Forums

Software Pulse Amplitude Demodulation

Started by berriferous July 17, 2005
Hi all,

I am doing a simple demodulation of a sampled RF signal.  I set my
digitizer at a fixed 1 GS/s.  I don't know what my carrier will be, but it
will be less that 200MHz.  The signal is pulse amplitude modulated, and I
want to recover the original signal.  To do this, I have applied a Hilbert
transform, and then derived the magnitude which gives me a good
approximation of the encoded signal. The problem is I get a fair amount of
noise riding on the original signal. Does anyone have ideas how to reduce
the noise from the original signal?  I have thought about filtering, but
since I can never be sure of the original bandwidth of the encoded signal,
I am worried about destroying the data.  Any ideas would be helpful... 

-gberry




		
This message was sent using the Comp.DSP web interface on
www.DSPRelated.com
"berriferous" <gberry@ieee.org> wrote in message 
news:78mdneaoKpjpbUffRVn-gw@giganews.com...
> Hi all, > > I am doing a simple demodulation of a sampled RF signal. I set my > digitizer at a fixed 1 GS/s. I don't know what my carrier will be, but it > will be less that 200MHz. The signal is pulse amplitude modulated, and I > want to recover the original signal. To do this, I have applied a Hilbert > transform, and then derived the magnitude which gives me a good > approximation of the encoded signal. The problem is I get a fair amount of > noise riding on the original signal. Does anyone have ideas how to reduce > the noise from the original signal? I have thought about filtering, but > since I can never be sure of the original bandwidth of the encoded signal, > I am worried about destroying the data. Any ideas would be helpful... > > -gberry >
Hello, Have you tried taking an HP signal generator and applying a simple 200MHz unmodulated carrier and seeing what the noise is like? One of the things you will want to know is what is the phase noise of your mixing oscillators? So if you work with a strong signal from a good source then you can see where your noise is coming from. Also do you know the noise figure for your receiver? Calculate the Johnson noise and add in your noise figure. How much above that is your signal? At room temperature your noise power for 500MHz bandwidth is around -86 dBm. Clay
>Hello, > >Have you tried taking an HP signal generator and applying a simple 200MHz
>unmodulated carrier and seeing what the noise is like? One of the things
you
>will want to know is what is the phase noise of your mixing oscillators?
So
>if you work with a strong signal from a good source then you can see
where
>your noise is coming from. Also do you know the noise figure for your >receiver? Calculate the Johnson noise and add in your noise figure. How
much
>above that is your signal? At room temperature your noise power for
500MHz
>bandwidth is around -86 dBm. > >Clay >
Hi Clay, I am doing all the demodulation in software... so I really don't have mixing oscillators per say. I worry a little about jitter in my sampler, but I think it is stable as I can make it. I could just apply a CW wave and get my noise floor as you suggested. I haven't done exactly that, but I have characterized the noise on a system. I do have some noise peaks due to the system which aren't pretty. I suspect system noise is creeping in to my signal. I was doing experiments with -10 dBm, (if I remember right) so I should be up out of the thermal noise. I also did an FFT and looked at the noise on the transformed signal. I had a little spike at my carrier frequency and some other spikes farther up in frequency. I tried doing some FIR low pass filters on it, but it didn't help much when I filtered off above 100MHz. It cleaned up some when I dropped down below 50Mhz, but that is really taking a chunk out of the encoded bandwidth. I was wondering if I could implement some sort of averaging based on multiple decoded waves. I was thinking that would take out at least the random noise. The big problem would be figuring out what the pulse frequency is. Typically it is far below the carrier. (This all assumes the pulse train is repetitive in a fixed time window) Anyway, thanks for the suggestions. I'll keep hunting for the noise...
> >
This message was sent using the Comp.DSP web interface on www.DSPRelated.com