An Example of USB Power Delivery

May 16, 2017 by Jason Cerundolo

Let's take a look at what it takes to get almost 100W of power for a USB-C Easy Bake Oven. In this post, I'll go through the detail of the transactions on the Configuration Channel (CC) when I plug in my oven.

I borrowed a fancy Teledyne LeCroy Mercury T2C USB HS/PD Protocol Analyzer to capture the traffic. You can follow along for free with their software and my saved session. Here is the .usbraw file as well. Once you have everything installed and open the file, you should see this.

The Left device is the Apple MacBook Pro charger. The Right device is my USB-C breakout board. The first packets are from when I plugged in each of the cables. The charger identifies as a Source with a Rp pull-up on CC1. Then my board identifies as a Sink with a Rd pull-down on CC1. Once the connection is established, the charger modulates its current on CC1 to advertised USB Type-C Current at 1.5 Amp. At this point, my board is allowed to draw 1.5 A at 5 V. To get a different voltage requires some transactions using USB Power Delivery.

The charger starts by sending unsolicited Source_Capabilities messages to inform my device about its capabilities as a power source. My firmware ignores these messages, and sends a Get_Source_Cap (packet 86) to request the same information. When the charger sends it again, my board replies with a Request message (90) for the highest available power, which in this case is 5V, 2.4A. The charger sends an Accept message, then a few milliseconds later sends a PS_Ready message to let my board know it can now draw up to 2.4A. Note that every message is followed up with a GoodCRC message.

After that is all settled, one of the devices starts interrogating the cable plug (packets 96-100). SOP' is the start of packet symbol to address the cable plug. All the packets until now have been one device to another using the SOP symbol. As for which device is talking to the cable, it must be the charger. The two devices have an explicit USB PD contract, so only the Source is allowed to communicate with SOP' and SOP'', as noted in Section 4.9 of the Type-C spec. The cable responds with a Structured Vendor Defined Message (VDM). The data objects describe its vendor name, cert status, product ID, USB SuperSpeed support, max Vbus current, latency, and few other miscellaneous bits of data.

Now that the charger knows that the cable can support 5 Amps, it sends a new Source_Capabilities message. In addition to 5V, 2.4A, it now includes 9V, 3A and 20V, 4.3A. Note that the 5V fixed supply is first, as required by the spec. Then all other options are listed in increasing order of voltage, in this case, 27W then 86W. Upon receiving the new capabilities, my board requests the highest possible power. Just like before, the charger sends an Accept message followed by a PS_Ready message.

Finally, the charger sends a Structured VDM to find out more about the device it's charging. Unfortunately, I haven't implemented that part of my firmware, so there's no reply.

That's the end of the transactions. After this point, the Easy Bake oven has the power it needs to run the heater. The Arduino inside is using its on-board regulator to convert the 20V down to 5V to power its microcontroller. In this system, there is no need for further communication. In a more complex version, the oven could send a request for less power when it knows it doesn't need to run the heater. This would allow the source to reallocate that power to other devices. This charger has only one USB-C receptacle and is only connected to my board and the wall, so there's no need for anything like that.

If you want to learn more, I recommend going through the captured transactions with the USB Power Delivery spec open. If you right-click on the Packet number, you will see an option to inspect the raw data. You can select between looking at the 5-bit data that was transmitted on the wire, or the 4-bit nibbles it represents. In 4b mode, you can go through each bit of each message and compare it against the tables in Section 6 of the USB PD spec.

If you want to play around with USB-C yourself, you can buy one of my board on Tindie. My firmware libraries are on github for the FUSB302 and USB PD. Please feel free to contribute pull requests or suggestions. I'm always open to help making USB-C more accessible.

USB-C Easy Bake Oven

March 14, 2017 by Jason Cerundolo

I modified an Easy Bake Oven to run off USB-C. The obvious question is "Why?", but I prefer to ask, "Why Not?". The USB-C spec allows for 100 Watts of power to be transferred through the connector, and that is the power rating for the oven, so it should work.

I found one of the new models of the Easy Bake Oven on Amazon. The first step was taking it apart to figure out how it worked. The hardest part was getting the right tool to access the screws down extremely long counterbores. Eventually, I had to get a Torx T10 on a long T-handle, at least six inches. Then I used a flashlight and carefully guided the tool into the screw.

Once I had apart, I took a look at the insides. Overall, the design is very simple. These new models have moved on from incandescent bulbs and now use a nichrome wire as the heating element. The nichrome wire is connected through a switch and across mains. Part of the nichrome wire is used as a resistor divider to power the light. The light assembly has an LED, some current limiting resistors, and a Schottky diode anti-parallel to the LED. The diode/LED pair receives a stepped-down AC voltage when the heating element is on.

The major modification I had to make was to the heating element. USB-C spec allows a maximum voltage of 20V, whereas mains is closer to 120V. Rather than step up the USB-C voltage, I instead added some wires to the heating element to divide it into six equal segments, then wired all six segments in parallel. This lowers the resistance by a factor of 36, so at one-sixth the voltage, the current increases by a factor six. Thus, the power is about the same. It's hard to solder to nichrome wire, so I just wrapped the copper around the nichrome at least once at each junction. After my modifications, the cold resistance dropped from 190 Ohms to 6 Ohms, which means I wasn't perfect in dividing up the nichrome wire, but it should be close enough. I tested the newly modified element on its own with a bench power supply. At 20V constant voltage, it was drawing about 3.75A, or about 75W total. This was a bit lower than the rating printed on the case of the oven, but it's good enough for this demonstration.

I made a few other minor modifications. I changed the LED assembly to run off 5VDC by replacing the resistors with a lower value. I also epoxied my USB-C breakout board onto the metal panel that previously held the strain relief for the mains power cable coming into the unit.

Until recently, one of the hardest parts of this project was finding a suitable USB-C power source. The Google Play Store has a 60W charger, but I could not find a 100W version anywhere. Fortunately for this project, Apple has released an 89W charger for their new MacBook Pros. The $89 charger does not come with a cable. That's an extra $20. The cable has to be rated for 5A, unlike most cables, which can handle 3A. Hopefully prices will come down as these products become more popular.

For the electronics, I used my USB-C breakout board with the FUSB302B PHY and an Arduino Uno. I wired I2C plus interrupt between the two. I connected VBUS from the breakout board to VIN on the Arduino to power it. Then, I connected +3V3 from the Arduino to the VDD on the breadout board to power the FUSB302B, as well as +5V to V_pullup on the breakout board. I also connected VBUS to the switch, then to the modified heating element and back to GND. To make the connections easier, I crimped spade connectors onto jumper wires. Finally, I plugged the modified light into pin 13 on the Arduino.

The code I used is the library I've been porting from Google's Chromebook open source repository. I set the max allowed voltage for my board to 20V and the max power to 100,000,000 microWatts (100W), so it will request everything the charger can provide. I also programmed the Arduino to flash the light at startup, turn it off while USB-C negotiation was ongoing, then turn it on when it successfully negotiated for at least 75W.

Once everything was connected, I ran some tests before I screwed the unit back together. I flashed the Arduino with the firmware and plugged in the USB-C cable. Everything looked good. The Kill-a-Watt noted a slight increase in power, up to 5W. The voltage of VBUS measured 20V. Then I turned the switch to power the heating element. The Kill-a-Watt showed about 80W of power being drawn, which is what I expected. I measured the cooking chamber temperature with a thermocouple, and it got up to 300F after about 15 or 20 minutes. Best of all, nothing caught fire.

Not sure how you get 12 cookies out of the package...

Finally it was time to put everything together and run a final test. I tightened up a couple of screws and ran some mechanical fit-check tests, passing the metal tray through the unit to make sure I had everything aligned. That worked, so I tightened the rest of the screws. It was the moment of truth. I plugged the USB-C cable in. The light on the front flashed, then went solid. While the unit was preheating, I put together the cookie dough mix that came with the unit and prepared the baking tray. After 20 minutes, I figured the oven was as hot it was going to get. I put the cookies in for the recommended 9 minutes. However, when I pulled them out, they seemed a little undercooked. Another 5 minutes did the trick.

I consider this project a complete success. During this project, I had to touch on a variety of aspects of the USB-C spec. Getting close to 100W involves USB Power Delivery communications, electronically marked cables, and some thought in board layout to handling the higher than normal amps. I had to work with over 4000 lines of code and thousands of pages of specifications. I also had to redesign my breakout board to better handle the power. In the end, I hope this project can help other people understand USB-C more deeply. Plus I get to eat some cookies.

USB-C for Engineers, Part 3

February 02, 2017 by Jason Cerundolo

Check out Part 1 for an overview of all the USB3.1 specs, and Part 2 for more details on the USB-C connector.

USB Power Delivery is a new specification that enables new functionality for the USB-C connector. In this part, I'll go over the benefits of the specification and details of its implementation.

USB Power Delivery Rev 3.0 is included in the USB 3.1 specification, which you can download from here. The spec has three major purposes.

Enable a much wider range of power options
Provide a side-band channel for standard and vendor defined messaging
Allow the use of Alternate Modes.

Power

Under the USB-C specifiction, without Power Delivery, the maximum power allowed is 15 W. Additionally, the only allowable voltage for Vbus is 5 V. With power delivery, the maximum allowed power increases to 100 W with a maximum voltage of 20 V. A Power Delivery Explicit Contract overrides any other means of determining power levels, as shown in the USB-C spec.

In addition to fixed voltage supplies, there a few other options (section 7.1.3). Variable supplies are "very poorly regulated Sources". Battery supplies expose a direct connection to a device's battery. Programmable supplies expose a well regulated, but adjustable voltage output. Regardless of its other capabilities, any Source must provide a fixed 5V supply.

Side-band Communication

Vendor Defined Messages (VDMs) allow devices to exchange information not defined by the USB specifications. There are structured and unstructured VDMs. Unstructured VDMs provide 14 bits in the header to use for your own purpose, along with up to an additional six objects, each 32 bits long.

Structured VDMs are used to send information about, and agree on, Alternate Modes. For example, under DisplayPort over USB-C, Hot Plug Detect (HPD) is sent as a structured VDM.

Alternate Modes

Alternate modes let you use some of the pins in the USB-C connector for your own purposes. The most common alternate mode is DisplayPort over USB-C. A spec for HDMI over USB-C has also been released. Modes are distinguished with the Standard or Vendor ID (SVID), a unique 16-bit number assigned by the USB-IF (section 6.4.4 of USB-PD spec).

For a full featured cable, you can reconfigure the four SuperSpeed pairs and the Side-Band Use (SBU) pair. If the device has a "captive cable" (cannot be unplugged) or has a "direct connect application" (plug orientation is otherwise assured), three more pins are available. These are the two pins opposite the USB2.0 D+/D- lines and Vconn. You can learn about the connector requirements in section 5.1 of the USB-C spec.

BMC Communication

As of Rev 3.0 of the USB Power Delivery specification, only Biphase Mark Coding (BMC) over USB-C is supported. The physical layer is defined in section 5 of the spec. BMC uses a single wire to communicate between two devices, up to two cable plugs, and up to two debug devices. All communication is half duplex with collision avoidance and 4b5b encoding for DC-balance. The data rate is about 300 kbaud. CRC32 is used to ensure data integrity.

BMC is a version of Manchester coding. There is a transition at the start of every unit interval, and a second transition to indicate a 1. A preamble is used to train the receiver on the exact bit rate. This is followed by a sequence of K-codes that form a Start of Packet (SOP*), which is essentially an address. Then the data is sent, along with a CRC, then finally an End of Packet (EOP).

The SOP* can be one of a few options. SOP addresses the other device, and only Power Delivery Capable Sources or Sinks can respond. SOP' (aka SOP Prime) is used to communicate with the cable plug, such as in the case of an electronically marked cable. SOP'' (double prime) is the other cable plug. SOP'_Debug and SOP''_Debug are undefined, but I'm assuming they will be used for debugging purposes based on the name.

For symbol and bit ordering, everything is least significant first. For an SOP* sequence, K-code 1 is sent first. For each K-code, bit 0 is sent first. When transmitting 16-bit headers or 32-bit data objects, the least significant byte is sent first, and the first bit of the 4b5b nibble transmitted is bit 0.

Section 5.6.2 provides the details of the CRC-32 calculation. I've used this calculator to double check CRCs I've received. Be sure to select Hex and reorder your bytes if needed.

USB-PD Protocol

Everything that happens with USB Power Delivery happens with messages. There are three types of messages.

Control Messages (no payload)
Data Messages (short payload, up to 240 bits)
Extended Messages (bigger payload, up to 2080 bits)

All messages have a 16-bit header. The structure of the header is defined in section 6.2.1. A control message is just this 16-bit header. If bit 15 is set, the message is an extended message. If bits 14..12 are not zero, it is a data message. Otherwise, it is a control message. The other bits of the header determine the current power and data roles of the sender, specification revision used, and a 3-bit rotating message id. The bottom four bits are the message type. Depending on if the message is control, data, or extended, you will need to check a different table.

Most of the control messages (section 6.3) are self explanatory. The most common control message is GoodCRC. It is sent in response to every correctly received message, and has to start within 195 microseconds. Get_Source_Cap is used to request the power source capabilities of the other device. PR_Swap, DR_Swap, and VCONN_Swap are used to swap power, data, and Vconn providing roles, respectively. PS_RDY is used to indicate the power supply is ready.

The data messages (section 6.4) are more complicated. Each data message has at least one 32-bit data object following the header. Capabilities messages are used to share the port's options for power. This must include a 5V fixed supply, which is always the first option, and up to five additional options. Fixed supplies are first, lower to highest voltage. Next are battery supplies by minimum voltage, then variable by minimum voltage, then programmable by maximum voltage, all lowest to highest voltage.

The Request data message is used by a Sink to request a power option from the Source Capabilities list. The Sink then provides some information about how it will use the supply, including if it supports USB communications, USB Suspend, and its maximum and nominal current or power usage.

Vendor Defined Messages were discussed above. BIST is used to enter Built-In Self Test mode. Battery_Status and Get_Country_Info are self-explanatory. Alert is used to indicate a change of status.

Extended Messages (section 6.5) are new to Rev 3 of the spec. They are used to send a lot more information than is possible with just Control and Data messages. This can include hardware and firmware version IDs, manufacturer strings, additional battery capabilities, security information and authentication, and even firmware updates.

Implementation

I've been putting together some code to implement the USB Type-C and USB Power Delivery specs. As of the time of this writing, it is still early days. I started with the Google Chromebook code base and ported it to C++. After 4000 to 5000 lines of code, including a 1200 line state machine with 29 states, I have something that might be a bit messy, but is starting to work.

Last week, I used this code to get 14.8 V out of my Macbook charger, using my FUSB302 breakout board. The code ran on an Arduino and used 70% of the program space. I've been pulling the Arduino-specific functionality out into separate functions. My intention is to create a more standalone library suited for embedded development on a variety of platforms. Any help and suggestions are appreciated. You can find me on Twitter, among other places.

USB-C for Engineers, Part 2

January 12, 2017 by Jason Cerundolo

This part will focus on the USB-C connector itself. Check out Part 1 for an overview of all the USB3.1 specs.

The USB Type-C connector is slightly larger than the micro-B connector. It has 24 pins in a radially symmetric pinout, making its orientation reversible. Unlike previous versions of the USB connectors, there is no physical distinction in the plug depending on the functionality supported by the port or plug. USB-C does it all.

The Type-C Spec is part of the USB3.1 Spec, as explained in Part 1. You can download the full USB3.1 specification from usb.org. Follow along with the USB Type-C Specification Release 1.2 in the USB Type-C folder.

Pinout

Taking a quick look at the pinout, the radial symmetry is obvious. The GND pins are always on the outside, and the VBUS pins are always four in from the outside. GND, VBUS, DP, DM, and SSTX/SSRX are all familiar from the USB3.0 spec. The new pins are CC, VCONN, and SBU.

Section 3.2.3 of the USB Type-C Spec (page 55) lists the pinout in more detail, but here's a summary.

VBUS - provides power to the sink
DP/DM - USB2.0 communication, up to High-Speed USB (480 Mbps)
SSTX1/2, SSRX1/2 - SuperSpeed transmit, differential pairs, usually twisted pairs in the cable
CC - Configuration Channel used to configure the connection and send Power Delivery messages
VCONN - Connector power to power active cables and accessories
SBU - Sideband Use, basically extra wires used in Alternate Modes

Traps for the Unsuspecting

The SuperSpeed Tx and Rx lines swap in the cable. So do the SBU wires. That means that SSTX1 on one side is connected to SSRX1 on the other side. Similarly, SBU1 is connected to SBU2 on the other side. You cannot use the mux to fix this. I speak from personal experience.
The differential pairs will have a 90 Ohm differential impedance. If you are using an alternate mode, make sure it can handle 90 Ohm +/- 5 Ohm (section 3.7.1)
Cables can be electronically marked. This means there is a microcontroller inside one of the plugs, connected to the CC line. It is responsible for reporting the capabilities of the cable. "All USB Full-Featured Type-C cables shall be electronically marked." (section 4.9). Cables that only support USB2.0 do not need to be marked.
Be careful powering anything off of VBUS. With USB Power Delivery 2.0, VBUS can go up to 20V. Make sure this doesn't violate voltage ratings in your circuits. Benson Leung has had a few incidents in his famous Amazon reviews. Consider making anything connected to VBUS or CC over-Volt tolerant.
Also be careful powering anything from VCONN. You might not have it. Or you might have to provide it. Check in with section 4.4.3 of the spec.

Muxes

To support the reversible pinout, and make things easier for the user, each USB receptacle is required to have "the functional equivalent of a switch in both the host and device to appropriately route the SuperSpeed TX and RX signal pairs to the connected path through the cable." (section 4.5.1.1) The USB spec leaves the implementation up to the designer. For USB2.0, you can just short both possible positions together. For the SuperSpeed pairs, this usually means you need a mux.

Figuring out which way the cable is connected is done through the use of the CC line. This wire is always in the same location on the plug, and it can only be connected to one of two pins in the receptacle. The location opposite CC in the plug is VCONN. This is why the plug has CC and VCONN, but the receptacle has CC1 and CC2.

Using the CC line

The Configuration Channel is used to determine plug orientation, communication device roles, power capabilities, and send Power Delivery messages. Section 4.5 of the spec describes the details about this line and its uses.

To communicate device roles, pull-up (Rp) and pull-down (Rd) resistors are used. In reality, these will likely be current sources and sinks. In general, devices that once had a Type-A port will use the pull-up, and devices that once had a Type-B port will use the pull-down. In the below example, the source looks for a drop in the voltage on either CC1 or CC2. The pin that drops with an Rd pull-down is connected to the CC line. When it sees that, it can then provide power to VCONN and VBUS.

Meanwhile, the sink is monitoring for either CC1 or CC2 going higher in voltage. Then it can activate a pull-up on the other pin to read the value of Ra (pull-down for accessory).

Type-C Current

Even without supporting USB Power Delivery, it is possible to get as much as 15W through the USB connector. VBUS is still limited to 5 Volts, but the current can be as high as 3 Amps. This is done with analog signaling of voltages on the CC line. Essentially, the source changes its value of the Rp pull-up resistor to set the voltage of the CC line within certain ranges. Section 4.6.2.1 describes the details, and section 4.11 has the parameters.

USB2.0 Designs on Type-C

Supporting Type-C from an existing USB2.0 design is straightforward and cheap (except for the connector). Basically you tie the DP pins together, tie the DM pins together, and add one pull-down resistor to each CC pin. This new Type-C device will identify as a data and power sink, use the default USB2.0 power of 500 mA, and work in either orientation of the plug.

Type-C Connector Designs

In the last year alone, many new designs of USB-C connectors have come onto the market. Most of the major players have a good line-up. Here's a few parameters to consider, beyond the usual.

USB2.0 vs USB3.0 vs USB3.1 - Usually Type-C connectors are rated for a max data rate. USB3.1 supports up to 10 Gbps, USB3.0 would be 5 Gbps, and USB2.0 is 480 Mbps on the DP/DM pins. USB2.0 might not have as much shielding.
Right-angle vs Vertical Mount - Fairly self-explanatory. Be sure to have a plan to mechanically support the vertical mount connector.
Dual-SMT or Hybrid - Some connectors have two rows of 12 SMT pads each. The hybrid connectors have a outer row of SMT pads and an inner row of thru-hole pins. Beware that usually the thru-hole pins are designed for 0.6-1.0 mm thick PCBs.
Current and Voltage Rating - Gone are the days you can rely on USB connectors supporting the current and voltage you expect. If you need more than 3.0 Amps or 5.0 Volts, take a close look at the current and voltage ratings.

More Info

Here's a few links to check out with more detail.

If you want to experiment with Type-C at home, you can buy an Fairchild FUSB302B Power Delivery PHY breakout board from my Tindie store and download the library I put together from GitHub. Just add an Arduino and you're good to go.

Part 3 will take a closer look at USB Power Delivery and the BMC signaling uses on the CC line to set up features like higher voltages on VBUS and Alternate Modes.

USB-C for Engineers, Part 1

December 30, 2016 by Jason Cerundolo

USB Type-C (USB-C for short) is a new connector that promises wonderful things, including reversibility, 100W of power, and 20 Gbps data transfer. However, I've noticed a lot of confusion about this new standard and how it relates to the rest of the USB ecosystem. I've been having to learn all about USB-C for my day job, and I'd like to share what I've learned and help other engineers design USB-C into their products.

The first point of confusion with USB-C is with the variety of other specs that came out at the same time, including

USB 3.1
USB SuperSpeed and SuperSpeedPlus
USB Type-C
USB Power Delivery 2.0

In this post, I'd like to go over what each of these specs are how they interrelate. In future posts, I'll go into more detail on each one.

USB 3.1

USB 3.1 is a specification published by the USB Implementers Forum. This is the overall standard for the next generation of USB devices and cables. Inside is everything you need to connect to the latest USB hosts and devices. It includes by reference the other specs listed above. If you download the spec from the USB-IF website (here), you will receive the following documents in a zip file.

USB 3.1 Rev 1.0
USB Type-C Rev 1.2 or later
USB Power Delivery Rev 2.0 or later
USB Port Controller Rev 1.0 or later
various other specifications, agreements, and redline versions of the above

USB SuperSpeedPlus

USB SuperSpeedPlus refers the latest USB data bus. In this case it operates at 10 Gbps/lane. SuperSpeedPlus is used to refer the new features that distinguish it from SuperSpeed (not Plus). USB 3.1 uses "Enchanced SuperSpeed" as a general term to refer features common to both. To make matters more confusing, SuperSpeed is also referred to as USB 3.1 Gen 1. Although this is usually more specifically referring to the physical layer.

The blue Type-A plugs you see add five wires to support the new USB 3.1 bus. There is one differential pair each for receive and transmit, e.g. full duplex, as well as an extra ground.

USB Type-C

USB Type-C is the connector itself. This name follows from the USB 2.0 connectors of Type-A (host side) and various forms of Type-B (device side), including standard B, mini-B, and micro-B. With USB Type-C, there is no distinguishing the power and data role based on the physical connector. Both devices and hosts, sources and sinks, will have the USB-C receptacle. USB-C cables have the same plug on both ends.

The spec includes mechanical, electrical, and some functional details on how to use it. Some of these functions include signaling data and power roles, available power, and whether or not a device is an accessory. The USB Port Controller Specification provides a common interface to ICs that are directly connected to a port. The goal of this spec is to ease the development of software.

USB Power Delivery

USB Power Delivery is where most of the revolutionary features are. "Power Delivery" is a bit of a misnomer, because it does so much more. This is the spec that describes how to swap data roles, swap power roles, move Vbus to different voltages, and use the pins on the USB-C connector for other purposes like DisplayPort. This is what lets you plug your laptop into your monitor to present slides while recharging and running a full USB 2.0 hub, all with one cable.

Technically Type-A and Type-B connectors can use Power Delivery, but I've never seen it implemented. The spec that lets your phone charge at high speed if D+ and D- are tied together is the USB Battery Charging Spec. Power Delivery will likely be a de facto replacement for Battery Charging as more phone go to USB-C, but it is not a direct replacement.

Conclusion

I hope this helps clear things up. The important take away is that USB 3.1 is the overall spec, and each of these new features and connectors are parts of that spec. In part 2, I will go into more details on the Type-C spec.