This is my first time laying out ethernet and USB. The point of the board is to provide external connections from internal ports via a cable harness. The board is a standard 1.6mm 2-layer without impedance control . I’ve tuned the lengths of each pair and used the exact widths/separation from the calculator. The external USB connector shield is connected to ground. The longest USB pair is ~40mm while the longest ethernet pair is ~35mm. Solid ground plane on layer 2.
The USB cables will be off-the-shelf pigtails with a short <25cm extension. The extension will be a 10 conductor cable for all the USB and 12V power. The ethernet cables/connectors will be standard non-shielded.
I hear USB controllers are very tolerant and I’m only looking to use low speed serial-USB converters anyway. Ethernet is 10/100 only unshielded. Thoughts?
If you only use Base 100 Ethernet and low speed USB (which is only 1.5 Mbaud/s) i don’t see a significant problem for signal integrity.
3 Minor changes i would made:
The USB pairs have probable a gab between D+ and D-. This is not ideal. How long is that cable? What kind of Cable? Changing it to 2 connects that are next to each other in the cable would be better. (If i am correct that there is a gap).
You could route one of the Ethernet pairs, the lower on in the picture, in a different way. In a way that one end goes through the front and the other to the back. This way they will have a similar length without length tuning. And you could route it a bit better so they are longer closer together.
Add a GND plane on the top. Not that important since you only have connectors.
But it should work without that changes.
What i don’t see:
The track sizes and clearance in between.
You shematic. I don’t know if the pinning is correct.
I made a mockup of the interconnect cable and a USB flash drive worked even through terminal blocks and partially unshielded cable segments using just ordinary 4 conductor cable for the shorter segment (see wiring diagram below).
Not sure what you mean by “gap”. Both USB differential pairs are beside each other on the 10pin connector. I will be using shielded USB cable throughout the cable harness. Although the shield will not be continuous because of the 10pin connectors.
Routing it that way only makes the Rx pair about the same total length as the Tx pair. It does nothing to make Rx- naturally longer or Rx+ shorter and the total length is only about 3mm less than before. Unless I don’t have enough imagination which is possible. Still it seems a better route.
Not sure about that myself. But from what I have read just having the solid ground plane underneath is sufficient and possibly preferred.
The impedance calculations are in the pictures in the original post. If I understand correctly the calculations are based on a standard 2 layer board with a solid ground plane underneath.
The 0.1525mm separation is just slightly above the minimum for most PCB manufacturers and keeps the cost down. Increasing the separation increases the trace width.
I just completed the cable harness wiring diagram. The 10pin connectors are smaller 3.0 pitch versions of the Molex type connectors you find in a PC. For reasons the connector has to be as small as practical. Thus why I can’t make the shields continuous. [EDIT] I originally intended that the short extension be an ordinary 10 conductor cable. But I decided to use individual cables. So the USB differential pairs will be continuous except for the shields. The longer segment is 48" so I though that the cable assembler could just use a 6’ pigtail cut to length and use the excess for the shorter segment. Also, one has to assume at least a 36" cable will be plugged into the external USB ports as well.
I should mention that the enclosure is metal. The USB receptacle is shielded whereas the Ethernet is not. I’ve connected the USB shield and mounting holes to ground which will connect to the case via the screws. Thanks for your comments. I feel this is a good starting point for the first version anyway.
You use pin 10 and 8. Isn’t there a the wire for pin 9 in between?
For the RX_D2 pair, see attached image. This way you don’t need much length correction and the tracks from the pad to the differential pair / start of the other track is shorter.
Not sure about that myself. But from what I have read just having the solid ground plane underneath is sufficient and possibly preferred.
For 2 layer boards, they have a larger gap. So a second power plane is better. High frequency AC signal look for the closest return path, and if this is a track that is nearer than GND, this track is used.
The longer segment is 48" so I though that the cable assembler could just use a 6’ pigtail cut to length and use the excess for the shorter segment. Also, one has to assume at least a 36" cable will be plugged into the external USB ports as well.
If you would use metric units, it would be much easier to understand and be much more professional.
The USB receptacle is shielded whereas the Ethernet is not
What do you have for requirements? We, using a 3m Ethernet cable go outside the machine, had problems shown during EMC testing as they injected a high pulse into the Ethernet cable via capacity coupling. After we connected the Ethernet port shield to earth it got away.
“If you would use metric units, it would be much easier to understand and be much more professional.”
I guess that would depend on which country you are in! Whilst Europe is 100% metric, the UK is half-way, and the Americans have no idea about metres or millimetres! I have NO problems understanding 48" or 6’ (I’m in England!)
(and if you buy stuff, did you ever wonder why cables are sometimes “1.8 metres”? Yup, 6 feet!)
And a football net is 24 ft. wide and 8 ft. high, and aeroplane altitude is always in feet and DIP footprints are Imperial decimal: 16 pin DIP is .7 in X .3 in.
Engineers should be versatile and willing to adapt.
Oh, and not to forget, the foliage span on the tree in my avatar is two chains… the span of two cricket pitches!
It is still more professional to use metric units, no matter where you are located.
And no, most 1.8m cables are not 6 foot, even when they are labelled with both, they are 1.8m most of the time. Besides that, it is still unprofessional to use British imperial units, no matter who is doing it.
aeroplane altitude is always in feet and DIP footprints are Imperial decimal: 16 pin DIP is .7 in X .3 in.
This wasn’t always the case. aeroplane industry was using metric, but some people where to stupid and we are not stuck with this mess.
Engineers should be versatile and willing to adapt.
No. I won’t to it. It is stupid to not use metric and i refuse to use anything else. If you want your problems be solved, use metric or ask someone else to help you or work for you.
( i would be willing to adopt the Planck System. But that is not the topic at hand).
I’m not arguing which is better. I have an open mind and had to learn and use both in my youth. My country converted from Imperial to Metric in my teenage years so I am familiar with both and use both.
If the footprint for a transformer or IC is obviously Imperial, I use an Imperial Grid. If the footprint for a modern IC or transformer is Metric, I use a metric grid.
Rarely in my working career would I use either Metric or Imperial exclusively on a PCB, usually I would be swapping from one to the other, as the component required, to make life easier for me.
A good example are the Arduino boards.
Why would I bother trying to multiply 2.54 x 18 Metric for socket length when I can use .1 x 18 Imperial for the same result. Which is easier???
Kicad, like all good programs, supply both measuring systems, so that is what I use… both.
“British imperial units”
They are called “United States customary units” in America and are (generally) the same thing. The problem is that a LOT of stuff gets designed in America.
“It is stupid to not use metric”
If you are in America, I suggest you would be “stupid to use metric”!
In UK, Canada too, we end up having to use both. At a pub, draught beer comes in pints, but in a bottle or can it’s in ml!
Having said all that, it’s nice to see such arguments coming from a European, rather than an American!!
It took me some time to find what ‘hos’ unit is.
In written in 2014 TPS54061 datasheet Error amp gm was specified as having a value 108 μMhos. I have read it as micro-Mega (so canceling each other) hos.
It’s weird!
I was brought up with both systems, but I still have a couple of hang-ups:
A persons height, I still have to convert to ft. & ins. not cms.
A persons weight: stone & lbs. not Kgs.
Car fuel economy: MpG.(and not US gallons, real gallons ) not Lp100Km.
I have NO problems understanding 48" or 6’ (I’m in England!)
) not Lp100Km. 