In the context of DAB, FDM refers to the manner in which the modulated sub-carriers are assembled across the allocated frequency range.

[edit] Modulation type

DAB uses a digital modulation type known as differential quadrature phase shift keying (DQPSK), which is an incoherent modulation scheme. DQPSK differs from the more common quadrature phase shift keying (QPSK) in that the modulated carrier phase for the current symbol being detected depends on its phase relative to that phase detected for the previous one. In QPSK it is just the absolute phase of the modulated carrier that determines the associated symbol. A differential modulation scheme can be more resillient to the typical fading scenarios of DAB. The modulation scheme also incorporates a form of Gray coding in that only one bit changes on moving from one symbol state to an adjacent one. For a constant phase progression, the consecutive set of symbols are represented by the bit pairs 00, 01, 11 and 10.

[edit] Time interleaving

DAB uses data buffering which enables the data symbols to be transmitted over the RF path in a different time-order than they were generated the audio source (studio). At the receiver they are re-assembled and returned to the original time-order before conversion back to analog signals to feed the receiver audio output. This process is called time interleaving. Typical multipath interference experienced in a moving vehicle is regular over time so an intelligent choice of time interleaving to some degree 'averages' out the resulting error bursts over time. This data buffering and other processing contributes to a delay, typically of a few seconds, between the studio source and the receiver. This is much longer than the equivalent delay for am FM broadcast channel which would typically be a fraction of a second. For most broadcasts such a delay would be unimportant but it does mean that, for example, real-time reference signals for setting clocks such as those re-broadcast by the BBC on DAB from their national FM service are actually quite inaccurate.[citation needed]

[edit] Frequency interleaving

DAB also uses frequency interleaving, a similar technique to time interleaving but applied to the sub-carriers centre frequencies in the RF spectrum instead. The data stream from the studio is deliberately not modulated serially onto sub-carriers across the frequency range, but instead in a more random way. Multipath and other forms of selective fading generally affect a relatively narrow part of the RF multiplex bandwidth at any one time so frequency interleaving would tend to average out 'bursts' of errors resulting from these.


[edit] DMT vs OFDM

On a somewhat related topic, i question whether the redirect to OFDM from DMT (http://en.wikipedia.org/w/index.php?title=Discrete_multitone_modulation&diff=next&oldid=10618564 ) is optimum. Certainly DMT and OFDM share many characteristics, but the common use of the DMT term in wireline, as well as in the standards reference in ANSI T1.413 http://en.wikipedia.org/wiki/ANSI_T1.413_Issue_2 suggest they should remain distinct topics. Particularly as there are non-trivial differences between the two techniques, that also in fact form some of the basis of T1.413 - such as the Cioffi/Amati patented bit loading / tone swapping algorithm which allows better throughput in copper specific interference such as the ISI found in bridged taps. Because in copper the noise is stationary and the channel is time invariant, DMT is much better able to adapt to the communication medium than OFDM would. In any case, this may not be the best place to discuss the validity of a redirect, but because it does involve technical details relevant to this article, I felt to introduce the idea for comments here first. Duedilly 23:47, 10 February 2007 (UTC)

To add some further detail to the consideration of differences between OFDM and DMT (and whether DMT should redirect to OFDM), here are more distinction points:
Duedilly 19:48, 11 February 2007 (UTC)
There is nothing about OFDM which precludes it from being baseband, wireline, or pre-equalised/rate-adaptive. It would be a bit like trying to state that 802.11a is distinct from OFDM because it uses PRBS-based scrambling, or that DVB is distinct from OFDM because it uses scattered pilots. IMHO, "DMT" really just describes a particular implementation, which just so happens to be baseband, wireless and pre-equalised and rate-adpative. Oli Filth 22:09, 11 February 2007 (UTC)
DMT is discussed in the section "Adaptive transmission", as well as in the ADSL section. I would like to encourage to you elaborate on the bit loading/tone swapping algorithm there. A VDSL section should also be created. A separate DMT article would not be a good idea. Mange01 07:37, 14 February 2007 (UTC)
Thanks Oli and Mange. I notice there is an entry for ITU_G.992.1 and wonder why DMT would not more likely redirect to that? (or even to an ANSI T1E1.413 entry, which predates and was the basis for the adopted ITU standard). I think your idea though to further clarify the ADSL section is good, though obviously not all ADSL is FDM - technologies like CAP and DWMT and other DFT implemented filter bank modulation schemes, (both perfect, and particularly the imperfect, due to frequency overlap with low cross-channel interference which are decidedly NOT like PO-FDM) have a rightful place of discussion near DMT - which would likely not be appropriate in this entry. I agree that these may all have evolved from a common ancestor and have filled different niches, but am not yet convinced of the value of reducing them to a single OFDM entry. I am traveling soon, but will look forward to discussing specifics and considering the best approach for further elucidation of these topics, and will also look to get some feedback from Cioffi and others involved in ANSI DSL standards to help. We might also consider adding discussion of expected "multicarrier" CDMA implementations of OFDM using spread spectrum. Duedilly 19:43, 16 February 2007 (UTC)

[edit] Sub-carrier

What is meant with a sub-carrier in OFDM and what is its difference compared to a normal carrier? --Abdull 10:40, 6 March 2007 (UTC)

[edit] More explanation of the cyclic prefix purpose

I was wondering about this, won't it be nice if we also show that the cyclic prefix serves to make the effect of the channel become circular convolution for the OFDM symbol? Then, we mention that circular convolution becomes just multiplication when the DFT is taken.

Something like The OFDM symbol [d_0, d_1, \ldots d_{N_c - 1}]^T is prefixed with the L − 1 length cyclic prefix and becomes [d_{N_c - L + 1}, d_{N_c - L + 2} \ldots d_{N_c - 2}, d_{N_c - 1}, d_0, d_1, \ldots d_{N_c - 1}]^T. Then, after convolution with the channel, which happens as

y[m] = \sum_{l = 0}^L h_l x[m - l] \quad 0 \le m \le N_c

which is circular convolution, as x[m − k] becomes x[(m - k) \mod N_c]. So, taking the DFT, we get:

Y[k] = H[k]\cdot X[k].

Of course, the noise etc. has to be accounted for. Note that another thign one could mention is that the distribution of the noise, if it is isotropic complex Gaussian, remains identical under the DFT operation.

Just my suggestions to make it more clear... Kumar Appaiah 12:32, 11 March 2007 (UTC)

I suggest that some of this stuff would be well worth putting in the Cyclic prefix article, which is essentially just a stub right now. I keep meaning to flesh that article out myself, but I never get round to it. I don't think we want any more maths in the OFDM article, as it's already overly-long. Oli Filth 13:05, 11 March 2007 (UTC)
Oli, thanks for the suggestion. I am heading to Cylic prefix to make a start! Kumar Appaiah 14:24, 11 March 2007 (UTC)
Retrieved from "http://en.wikipedia.org../../../o/r/t/Talk%7EOrthogonal_frequency-division_multiplexing_f84e.html"