Nifty Type-C Features and Optimizations

Transcription

1 Nifty Type-C Features and Optimizations Mark Hayter & David Schneider Hardware Engineering, Chrome OS, Google Inc Taipei, Taiwan October 24 25,

2 Chromebook/USB-C: State of the union is good All new chromebooks (even low cost ones) have USB-C TM Power sink: wide input range (5-20V on most, 5V-PDP limit on others) Power source: min 7.5W, some 15W USB Data DP Alt mode video Most devices have 2xUSB-C: normally one on each side Good to also have USB-A if it fits in the device Still mostly USB PD 2.0: focus has been on adoption and cost Just started on PD 3.0: FRS, PPS, Power adapter info, Battery info 2

3 Nifty New USB-C Features Feature: Fast Role Swap Architecture: Mux and TCPC integration Architecture: Power Integration Part Feature: Programmable Power Source Feature: Vconn-powered Devices (VPDs) 3

4 Fast role swap and TCPC integration drive the need for a power front-end chip 4

5 Fast Role Swap Fast Role Swap added to USB-C / PD 3.0 to improve user experience Problem (in 1000 words): Solution: FRS Detect on accessory Signal on CC Device switches sink to 5V source in 150µs Protocol catches up Sinking DFP Post-FRS DFP Multi-port accessory Multi-port accessory AC adapter Bus-powered USB device AC adapter Bus-powered USB device Vconn makes surviving the transition a bit simpler, but it s not mandatory Supporting parts becoming available for laptop and accessory 5

6 Integration already here: IT8320 ITE integrated in EC 2x TCPC Non-integrated SoC ITE 8320 EC with dual TCPC Buck-boost Charger 5-20V 3A in/out 5V 3A VBUS 4-lane DisplayPort MLx, AUX SSTX/RX Passive USB-C Mux CC/Vconn (1.5W) SBU Four SS pairs USB Type-C USB 3.1 Gen 1 DM/DP BC1.2 Detect DM/DP 6

7 Integration already here: RK3399 Rock-Chip integrated in SoC Superspeed Mux TCPC Android makes good use USB PD in kernel RK3399 USB-C port (chip has 2 ports) EC TCPC Buck-boost Charger 5-20V 3A in/out 5V 3A VBUS 5V 0.3A 4-lane DisplayPort MLx, AUX SSTX/RX Muxing PHY CC/Vconn (1.5W) SBU Four SS pairs USB Type-C USB 3.1 Gen 1 DM/DP BC1.2 Detect DM/DP 7

8 Integration already here: RK3399 Rock-Chip integrated in SoC Buck-boost Charger ChromeOS today controls TCPC from the EC Mux is controlled from kernel RK3399 USB-C port (chip has 2 ports) EC TCPC 5-20V 3A in/out 5V 3A VBUS 5V 0.3A 4-lane DisplayPort MLx, AUX SSTX/RX Muxing PHY TCPC CC/Vconn (1.5W) SBU Four SS pairs USB Type-C USB 3.1 Gen 1 DM/DP BC1.2 Detect DM/DP 8

9 Consequence: move more code to OS kernel Today on Chrome OS EC runs TCPM, Power policy and uses TCPC Integrated mux is already controlled by kernel: system up when mux useful Good division for laptop where EC controls power states and battery charge Different from Android phones, which have USB-PD in the kernel Heavily limited code space ECs are disappearing as device form factors converge and prices drop Many factors point to TCPM ending up in the OS kernel Needed as the number of Quirks increases Needed for PDFU and Auth where internet searches may be involved Allows integrated TCPC and Muxes to be used easily Allows for much more involved power management between ports Downside: complexities handling low-power chargers and low battery 9

10 Consequence: Type-C Power Integration Part Frontend chip that sits next to the port Power path to/from charger Power path for 5V out FRS detect, charge in -> 5V out Vconn load switch Need 5V low Rds(on) process Present dead battery pulldown Protect host CC from > 3V Allows TCPC in process without 5V Also protects from CC-Vbus short Several vendors sample now/soon 3.xV 3.xV 5-20V 3A in/out w/discharge 5V 1.5A/3A out FRS Vconn out DeadBat (TCPIP?) VBUS CC1 CC2 USB Type-C 10

11 Programmable Power Source is for laptops too! 11

12 PPS: Laptops have the same issue as phones PPS added to USB-C / PD to allow sink to control source voltage Many think this is for phones -- but it is for laptops too! Problem: laptops are getting thin (thus USB-C) and thermally limited The CPU can max out the thermal dissipation of the enclosure In practice 15V -> 8V buck-boost is 90-92% efficient, at 40W get ~4W heat The CPU must reduce power by 4W Connecting charger causes slower machine Pixel-C tablet had this problem too Solution was to switch to 5V PDO (15W max) when screen is turned on 12

13 PPS can work for 2S Laptops Solution for 2S devices: Buck-boost for existing 5/9/15/20V adapters Super-efficient switched capacitor 2:1 conversion ~98% efficiency measured PPS to get 2x voltage needed to charge battery (~12-18V) Allows use of 3A cables for ~40W charging (1C) Charger chip has buck-boost and switched capacitor Control loops match, switch-over seamless Parts for 1S (phones) are becoming available, no 2S yet Hopefully by next year we can show data! Vbus Buck-Boost Vbus 2*Vbat Vbus=2*Vbat 2:1 Switched cap Vbat 13

14 Vconn-Powered USB Devices 14

15 Does this look familiar? Phone Accessory 3.2~4.5V 5V Boost (discrete or battery charger) 5V VBUS 3.3V LDO 1.8V LDO USB device µc This is how many USB accessories (especially compact audio devices) work with phones today Generously assuming boost is ~90% efficient, USB device s core power is coming in at 33% efficiency. Yikes! A USB2 device at 0.5A could consume 2.7W at battery 15

16 Ideal 3.2~4.5V Phone 3~5.5V Vconn 3.0V LDO 1.8V LDO Accessory USB device µc Now we re at 40~56% efficiency, without really changing the accessory Limiting maximum current in this mode to 100mA brings the worst-case battery load to 450mW (320mW at low battery) Your headphones can no longer ruin your phone s battery life! 16

17 VPDs enable the ideal, but keep compatibility VPD must be able to run from both VBUS ( V) and Vconn (3-5.5V), so it works with any kind of host VPD adds an emarker (simplest PD silicon possible) to advertise its capability. Host can probe SOP to find out it can remove VBUS Vconn is only available on cable plugs, so accessory must be direct-attach or have a captive cable Optimizing power for VPDs in a phone (or any other PD-capable source or DRP) does not require additional hardware! VPD DiscoverIdentity VDO is only defined in PD3, but PD2 devices can safely use PD3 just to query emarkers (minor header change) 17

18 VPDs enable something else (Soon ) The VPD architecture leaves room for adding charge-through to a future Type-C / PD spec This charge-through approach would be dramatically simpler and cheaper than what s required of multi-port adapters today Most DRP implementations out now would be able to support it with a software update...but it s not fully specified yet, so I ll have something to talk about next year 18

19 Q&A 19