After the previous post on what is inside a 5G NR CORSET and DCI messages, let’s have a look at how downlink assignments and uplink transmit opportunities for a device are signaled in a DCI. As you might remember from the previous post, the DCI contains, among other information, information on the frequency domain resource assignment and time domain resource assignment.
The Frequency Domain
Let’s have a look at the frequency domain assignment first. This works pretty similar in NR as in LTE and there are again two possibilities: One option for the network is to use a bitmap to indicate which Resource Blocks (RB) in the frequency domain are assigned to the device. The second option is to send the number of the first resource block to be assigned and the number of consecutive RBs that follow on the frequency axis.
The Time Domain
In LTE, there is no scheduling time domain because there are exactly two RBs in a subframe. If a mobile device gets an assignment on the frequency axis, both RBs are assigned to it. In 5G NR, this has become more flexible. A start and length value indicates where the downlink or uplink assignment starts in a slot and how many symbols are included. Assignments must not cross a slot boundary, which means that at most 14 consecutive symbols can be assigned on the time axis. With a sub-carrier spacing of 30 kHz, which is typical for a carrier in the 3.5 GHz band, a grant can span 0.5 ms at most. It is possible, however, to not only schedule resources in the current slot but also in slots that follow.
Another thing that is significantly different compared to LTE is the resource assignment timing. In LTE, all assignments for uplink resources are applicable only 4 sub-frames later. This means that there is a delay of 4 ms.
This is applicable for both the downlink and the uplink. With empty buffers and processing overhead, the round trip delay of the LTE air interface is thus around 8-10 ms in practice today. On the NR interface, however, uplink assignments apply to the same next uplink slot in case of TDD which means that the air interface delay is significantly lower. An additional speedup is that resources can be scheduled every 0.5 ms in the 3.5 GHz band instead of every 1 ms as on the LTE air interface.
These gains can only be realized, however, when data is only transferred on a 5G frequency band that uses 30 kHz sub-carriers. In practice, however, downlink data transfers usually employ a ‘split bearer’, i.e. data is transferred on LTE and NR. In the uplink direction, it is not uncommon to use LTE for data transfers due to the lower frequency band typically employed there when a 3.5 GHz NR carrier is added to an LTE connection.
Asynchronous HARQ Ack/Nack
Another difference I would like to point out is that in NR, the HARQ (Hybrid Automatic Repeat Request) mechanism is not synchronous as in LTE, but asynchronous. This means that there is no fixed time between transmitting the data on one side and expecting an acknowledgement from the other side. Instead, the DCI message contains information when and where to transmit or expect an acknowledgement for a data block.
For those who would like to take a look at the nitty gritty details head over to 3GPP TS 38.214.
The takeaway from this post is that downlink assignment and uplink transmit opportunities are scheduled in blocks that are flexible on the frequency and the time axis. Latency is significantly shorter compared to LTE because assignments can be made for the same slot instead of 4 sub-frames in the future. This only applies however, if data is transmitted only on an NR carrier, which is typically not the case today.