Questions regarding impedance matching of USB

Hi,

I’m working on a 4 layer PCB breakout for WT41u-E which has both a UART and a USB 2.0 interface (Full speed 12Mbits/s)

I’m thinking about ordering from JLCPCB, see their stackup.

Now the pads on the WT41 looks randomly placed… the USB+ and USB- have an orientation which require the traces to move through the inner module area, otherwise they cross each other (see image)

I’ll be very thankful if you’ll help with the following:

  1. How can I match a 90 ohm impedance on the USB differential pair? The traces start on the first layer, go through an ESD protection IC then through resistors (which should do the matching, according to the WT41 datasheet), then through vias to the forth layer and back to the first and only then they are connected to the USB pads of module…

  2. Is it ok to route the USB traces in the 4th layer? the 4 layers are 1:Signal 2:GND 3:Power 4:Signal

  3. Is it ok to place the VIAs and the traces under the module?

  4. Bonus question - the UART pads are also randomly placed, how bad is it if the traces length are not the same?

Thanks!

If the module would be at bottom while USB connector at top then they wouldn’t be crossed :slight_smile:
I have done only few PCBs with USB, but not the fastest one and only one of them is a product to be sold (others are just our tools used for our own needs). So don’t assume that I’m right in my opinions.
I think - the shorter connection the less important impedance matching.
I would rotate the module to have USB pins down. I would crossed them at resistors placing them horizontally.
I understand UART as RS232 not RS485. There you have not symmetrical signals - one line TXD and one RXD. Each signal comes from different source. Difference in their length is absolutely not important. If it would be RS485 then A and B lines should have the same length, but even then it is not so critical.

1 Like

This is USB FS (12MB/s). In practice a wet noodle will suffice.

2 Likes

Good point regarding the connector at the bottom, thanks for showing me the light :slight_smile: here, any idea why UART TX and RX are on opposite sides?

Just to clarify, are you suggesting to ignore the impedance of the USB traces? what about the ESD?

To do it properly, you should calculate microstrip width/spacing for differential impedance of 90 ohm. And when changing reference planes, you should have a via (or decoupling cap for GND <-> VCC) nearby.

To do it properly, you should calculate microstrip width/spacing for differential impedance of 90 ohm. And when changing reference planes, you should have a via (or decoupling cap for GND <-> VCC) nearby.

Will be great if you can point me any good resource that explain this.

https://www.cypress.com/file/144296/download

https://www.ti.com/lit/an/spraar7h/spraar7h.pdf

1 Like

Thanks, looks like a great resource.

I take this to mean that it doesn’t matter because it’s not high speed. I think that’s true. Of course it would be good to practice high speed signal handling with USB2, but it’s not necessary for functionality. USB3 with higher speed would be a different matter. This design looks better than any of mine, although I have always had serious space restrictions.

D+/- can’t be swapped in the MCU which is intentional because it doesn’t matter if you have to cross them. In real high speed signal pairs (like USB3) the pins are swappable to avoid the need to cross the traces.

Don’t ignore ESD. Give it better GND connection if possible. If it has more impedance for the peak current than the signal path has, it’s useless or at least less effective. Now it has only one narrow track/via for GND connection. But VBUS is connected to somewhere while having no ESD?

Interesting regarding the swapability of USB+/-.
Regarding ESD, nice spot, I increased the GND via and track width, do I have anything else that I can do there?

VBUS is connected straight to a 10BQ040 diode and then to a TLV70233DBVR LDO, which suppose to have some sort of ESD protection, not sure if that is enough though.

If it is 12MHz then yes. I think connection having in total 2…3 cm (including ESD protector) can be done really anyway.

15 years ago I have bought the home piezoelectric gas ignitor. I made internal sparking pin a bit longer and out second end with a wire and I am using it to check ESD resistance of devices. Using it I hang-up the device which passed ESD test in lab so I assume that my test is stronger.
You must assume than a spark from ESD gun traveling along surfaces (like pcb surface) can reach even 15mm length. When I had a small distance between chink (from dictionary not sure) of the plastic case and PCB I made around PCB the thin track (with small gap to not have a full coil) assuming its task is to trap ESD not allowing it to travel through my main GND zone (when ESD travels through GND zone I would expect short (<ns) voltage peaks even higher then 100V between two points at GND as its impedance is not 0). I then connected that track to USB connector GND.
I’m not sure if it is the good solution (didn’t tested it) but it agrees with my intuition.

1 Like

Wrong assumption. You should have ESD protector (transil) at power input from USB.

I just forgot to connect VBUS to USBLC6-2SC6, now it should be ok.

Neat idea, cheaper than the official gear. ESD standards test up to 15kV for air discharge for “extreme” environments. I think we test up to 20kV to give some margin. According to wikipedia, a gas lighter generates 800V, but normal rule of thumb is 1kV per 1mm spark gap, so I would guess your lighter is generating 2kV or so.

However, the ESD standards typically use the “Human Body Model”, which discharges a 200pF cap through 1.5k resistor to simulate a human body. So the gas lighter might be providing more energy with higher rise times than the “standard” test.

I think you have a good point about ESD vs trace length matching. I have seen a lot of people here ask about trace length, for circuits like USB where it is not relevant. OTOH, the real life risk of devices to work reliably is poor ESD protection, and people rarely ask about that.

Certainly much more.
Normally in such gas lighter there is a gap about 1mm between internal pin and … cal it GND around it. So the 800V may be a right value.
I think there are 3 types of spark gas lighters:

  • solenoid powered form AC230 through the contact opened (sparks here) by this solenoid (we had such when I was young),
  • electronic step-up powered by AA batteries,
  • piezoelectric - energy you push piezoceramic material is converted into electricity.

I am speaking about the third one (the second (I have never had in my hands) could probably be damaged if you try to make the spark distance bigger). I made the center pin longer and put on it isolation going in front of surrounding GND.
I think (can check on Monday) that from point where isolation ends to GND is about 7mm. When spark don’t founds the way through something I wont to test it jumps here. During one press I can probably have 3 sparks here while pressing and 3 why depressing. If I put the top in a small distance to something connected to my GND (wire with banana plug from back of gas lighter) I have more sparks while one press and one depress.I can press it frequently having in fact a serie of 20 or more sparks during few seconds.
If it is 7mm I would assume I have 7kV. I could make that distance bigger but probably then the spark will find the way somewhere inside the gas lighter.
I hope description of what I have done is clear. If someone would like to see it I can make a photo but we are going out of KiCad subject.