I'm just looking in a small algorithm development where I've to find out CW Signals in my band of interest 140 - 180MHz (40MHz BW). I'm using 1024 point FFT (8us Frame size) with 125MHz sampling rate to find out a CW signal. If a signal is there for 122 contiguous FFT frames (1ms time period - 122*8us), then I'm deciding that as a CW signal. The logic is working fine if a continuous signal is there. But if I'm giving a pulse signal which is 2us pulse width with 4us PRI, then in each FFT frame will have same signal component and I'm declaring that also a CW signal. But actually that is pulsed signal. For pulsed signal which is more than 8us PRI, still my logic will work fine. But I don't want to declare a pulsed signal with less than 8us PRI also as a CW signal. And I can't compromise with my FFT frame size because of my sensitivity requirement. Can any body suggest an algorithm to eliminate this pulsed signal with my CW finding logic?
Firstly, I assume CW is something with "continuous" in it? Abbreviations can obscure a problem, so it is usually helpful to give the complete notion once.
If you are trying to find a continuous carrier, then do I understand you correctly in that you are detecting a pulsed signal - pulsed at exactly your fft frame rate or multiples thereof - as a continuous wave?
If this is only about detection, then you could accept a delay between two FFT frames. For every two consequtive frames you choose this delay randomly within certain bounds. So your detection frames jitter (a lot). This would eliminate your problem since the carrier would occur only in some frames.
Another method would be to measure the bandwidth of the carrier - a pulsed carrier being switched on and off within one frame would have a larger bandwidth than a continuous carrier. But then this is only detectable for unmodulated narrow-band carriers..