1

I understand that some reversible connectors already rectify symmetric power pins (not USBC.) Generally all data moves over balanced pair wires. Couldn't the balanced pairs also be simply rectified? This would add some electronics, but halve the pin count and surely increase robustness of the connector and reduce the complexity and cost of the cables.

Is rectifying a high-frequency communications channel just impossible for the lag/skew it introduces?

The USB-C solution just seems to be so delicate and overcomplicated - then again I've not seen any mechanical problems compared to loads with the previous micro USB.

JRE
  • 71,321
  • 10
  • 107
  • 188
J Collins
  • 585
  • 2
  • 12
  • "Couldn't the balanced pairs also be simply rectified?" They're typically not AC signals - they are always positive with respect to ground. Can you [edit] to give an example of a connector standard that you are thinking of? – Transistor Jan 09 '24 at 14:33
  • USB-C is complicated, because it needs to support data (USB2, USB3 up to tens of gigabits), power (100W or even more, that's 5A), method of negotiating what's happening (CC pins), and has alternate mode support for Thunderbolt, Displayport, and HDMI. It's not overcomplicated. – Justme Jan 09 '24 at 15:06

2 Answers2

2

I believe you're describing a port for a balanced bus with a pinout such as this:

  1. Gnd
  2. D+ or D-
  3. +5v
  4. D- or D+
  5. Gnd

If so, yes, that is technically possible. A specialized IC would include a demultiplexer (four bidirectional MOSFET switches, for a total of 8 MOSFETs) to direct pin 2 to either the D- or D+ line, and pin 3 to the other line. The demultiplexer would flip one way or the other based on sensing the voltage on pin 2 vs pin 4. If idle, the port would try one polarity, send out a message and see if it is acknowledged. If not, it would switch polarity and try again.

(We would not use rectification because diodes would reduce the signal level and demodulate any AM interference as the cable acts as an antenna.)

The plug would have just five contacts on just one face. The receptacle would also have 5 contacts, but they would be forks, so that they contact the plug's contacts regardless of its orientation.

Now, a connector with 5 fork contacts is not that much cheaper than a two-row connector with 10 contacts, 5 at the top and 5 at the bottom. With such 10-contact connector there would be no need for a demultiplexer because pin 2-top would be connected to pin 2-bottom through a PCB trace. That saves the cost of a demultiplexer.

Extending the concept even further, why not also have a 2-row plug? Now you have twice the current handling (because there are 2 pins in parallel for each line) and twice the reliability (because even if one side is not making good contact, the other side is). Now you're back to the USB-C solution. Which doesn't sound so back after you consider all the above, doesn't it?

Davide Andrea
  • 21,104
  • 6
  • 34
  • 73
  • 1
    Couldn't one just derive the polarity from the data, e.g. by knowing that the protocol always starts a packet with a certain bit pattern? – PMF Jan 09 '24 at 14:45
  • @PMF: That is done in some communication systems such as PCI express. During TS1 and TS2 (part of the LTSSM) the receiver calculates the running disparity of the ordered sets to determine if link pair reversal is necessary. – Peter Smith Jan 09 '24 at 14:51
  • I agree, there is no point in muxing if data is specified to be transferred polarity-free to begin with. USB-C needs duplicating the USB2 bus to always have correct polarity. And USB3 pins need to be muxed to correct places but it should be polarity free, many implementations allow swapping the SSRX+ and SSRX- pins. – Justme Jan 09 '24 at 15:01
  • How would full bridge rectification reintroduce the common mode interference? (If that is what you mean). As for power, just rectify it too and use only two pins. As for the mechanical design I could see the receptacle keeping the two rows of contacts and simply halving the density. Any pair is just made up of its matching pair in rotational symmetry. – J Collins Jan 09 '24 at 15:22
  • @J Collins A rectifier, together with stray capacitance on the far side from the connector, produces a DC voltage proportional to the level of the noise picked up by the cable. In other words, an AM demodulator. Any sources of RF (noise, transmitters, etc.) will create a DC level that interferes with the data signal. – Davide Andrea Jan 09 '24 at 15:31
  • ESP32 processors have an optional inversion stage in their UARTS so for RS485 inversion correction is as simple as flipping a few bits in a hardware register – Jasen Слава Україні Jan 09 '24 at 20:52
  • 1
    Actually this would not work. USB polarity cannot be detected from the pins when they are plugged in, because host keeps both pins low and device sets one of the pins high, depending on if it is a low-speed or full-speed device. @JasenСлаваУкраїні Sure it inverts the logic level but physical biasing on RS485 pins will still be wrong, so you will get wrong logic level on idle or fail-safe bias conditions, it would not work if the RS485 bus is swapped. – Justme Jan 09 '24 at 21:23
  • it works, I do it. so bias externally – Jasen Слава Україні Jan 10 '24 at 21:53
1

The USB legacy signals (D+, D-) are not 'balanced pairs' per se (although that's how they work in HS mode.) They're not symmetrical. The startup condition of the lines is single-ended, not only to sense LS vs. FS, but also some symbols used in LS and FS care about which pin is which. (Since HS first enumerates in LS or FS, it's implied that HS also depends on not-flipping D+/D-).

A deeper dive into why that is the case, here: In a USB cable, is it OK to swap the D+ and D- wires?

So the legacy D+/D- signals need to be flipped to the 'right' orientation with switching (not rectification) to function properly. The USB C answer is to simply duplicate the signals, which adds pins. For a custom, minimum-pins connection you'd need to detect the cable orientation somehow and use high-speed switches to flip the D+/D- signals to the right place.

On the other hand, the SS pairs can tolerate a polarity inversion, it's part of the spec. This helps deal with layout concerns. These can be flipped. But, in all cases the TX / RX have to connect to each other. More here: Swapping USB3 lines

It's permissible for upstream-facing ports (that is, devices) to only support the SS pairs and have no D+/D-, so this could be a symmetric connector.

The long-term USB evolution plan is to eventually drop legacy pins altogether. But as it stands now, downstream-facing ports must always support both legacy and SuperSpeed, so such a port needs to detect cable orientation or duplicate D+/D-.

hacktastical
  • 53,912
  • 2
  • 49
  • 152