discuss-gnuradio
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Discuss-gnuradio] Direct ADC for Quadrature Demodulator


From: Joshua Lackey
Subject: Re: [Discuss-gnuradio] Direct ADC for Quadrature Demodulator
Date: Mon, 22 Mar 2004 07:45:11 -0800
User-agent: Mutt/1.2.5.1i

Quoting Radio Dev (address@hidden):
> Thanks a lot for Eric's comment -- "Should work fine."
> 
> I am hoping to use Joshua Lackey's Linux driver for the Datel ADC.
> 
> Now I am working on a Windows2000 with Datel's APIs for Win32.
> Datel ADC can work on DMA mode and FIFO mode.
> DMA -- stream continuously  (MAX 30MHz sampling rate, must use FIFO memory
> when samples > 32M)
> FIFO --  not continuously (MAX 40MHz sampling rate, few samples one time
> depending on FIFO memory size).
> Using DMA mode, the sampling rate will be 30MHz,  and seems enough fast
> (>2x10.7MH).

The issue is the bus speed.  I can stream ~32Msps (512Mbps) reliably
with my driver over the PCI bus.  If the CPU is doing serious signal
processing, 32Msps gives quite a few FIFO overruns.


> The problem is band width.
> When IF signal has a band width < 4MHz,  the sampling rate will be enough,
> i.e. 30MHz > 2x(10.7+4)MHz. Dangerous but it works.
> When IF signal has a band width > 5MHz,  the sampling rate will not be
> enough, i.e. 30MHz < 2x(10.7+5)MHz. Can not work.
> 
> Did I have a proper right understanding?

Not quite.  Say we're sampling at 32Msps.  This allows us to directly
identify frequencies between 0 and 16MHz.  Your AR5000 outputs an IF at
10.7MHz and, IIRC, has a bandwidth of 10MHz there.  Which is to say it
has a filter which is centered at 10.7 and has 10MHz bandwidth -- the
filter passes frequencies between 5.7MHz and 15.7MHz.  That means we get
the entire 10MHz from your frontend.  (Actually, 10MHz is just the 3db
points and so you can see even more but it starts to get real close to
the noise floor.)

There are quite a few details I'm glossing over of course.  The main one
being that there really aren't (personal) computers out there that can
do something meaningful with that bandwidth.  With the signal processing
I'm doing, I find that I can barely use even 1MHz of the bandwidth
before I start getting serious FIFO overflows.  (That is, I use a
frequency translating filter to mix the signal down and, at the same
time, decimate it to 1MHz bandwidth.  Even then I sometimes overflow.)

-- 
Joshua Lackey, PhD. -- address@hidden




reply via email to

[Prev in Thread] Current Thread [Next in Thread]