[USRP-users] USRP X300/X310 reference daisy-chain

Marcus Müller marcus.mueller at ettus.com
Fri Oct 17 06:13:18 EDT 2014


One thing I forgot to mention: the ref clock signal goes through a
jitter cleaner, which should reduce errors when daisy-chaining. (it's an
LMK04816, if you want to look up the datasheet).

Marcus

On 17.10.2014 11:55, Marcus Müller wrote:
> Hi Rob,
>
> Michael is right, the Ref out is buffered (it's U530, which is a FIN1002
> LVDS->LVTTL converter/buffer), and the signal is basically just an
> output of the internal 10MHz clock.
> So what you get by daisy chaining is USRPs that shouldn't exhibit
> significant frequency offset[1]. What you won't get is
> delay-compensation, since the X300 can't know your cable length. Also,
> you'll get accumulated error -- which might not be a big problem for the
> overall system, because the PLLs generating the ADC clocks on the
> motherboards and the LOs on the daughterboards don't have an infinite
> control loop bandwidth, which will antagonize random variations of pulse
> duration, but it surely won't increase system performance.
>
> Personally, I find daisy chaining to be an awesome concept when it comes
> to synchronizing "a few" devices, because you can do that without an
> external clock distributor, knowing that errors introduced can't be that
> serious. When it comes to synchronizing a lot of devices, I'd say a
> clock distributor is the easiest way to go, on the fact alone that you
> can take out any X300 without breaking clocking.
>
> So, as usual, I'm curious [2], so: what setup do you have in mind?
>
> Greetings,
> Marcus
>
> [1] let's assume, for gross simplification, that you're not moving the
> x300s relative to each other at high speeds while having them connected
> via coax, so we can neglect Doppler ;)
> [2] ok. I'm a nerd. Stuff me in a room with a rack full of cool devices
> and I'm happy for a while. That might explain the curiosity when it
> comes to connecting numbers of X300s...
> On 17.10.2014 02:36, Michael West via USRP-users wrote:
>> Hi Rob,
>>
>> Characterization of the daisy-chaining for the X3x0s has not been
>> completed, so it is not possible to say at this time.  Officially,
>> daisy-chaining is not yet supported.
>>
>> That said, I believe there is buffering on the lines so, in theory, an
>> unlimited number of devices can be daisy-chained.  Some propagation delay
>> from device to device is expected (varies dependent on cable length), so
>> the limitation is primarily dependent on how tightly all the devices need
>> to be synchronized for a particular application.  The bottom line is that
>> daisy-chaining is still experimental, so you will have to try it to see if
>> it works for you.
>>
>> Best regards,
>> Michael E. West
>>
>> On Thu, Oct 16, 2014 at 3:30 PM, Rob Miller via USRP-users <
>> usrp-users at lists.ettus.com> wrote:
>>
>>> Hi -
>>> Does anyone know a practical maximum for the number of USRP X300/310s that
>>> can have their references (10MHz/1PPS) daisy-chained together.
>>> Best,
>>> Rob
>>>
>>>
>>> _______________________________________________
>>> USRP-users mailing list
>>> USRP-users at lists.ettus.com
>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
>>>
>>>
>>
>> _______________________________________________
>> USRP-users mailing list
>> USRP-users at lists.ettus.com
>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
>





More information about the USRP-users mailing list