discuss-gnuradio
[Top][All Lists]
Advanced

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

Re: [Discuss-gnuradio] Success (kinda)


From: Eric Blossom
Subject: Re: [Discuss-gnuradio] Success (kinda)
Date: Tue, 23 Sep 2003 22:10:06 -0700
User-agent: Mutt/1.4.1i

On Wed, Sep 24, 2003 at 01:27:27PM +1000, Alan Gray wrote:
> 
> Ok, I've made a bit more progress toward troubleshooting FM
> reception with my system. John Gilmore suggested that I write
> a bunch of data to file and post it online for others to take
> a look at, I don't have a fast disk system yet, but I thought
> I'd give it a shot anyway.
>
> To cut a long story short, I ran the data through fm_demod1
> just to see what it sounded like. The audio produced is much
> clearer than when I try to read straight from the ADC, it's
> almost of listenable quality though it sounds as though it has
> been sped up a little (very slightly fastforwarded, it that
> normal? Or possibly a result of a slower disk system?)

Nope.  That's not normal ;-)

> So it seems I have an issue with the way fm_demod1 is reading
> data in. I'm currently not using any double buffering, so my
> first thought is that there are "gaps" occuring in the data
> I'm reading in, though it is a little odd because I'm using
> the exact same method of reading in to write data to file,
> maybe a little more time spent in processing?

You're definitely using more cycles running fm_demod1 then just
writing to the disk.  Does your card's i/o library indicate when there
are underruns in the data?

> I'm currently attempting to implement a double buffering
> system, and it appears though it appears to have had little to
> no effect on the sound quality as yet, so I'm a bit stumped as
> to what to do next.

FWIW, We the run 4020 driver with 16 MB of buffer space.  This gives
us about 400 ms of slack and allows us survive the vagaries of the
GNU/Linux scheduler.

> Results thus far are encouraging enough
> though, I might know better if I can get my hands on a fast
> disk system. Any suggestions as to what else I could look at?

If you've got a bunch of memory, you might try acquiring a bunch of
samples to memory, then when memory is filled, write it all to disk.
I suggest allocating the memory (say total memory - 100MB), then
touching each page to ensure that it's resident, then doing your
acquisition, then writing the result to disk.

For building a cheap fast disk system, I suggest two of Western
Digital large IDE drives.  I haven't checked in a while, but the one's
I've been using (about a year old, 120GB) are pretty fast.  Use the
linux LVM (logical volume manager) to stripe them RAID 0.  I think I
used a 32K block size.  I had to use ext2 filesystems instead of ext3
to get them to stream without choking, but then I was able to sustain
> 50 MB/sec until the disks filled.  FWIW, my system also has a 66
MHz, 64-bit IDE controller, but you may be able to get away without
it.  Using 3 10K RPM IBM SCSI drives, I was able to sustain 110
MB/sec, but the Measurement Computing 4020 card didn't get along
nicely with the SCSI controller, so that didn't work out so well.

Glad to hear you're making progress!

Eric




reply via email to

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