USB C connector traces diffrential pair matching

[Veketti]

Dear All,

I'm planning to integrate logging to one of my designed devices by adding FRAM and CP2102N USB2Serial -converter. Nowadays I think anything else than USB C is outdated so planning to go with that connector. However as USB-C connector can be plugged either way, all pins are duplicated and I can't find connector which has both side pins internally connected (would appreciate to know if such exists?). So this leads to problem which I can't figure out how can I create the traces so that the diffrential pairs are matched? For example this connector that I'm planning to use Molex 2169890001 has D+ pins A6 and B6 which are on opposite sides and D- pins A7 & B7 which are also on opposite sides, but also separated by one pin and you couldn't even route trace between them. Could somebody explain how these data lines should be correctly routed? Do I need to match A7 -> CP2102N D- pin the same as B7 -> CP2102N D-? Example picture would be also highly appreciated. Thank you!

[gamalot]

Since the CP2102N is just a 12Mbps USB 2.0 Full-Speed device, you really don't have to worry about all that trace length matching for the differential pair.

[Whales]

(Assuming the 12MBit is correct)

It's worth getting a vague understanding of the dimensions of physical error that your signals will care about.

First start by getting an understand of the "length" of your signals in physical terms (meters).

λ = v / f
wavelength = velocity / frequency

Assume velocity = speed of light in vacuum = 3 *10^8 meters per second = 3e8 as shorthand most calculator software.  Really it should be speed of light in a pcb (which will be some fraction of this like half or a quarter) but I'm lazy and this is only a rough estimation technique.

A 12MBit/s signal will want more than 12MHz of bandwidth to work reliably.  Let's a assume a 10x factor (way overkill), so f = 120e6

wavelength = 3e8 / 120e6 = 2.5 meters

You don't want your errors to be near in size of the wavelength of your signal.  Let's assume another 10x factor difference, therefore you want your errors to be below 0.25 meters. 

That's massive.  Your track layout errors are unlikely to get anywhere near 25cm in size.  You may as well just get your differential pair within 1cm of the connector and then tee off random wires to the pins however is easiest.

[tooki]

For example this connector that I'm planning to use Molex 2169890001 has D+ pins A6 and B6 which are on opposite sides and D- pins A7 & B7 which are also on opposite sides

Well yeah, you’ll find that basically all USB-C connectors’ pinouts are like that, because the USB-C connector has rotational symmetry — that’s why you can insert the cable either way. It’s impractical to make the PCB pins come out in an order other than the order of the contacts inside the connector, so the PCB pins will also exhibit the rotational symmetry.

but also separated by one pin and you couldn't even route trace between them.

Have you reduced your clearances to the minimum the PCB manufacturer allows? You might be able to route between them then. You can do one trace on one layer and the other trace on another layer.

SMD versions anre easier to lay out, since you can run traces below the pads.

See the attached photo, copied from https://www.3dphotonix.com/post/usbcmux

Alas, the connector style you chose (the upright right-angle or “vertical flag” type) is one of the more exotic styles, with very few options. But I did find a few that are SMD:
LUB-C128D-007-G-19
USB4205-GF-A
USB4210-GF-A

Any of those would be easier to route.



Have you checked out how other products (especially open-source ones where you can download and inspect the PCB layout files) handle this?

FWIW, for USB 3 and higher speeds, you don’t even bother to try; instead, you use a multiplexer (MUX) chip and route each side separately. (You can see this in the image, and that topic is what the page I copied it from is about.)

[gamalot]

For the connector, I have this one in my LCSC's cart.

https://www.lcsc.com/product-detail/C22383838.html

[magic]

First start by getting an understand of the "length" of your signals in physical terms (meters).

λ = v / f
wavelength = velocity / frequency

Assume velocity = speed of light in vacuum = 3 *10^8 meters per second = 3e8 as shorthand most calculator software.  Really it should be speed of light in a pcb (which will be some fraction of this like half or a quarter) but I'm lazy and this is only a rough estimation technique.

It's called "velocity factor" and for cables it's said to be on the order of 0.6~0.8.

It often use 20cm/ns because it's easy to remember and convenient in calculations.
Speed of light is "one foot per nanosecond", by the way. Or close enough.

[Veketti]

Thank you all, this was eye opening. Especially Tooki's comment made me realize how stupid I was choosing the th -version. Initially I thought that I need the most robustness so I chose the th version. However it seems even (some of) the SMD versions have TH the housing so they most probably are equally robust. Indeed looking at the SMD version pad layout it's much easier to route.

I noticed that however those TH versions mentioned, doesn't have protruding pins which could be hand soldered by soldering iron. So my next question is, is there smd version of USB-C connector which has extended pins which can be soldered by soldering iron and housing be TH for robustness? Orientation doesn't really matter.

[gamalot]

They usually aren't designed with extended pins, maybe you can try to make the pads a bit longer so it's easier to solder with an iron.

Honestly, the best way is to avoid using a soldering iron. For components with plastic parts, I wouldn't recommend a hot air gun either.

I will go with a heating plate, it’s the simplest and fastest option to me.

[tooki]

Thank you all, this was eye opening. Especially Tooki's comment made me realize how stupid I was choosing the th -version. Initially I thought that I need the most robustness so I chose the th version. However it seems even (some of) the SMD versions have TH the housing so they most probably are equally robust. Indeed looking at the SMD version pad layout it's much easier to route.

I noticed that however those TH versions mentioned, doesn't have protruding pins which could be hand soldered by soldering iron. So my next question is, is there smd version of USB-C connector which has extended pins which can be soldered by soldering iron and housing be TH for robustness? Orientation doesn't really matter.

They exist, for example Molex 2171790003. But I wouldn’t worry about this — the kind with the short pins are still easy to solder by hand: just solder them from the top.

[gamalot]

For your application, I'd recommend model TYPE‑C‑31‑M‑12. I believe this one is the easiest to solder.

https://www.lcsc.com/product-detail/C165948.html