Reply by Randy Yates August 25, 20052005-08-25
"Paul Solomon" <psolomon@tpg.com.au> writes:
> [...] > One thing that I found was that if I decimate (after the quad mixing and low > pas filtering) any more than 8, then I can no longer reliable demodulate the > signal using a cordic / differentiator approach. I was hoping to be able to > decimate by ~64 to get down to 1.25MSPS or less as the FM signal is > bandlimited to ~ 300kHz.
Paul, Is it possible that your decimation filters are too "loose" and thus allow too much aliasing at high decimation ratios? -- % Randy Yates % "How's life on earth? %% Fuquay-Varina, NC % ... What is it worth?" %%% 919-577-9882 % 'Mission (A World Record)', %%%% <yates@ieee.org> % *A New World Record*, ELO http://home.earthlink.net/~yatescr
Reply by Eric Jacobsen August 25, 20052005-08-25
On Thu, 25 Aug 2005 22:25:10 +1000, "Paul Solomon"
<psolomon@tpg.com.au> wrote:

>Hi, > >Sorry for all the questions all the time, however I am still working on this >FM demod design and My shiny new stratix II dev board arrived the other day >as well as a long needed copy of Rick Lyons Understanding DSP book. So I >have started reading the book and started the design again from scratch as >my first draft was just to prove concept and I took alot of shortcuts along >the way, so now I would lke to do it properly. > >I read chap 1, 2, and 8 (Quadrature Signals) As I am attempting a quadrature >demod, and I realised that I may be doing something wrong so I wanted to >check with you guys. > >In chap 8, it states "one drawback of this quadrature sampling with digital >mixing process: the fs sampling rate must be four times the signals fc" pg >359 second edition.. > >Is this true? I have been sampling a 88-108 MHz bandpass signal at 80MSPS >and then doing a quadrature demod with a nco generating sin and cos at >25.7MHz. however (to shift the equivelent of 105.7 down to 0) then filtering >and decimating. however 25.7 * 4 > 80MSPS so should I be comming unstuck >here??
It sounds like you're doing an appropriate IF sampling process, and as long as the anti-aliasing filtering is working you should be set. There's nothing wrong with building a system as you've described if there is a bandpass IF filter for antialiasing, or if there is NO other energy than the desired, bandlimited signal. The 4*fc notion in Rick's book is only a restriction if you use the fs/4 sampling trick to downconvert the signal when multiplying by only 0 and +/-1. Otherwise it doesn't really hold. What you're doing is bandpass sampling as shown in Fig. 8-24.
>One thing that I found was that if I decimate (after the quad mixing and low >pas filtering) any more than 8, then I can no longer reliable demodulate the >signal using a cordic / differentiator approach. I was hoping to be able to >decimate by ~64 to get down to 1.25MSPS or less as the FM signal is >bandlimited to ~ 300kHz. > >Any thoughts on this, in particular, should I be upping the initial sampling >rate st that I maintain the 4 * fc relationship for the quadrature mixing?
Your sampling scheme looks sound to me. If you're having signal tracking problems you may be getting distortion from your decimating filter or something like that. The sampling process itself isn't the issue. Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
Reply by Bhaskar Thiagarajan August 25, 20052005-08-25
"Paul Solomon" <psolomon@tpg.com.au> wrote in message
news:430db8a7$1@dnews.tpgi.com.au...
> Hi, > > Sorry for all the questions all the time, however I am still working on
this
> FM demod design and My shiny new stratix II dev board arrived the other
day
> as well as a long needed copy of Rick Lyons Understanding DSP book. So I > have started reading the book and started the design again from scratch as > my first draft was just to prove concept and I took alot of shortcuts
along
> the way, so now I would lke to do it properly. > > I read chap 1, 2, and 8 (Quadrature Signals) As I am attempting a
quadrature
> demod, and I realised that I may be doing something wrong so I wanted to > check with you guys. > > In chap 8, it states "one drawback of this quadrature sampling with
digital
> mixing process: the fs sampling rate must be four times the signals fc" pg > 359 second edition.. > > Is this true? I have been sampling a 88-108 MHz bandpass signal at 80MSPS > and then doing a quadrature demod with a nco generating sin and cos at > 25.7MHz. however (to shift the equivelent of 105.7 down to 0) then
filtering
> and decimating. however 25.7 * 4 > 80MSPS so should I be comming unstuck > here??
The comment in the book is related to the 'trick' of not needing a local NCO after the A/D if fs = 4*fc since the cosine and sin components work out to be ones and zeros. What you are doing should work just fine (to bring your band of interest to baseband) - can't comment on what you do with it after that since I'm not familiar with the various techniques and limitations of FM demod algorithms. Cheers Bhaskar
> > One thing that I found was that if I decimate (after the quad mixing and
low
> pas filtering) any more than 8, then I can no longer reliable demodulate
the
> signal using a cordic / differentiator approach. I was hoping to be able
to
> decimate by ~64 to get down to 1.25MSPS or less as the FM signal is > bandlimited to ~ 300kHz. > > Any thoughts on this, in particular, should I be upping the initial
sampling
> rate st that I maintain the 4 * fc relationship for the quadrature mixing? > > Regards, > > Paul Solomon > >
Reply by Paul Solomon August 25, 20052005-08-25
Hi,

Sorry for all the questions all the time, however I am still working on this 
FM demod design and My shiny new stratix II dev board arrived the other day 
as well as a long needed copy of Rick Lyons Understanding DSP book. So I 
have started reading the book and started the design again from scratch as 
my first draft was just to prove concept and I took alot of shortcuts along 
the way, so now I would lke to do it properly.

I read chap 1, 2, and 8 (Quadrature Signals) As I am attempting a quadrature 
demod, and I realised that I may be doing something wrong so I wanted to 
check with you guys.

In chap 8, it states "one drawback of this quadrature sampling with digital 
mixing process: the fs sampling rate must be four times the signals fc" pg 
359 second edition..

Is this true? I have been sampling a 88-108 MHz bandpass signal at 80MSPS 
and then doing a quadrature demod with a nco generating sin and cos at 
25.7MHz. however (to shift the equivelent of 105.7 down to 0) then filtering 
and decimating. however 25.7 * 4 > 80MSPS so should I be comming unstuck 
here??

One thing that I found was that if I decimate (after the quad mixing and low 
pas filtering) any more than 8, then I can no longer reliable demodulate the 
signal using a cordic / differentiator approach.  I was hoping to be able to 
decimate by ~64 to get down to 1.25MSPS  or less as the FM signal is 
bandlimited to ~ 300kHz.

Any thoughts on this, in particular, should I be upping the initial sampling 
rate st that I maintain the 4 * fc relationship for the quadrature mixing?

Regards,

Paul Solomon