The inputs to the FFT for each subcarrier are the sine and cosine of the phase of the signal. This is usually done in a lookup table. However, you're right that a FFT is inefficient for 9 subcarriers. Each could be generated by an accumulator and a sine/cosine lookup table. The input to the accumulator is the frequency of the subcarrier and the phase can be altered by adding to the output of the accumulator (modulo table size). Receiving would be done by multiplying the incoming signal by the sine and cosine of each subcarrier center frequency. The in-phase and quadrature outputs are then accumulated over the sample period.
The phase change has to be done at the beginning of each sample period. This actually generates multiple sidebands as the modulating signal is a square wave. However, all sidebands except the first sideband cancel out if the subcarriers are evenly spaced. 73, John KD6OZH ----- Original Message ----- From: Rud Merriam To: digitalradio@yahoogroups.com Sent: Monday, October 22, 2007 15:40 UTC Subject: RE: [digitalradio] Re: OFDM Proposal Thanks, John, for this reference and the other responses. I do keep an eye on Phil's work and comments. Time for a direct question: How do you modulate an OFDM sub channel using an IFFT? Or is that not the way to do it in practice? I have experimented with Excel. Looking at the FFT of a cosine wave I get a nice solid single frequency bin. Doing a modulation of it with a phase change in the middle I get a number of bins which looks impractical to implement. A value in a single bin then running an IFFT generates a nice cosine. I can kind of make a PSK modulate signal by copying 3 values from the FFT experiment above but other attempts generate a mess. <g> Again, that does not seem like a feasible approach. Musing about it while going to sleep I got thinking about another approach based on the observation that the sign of a value in the complex number controls the phase of the start of the curve. The process is: Generate a "symbol" If a phase change is needed change the sign in the bin Generate the next "symbol" The trick is that the "symbols" are offset by 1/2 the timing period, i.e. the start of the 2nd symbol is actually the midpoint of the 2nd symbol. This works because in OFDM the symbol period contains complete cycles of the waves. Possibly using the IFFT for an HF OFDM signal is inefficient, especially when working with a 500 Hz bandwidth signal. The 62.5 baud suggestion you made only using 9 tones so generating them directly would not be a CPU intensive process, especially using table lookups. Rud Merriam K5RUD ARES AEC Montgomery County, TX http://TheHamNetwork.net -----Original Message----- From: digitalradio@yahoogroups.com [mailto:[EMAIL PROTECTED] On Behalf Of John B. Stephensen Sent: Monday, October 22, 2007 7:40 AM To: digitalradio@yahoogroups.com Subject: Re: [digitalradio] Re: OFDM Proposal Look at KA9Q's web site, especially http://www.ka9q.net/code/fec/, for FEC software. 73, John KD6OZH