BURX 26MHz clock

GNU Radio
1

Hi,

Probably a stupid question…
By connecting the BURX’s 26MHz clock to the USRP2’s 100MHz ref. input,
as
described in the source, does this pull the mainboard clock to 26MHz, or
does it
just phase lock to it? i.e. if I collect samples with usrp2_rx_cfile,
these will
all be samples at 26MHz?

Thanks
Dave

2

On Thu, Aug 19, 2010 at 12:17 PM, David E. [email protected]
wrote:

Hi,

Probably a stupid question…
By connecting the BURX’s 26MHz clock to the USRP2’s 100MHz ref. input, as
described in the source, does this pull the mainboard clock to 26MHz, or does it
just phase lock to it? i.e. if I collect samples with usrp2_rx_cfile, these will
all be samples at 26MHz?

On the USRP2, there isn’t any way to provide an external clock source
(at least not a straight-forward way…you can always unsolder the 100
MHz oscillator etc). So if you’re interested in using the BURX’s
on-board 26 MHz TCVCXO, this would clock the BURX board only, leaving
the A/D converters still running at 100 Msamples/sec.

The USRP2 can phase lock to an external 10 MHz reference, and it is
also possible (with firmware modifications to the aeMB code) to phase
lock to a reference clock frequency other than 10 MHz (this has been a
topic of discussion on the mailing list over the last week or so). So
you can access the 26 MHz TCVCXO output of the BURX board through the
U.FL connector, and route it to the SMA input on the USRP2 where the
10 MHz reference typically goes, which will phase lock the two clocks.
But you will need to tweak the aeMB firware to make this happen.

–
John O.
CEO/System Architect
Epiq Solutions

3

Thanks John,

This is how I understood it, but I suppose my real question is (not
being a dsp
expert), why would you want to do this?

I assumed that being a 26MHz clock, this would be ideal for GSM rate
signals?

Cheers,
Dave

----- Original Message ----
From: John O. [email protected]
To: David E. [email protected]
Cc: GNURadio [email protected]
Sent: Thu, 19 August, 2010 18:39:30
Subject: Re: [Discuss-gnuradio] BURX 26MHz clock

On Thu, Aug 19, 2010 at 12:17 PM, David E. [email protected]
wrote:

Hi,

Probably a stupid question…
By connecting the BURX’s 26MHz clock to the USRP2’s 100MHz ref. input, as
described in the source, does this pull the mainboard clock to 26MHz, or does
it
just phase lock to it? i.e. if I collect samples with usrp2_rx_cfile, these
will
all be samples at 26MHz?

On the USRP2, there isn’t any way to provide an external clock source
(at least not a straight-forward way…you can always unsolder the 100
MHz oscillator etc). So if you’re interested in using the BURX’s
on-board 26 MHz TCVCXO, this would clock the BURX board only, leaving
the A/D converters still running at 100 Msamples/sec.

The USRP2 can phase lock to an external 10 MHz reference, and it is
also possible (with firmware modifications to the aeMB code) to phase
lock to a reference clock frequency other than 10 MHz (this has been a
topic of discussion on the mailing list over the last week or so). So
you can access the 26 MHz TCVCXO output of the BURX board through the
U.FL connector, and route it to the SMA input on the USRP2 where the
10 MHz reference typically goes, which will phase lock the two clocks.
But you will need to tweak the aeMB firware to make this happen.

–
John O.
CEO/System Architect
Epiq Solutions

4

On Fri, Aug 20, 2010 at 8:47 AM, David E. [email protected] wrote:

Thanks John,

This is how I understood it, but I suppose my real question is (not being a dsp
expert), why would you want to do this?

For a system with only a single USRP2 in it, there isn’t much benefit
to having phase lock between the BURX board’s 26 MHz TCVCXO and the
USRP2’s 100 MHz oscillator. However, in a system with more than one
USRP2, it is possible to have the 26 MHz TCVCXO from one BURX board
feed additional BURX boards such that they all have a common
phase-locked reference clock. This requires the appropriate
amplification/distribution of the 26 MHz signal but it can be done.

The primary benefit to using the BURX’s 26 MHz TCVCXO is when using
the USRP1, since you actually get the entire system clocked from a
single clock source. This includes the A/D converters. In addition,
the BURX board also provides the ability to warp the TCVCXO to adjust
for clock error, and this correction translates to the sample rate as
well which is what you want for most RF receivers.

When we started the design of the BURX board, we had (mistakenly)
assumed that we would be able to clock the USRP2 from our on-board
oscillator, just like on the USRP1. There is always the option of
un-soldering the 100 MHz reference and piping in our 26 MHz reference
clock, but this is a non-trivial modification.

I assumed that being a 26MHz clock, this would be ideal for GSM rate signals?

Indeed, in either a USRP1 modified to accept an external clock, or a
USRP2 that has undergone the non-trivial modification mentioned above,
having the A/D converters run at a multiple of 13 MHz reduces the
computational load on the host machine.

–
John O.
CEO/System Architect
Epiq Solutions