5G NR URLLC 1ms latency calculation O-RAN split 7.2
NR URLLC RTT 1ms is achievable in the real deployment scenario?
When we talk about to achieve 3gpp define URLLC 1ms RTT target, main problem is to achieve 1ms latency in real deployment scenario.
Here are some basic understanding :
+ For URLLC target is 0.5ms both in DL & UL.
+ Modem processing times, radio Transmission Time Interval (TTI) and an averaged contribution from Hybrid Automatic Repeat Request (HARQ) retransmissions all contribute to the user plane latency.
+ URLLC will require much more than lower latency in the radio domain. Edge computing, core network functionality, and even public cloud functions will need to be tied together and orchestrated to achieve end-to-end lower latency and reliable services.
+ NR defined Mini slot concept, Non-slot (Mini slot) based scheduling (The mini-slot length of 2, 4 or 7 symbols is recommended by 3GPP specs). For URLLC, at least 2 is supported.
- For URLLC, at least 2mini slots supported.
Now let’s move to ORAN specification for user plane latency components.
For URLLC use case: Considering UE at cell site, O-CU-UP resides in Edge cloud, O-DU can be located to Edge cloud & Edge cloud near at cell site.
TT1- Transport Delay between Cell Site & Edge Cloud
TUE -Delay Through the UE SW and HW Stack
TAIR -Transport propagation delay over air interface
TE1 -Cell site Switch/Router Delay
TRU- Delay through the O-RU User Plane
TE2 -Edge Site cloud fabric delay
TDU- Delay through the O-DU User Plane
TCU-UP- Delay through the O-CU User Plane.
Here we are considering only RAN elements processing time.
For O-CU-UP processing delay, DU processing delay- If they are located at same site so delay is very minimal in microseconds. We can bypass this.
DU processing delay- Vendor specific encoding time & other in us.(microseconds)
Fronthaul delay for 7.2x split- As per ORAN 100 us is FH delay, it depends on your FH requirement, distance etc.
RU processing delay- depends on RFSOC used, as per Xilinx webinar discussion approx. 85 µs in DL and 135 µs in UL.
Air interface processing time- If we implement mini slot concept so consider 2 ofdm symbol timing as per SCS 15,30,60 khz as per numerology figure in this post.
UE processing time- At UE side, its depend on chipset specification but it includes encoding, decoding etc.
To achieve this 1 ms target MNO should work on these aspects- mini slot config, SCS 15,30,60 khz for FR1,RU physical location, MEC deployment.
Thanks for reading!!.
References-
. R&S whitepaper
. ORAN specification Cloud Architecture & deployment scenarios
. Xilinx-Keysight webinar
. 5G_Americas_URLLLC_White_Paper