Hello all,
In this post I would like to discuss about the OFDM numerology implemented in 5G NR and the corresponding frame structure.
For those of you who are not aware, I am laying out an overview of OFDM implemented in 4G as below. If you are familiar with it, you can skip to next section.
OFDM in 4G:
In this post I would like to discuss about the OFDM numerology implemented in 5G NR and the corresponding frame structure.
For those of you who are not aware, I am laying out an overview of OFDM implemented in 4G as below. If you are familiar with it, you can skip to next section.
OFDM in 4G:
- OFDM is a form of frequency division multiplexing technique that uses a large number of narrow band closely spaced sub-carriers to carry data.
- What is sub-carrier?
- it is basically a sinc shaped waveform. A short pulse in time domain is equivalent sinc shaped waveform in frequency domain. It is better understood with the below diagram.
- The subcarrier spacing equals the inverse of the symbol period. => Subcarrier Spacing = 1/T.
- In LTE, the sub-carrier spacing is chosen as static - which is 15kHz. How they arrived at 15kHz is kind of tricky. It is based on the coherence bandwidth & coherence time of the channel. The goal is to effectively combat frequency selective fast fading.
- Coherence bandwidth is a statistical measurement of the range of frequencies over which the channel can be considered "flat“.
- Coherence time is the time duration over which the channel impulse response is considered to be not varying.
- So, basically we need a waveform which has a low sub-carrier spacing over which the fading can be considered as flat and at the same time the symbol duration should be less than coherence time so that only slow fading occurs. For 4G frequencies, after a lot of simulations, they arrived at 15kHz.
- In OFDM, these sub-carrier are tightly placed together so that the zero crossings of one carrier will coincide with the highs of adjacent carriers. That's how they maintain orthogonality.
- OFDM signal generation. The snippet has been taken from a QCOM paper.
OFDM numerology in 5G:
Similar to 4G, 5G too uses OFDM. But 5G supports multiple subcarrier spacings. Unlike 4G, 5G is expected to support wide bandwidths (upto 400MHz). If we stick to the same 15kHz spacing, it needs an IFFT which should support 20k+ (check the diagram above for OFDM signal generation). This places a lot of computing burden and the system too becomes complex. There are also applications which need comparatively less symbol duration (low latency applications like real time gaming etc). So, a new parameter μ is introduced as part of 5G OFDM numerology. Based on this parameter, subcarrier spacing changes and accordingly the symbol duration. As the subcarrier spacing increases, symbol duration decreases and vice versa.
So, as μ ranges from 0 to 4, subcarrier spacing can change from 15kHz, 30kHz, 60kHz, 120kHz & 240kHz. Only the 60kHz subcarrier supports extended cyclic prefix.
There is one more reason why this increased sub carrier spacing is needed. As the mobile is moving with respect to transmitter, some Doppler frequency shift is seen at gNB. This Doppler frequency shift increases with frequency of operation. For 5G frequencies (say for FR2 which ranges from 24GHz to 52.6 GHz), the Doppler frequency shift is comparatively high. Since OFDM is prone to such frequency errors, measures are needed to combat this increased Doppler frequency shift. The increased sub carrier spacing helps here.
Also, when I was reading this, I asked this question to myself but couldn't get a proper answer. What could be the reason for enabling extended CP only for the subcarrier spacing 60kHz? If anyone of you know, please do let me know in comments.
Frame structure:
From the above sinc pulse, you can observe that the subcarrier spacing has relation to symbol duration. Basically, as the subcarrier spacing increases, symbol duration decreases. So, the OFDM numerology has dependence on frame structure. Here is an interesting diagram I found from a keysight whitepaper.
As you can see, symbol duration decreases with increase in subcarrier spacing. This results in more number of symbols with less duration in one subframe. the lower TTI helps lower latency requirement.
There is one more reason why this increased sub carrier spacing is needed. As the mobile is moving with respect to transmitter, some Doppler frequency shift is seen at gNB. This Doppler frequency shift increases with frequency of operation. For 5G frequencies (say for FR2 which ranges from 24GHz to 52.6 GHz), the Doppler frequency shift is comparatively high. Since OFDM is prone to such frequency errors, measures are needed to combat this increased Doppler frequency shift. The increased sub carrier spacing helps here.
Also, when I was reading this, I asked this question to myself but couldn't get a proper answer. What could be the reason for enabling extended CP only for the subcarrier spacing 60kHz? If anyone of you know, please do let me know in comments.
Frame structure:
From the above sinc pulse, you can observe that the subcarrier spacing has relation to symbol duration. Basically, as the subcarrier spacing increases, symbol duration decreases. So, the OFDM numerology has dependence on frame structure. Here is an interesting diagram I found from a keysight whitepaper.
As you can see, symbol duration decreases with increase in subcarrier spacing. This results in more number of symbols with less duration in one subframe. the lower TTI helps lower latency requirement.
- One radio frame has a duration of 10 ms and is divided into 10 subframes with the duration of 1 ms each.
- 1 Slot = 14 OFDM symbols (normal CP, otherwise 12 symbols). Forget about the slot in LTE literature. The Slot in 5G means different than that of 4G.
- Concept of Mini slot:
- The motivation for a mini-slot is to support URLLC traffic.
- Mini-slots have the duration of 2, 4, or 7 OFDM symbols and can start at arbitrary symbols relative to the symbol 0 of each sub-frame.
- Slots and Mini-slots are the basic scheduling unit. However, 5G NR also supports scheduling time of a partial slot - the How part, we'll discuss later.
- And then there are self-contained slots to support both DL & UL in one slot.
Will add more details in future posts.
No comments:
Post a Comment