Re: [Discuss-gnuradio] Volk machine

[Nick Foster]

What was the size of the recorded file on the RAM disk? Are you seeing "O" indications (for overflow)?

--n

On Mon, Dec 2, 2013 at 1:59 PM, Paul B. Huter <address@hidden> wrote:

Nick:

I tried downsampling the 25MHz to 10, but still was not recording the whole time I ran (about 10 seconds, with only a couple seconds of playback). That was when I tried using a RAM disk, albeit at 50 downsampled to 30. I am using the following for my disk:

mount -o size=1G -t tmpfs none /mnt/tmpfs

I then store my data file to /mnt/tmpfs

Am I doing that wrong? Or is there some other explanation for not storing data (other than storage size)?

Again, all feedback and help is appreciated.

Nick Foster <address@hidden> wrote:

Paul,

30MHz is a big chunk of data to be streaming to anything. A 4GB ramdisk will be full in 30 seconds at this rate. Do you really not know a priori where in the whole 0-30MHz spectrum your signal will be?

I notice now in your first post that you're streaming 25Msps with a low-pass filter to 10MHz -- if this is the case, why use a 25Msps rate in the first place? You can make the USRP work for you by asking for 10Msps (or less, if you know you need less bandwidth) -- it'll handle the downsampling and filtering.

--n

On Mon, Dec 2, 2013 at 1:32 PM, Paul B. Huter <address@hidden> wrote:

I finally got around to trying writing to RAM, and the result is worse - my replay FFT is static.

I am trying to record a chunk of spectrum (in this case the shortwave chunk, 0-30MHz) and then go back and look at small pieces to find my specific data. If someone can provide insight into how to do this, I would appreciate it.

Thank you all for all the assistance to date.

Paul B. Huter

On Nov 25, 2013 9:38 AM, "West, Nathan" <address@hidden> wrote:

I agree with Nick: that VOLK stuff is all expected behavior. If you're
trying to write to a file at high rates you should look in to using a
ramdisk/tmpfs. You'll be limited by how much RAM you have rather than
IO speed.

However, based on your other threads I wonder if you've taken Tom's
recent suggestion to just lower your input sampling rate? If you're
only interested in ~1MHz bandwidth you shouldn't be sampling at 50
MHz.

-nathan

On Mon, Nov 25, 2013 at 12:08 AM, Nick Foster <address@hidden> wrote:
> Your file sink only records a few seconds of data because your hard drive
> can't keep up, not because of any problem with Volk. The Volk machine being
> used does not indicate which particular architecture is used for each kernel
> -- that isn't printed at runtime.
>
> --n
>
> On Nov 24, 2013 9:58 PM, "Paul B. Huter" <address@hidden> wrote:
>>
>> When running with a USRP source at 25M and a low pass filter down to
>> 10MHz, I get something saying "Using Volk machine: sse4_a_64", and my file
>> sink only records a couple seconds of data. I ran the volk_profile script,
>> but still get the same result. The script returned something other than
>> "sse4_a_64" as the best volk to use. My GNU Radio seems to have trouble
>> reading configuration files, so is there a way to manually point to the volk
>> parameter to use when I load GNU Radio?
>>
>> Paul B. Huter
>>