[hpsdr] Blackfin

Chris Albertson chrisalbertson90278 at yahoo.com
Thu Jul 12 12:36:23 PDT 2007


--- Frank Brickle <brickle at pobox.com> wrote:

> ***** High Performance Software Defined Radio Discussion List *****
> 
> J.D. Bakker wrote:
> 
> > The Blackfin (and many other DSPs) have direct interfaces for high 
> > speed low latency data converters, which is harder to achieve with 
> > PC-standard interfaces such as USB or PCI.
> 
> Isn't that exactly what Janus/Ozy are providing via USB?

No, the Janus/Ozy buffers the data on it's side of the
USB cable until the block is "full" and then sends the blocks.
A low latency design would move the data synchronously as
it was clocked off the ADC chip.  

On the PC side there is an undetermined latency inside Windows while
it (a) notices there is a new block of data available,
(2) Saves the context of whatever process was running, (3) reads the
block of data into RAM from the interface and (4) Marks the
SDR process that was waiting for the block of data as "runable"
and then returns to the scheduler so that it can run "the highest
priority runable process".  All this would add some milliseconds

This is not optimal if low latency is a goal.

With an on-card processor data can be clocked directly into
the processor's memory as the data comes off the ADC chip.

In either case however there is some unavoidable latency due
to the nature of the FFT.  For example if you are computing a
1024 point FFT you have to wait for 1024 samples before you
can start.  With relativly low sample rates (48 to 192 kilo
samples per second) this  unavoidable latency may dominate so
the latency in the USB interface and buffers may not matter as
much, I'd have to measure it to know but my guess is that the
USB design no more than doubles latency.  A lot depends on the
FFT block size.

Was it a mistake?  Certianly not because the design goal was to
build an interface to a PC so existing PC based software could be
used.   But if you have a different goal and are willing to start
a "clean slate" design you might do things differently.  The point is
that you can only evaluate a design relative to it's goals



Chris Albertson
  Home:   310-376-1029  chrisalbertson90278 at yahoo.com
  Office: 310-336-5189  Christopher.J.Albertson at aero.org
  KG6OMK/AG


      ____________________________________________________________________________________
Luggage? GPS? Comic books? 
Check out fitting gifts for grads at Yahoo! Search
http://search.yahoo.com/search?fr=oni_on_mail&p=graduation+gifts&cs=bz

 1184268983.0


More information about the Hpsdr mailing list