вторник, 1 июня 2010 г.

OFDM features and nuances: Models of OFDM signals and possibilities of SA and OCG new versions.

OCG update to version
SA update to version

Both OCG and SA are modernized.

The matter is that we will refer on OCG, we will consider various models of the signals, and any interested person should have possibility to repeat or check up everything what we are doing. Old version OCG hadn’t abilities to provide it.

From the beginning of OCG creation, the main program operating mode was "Synthese" mode. "Calculate" mode had rather relative character, owing to limitations, which were put in there initially, and it's main purpose was demonstrating of that the synthesized signal corresponds to settled parameters.

In version OCG possibility to form signals of three types is added.

We have detected at least three models of OFDM signals. All three models are presented widely enough, and we see the necessity of synthesis of any of these models. As the last several months, we received a lot of questions about the problems of OFDM analysis, the part of these questions is directly linked to type of model of a signal. Let's talk about it.

Conditionally, OFDM signals can be divided into two main types, by the way of their channelization (method of channel forming). However, by the way of creation of the working channels and the pilot tones, we can observe three models. Actually there are can be more models, but we will consider the most key basis models.

Model A:
All channels are formed "as is", including the pilot tones. In this case, the pilot tone cannot be selected randomly and settled from the limited number of suitable candidates. The typical representative of this model is signal WINDRM 51-Tone COFDM Modem

Model B:
All channels are formed as potential pilot tone, any channel can be settled as the pilot tone. The typical representative of this model is CIS 12-tones PSK-2,4 120(240) bps per channel .

Model C:
Has mixed type of creation, all channels are formed "as is" by "A" model. But the pilot tones are formed specially by "B" model. In this model, any channel or channels can be settled as the pilot-tone. The typical representative of this model, signal MIL-188-110B-39 tone .

It is necessary to mark that:
model B is characteristic for signals CIS
model C for NATO signals

But, of course this is rather relative classification, although it is steady enough.

I cannot tell that this is iron rule, not so many signals has been passed through our hands to make such conclusions. Nether less, there is no special problems, to generate this or that model by the standard algorithms FFT/IFFT.

Moreover, signal MIL-STD-188-110A 16 Channels is generated by the model B. Most likely, as a hypothesis, signals of models A and C, are formed with using FFT/IFFT algorithms of dimension 2^n, and signals of model B are formed without this limitation. But I will repeat, there is no problems to generate any model in any dimension FFT/IFFT, OCG does it. :-)

Accordingly in SA, at the analysis, there is the possibility of a choice of this or that type of creation of the channel. There are two types as it has already been marked. These are three (and more) models of signals can be generated on the basis of these two types.

The classical analysis does not provide authentic definition of modulation mode/manipulation in the channel, even if you will select somehow the channel more or less purely. This phenomenon is directly linked to type of OFDM signal, and it is better to consider it on an example of synthesis of the signal with a good rating/spacing of channels, for example CIS-12 tone.

But, preliminary, it is necessary to consider some important points.

Parameters CIS-12 tone are well-known, these are LU = 36, LG = 24. These values characterize OFDM signals, and these values are the ultimate goal of analysis.

One more great value is even more universal k = LG/LU, this coefficient is the universal constant for OFDM with CP, and it allows to receive all necessary, for implementation/demodulation of a concrete signal and-or concrete conditions.

Certainly in the real signals, there is at least one additional moment: this is the quantity of used channels, but it is secondary parameter. The knowledge of quantity of channels, without knowledge of coefficient k, does not speak about anything, while only knowledge of k value, almost completely characterizes the signal.

I absolutely meaningly ignore such "the important" values, as sampling rate and-or clock frequency of manipulation. The matter is that these values, in the conditions of a priori uncertainty of the analyzed signal, especially on records, have no special sense. Because they can be easily distorted, and do not represent the real facts at all, while k is a fundamental constant for OFDM, and does not depend on external factors.

Let's come back to practice. Basically, everything that is necessary is already known. It is possible to start study distinctions among the models.

Let's synthesize OFDM signal, model A, with parameters LU = 36, LG = 24. But for descriptive reasons all channels we will settle as the pilot-tones. There is no sense to take many channels, 6-7 will be enough.

It is well visible that, despite the fact of an absence of manipulation in channels, the channels on the spectrum look as if manipulation is present. The exception makes every third channel. This is absolutely normal phenomenon, characteristic for OFDM signals, which channels are formed "as is". The phenomenon is linked to including CP, which is equal LG, and all this has a close connection with coefficient k. As it is easy to calculate, the coefficient k is equal in this model 24/36 = 2/3 = 0.66(6), and k*3 there is precisely/exact integer value, that gives the period(3) of pure/clear pilot-tones on the spectrum.

This phenomenon, does not leave chances for correct definition of manipulation/modulation in the channel by the classical method, even in case of complete selection of the channel.

Now we will check this up.

Let's switch off in current model all channels except the third, and after synthesis we will select it. As it will be one, it can be easily selected.

Now it becomes absolutely clear - we can detect, everything, besides, that is actually presents. As we remember, manipulation in the channel is not present, the channels are settled as the pilot-tones.
However, introduction/including of CP provides exact phasing in the symbol, it does not guarantee saving of this phasing among characters/symbols at all. And the malfunction of intercharacter/symbol phasing leads to such results for the classical analysis.

We synthesize now the initial signal both as model A and model C, for matching. We will set the third channel as the pilot-tone, the others we will leave as working ones.

In model C, at the expense of saving of intercharacter/symbol phasing on channels, which are settled as the pilot/tones, we obtain possibility to select as the pilot any channel. But in this model, on the working channels intercharacter/symbol phasing is broken with CP introduction/including, same as in model A, that does not give guarantees in definition of manipulation mode by the classical methods.

We will not synthesize model B, interested persons are able to do it, but I will mark that in this model, intercharacter/symbol phasing is saved for all channels at once by default in the course of signal’s creation. And for this type of signals, the classical analysis allows to define precisely and successfully manipulation mode in the channel, in case of its qualitative selection is finite.

And here is the most interesting.

Hardly possible that the model B was oriented to make the analysis of this type of signals easier, and the models A and C to make it more difficult, of course no. But there are two different approaches in solution of the same tasks.

The problem is that even having on hands special OFDM analyzer, without having possibility to consider these different approaches in creation of the signals, or simply without knowing about types of OFDM signals, chances of successful analysis are not great. In this case, problems with the following tasks are guaranteed:

- with definition of manipulation mode
- with pilot/tones
- with detection of nonexistent phase shifts between channels etc., etc.

In this situation, everything may be good on one signals, while on others it will become worse than ever.

In the last SA version, OFDM mode supports two modes of channelization.

Mode A: demodulation of the selected channel is realized "as is" (by model A)
Mode B: demodulation of the selected channel is realized in a special mode (by model B).

As it is possible to select any mode on any channel, possibility of the analysis of all considered above models of signals is provided.

Ignoring or absence of knowledge about existing features of creation of signals OFDM, cause gross errors in the analysis. For example like this. :-)

Good Luck~