New USB Pmod design

Projects Pmods
1

Hey everyone,

At the end of the last Electronics Let’s Play stream I promised that I will commit the USB Pmod schematic and post it here for discussion. It took longer than I expected but here we go.

As I explained at the beginning of the stream the design goals are:

  • Provide electric connection of a USB-A and/or USB-C interface to an FPGA.
  • Allow the use of the connectors individually as well as passthrough for sniffing/analysis applications.
  • Be able to serve as source and sink devices.
  • Allow the use of Luna USB core and the ViewSB USB analysis software stack.

This is how the schematic looks like at the moment:

usb-sch
usb-sch3508×2483 815 KB

You can find the design files in the icebraker-pmod GitHub repository.

Credit goes to Kate Temkin for the idea and direction. For reference here is her napkin schematic:

schema2
schema21106×994 126 KB

And I have lifted the pull-up/-down networks from Konrad Beckmann’s USB Pmods that you can find here: https://git.xil.se/kbeckmann/pmods/src/branch/master/usbc_pmod/usbc_pmod.pdf

Since the stream it has been pointed out on Discord that we should reassign one of the pins used for the USB-A connector pullups to the USB-C connector. It is more useful to have USB LowSpeed pullup capability on the USB-C connector than on the USB-A connector that should only be used as an upstream port. :wink:

If you have any additional suggestions and ideas. Let me know. I will keep updating this thread as we make progress with this project. :smiley:

Cheers,
Piotr

2 Likes
2

Quickly noting some possible typos, so they don’t get lost on Discord:

  • (top right) USB_CC (likely USBC_CC)
    • maybe the two PU/PD circuits for CC should have USBC_CC (visually) connected together?
  • (top left) USBA_PUPD and USBC_PUPD (likely USBA_DP_PU and USBA_DM_PU)
    • although USBA_D{P,M}_PU might want to be named PUPD instead of PU since they both have complete PU/PD circuits, not just pull-up?
1 Like
3

So I ended up nerdsniping myself into building a calculator for the diode PU/PD circuitry (it’s on JSFiddle so you can also view/edit source code).
I kinda wanted to integrate the circuit it assumes, into the page, so it’s overall clearer about what’s going on, but I wasn’t sure about how to proceed with that (suggestions welcome).

It makes these assumptions:

  • one side pulls up, the other side pulls down, on the same line
  • any cable resistance between the two sides, or input impedance connected to that line, is negligible
  • therefore, the pull-up and pull-down combine to form a voltage divider, producing the “idle voltage”

I’ve more or less confirmed the voltage divider aspect for the USB D-/D+ case, thanks to this paragraph in the “Resistor ECN” document in the USB 2.0 spec:

These alternative means must produce an idle voltage (VIHZ) on D- or D+ of between 2.7 V and 3.6 V when terminated with a pull-down resistance of 15 kΩ±5% to ground.

Relevant (to this pmod) standard values:

D-/D+ CC
Rpull-up 1.5kΩ 36kΩ
Rpull-down 15kΩ 5.1kΩ
Vidle 3V 0.409V

Trying to replicate the standard Vidle isn’t easy with a diode drop, on the side closest to either 0V (down for CC) or 3.3V (up for D-/D+). In fact, a 800mV diode in a pull-up circuit will drop from 3.3V to 2.5V, which is already below the 2.7V minimum in the USB spec (for D-/D+ idle voltage).

To be clear, I’m not saying it won’t work, just that we might want to avoid diodes with a large forward drop in this kind of circuit, for maximum compatibility.

If 220mV diodes (Schottky?) are an option, they would allow reproducing the exact Vidle, with the following resistor values (warning: I haven’t looked at current or dissipated power at all):

D-/D+ CC
Rpull-up 0.4kΩ 33.26kΩ
Rpull-down 13.9kΩ 2.36kΩ

It’s still possible to use 800mV diodes for D-/D+ pull-down or CC pull-up (which have more voltage margin), I suppose it mostly depends on price and other characteristics I haven’t looked at.

If the 400Ω pull-up for D-/D+ seems a bit too low, it can be increased to any value below 2kΩ without going out of spec (when using a 220mV diode, that is - diodes with a larger drop require smaller resistors).

EDIT: it’s definitely too low, @tnt pointed out on Discord that I’m not accounting for the output driver impedance, and its ability to pull D-/D+ down to a logic low (below 0.3V for low/full-speed).
I’ll try to get that in somehow, probably by finding the max output impedance that would (barely) work for the standard PU/PD values and then compute what logic low/high voltage such a driver would be able to produce.
Preliminary testing shows that one needs at least 1.4kΩ pull-up (for 220mV diode drop), 400Ω is nowhere near enough to allow worst-case output driver to pull a logic low (best-case is plausible, based on the spec, but we want it to work with all spec-compliant hardware not just a small subset).
And I haven’t looked at all at the requirements on the CC lines.

2 Likes
4

(new comment because I’m limited to one link per comment, as a new user)

I just tried the D-/D+ values in CircuitJS, and while the math seems to be sound, to get a 220mV drop from the diodes I had to put in some arbitrary “forward drop at 1A” value.
CircuitJS tries to model diodes accurately and it seems like the voltage divider only draws 200μA through them, leading to lower voltage drop.
This might work to our advantage, but I don’t know how hard it would be to incorporate into the calculator.

1 Like