discuss-gnuradio
[Top][All Lists]
Advanced

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

Re: [Discuss-gnuradio] Custom block overflow


From: Marcus Müller
Subject: Re: [Discuss-gnuradio] Custom block overflow
Date: Wed, 05 Feb 2014 13:11:16 +0100
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:24.0) Gecko/20100101 Thunderbird/24.2.0

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ok, I think don't know why it overflows, but I get a better feeling
for your system :) thanks.

So: You have four UHD sources in parallel, feeding your block's four
input streams simultaneously, is that right? So, in that case, you
have a total sample rate of 2MS/s, which is more in regions where PCs
might be underpowered. You have to do a covariance matrix every 16µs
on average; this is quite a lot, possibly, though I have no idea of
the complexity that involves when using armadillo.

Maybe, you are I/O bound; do you write data to your hard drive or do
something similar after processing?

I think Martin's approach makes very much sense. Could you take the
four USRP sources, do a "streams to vector" and save that using a
file_sink? Do you still get overruns then?

Then, using file source and a vector to streams, could you process
your data offline and see if the would be sufficient for real time
processing?

Greetings,
Marcus



On 05.02.2014 10:55, sarankumar wrote:
> Hi, There is no throttle block in my flow graph. The USRP source is
> directly connected to the custom block. In the custom block, I form
> a 4x8 matrix with 4 being number of USRP's and 8 is the number of
> samples from each of them. Then I do cross co-variance of the
> matrix(4x8) with its transpose(8x4). The resulting 4x4 matrix is
> inverted. I use Armadillo linear algebra package for the
> co-variance and inverse operations. The computer is equipped with 8
> Gig ram and gigabit ethernet card.
> 
> Regards, Saran On Wed, Feb 5, 2014 at 2:56 PM, Marcus Müller
> <address@hidden> wrote:
> 
> Also, since this is a very common mistake: Maybe you have a
> throttle block in your flowgraph. If that's the case, remove it.
> 
> Greetings, Marcus
> 
> (Um, how do you invert a 4x8 matrix?)
> 
> On 05.02.2014 03:52, Martin Braun wrote:
>>>> On 04.02.2014 18:46, sarankumar wrote:
>>>>> Hi all, I am encountering overflow errors in my custom
>>>>> block in GRC even for low USRP sample rates of 500 Ksps.
>>>>> The flow graph goes like 4 USRP's connected to the custom
>>>>> block. The block does matrix correlation and inverse. The
>>>>> size of the matrix is 4x8. Any help? Regards, Saran
>>>> 
>>>> Your flow graph is probably running too slowly. You could
>>>> record IQ data and try w/ offline data.
>>>> 
>>>> MB
>>>> 
>>>> 
>>>> _______________________________________________
>>>> Discuss-gnuradio mailing list address@hidden 
>>>> https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
>> 
>> _______________________________________________ Discuss-gnuradio
>> mailing list address@hidden 
>> https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
>> 
> 
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJS8ipkAAoJEAFxB7BbsDrLYfMH/itHwES1O7cXxd5iTXjHmfkv
h1T2JEhVIVSLTN2s419Fh7t4iD0fW2DXPzQjOAEuqSlatmB+S2u4O7NnKUtuQURe
0CJQ3rybow/2bPksxufaZRja9703S1zhFILoAapsxfGgklcf61uw39fRjhicMx65
BgAQJHt6zMUmKMVVP9t7Rxyy+0QpzNge7H2PiMxZIrPuyWCOSKBhIabLdazfs420
P8Ko/TqEl0Mm9SVDTionlpx8dkfyUaVmtMahCalFqdgHtBiB4Ji6IIz/VPlJ7BXO
29DIHbJaZUf+OsrdkxmFRUx4R4vVmfQHGiEanMKiLLTrs4ffusQtPKtxqc5MUKM=
=szn5
-----END PGP SIGNATURE-----



reply via email to

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