Preamble is 1010..., not 0101... as specified in the USB PD spec

[eliasnaur]

Thank you for this repository of code for communicating with USB power supplies. I working on my own implementation, but so far failed to make the power supply recognize my GoodCRC and Request messages.

I'm looking at your code as reference, and ran into a bug or misunderstanding regarding the PD packet preamble.

In the encoding of the preamble,

usb-pd-arduino/src/PDMessageEncoder.cpp

Lines 90 to 92 in a9765e7

	// preamble
	for (int i = 0; i < 16; i++)
	buffer[i] = 0b10110100;

there are 16 chunks of "1010" (in BMC). However, the USB PD specification section 5.6.1.1 states that "The Preamble Shall consist of a 64-bit sequence of alternating 0s and 1s. The Preamble Shall start with a "0" and
Shall end with a "1"." This pattern is also what I see when receiving Source_Capabilities messages in my own implementation.

As such, I had expected

    // preamble
    for (int i = 0; i < 16; i++)
        buffer[i] = 0b00101101; // "0101" in BMC.

Your code obviously works, and mine doesn't. What did I miss?

[manuelbl]

There are two layers to the protocol as shown in this illustration from the USB PD specification (revision 3.1, version 1.8)

At the data layer, the preamble is indeed 010101... However, at the BMC level (signal you would observe on the oscilloscope), 0 becomes a full clock cycle at the same level and 1 becomes a transition from 0 to 1 (or vice versa) in the middle of the cycle, plus a transition at the end of each cycle for both cases.

In my code, 1 bit is used for half a clock cycle. So the pattern 0b10110100 represent 4 data bits at the BMC level (and in little-endian representation; so it's transmitted from right to left). At the data layer it would be 0 1 0 1. This pattern is repeated 16 times to make up the 64-bit sequence. The resulting BMC bits are transmitted at twice the nominal clock rate.

I don't know your code and if you need the pattern at the data layer or the BMC layer. But it would explain the difference.

[eliasnaur]

Thank you very much for the detailed answer. I had missed you read the data in little-endian such that your 0b10110100 pattern indeed matches the preamble.

I also got my own implementation to work. The issue was that "USB PD protocol is very timing sensitive" :)