Re: [Discuss-gnuradio] timestamp tags and set_min_output_buffer

[Tim]

On 10/09/2013 02:59 PM, Juha Vierinen wrote:

Hi,

I'm writing a block that always outputs the same number of items as it gets in. 

The block is doing FFT on a fixed length vector, so I always need nfft samples in and nfft samples out. I've achieved this (I think) by using: set_min_noutput_items(nfft) and set_max_noutput_items(nfft). However, when I increase nfft to 8192, I get a message about not enough buffer space. I tested various things and ended up calling  

set_min_output_buffer(2*nfft)  as just nfft wasn't enough. It all seems to work nicely, except that the rx_time tags are screwed up -- I get a nearly random offset in the beginning, even though there are no dropped packets on the USRP. 

Is this a bug or a feature? What is the correct way to increase the value of noutput_items without breaking rx_time tags?

juha

_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio

Juha,

An easier way to achieve this is, is calling set_output_multiple(nfft) in the block constructor - which assures noutput_items = n*nfft where n >= 1
then if its a sync block without the relative rate or forecast being change, it will also assume it also needs nfft items each work call
set_min_noutput_nitems and set_min_output_buffer are bigger hammers that probably shouldn't be used in this case unless you have really big vectors - I think by default there should be no problems with an output multiple <= 8k
are you consuming nfft each time in your current  block?  If it is providing you nfft+extra samples each time through work you would get strange offsets if you simply consumed noutput_items each time.

-Tim