discuss-gnuradio
[Top][All Lists]
Advanced

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

Re: [Discuss-gnuradio] File Sink Output Size Problem


From: Pavan Yedavalli
Subject: Re: [Discuss-gnuradio] File Sink Output Size Problem
Date: Thu, 8 Sep 2016 19:08:15 -0700

Hi Kevin,

Thanks for getting back to me. I guess that makes sense. The only thing I was planning to use was the scipy.fromfile() function, in hopes that it would read out the binary data into a 1024-point vector, but I guess that was the wrong assumption.

Maybe the Head block is a good option - it looks like I should put that in between the complex_to_mag and the file_sink blocks, right? And there are two values, num_items and vec_length. By the documentation, it seems like num_items should be 1024 to match the complex_to_mag's vector length of 1024; however there is a mismatch there, as I get the error "Source IO size 4096 does not match sink IO size 4." I'm not sure what the problem is though. 

And based on your description, vec_length should be 1 and the file sink's length should also be 1; those do not give any problem. Thanks for the help again.

On Thu, Sep 8, 2016 at 6:43 PM, Kevin Reid <address@hidden> wrote:
On Thu, Sep 8, 2016 at 6:34 PM, Pavan Yedavalli <address@hidden> wrote:
My current flowgraph consists of a signal source (cosine) -> stream_to_vector of size 1024 -> forward FFT of size 1024 -> complex_to_mag of size 1024 -> file sink of size 1024 (unbuffered OFF and append set to overwrite). 

After I run this flowgraph for a few seconds and manually stop it, I use scipy.fromfile(open("filename"), dtype=scipy.float32) to retrieve the values from the binary file. However, the length of this output vector is some integer multiple of 1024 every time. Shouldn't it always be 1024 because I'm overwriting the file? I noticed that the longer I run the flowgraph, the larger this value is, so it seems like it may not be overwriting? Or perhaps there is something else I need to do make sure that my output is always of length 1024 with just the magnitudes of each sample?

The Vec Length field on the file sink doesn't mean the size of file it writes; it means the size of each stream item it writes (as a multiple of the input type's size). In your case each stream item is 1024 floats, which means exactly that your file will always contain some number of 1024-float items, as you have observed.

There is nothing in GNU Radio that will repeatedly overwrite a file as the flow graph runs — I suggest you find another way to accomplish your goal. What were you planning to use to read this file?

(If you just want a single FFT, you can add the Head block in your flow graph, set to 1, and the flow graph will exit after writing that one.)



--
Pavan

reply via email to

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