There was a change related to how get_time_now was implemented in UHD 4.0 but that was improved in UHD 4.2. I would expect the behavior in 4.2 to be more like 3.15 and earlier.
Wade On Wed, Oct 12, 2022 at 3:24 AM zhou via USRP-users < usrp-users@lists.ettus.com> wrote: > Hi Marcus, > > Different OS/kernels in the two systems. > In old system with UHD 3.10, the OS/kernel are Ubuntu 16.04.6 LTS > (GNU/Linux 4.4.0-176-lowlatency x86_64) > In new system with UHD 4.1.0, the OS/kernel are Ubuntu 18.04.5 LTS > (GNU/Linux 5.4.0-80-lowlatency x86_64) > > > > On Tuesday, 11 October 2022 at 23:11:50 BST, Marcus D Leech < > patchvonbr...@gmail.com> wrote: > > > Is it the same OS/kernel version in both cases? > > Sent from my iPhone > > On Oct 11, 2022, at 5:53 PM, zhou <hwz...@yahoo.com> wrote: > > > Hi Marcus, > > I haven't tried UHD 4.2 or 4.3. I see ULLL sometimes in new system, so > actually I am thinking reverting UHD back to the 3.10. > > > > > On Tuesday, 11 October 2022 at 21:47:30 BST, Marcus D. Leech < > patchvonbr...@gmail.com> wrote: > > > On 2022-10-11 16:23, zhou wrote: > > Thanks Marcus. Yes, I use the get_time_now function to read time inside > USRPs. > > I also think the newer UHD versions should be better, but the measurements > tell differently. The queries to USRPs are sequential. > There are 16 USRPs using UHD 4.1.0. I query them sequentially and the time > intervals one example is as follows (unit: ms): > 1.683691 > 1.606288 > 1.629145 > 1.790804 > 1.685894 > 1.736507 > 1.471674 > 1.424962 > 1.460004 > 1.403206 > 1.435221 > 1.394569 > 2.043707 > 1.509543 > 1.853792 > > In old system, 24 USRPs use UHD 3.10. The time intervals in one example is > (unit: ms): > 0.385173 > 0.296745 > 0.284212 > 0.273682 > 0.278543 > 0.274327 > 0.279519 > 0.274441 > 0.276693 > 0.296875 > 0.275212 > 0.307623 > 0.309554 > 0.28233 > 0.28234 > 0.275119 > 0.283057 > 0.277138 > 0.279357 > 0.276074 > 0.277696 > 0.276335 > 0.280675 > > > Though there is some variance in intervals in both systems, the difference > between old and new systems are significant and reliable. > Could you please pass my case to the R&D team to confirm? > > Thanks, > Hongwei > > > > On Tuesday, 11 October 2022 at 18:35:24 BST, Marcus D. Leech > <patchvonbr...@gmail.com> <patchvonbr...@gmail.com> wrote: > > > On 2022-10-11 12:15, zhou via USRP-users wrote: > > Hi, > > Some of our USRPs are using UHD 3.10 because they are in old systems. and > I am also using UHD 4.1.0 in some other USRPs in new system. The USRP > products are all NI USRP 2944 (X310). > In my applications, I need to use PPS signal to synchronize multiple > USRPs. After applying PPS signals, I read back the time in USRPs one by > one. Because of network delay, there is some difference between the > readings. However, the difference is much bigger in UHD 4.1.0. > > The interval between two USRPs using UHD 3.10 is about 0.2ms while it is > about 1.4ms in UHD 4.1.0 > > Does this mean that UHD 4.1.0 is slower than UHD 3.10? > > Thanks for any suggestion, > > Hongwei > > Have you tried UHD 4.2 or UHD 4.3? > > I'll point out that the performance-critical aspect is the streaming > performance. The latency on get_time_now() shouldn't matter > that much, unless you're trying to use it for synchronization, and you > really should be doing that. That's what set_time_next_pps() > and friends are for. > > > _______________________________________________ > USRP-users mailing list -- usrp-users@lists.ettus.com > To unsubscribe send an email to usrp-users-le...@lists.ettus.com > > _______________________________________________ > USRP-users mailing list -- usrp-users@lists.ettus.com > To unsubscribe send an email to usrp-users-le...@lists.ettus.com >
_______________________________________________ USRP-users mailing list -- usrp-users@lists.ettus.com To unsubscribe send an email to usrp-users-le...@lists.ettus.com
