10 Steps to Determine 3G/4G IP Data Throughput Michael Lawton Wireless Product Planning Engineer Marv Wagner Wireless Applications Engineer Slide 1 Agenda • Introduction • 10 Steps to Data Throughput Testing – building up complexity • Case studies – “peeling back the onion” • Summary Slide 2 Technology Drivers for Wireless Networks 120 100 Mbps 80 60 • Higher speed, 40 • Lower latency 0 20 Voice Data 800 700 600 500 400 300 200 100 0 ms • All IP • Convergence (Radio Access and Core Networks) – LTE and IMS • Interworking E2E IP Throughput Testing is a key performance test which aligns with these technology drivers Slide 3 It’s All About More Data, Faster! • Mobile penetration continues to grow: > 5 billion subscribers worldwide – more than 70% penetration* =1 exabyte! • Mobile data traffic is growing exponentially caused by growing number of mobile devices such as tablets and smartphones accessing high-bandwidth applications. VoLTE • More spectrum is being made available • In addition to subscriber growth, there is parallel growth in cellular peak data rates Source: LTE World Summit presentation 2011 LTE-Advanced W-CDMA 384 kbps HSPA HSPA+ 42 Mbps 14 Mbps 21-168 Mbps LTE 150-300 Mbps 1 Gbps Growth in cellular peak data rates (theoretical) showing more than 2500 times higher data rate over a period of 10 years * Note some users have multiple subscriptions Slide 4 IP Multimedia System (IMS) Convergence Legacy IMS File Find Share Me Voice Chat Video Now Multiple vertical solutions Slide 5 RCSe IP/ IMS arch. IP SMS Voice Core Core Networks Core Networks Core Networks Networks Multiple Multiple Access Multiple Access Multiple Networks Access Networks Access Networks Networks Contacts/ Presence Applications Applications Applications Applications IMS VoLTE 2013 and on … All IMS/IP Multiple Multiple Access Multiple Access Multiple Networks Access Networks Access Networks Networks Traditional Data Channel Testing Methods Physical layer testing – Benefits: Verifies coding and basic performance of L1 – Issues: Does not include higher layers, signaling, or apps Standards-based testing – Benefits: Industry standard, repeatable, required for conformance – Issues: Does not include apps, limited configs tested, ideal conditions Often does not match real user experience Slide 6 Traditional Data Channel Testing Methods Field testing – Benefits: Real world conditions, can include apps – Issues: Not repeatable, often requires travel, difficult to troubleshoot and time consuming Proprietary test systems – Benefits: Repeatable test scenarios, in house 24x7 access – Issues: Requires large investment $$ and time plus dedicated staff Slide 7 E2E IP as a Measurement • Benefits – A simple measurement to make yielding quick results – Tests a key performance parameter vs a headline theoretical limit – Is a stress test that tests the complete phone – Excellent at finding if you have a problem • Issues – Not so good at isolating what your problem is! – Sometimes finds problems with the test and not the phone Slide 8 Agenda • Introduction • 10 Steps to Data Throughput Testing – building up complexity • Case studies – “peeling back the onion” • Summary Slide 9 E2E IP … 10 Step Plan, building up complexity 1. Will my device connect? 2. Do I have a good quality transmitter? 3. Do I have a good quality receiver? 4. Can I achieve max E2E tput under ideal conditions with UDP 5. What about with TCP and simultaneous UL/DL? 6. What happens if I try real application? 7. What happens under non-ideal conditions? 8. Is it robust? 9. Does it work closed loop? 10. How good is my battery life? Slide 10 Step 1: Will my device connect? Synchronize with Downlink of serving BS Decode PBCH Decode PDCCH/PDSCH to get SIB data Slide 11 UE sends PRACH using Zadoff Chu Sequence If no response UE retransmits with higher power BS responds addressing MS with the preamble identifier and providing an RA-RNTI BS sends timing alignment BS provides UL grant allocation using TCRNTI Security, bearer establishment, and IP Scan for downlink channels UL Power ranging & Random Access Sync to DL and decode broadcast info Power on SIB info provides capability info UE/eNB exchange RRC connection Request/Setup messages AAA exchanges with MME UE/eNB exchange RRC connection Reconfig/Complete messages Attach complete Establish Default bearer Obtain IP address 1. Will my device connect? Protocol test Protocol Logging and Analysis Software (N6061A) Message Editor Software (N6062A) NAS RRC Script Layer 3 RRC/NAS scenarios DL • Is the UE able to sync to the DL? • Can I get through the connection set-up • Can I ping my UE? • If not take a log and de-bug message exchange • Make edits as required with Message editor UL NAS NAS RRC IP RRC IP PDCP PDCP RLC RLC MAC MAC PHY PHY RF Slide 12 UE 2. Do I have a good quality Transmitter? RF test • High data throughput testing relies on good quality UL transmissions • Look for the following:– Ensure you have appropriate power and attenuation settings – High EVM for high order modulation schemes – High EVM at the band edge – Spurs both in band and out of band – Linearity issues/ spectral growth – Switching transients, LO settling time DL – Repeat tests with any “other” radio’s active UL Slide 13 UE 3GPP Tx Measurements Test case Number 6.2.2 6.2.3 6.2.4 6.2.5 6.3.2 6.3.3 6.3.4.1 6.3.4.2.1 6.3.4.2.2 6.3.5.1 6.3.5.2 6.3.5.3 6.5.1 6.5.2.1 6.5.2.1 A 6.5.2.2 6.5.2.3 6.5.2.4 6.6.1 6.6.2.1 6.6.2.2 6.6.2.3 6.6.3.1 6.6.3.2 6.6.3.3 6.7 Slide 14 3GPP 36.521 Test Case Description UE Maximum Output Power Maximum Power Reduction (MPR) Additional Maximum Power Reduction (A-MPR) Configured UE transmitted Output Power Minimum Output Power Transmit OFF Power (Covered by 6.3.4.1) General ON/OFF time mask PRACH time mask SRS time mask Power Control Absolute power tolerance Power Control Relative power tolerance Aggregate power control tolerance Frequency error Error Vector Magnitude (EVM) PUSCH-EVM with exclusion period Carrier leakage In-band emissions for non allocated RB EVM Equalizer spectrum flatness Occupied bandwidth Spectrum Emission Mask Additional Spectrum Emission Mask Adjacent Channel Leakage power Ratio Transmitter Spurious emissions Spurious emission band UE co-existence Additional spurious emissions Transmit intermodulation UL RF Measurements Constellation Channel Power Sub-carrier flatness SEM ACLR EVM vs symbol Slide 15 3. Do I have a good quality receiver? • High Data throughput testing relies on good a quality receiver • Look for the following:– sensitivity for different modulation schemes – Max input level performance – susceptibility to interference (simultaneous UL/DL, other radios, spurs from digital board, …) Slide 16 3. Do I have a good quality receiver? Slide 17 Rx Measurements Requires SS Requires SA Section 7 Receiver Characteristics 7.3 Reference sensitivity level 7.4 Maximum input level 7.5 Adjacent Channel Selectivity (ACS) 7.6.1 In-band blocking 7.6.2 Out-of-band blocking 7.6.3 Narrow band blocking 7.7 Spurious response 7.8.1 Wideband intermodulation 7.9 Spurious emissions Slide 18 Supported Supported Supported Supported Supported Supported Supported Supported Supported Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Y Y Y Y Y Yx2 Y 4. Can I achieve max E2E Tput under ideal conditions with UDP? E6621A PXT No Acks reqd at IP layer Tput/BLER DL DL data tput controlled by iperf UL Received DL data tput for radio link • iperf used to provide UDP data stream and measure received throughput • No IP level ACKs required • Measure results vs modulation/coding scheme • Fluctuating BLER may indicate RF issues • Sudden loss of data may indicate memory loss issues Slide 19 DL Data Throughput for TD LTE (20MHz channel, 2x2 MIMO, UL/DL config 5, special subframe config 6) 120 Mbps 100 80 Theory 60 MAC meas E2E IP meas 40 20 0 5 10 15 20 25 27 MCS Slide 20 Measurement Technique: UDP vs FTP (TCP) UDP FTP + Unacknowledged + Simulates real-world file transfers + removes flow control complexity + Transferred files can be viewed and/or compared + removes higher layer acks + Less susceptible latency - Not the full story for file transfers - Not suitable for used in shared networks Slide 21 - Adds flow control complexity - Add higher layer acks and retransmissions - TCP Control algorithms sensitive to multiple parameters - Test system configuration can affect results 5. Can I achieve max E2E tput under ideal conditions with TCP? E6621A PXT Acks reqd at IP layer Tput/BLER DL DL data tput controlled by TCP flow control UL Received DL data tput for radio link • TCP adds higher layer support for error detection, re-transmissions, congestion control and flow control • TCP flow control algorithms interpret “lost” packets as congestion • Careful consideration of parameters such as window size, number of parallel process, segment size etc. need to be considered Slide 22 TCP “Flapping” The same file FTP’d 10 times. 9 times rate is flat and consistent. 1 time there is a TCP slow-start as the flow control algorithm responds to an error – this is known as a TCP “flap” Slide 23 6. What happens if I try a real application? … (Voice, video, ftp …) • This should not add too much complexity VOICE • Most IP applications will typically use UDP or TCP VIDEO E6621A PXT Tput/BLER DL UL Slide 24 Received DL data tput for radio link 7. What happens under non-ideal conditions? AWGN OCNG • Typically fade the DL and use robust UL • Perform test mode and E2E testing • Measure MAC (BLER & Tput) and IP layer throughput • Use TCP with care! CHANNEL EMULATOR Slide 25 8. Is it robust? … • E2E IP tests PHY, MAC, PDCP, and IP layers all working together at full rate • Check processor can handle multiple real time activities – add SMS and voice calls during E2E IP • Check there are no memory overflow/leakage issues Slide 26 9. Does it work closed loop? 64QAM MCS 17-25 AUTO QPSK AUTO MCS 0-9 DL MCS 16QAM MCS 10-16 RI 2 1 AUTO 0 3 2 PMI 1 CHANNEL EMULATOR CQI PMI RI • BLER/Tput Testing • Supports Test Mode and E2E Testing Slide 27 10. How good is my battery life? 8960: •2G/3G BS emulation PSU: •Current monitoring Server: •FTP server •UDP Server •Apache HTTP server •MMS/SMS server Slide 28 Client: •Interactive Functional Test (IFT) •Wireless Protocol Advisor •License Keys •Modem drivers Agenda • Introduction • 10 Steps to Data Throughput Testing – building up complexity • Case studies – “peeling back the onion” • Summary Slide 29 Automated Measurements Give Repeatable 21Mbps Results! Occasional TCP “flapping” (2 of 23) Simul UL/DL FTP DL UDP DL FTP Simul UL/DL UDP UL UDP UL FTP Consistent UL UDP Over-flood issues Slide 30 Device Performance: MIPS Matter! MAC-d PDU Size Comparison with UDP PDU = 656 bits, Average = 792 kBytes/s (6.34 Mbps) 656 bit PDU PDU = 336 bits, Average = 497 kBytes/s (3.98 Mbps) This comparison was made with a very early Cat 8 HSDPA phone. When the MAC-d block size is smaller, the device doesn’t have the MIPS to sustain the high rate. Slide 31 336 bit PDU UDP FTP UDP FTP Cat14 (21Mbps) Devices – Better the second time around Slide 32 Dual-Carrier HSDPA (42Mbps) – Diversity Matters! Throughput for Rx Power = -20dBm to -85dBm 2nd RF Connector terminated (or connected via a splitter) Same device with 2nd RF connector left “floating” Slide 33 Dual-Carrier HSDPA (42Mbps) – Diversity Matters! Throughput for Rx Power = -20dBm to -85dBm Same device with 2nd RF connector left “floating” Device with only 1 RF connector available Slide 34 Not All HSDPA Cat 6 Devices Have the Same Throughput TCP Flap Slide 35 Not All HSDPA Cat 6 Devices Have the Same Throughput It’s really hard to tell when individual FTP transfers start and stop (there are 10 on this screen capture). You can see the “clean” FTP is the high data rate on the Ch 10677 trace. Slide 36 Not All HSDPA Cat 6 Devices Have the Same Throughput Slide 37 Data Throughput Across Input Power Level Data Throughput vs RF Input Level 6.00 Ch 10637 5.50 Data Throughput (Mbps) 5.00 4.50 Classic digital modulation performance. Consistent performance up to sensitivity threshold with very quick rolloff at power levels below threshold. 4.00 3.50 3.00 2.50 2.00 -100 -90 -80 -70 -60 Received Power (dBm) Slide 38 -50 -40 Data Throughput Across Input Power Level Data Throughput vs RF Input Level 6.00 Ch 10637 5.50 Ch 10562 Data Throughput (Mbps) 5.00 4.50 4.00 3.50 Data throughput roll off at high input, low band edge 3.00 2.50 2.00 -100 -90 -80 -70 Received Power (dBm) Slide 39 -60 -50 -40 Data Throughput Across Channels and RF Input Levels Ch 10562 Data Throughput vs Channel and RF Input Level - Phone A Ch 10587 Ch 10612 6.00 Ch 10637 Ch 10662 Ch 10687 5.50 Ch 10712 Ch 10737 Data Throughput (Mbps) Ch 10762 Ch 10787 5.00 Ch 10812 Ch 10837 4.50 4.00 3.50 Very consistent sensitivity across all RF Channels. 3.00 2.50 2.00 -100 -90 -80 -70 -60 Received Power (dBm) Slide 40 -50 -40 Data Throughput Across Channels and RF Input Levels Data Throughput vs Channel and RF Input Level - Phone B 7.00 Ch 10562 6.00 Data Throughput (Mbps) Ch 10585 Ch 10608 5.00 Ch 10631 Ch 10654 Ch 10677 4.00 Ch 10700 Ch 10723 3.00 Ch 10746 Ch 10769 Ch 10792 Sensitivity Issue at high band edge (approx 10dB) 2.00 Ch 10815 Ch 10838 1.00 0.00 -100 -90 -80 -70 -60 Received Power (dBm) Slide 41 -50 -40 Agenda • Introduction • 10 Steps to Data Throughput Testing – building up complexity • Case studies – “peeling back the onion” • Summary Slide 42 Summary • E2E IP Data throughput is a very useful measurement which stress tests the device against a key specification • The measurement is good for finding problems • Troubleshooting the problem requires you to peel back the onion • We have looked at examples of E2E IP issues found testing 3G/4G commercial UEs Slide 43