7/20/2005 Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [Throughput calculation discussion] Date Submitted: [20 July, 2005 Source: [James Gilb] Company [SiBeam] Address [840 W. California, Suite 110, Sunnyvale, CA 94086] Voice:[858-484-4339], FAX: [858-484-4339], E-Mail:[last name at ieee dot org] Re: [TG3c requirements document] Abstract: [Calculating the theoretical throughput of the system is important in comparing PHY proposals. this document provides a simple method for calculating the throughput. ] Purpose: [Provide a method for calculating throughput at the PHY SAP.] Notice: This document has been prepared to assist the IEEE P802.15. It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release: The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P802.15. James Gilb, SiBeam

Elements of frame duration <month year> doc.: IEEE 802.15-<doc#> 7/20/2005 Elements of frame duration Frame duration consists of: TP: Duration of the preamble TPheader: PHY header duration TMheader: MAC header duration THCS: Header check sequence Tdata: Data duration, including any coding TFCS: FCS duration Each frame is separated from the preceding frame by an inter-frame spacing (IFS) SIFS: One of the DEVs is changing TX to RX or RX to TX MIFS: None of the DEVs need to change TX or RX state James Gilb, SiBeam <author>, <company>

Overhead depends on ACK Policy 7/20/2005 Overhead depends on ACK Policy Imm-ACK Policy Frame SIFS ACK SIFS Frame SIFS ACK SIFS Frame SIFS ACK SIFS Frame SIFS ACK No-ACK Policy Frame MIFS Frame MIFS Frame MIFS Frame MIFS Frame MIFS Dly-ACK Policy Frame MIFS Frame MIFS Frame MIFS Frame MIFS Frame SIFS Dly-ACK SIFS Implied-ACK Policy Frame SIFS Response + ACK SIFS Frame Response + ACK SIFS Frame SIFS James Gilb, SiBeam

Imm-ACK calculation Time to send data Time for frame SIFS ACK 7/20/2005 Imm-ACK calculation Time to send data Time for frame SIFS ACK Throughput = (number bits/frame)/(Frame time + 2*SIFS + ACK time) James Gilb, SiBeam

No-ACK calculation Time to send data Time for frame SIFS 7/20/2005 No-ACK calculation Time to send data Time for frame SIFS Throughput = (number bits/frame)/(Frame time + SIFS) James Gilb, SiBeam

Dly-ACK calculation Time to send data n*Time for frame 7/20/2005 Dly-ACK calculation Time to send data n*Time for frame n-1*MIFS between n frames SIFS ACK Throughput = n*(number bits/frame)/(n*Frame time + (n-1)*MIFS + 2*SIFS + ACK time) James Gilb, SiBeam

Implied ACK calculation 7/20/2005 Implied ACK calculation Assume TX and RX frames are same size Time to send data Time for frame SIFS Throughput = (number bits/frame)/[2*(Frame time + SIFS)] Bi-directional throughput = (number bits/frame)/(Frame time + SIFS) James Gilb, SiBeam

Key issues in overhead SIFS time MIFS time PHY preamble 7/20/2005 Key issues in overhead SIFS time Switch circuits, wideband is fast Is the frame OK? Demodulation and FEC shall have completed. MIFS time Don’t necessarily have to complete frame demodulation and FEC. PHY preamble How long to sync, equalize, AGC, frequency offset, phase sync, etc. James Gilb, SiBeam

Not an issue for high rates 7/20/2005 Not an issue for high rates Assume 500 Mb/s data rate PHY header: 4*8/5e8 = 64 ns MAC header: 10*8/5e8 = 160 ns MAC HCS: 2*8/5e8 = 32 ns MAC FCS: 4*8/5e8 = 64 ns Frame size? In practice it will be hard to get large enough frames. 20 k octets: 20*1024*8/5e8 = 327 us Requires aggregation, which is application dependent. James Gilb, SiBeam

Lost frames decrease throughput 7/20/2005 Lost frames decrease throughput Not every frame makes it. PER = 1-(1-BER)^(# of bits) For low BER, ~ (# of bits) * BER Real throughput (TP) reduced, depends on ACK policy No-ACK: TPactual = TP/(1+PER) Imm-ACK: TPactual = TP*(1-PER) See next slide Dly-ACK: Similar to Imm-ACK Implied-ACK: Similar to Imm-ACK James Gilb, SiBeam

Typical links have low PER 7/20/2005 Typical links have low PER Typically the link needs to have PER < 10 % No-ACK: TPactual = 1/(1+0.1) = 91% TPNo-ACK Imm-ACK: TPactual = (1-0.1) = 90% TPImm-ACK This will be roughly the same for all PHYs Recommend ignore PER for throughput calculations James Gilb, SiBeam

What else is left out Beacon time CAP time Guard intervals before CTAs 7/20/2005 What else is left out Beacon time CAP time Variable, can be zero for a given PHY Guard intervals before CTAs Only integer number of frames in a supeframe Not all space filled Allocations for other devices Many applications require a reverse data channel James Gilb, SiBeam

Theoretical Maximum Throughput 7/20/2005 Theoretical Maximum Throughput TMT is the best you can get if you had infinite data rate 802.11 has discussed TMT CSMA/CA has serious issues with TMT Pre .11e, TMT for 802.11 was < 75 Mb/s 802.11e changes have improved this. TDMA has different limits No time for contention James Gilb, SiBeam