[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Discuss-gnuradio] Packet drop from Ethernet (A BIG PROBLEM)
From: |
Marcus Müller |
Subject: |
Re: [Discuss-gnuradio] Packet drop from Ethernet (A BIG PROBLEM) |
Date: |
Wed, 9 Mar 2016 13:14:59 +0100 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.5.0 |
Hi Nikos,
you're absolutely right with respect to it just being a bus bandwidth
limitation: Gigabit ethernet can only carry 1Gbit/s, and with 16 bit per
I and Q part, 31.25 Megasamples would be the maximum rate you could get
through; adding a bit of overhead (it's not that much, roughly 40B
overhead per 1450B payload, so roughly 2.7%), this gets reduced to
roughly 30 MS/s, and including the fact that an average NIC isn't always
using the cable to its fullest extent, that might be even less.
So another reason why this isn't a GNU Radio issue is that 30MS/s is not
a rate supported by the USRP: sampling rates need to be
master_clock_rate/N, N being a natural number. That master clock rate is
fixed to 100MHz on N210, so the only possible rates are
100MS/s (impossible over Gigabit Ethernet),
50MS/s (only possible with 8bit samples),
33.333MS/s (only possible with 8bit samples, but bad, because 3 is an
odd decimation, and anti-aliasing filter performance is much worse for
these),
25MS/s,
20MS/s,
and so on.
UHD will print a warning that the rate is impossible and that a "close
by" rate was used.
Best regards,
Marcus