[USRP-users] 10us+ sample delay between daughterboards

Marcus D. Leech patchvonbraun at gmail.com
Wed Jul 22 13:00:37 EDT 2020

On 07/22/2020 12:39 PM, Johannes Demel via USRP-users wrote:
> Hi all,
> I have an issue with multiple USRP streams. If they are on separate 
> daughterboards, but on the same motherboard aka USRP, those streams 
> are not time aligned. They show a time offset of more than 10us. I use 
> one USRP source block in GNU Radio and configure it with multiple 
> streams.
> At first, I observed this behavior with N310s but now I see it on an 
> X310 as well. I use a GNU Radio flowgraph where I set the clock to "PC 
> clock". I'd expect all streams to be in sync because they are handled 
> by the same object. At least that's what I hope.
> On an X310 I have 2 RX streams that observe a signal from across the 
> room. On an N310, I tried different configurations and I could observe 
> this delay between daughterboards. They are in sync for one 
> daughterboard.
> My flowgraph handles frequency offsets etc. But constant time offsets 
> between RX streams are problematic. And I suppose they should not be 
> there.
> I'd like to use GPSDOs at some point but I don't have a GPS signal in 
> the lab. And I don't have an octoclock or smth similar. But all of 
> this shouldn't matter since I only use on USRP for RX. Or am I missing 
> smth?
> USRP: N310/X310(with 2x UBX160)
> UHD: 3.15LTS
> OS: Ubuntu 20.04
> GNU Radio 3.9 (aka master)
> Cheers
> Johannes
Here's something you need to know about the time-keeping architecture in 
both X310 and N310.  There are, effectively TWO time-keeping

When you use "Sync to PC", those registers each get updated as a 
*separate* operation across the wire, so there can be (as you observe) a
   several-microsecond inconsistency between them.

The whole thing about the 1PPS pulse when doing time-setting is that no 
matter how many timekeeper registers there are, they all get
   set at the same instant, due to the 1PPS pulse processing.

More information about the USRP-users mailing list