We have decided to limit the bandwidth of the Spec Analyser on our 20 mtr
at Boonah (see www.seti.org.au) to 35 mhz. The DSP will be setup to provide a
of ~2 mhz channels that will be fed into some low cost PII celeron's that will
and other analysis at a resolution of at least 0.5 Hz. We chose this method of
it allowed us to use relatively cheap computers to number crunch as fast as
Although building a computer that would crunch a bandwidth of 100 MHz sounds great
in reality, the task is rather daunting. The system could suffer many of the
we get with the RF side at the moment, namely RFI. Computers manufacture heaps
of spurious signals from the oscillators, and many other processes that go on
the system. We are looking for very weak signals. Computers generate these in
quantities and in my experience, the lower the frequency that you have to work
with, the better.
Our plan is to number crunch these ~2 MHz channels at 0.5 Hz FFT Resolution over
sit for a time , say 10 minutes, then slide the receiver up or down in increments
of ~10 MHz.
This will ensure that we are always getting the max. use from the DSP. We
we could effectively monitor about 100 mhz total bandwidth over the period of 1 hr
while positioned on the target source. This will be enough to get a good picture
going on in this area of the sky and determine if we should do a more detailed
later on. The big advantage of this type of approach is that the 2 mhz channels
relatively free of spurious signals as the computers will be isolated from the
receiver and DSP.
We will feed the DSP output via co-ax to the fully screened and isolated computer
The walls of the computer room will be made of aluminium (cool room type with foam
and have a separate mains supply. I hope we've got it all covered????
Noel C. Welstead
SETIL volcor Brisbane Australia
> > Some further information. The specgrm2 FFT program has 30 instructions
> > in the basic operation, including 4 multiplies, 2 adds, and 2 subtracts
> > on the actual data, which gives order 20 Gflops, or 100 Gips; the processor
> > clock cycle count will obviously be several times this.
> So, let me turn this question around, to answer the question more directly.
> You let me know if I have this right. You are saying that approximately 300
> Gips is neccessary to FFT a 100 mhz bandwidth. Lets say a 300 Mhz pentium II
> processor has only about 10 percent of that capacity. Therefore, the upper
> limit of FFT capacity of a 300 Mhz pentium II should be about 10 mhz. This
> would leave no extra cpu cycles for other things. So, maybe 5 mhz of bandwidth
> would use about half of the CPU cycles of a pentium II. Do I have this
> correct? It seems plausible to me.
> John Marcus MD