RS485 stubs

I'm reading an app note from sti (AN1348), and they reinforced something that I have read somewhere else before. They reccomend that the stub that
connects the receiver to the twisted pair should not be longer than 15cm (6 inches). That's fine with me and I can understand why.
Now, I don't know if I'm complying with this reccomendation on my design, here's what I'm planning to do:
A twisted pair leaves the master device and connects (molex connector) to the first slave board, let's call it the input connector. From the input connector the signal is traced to the output connector (less than 1 cm apart) that may be connected to another slave board. At the end of this daisy chain I'm planning to connect a terminator resistance as suggested on the app note.
A/B signals on the slave board connect directly to this "bridge" between connectors.
The trace that connect A/B lines to the "bridge" will be very short, less than 1 inch. Can I consider this 1 inch trace as my stub? Are there any requirements on trace width to keep the 120Ohm impedance?
All distances in my design are very short (no wire exceeds 1 ft).
Cheers,
Padu
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

<snip>
Don't worry about it. The impedance is a function of signal to return distance and trace width, and depends heavily on how many layers your PCB is. Every impedance calc I've used is off by quite a lot when you measure the effective impedance of the actual PCB (traces are ALWAYS too narrow).
The reflections you'll get from the mismatch won't bugger up RS-485 too much, as it's a slow speed connection. How many layers is your board, and how far down is the first ground plane? I'll presume you're using .062" fiberglass substrate. I can give you an estimate, but your mileage WILL vary.
We generally don't even bother to match the impedance on the '485 traces, and don't have any trouble with up to 16 devices. And we use cheap flat cable, not twisted-pair. It's good for around 20 feet cables.
With 1" stubs, you probably won't be able to see the reflection unless you're using a 500MHz scope. Your '485 receivers sure as heck won't respond to it. The chips just aren't that fast.

That's about what we use, somewhere between 4" and a couple of feet between connectors, same ~ 1" connector <> transceiver. Likely not a problem.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
"Serious"

< snip >
That leads to another question. How do you actually measure impedance? On another project I'm having to design a trace that must have specifically 50 Ohms impedance. I've downloaded a utility from agilent that helps you calculate the impedance given certain parameters (signal frequency, pcb material, copper thickness, etc), but I don't know how to measure it.
Cheers,
Padu
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Generally, we use an H-P Vector Network Analyzer. It's not a tool that most folks have access to.
A cheaper method is to hook the system up and inject a high speed square wave into the input, then measure the destination with a properly grounded scope. If the impedance is off, the square wave will ring and overshoot/ undershoot (too low) or be rounded off (too high). When it's correctly matched, it'll be square. Of course, the source impedance that's injecting the signal and the destination must match your desired, or you're just fooling yourself.
I used that once to prove to a competitor that their output was WAY off the mark on output impedance (observed into a resistive load). It gave a very obvious difference on the two 'scope plots, 50 ohm cable & 50 ohm terminator versus 75 ohm cable & 75 ohm terminator. It was a quick and easy to duplicate test.
If your traces are quite short, the observed mis-match may be too small to measure.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
<Serious>

That's exactly the case, the trace is less than 1 inch.
To be more specific, this trace is the antenna connection of a GPS chip (RF_IN pin). On the datasheet they say that it must be 50 ohms (antenna, antenna cable, antenna connector and finally the trace that connects the RF_IN pin to the connector). The GPS is not working right now, and the manufacturer is saying it could be mismatch impedance. I've used the agilent software to calculate the impedance of my trace and it's giving me 70 ohms.
The most amazing thing is that this trace is on a PCB properly designed (eagle) and fabricated by a board house (pcbfabexpress). Before making this board, I had the same gps chip on dirt and cheap home made PCB and it worked fine!
I sent my board to the gps chip manufacturer and so far they didn't find why it's not working... let's see.
Thanks for the explanation though.
Cheers
Padu
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Padu wrote:

How fast are you planning on running the network? And where do you get a 10Mb/s serial port? RF termination issues do matter at 10MHz, but they're unlikely to be an issue down below 100KHz.
If you're running some more traditional serial port rate on RS-485, like 56Kb, you don't have to worry about this. The wavelength is way too long.
                John Nagle
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
"John Nagle"

Probably 115Kp. I don't know where you get idea I have a 10Mb/s serial port, because I don't. I have a USB to RS485 device plugged to my PC (the master in the network), and it says that depending on your cable length you can achieve speeds up to 1Mbps.
Cheers
Padu
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.