[Top][All Lists]

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

Re: [Discuss-gnuradio] sampling rate of tx_sampless.cc and rx_streaming_

From: Krishna S
Subject: Re: [Discuss-gnuradio] sampling rate of tx_sampless.cc and rx_streaming_sampless.cc
Date: Thu, 15 Apr 2010 11:51:10 +0530 (IST)

 It is for UMTS based standard LTE (20MHz bandwidth) which has 30.72Msps and this is beyond USRP2's instantaneous bandwidth (25MHz). I am simulating LTE (10MHz bandwidth) where I can halve the sampling rate of 15.36Msps. Reading the earlier mail threads and responses from Matt I understand that it is better to handle the sample rate converters at the software level rather than doing at the FPGA level in the USRP2 code.

LTE is based on OFDM and is not using any spectral shaping filter, instead the samples generated for OFDM symbols at the above rate are directly read/write into DAC/ADC. I could successfully do the above  by performing downlink and uplink using interpolation factor 16 which boils down to 6.25Msps but when I used the interpolation factor 6 (or 8) it gives the overflowing errors. I am using Quadcore PC (2.6GHz), can you suggest any other high end PC that can handle the above processing?

Krishna S      

--- On Fri, 9/4/10, John Orlando <address@hidden> wrote:

From: John Orlando <address@hidden>
Subject: Re: [Discuss-gnuradio] sampling rate of tx_sampless.cc and rx_streaming_sampless.cc
To: "Johnathan Corgan" <address@hidden>
Cc: "Krishna S" <address@hidden>, "gnu" <address@hidden>
Date: Friday, 9 April, 2010, 4:15 PM

> If what you really mean is that you have a host PC generated sample
> stream at 15.36 Msps and need to transmit it with the USRP2, then yes,
> you'd set the USRP2 FPGA interpolation to 6, then fractionally
> resample from 15.36 Msps to 16 Msps on the host.
> Depending on what the last DSP processing stage in your application
> is, you may be able to "fold" this resampling into the prior stage.
> For example, if the last step in your signal processing chain is a
> spectral shaping filter, like a root-raised-cosine filter, then you
> can reimplement this using a polyphase resampler and use the output
> filter taps there.  This would combine the filter and resampling
> operation in one block and eliminate the need for a very high CPU
> resampling block.
> Where does the figure of 15.36 Msps come from?

Just a guess, but that sounds like 4x the UMTS chip rate (which is
3.84 Mchips/sec).  If this is the case, there is an RRC filter at the
last DSP stage typically, so your suggestion for folding this
re-sampling in is spot-on.

John Orlando
CEO/System Architect
Epiq Solutions

Send free SMS to your Friends on Mobile from your Yahoo! Messenger. Download Now! http://messenger.yahoo.com/download.php
reply via email to

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