select number of Data bits out of generator,know Data Rate?

how many bits are needed to support 8-qam?

Hi,
I have one doubt.
If i want to simulate OFDM system for the 54 Mbps Data rate combination.

1)Is that Data Rate depends on number of input data bits?

2)I took 2048 bits i simulated total System its giving maximum 1/10^4 only as BER.Is it correct to do?

3)We are doing simulation by taking some data and some modulation technique.
From the simulation is it possible to calculate the Data Rate?How?Plz clarify this doubt.

Thanks in Advance
Tarakarpaveen

Hi,

probably I will not be able to answer your questions completely, but I hope I will give you some hints. This is because you are describing your system in a general way, but you questions are quite particular - but please don't take it as an offence.

1. It depends how you implement your system, in particular how you allocate resources to different users. There are two ways how to do that - allocate 50 Mbps OFDM channel for a single user at a time, or allocate same channel to several users at a time (some of the sub-carriers would be allocated to one user and some other sub-carriers - to another, etc)
In case your system is the first one and your choice is pure 50 Mbps then data rate is independent on input data bits. It should be rather dependent on channel conditions.

2. In order to assess BER of the system, you have to take A LOT of samples. The more samples you take - the higher accuracy. Resolution is one bit, so if you transfer only 2048 bits, then the maximum achievable accuracy is:
1/2048 = 0.5E-3

I assume that your result is very consistent. Furthermore transfer of 2048 bits over 50Mbps channel takes tiny amount of time - the rf channel could be very different next second, but the simulations ignore this fact.

3. It depend on what you mean by modulation technique. If it is QPSK or 16QAM - then it is not possible to say what is data rate. It is possible to state that QPSK carries 2 bits per symbol and 16QAM - 4 bits per symbol. In order to know data rate, it essential to know symbol rate.
Say you have 50 000 symbols per second and your modulation is QPSK - then data rate is 100Kbps. However if FER (forward error correction) is employed in the system, then data rate is much lower, as some of the bits are used for error correction purposes.

Systems are designed to have certain Data rates and simulation only tries to imitate that system. If you are building simulation of a system, then you have to know the data rate of that particular system. It is not the other way round.

I hope I haven't confused you even further...

Re: select number of Data bits out of generator,know Data Ra

Hi ,

Very very thanks for ur valuable reply since some of my doubts that are in dilemma are clarified clearly.

1) Here if we take 2048 bits ==>Max BER is 0.5E-3 .Can we go for more number of iterations while simulation of a Transceiver to improve BER?Is it correct way
to do?

2)While transmitting through the Rayleigh channel jsut i considered direct sum of Tx signal and delayed version of transmitted signals simply(convolved the transmitted signals with the rayleigh fading coefficients).But what i felt we are getting doppler shift effect due to multipath fading.please suggest me to overcome this drawback?

3)If i use 2X2 MIMO OFDM in case of 8-QAM i am getting flat curve for all Eb/No Values .Is it correct?and for others BPSK,QPSK,8-PSK and 4-QAM am able to get the rain fall graphs(BER vs Eb/No)


Thanks in Advance
Taraka

1) Of course more samples will provide more accurate BER, but keep in mind that your channel should be changing in a (pseudo)random way. If same 2048 bits are sent over the same channel, then it is not worth to repeat, since results from each simulation would be just the same.

2) I have been involved in UMTS simulations long time ago. I haven't built the channel model, but as much as I remember implemented by IIR filter.
The idea behind is that each tap (single path) gets spectrum distribution enlarged around center frequency. For certain shape of Dopler spectrum. This can be described by transfer function in frequency domain. Now inverse Fourier transform of the transfer function will give you function of IIR filter. The filter shouldn't be hard to build - you have to have one already for pulse shaping in transmitter (like root-rise cosine impulse filter.

3) I have in sufficient knowledge of MIMO and OFDM. Just quick remarks - QPSK and 4QAM is just the same modulation type. I am just curious what is 8QAM. I imagine that constellation of 8QAM should be close to 8PSK, which should be better for PA in transmitter.

BR, Gied

Re:select numberof Data bits out of generator,know Data Rate

hi,

Thank you

I find that the constellation of 8-QAM fully different from 8-psk it is like a rectangle constelation .I hav few doubts in MIMOalso.I have simulated MIMO OFDM of 2X2 .

1)I have done the simulation in baseband .I didn't simulate it at the center frequency of 5.2GHz .Is there any mistake if we do like this ?

2)plz explain in few words what we are exactly doing in the channel estimation part.Is it same as equalization of different?

3)Previous reply you have mentioned for 50,000 symbols per second modulation is QPSK means datarate becomes 100kbps.I can understand that ,In this what i understand was that 50,000 symbols must be generated from data generator.If it is so how to say we are generating 50,000 bits in one second from the random generator using in theMatlab .Please clarify this doubt.

Thanks in advance
Taraka

hi,

considering 8QAM and 8 PSK the difference is only in amplitude of 45+n*90 constellation symbols. If the amplitudes are made the same for all points, then 8QAM becomes 8PSK. Those 45+n*90 are very challenging for Power Amplifier in TX as it requires higher linearity.

1. If your baseband has the same properties of RF channel, then everything should be ok. Maybe signal distortion of frequency up/downconversio could be taken into account.

2. channel estimation determines transfer function (say T(x)) of RF channel, i.e. how tx signal is distorted by the channel. Then it is possible to build inverse function of RF channel transfer function, say I(x). If I(x) function is applied to received signal, then distortion effect of rf channel is cancel out. This operation is called equalization. In order to perform ch. estimation, certain tx info should be know by receiver beforehand, like pilot ch or midamble. There are plenty of ebooks in internet and probably here on ch. estimation.

3. The symbol rate of the system in example is 50k per second. However each symbol carries 2 bits, as there are 4 possible different states of QPSK. So, source bit generator has speed of 100kbps. Now the bits stream is demultiplexed in two parallel bit streams of 50 kbps. Then the symbol is created by taking two bit from the parallel streams (single bit from single stream at a time).

If modulation type of the system is 8QAM, then each symbol carries 3 bits at a time. In order to transfer 100kbps stream, only 33.3kilosymbol per s is neccessary for 8QAM.

Hope this helps, Gied

Re: select number of Data bits out of generator,know Data Ra

Thanks alot for ur valuable replies

We are using IFFT/FFT pair in OFDM.IFFT is for baseband modulation upto my knowledge.In my project review external observer asked me how IFFT causes baseband modulation and he explained like this.

1) If we apply FFT to a sine wave we are getting frequency components.How it gives modulation?

2)How you implement/interpret IFFT function physically?

3)we are getting serial data from the source before applying the IFFT we are dividing intoOFDM symobls,after that applying IFFT.My doubt is if we apply IFFT to serial data i.e without converting into OFDM symbols is same as IFFT applied after converting into OFDM symbols?

4)while before applying to the channel the data must be serial form.Is there any restriction like that ?(since we are adding SNR in awgn(signal,snr) in Matlab).

Thanking you
Taraka

Thank you,

hi,

refering to your first post :

according to monte-carlo simulation, to estimate a given BER probability, Pb , the required number of samples ,N, must satisfy the following condition :
N >> 1/Pb .

in simulation of digital comm. systems N>10/Pb is a good choise .

regards

hi,
although I have a vague idea why IFT is used in OFDM, i don't feel competent enough to explain that. As well my knowledge matlab functions is bad.