Wireless Communication Systems

Wireless Communication Systems

@CS.NCTU

Lecture 10: Rate Adaptation

Frequency-Aware Rate Adaptation (MobiCom’09)

Lecturer: Kate Ching-Ju Lin ( 林靖茹 )

Motivation

• The bandwidth supported in 802.11 is getting wider

⎻20MHz in 802.11a/b/g

⎻40MHz in 802.11n

⎻80-160MHz in 802.11ac

• 802.11 adopts OFDM, which partitions the wideband channel to subcarrier

• Frequency-selective fading

⎻Different subcarriers experience independent fading due to the multipath effect

⎻Different frequencies exhibit very different SNRs

⎻But the transmitter can assign one rate to the entire band

Frequency Diversity

• The SNRs of different frequencies can differ by as much as 20dB

• Different receivers prefer different frequencies

3

Key Features of FARA

• Allow a receiver to measure the SNR of each sub-channel

• Instead of assigning the same rate to the

entire band, allows each sub-channel to pick the optimal rate matching its SNR

54Mb/s

6Mb/s

SNR-based Adaptation

• Maintain a SNR-to-rate lookup table

• The sender transmits few symbols at the lowest bit- rate for all sub-channels

• The receiver selects the highest rate for each sub-

channel corresponding to the SNR of that sub-channel

⎻Discard the sub-channels if SNR is too low to support the

lowest rate ⁵

Rx-based Adaptation

• The receiver is in charge of

⎻Measuring the channel

⎻Selecting the rate

⎻Responding to the AP

• To decrease the feedback overhead, embed the rate information in ACK

• Perform some optimization to reduce the

size of the embedded information

FARA in Frequency-Aware MAC

• Further combine FARA with the frequency-aware MAC protocol to leverage frequency diversity

• Instead of communicating with one receiver at a time, serve N (2-5) receivers concurrently

⎻Randomly select N receivers with queued packets

⎻Assign each sub-channel to a proper receiver

⎻All the N receivers occupy the entire band

7

assign to the blue receiver assign to the green receiver

Performance

• Compare with SampleRate in 20MHz and 100MHz channel

location location

100MHz 20MHz

Throughput gain is especially large as

the band is wider

Wireless Communication Systems

@CS.NCTU

Lecture 10: Rate Adaptation

Predictable 802.11 Packet Delivery from Wireless Channel Measurements (SIGCOMM’10)

Kate Ching-Ju Lin ( 林靖茹 )

Motivation

• Again, different frequencies experience different channel condition  frequency-selective

• Why not FARA?

⎻Need hardware modification

Traditional SNR-based Adaptation

• SNR-based rate adaptation is usually inaccurate because we

⎻Assume frequency-flat fading

⎻Select the bit-rate based on “average SNR”

across subcarriers

• However, this will over-estimate the channel quality because

⎻A packet will fail to pass the CRC check even if

only a few bits are in error due to frequency-

selective fading

Traditional model: Packet SNR

• Traditional theory well maps the channel

condition (SNR) to the corresponding bit-error rate (BER)

⎻e.g., in BPSK

⎻But, this only work for a narrow band channel

• The average SNR over all sub-carriers is not a good representation of a wideband channel

⎻Why? The channel condition is not a linear function

⎻The losses in a few subcarriers would lead to packet

errors

Traditional model: Packet SNR

• Packet SNR: Average power of a link / Noise power

• Due to frequency-selective fading, a link could have a higher packet SNR, but also have a high bit-error rate

13

Packet SNR

Errors

Effective SNR (ESNR)

• Can we find a metric that can be used to

⎻Represent a wideband channel

⎻Estimate the BER of the whole packet

• Average SNR vs. Effective SNR

⎻Total power of a link vs. Useful power of a link

 Effective SNR (ESNR)

Packet SNR

Effective SNR

Effective SNR (ESNR)

• Benefits

⎻Can accurately estimate the packet delivery rate of packets

⎻Pick a single bit-rate that maximizes the packet delivery rate or the effective throughput in a wideband channel

• How to calculate?

⎻Reuse the theoretical channel model derived in the textbook

⎻Find the expected BER of a link

⎻Then, convert it back to the effective SNR

15

narrow-band SNR narrow-band BER

effective SNR packet BER

Effective BER and Effective SNR

• First calculate the average BER of a selected modulation k across all subcarriers i

• Convert it back to the effective SNR

16

BER

(): the inverse function BER

()

OFDM

Demodulator Deinterleaver Convolutional

Decoder Descrambler (0)

Received signal

MIMO Stream Separation

Separated signals for each spatial stream

(1)

Scrambled, coded bits

(3)

(2) Scrambled, interleaved, coded bits

(4)

Scrambled bits

(5) Received bitstream

Packet processing

Figure 3: The 802.11n MIMO-OFDM decoding process. MIMO receiver separates the RF signal (0) for each spatial stream (1).

Demodulation convertstheseparated signalsinto bits(2). Bitsfromthemultiplestreamsaredeinterleaved and combined (3) followed by convolutional decoding (4) to correct errors. Finally, scrambling that randomizes bit patterns is removed and the packet is processed (5).

Modulation Bits/Symbol (k) BERk(ρ)

BPSK 1 Q √

2ρ

QPSK 2 Q √

ρ

QAM-16 4 ³₄Q ρ/ 5

QAM-64 6 ₁₂⁷ Q ρ/ 21

Table 2: Bit error rate as a function of the symbol SNR ρ for narrowband signalsand OFDM modulations. Q isthestandard normal CDF.

likely to be variable, and simply knowing when the link starts to work isuseful information in practice.

802.11 Packet Reception. The model must account for the action of the802.11receiver onthereceivedsignal. Thisisacomplex pro- cess described in many pages of the802.11n specification [1]. Our challenge is to capture it well enough with a fairly simple model.

Webegin by describing themain steps involved (Figure3).

First, MIMO processing separates thesignals of multiple spatial streams that have been mixed by the channel. As wireless chan- nels are frequency-selective, this operation happens separately for each subcarrier. The demodulator converts each subcarrier’s sym- bols into the bits of each streamfromconstellations of several dif- ferent modulations (BPSK, QPSK, QAM-16, QAM-64). This hap- pensinmuchthesameway asdemodulating anarrowbandchannel.

The bits are then deinterleaved to undo an encoding that spreads errors that arebursty in frequency across thedatastream. A paral- lel to serial converter combines the bits into a single stream. For- ward error correction at any of several rates (1/2, 2/3, 3/4, and 5/6) is then decoded. Finally, the descrambler exclusive-ORs the bit- stream with a pseudorandom bitmask added at the transmitter to avoid data-dependent deterministic errors.

Modeling Delivery. We build our model up from narrowband de- modulation. Standard formulas summarized in Table2relateSNR (denoted ρ) to bit-error rate (BER) for the modulations used in 802.11 [8]. CSI gives us the SNR values (ρs) to use for each sub- carrier. For aSISO system, ρs is givenby thesingleentry in Hs.

In OFDM, decoding is applied across the demodulated bits of subcarriers. If we assume frequency-flat fading for the moment, then all the subcarriers have the same SNR. The link will behave the same as in our wired experiments in which RSSI reflect real performanceanditwill beeasytomakepredictionsfor agivenSNR and modulation combination. We can use Figure 1(a) to measure thefixedtransition points between rates and thus makeour choice.

Frequency-selectivefadingcomplicatesthispictureassomeweak subcarriers will bemuch morelikely to haveerrors than others that arestronger. Tomodel alink inthiscase, weturntothenotionof an effectiveSNR. This is defined as theSNR that would givethesame

error performance on a narrowband channel [18]. For example, thelinks in Figure2will haveeffectiveSNR values that arenearly equal because they perform similarly, even though their RSSIs are spread over 15dB.

TheeffectiveSNR is not simply theaveragesubcarrier SNR; in- deed, assuming auniformnoisefloor, that averageis indeed equiv- alent to the packet SNR derived from the RSSI. Instead, the effec- tive SNR is biased towards the weaker subcarrier SNRs because it is these subcarriers that produce most of the errors. If we ignore coding for themoment, then wecan compute theeffectiveSNR by averaging the subcarrier BERs and then finding the corresponding SNR. That is:

BEReff,k = 1

52 BERk(ρs) (1)

ρ_eff,k = BER^{− 1}_k (BER_eff,k) (2)

WeuseBER^{− 1}_k to denote the inverse mapping, fromBER to SNR.

We have also called the average BER across subcarriers the effec- tive BER, BEReff. SoftRate estimates BER using internal receiver state[28]. Wecomputeit fromchannel measurements instead.

Note that the BER mapping and hence effective SNR are func- tions of the modulation (k). That is, unlike the RSSI, a particular wireless channel will havefour different effectiveSNR values, one describingperformancefor eachof themodulations. Inpractice, the interesting regions for the four effective SNRs do not overlap be- causeat aparticular effectiveSNR valueonly one modulation will be near the transition from useless (BER ≈0.5) to lossless (BER

≈0). Whengraphsinthispaper arepresentedwithaneffectiveSNR axis, weuseall four values, eachin theappropriate SNR range.

For 802.11n, we also model MIMO processing at the receiver.

To do this we need to estimate the subcarrier SNRs for each spa- tial stream from the channel state matrix Hs. Although the stan- dard does not specify receiver processing, we assume that a Min- imum Mean Square Error (MMSE) receiver is used. It is compu- tationally simple, optimal and equivalent to Maximal-Ratio Com- bining (MRC) for a single stream, and near optimal for multiple streams. All of these makeit a likely choice in practice. The SNR of the i^th stream after MMSE processing for subcarrier s is given by ρs,i = 1/ Yi i − 1, where Y = H_s^H Hs + I ^{− 1} for i ∈ [1, N ] and N xN identity matrix I [27]. For MIMO, themodel computes theeffectiveBER averaged across both subcarriers and streams.

Coding interacts with the notion of effective SNR in a way that is difficult to analyze. One challenge is that the ability to correct bit errors depends on the position of the errors in the data stream.

To sidestep this problem, we rely on the interleaving that random- izes the coded bits across subcarriers and spatial streams. Assum- ing perfect interleaving and robust coding, bit errors in the stream should look no different from bit errors for flat channels (but at a

ESNR-based Rate Adaptation

• ESNR can be thought of the equivalent SNR of a wideband flat-fading channel

• Hence, now we are able to use ESNR to find the optimal rate by looking up the SNR-to- rate mapping table

17