Resistor and Ferrite on signal lines

[docmur]

Hello Everyone

So I was taking a PCB Design course a few years ago and my prof said something that I haven't forgotten about:

1) Put Ferrite Beads on all lines that leave or enter the board through connectors
2) Put 33 Ohm resistors on signal / data lines.

I understand why to use Ferrite but has anyone heard about the resistor thing?  I'm a little confused about what it will do. 

[Andy Watson]

It's an acknowledgement that the length of the signal path is becoming significant with respect to the wave length of the frequencies in use. To reduce reflections and generally improve signal quality the data lines should be treated as a transmission line and matched impedances used throughout. Although the lines are unlikely to be 33 ohms, it is better than zero ohms, and not so great as to significantly affect signal level.

[mmagin]

These sound like good rules of thumb for a certain area of design (impedances and bandwidth), but hopelessly vague for general use.

I've heard of resistors in the output of high speed logic at the start of a transmission line to reduce reflections, but in general, the proper transmission line parameters and termination depend on the logic family.

[AndyC_772]

33 Ohms (plus the output impedance of the logic driver) is a pretty good starting point in many cases, and certainly better than zero. I sometimes use 22 if I need very fast edges, or change them to 47 after a bad day at the EMC lab.

[Andreas]

The resistor will work good if there is only 1 driver and 1 receiver on the line and if the resistor is placed near the driver.

With long lines and multiple receivers the signal integrity for the receivers near the driver is compromised.
In this case the output voltage near the driver stays in the "forbidden range" until the reflected wave returns from the other end.

So the series termination is only a good idea for point to point connections.

With best regards

Andreas

[AndyC_772]

If you have one driver and multiple receivers, then layout topology becomes important.

You can still get a valid signal at every receiver if you route a separate trace from the driver to each receiver individually, and put a separate series terminator at the start of each trace. So, you end up with a group of resistors all close together next to the driver, and separate traces from each of those resistors to the loads.

If you want to route a single trace from the driver to each receiver in turn, then you'll need to consider parallel termination instead. With parallel termination, a resistor is placed between the input to the last receiver in the chain and GND, and it's this resistor that absorbs the incoming wave and stops it being reflected back. It's arguably a better topology for things like backplanes which can have many receivers on a single net, and if you put a terminating resistor at each end of the trace then you can also drive it from any point on the trace and still get a good signal at each receiver. The downside is the dc current that the driver needs to supply.

[ElectroIrradiator]

(...) The downside is the dc current that the driver needs to supply.

This problem can partially be circumvented by using a series RC combination for the terminator. The series capacitor blocks the DC load on the driver(s), yet is dimensioned to have a low impedance at the termination frequencies of interest.

[AndyC_772]

Yes, ac termination can work, though you then have to consider that each trace is a resonant L-C circuit, which probably isn't what you want. I'm not sure I've ever actually seen this technique used in practice.

[SeanB]

What about the old fashioned method used on SCSI buses of a series resistor connected to a termination voltage, or you can use the common and standard resistor arrays that had a series connected 220r and 330R with the centre tap being used as the line, with the 330R being 5V and the 220R being grounded.

[dannyf]

has anyone heard about the resistor thing?

Both are simple signal conditioning methods: the beads are easy to understand. The resistors essentially form a  low-pass filter with the capacitance on the input pin.

[Andreas]

Yes, ac termination can work, though you then have to consider that each trace is a resonant L-C circuit, which probably isn't what you want. I'm not sure I've ever actually seen this technique used in practice.

I have seen several systems working with A/C termination. It is especially useful when using CMOS drivers to save energy. If you want to do a optimized design you have to adapt the capacitor according to the transmission line length.

The resistors essentially form a  low-pass filter with the capacitance on the input pin.

This is only true for short line lengths. For electrical long line lengths the transmission line behaves ohmic.

With best regards

Andreas