I see, I'll try and play with the network parameters to see if anything changes.
The large packet strategy doesn't seem to be the best approach for what I am doing anyway, since I cannot reach 2048 samples. I'll also try to modify the K1N block to me more like a Keep-M-in-N block. Where the M packets are not discarded but accumulated. Hopefully the combination of this mod and a packet resize on the FFT side will allow me to make a 2048 samples long decimation and FFT. Your help is really appreciated. Lorenzo ________________________________ From: Rob Kossler <rkoss...@nd.edu> Sent: Tuesday, February 28, 2023 8:05 AM To: Minutolo, Lorenzo <minut...@caltech.edu> Cc: usrp-users@lists.ettus.com <usrp-users@lists.ettus.com> Subject: Re: [USRP-users] RFNOC packet size - Keep one in N Hi Lorenzo, Regarding the SPP=364, this sounds like it is being set according to your Ethernet transport MTU. Sometimes the NIC defaults to MTU=1500 rather than 9000. If I am right it means that even if you created a graph without K1N, you would still get SPP=364. If the Ethernet MTU is set to 9000, I would expect that the SPP would resolve to 2000. Regarding the 2048 vs 2044, you might be running into a limitation there. With the packet header bytes, the maximum payload length might not be able to reach 2048 - I'm not really certain. But, as a test, you could try 2000 (after you resolve the SPP=364 issue). Rob On Mon, Feb 27, 2023 at 6:36 PM Minutolo, Lorenzo <minut...@caltech.edu<mailto:minut...@caltech.edu>> wrote: Hi All, I am trying to build a firmware for an x300 device that uses the radio block and the Keep-1-in-N block using UHD 4.4. For my application I need the Keep-1-in-N block to operate in packet mode, on packets 2048 samples long. Before connecting the graph, I try to set the radio block to use packets of that length with the following command (C++): radio_control->set_property<int>("spp", 2048, 0); The first issue arises as, calling the function to check the spp, I read 2044 instead of 2048. After connecting the graph with radio RX->K1N->RX_streamer I check again for the radio spp and I get 364. Reading this page: https://kb.ettus.com/RFNoC_Frequently_Asked_Questions#What_is_the_SEP_buffer_size.3F I understand I could possibly use a larger packet by changing the CHDR width of the design. If I change the value in my YML file, I get Vivado to throw an error saying that 64 is the only value supported for the device. Running the design compiled with 64 bit CHDR width results in an even different packet size written on file. The questions are: 1. how can I make the radio to work with packets 2048 samples long in a x300? Does changing the SEP buffer size help? I'm currently using 8192 as buffer size. 2. why introducing the K1N block reduces the packet size to 364 samples? how do I change this behaviour? I can share the whole code/firmware if needed. Thanks in advance for the help. Lorenzo _______________________________________________ USRP-users mailing list -- usrp-users@lists.ettus.com<mailto:usrp-users@lists.ettus.com> To unsubscribe send an email to usrp-users-le...@lists.ettus.com<mailto:usrp-users-le...@lists.ettus.com>
_______________________________________________ USRP-users mailing list -- usrp-users@lists.ettus.com To unsubscribe send an email to usrp-users-le...@lists.ettus.com
