50 Ohm Single Ended Signal Impedances on PCBs

Question

I have a basic understanding of transmission line theory and how 50ohm traces are designed on a PCB. I was wondering what the origin of this 50ohm impedance is. Are single ended digital signals on ICs designed per this spec? It wouldn't make much sense to create 50ohm traces on a PCB if the driver and receiver ICs were not designed with these characteristic impedances so as to impedance match. Or is this a non-issue for low frequencies?

Answer 1

A rule of thumb is that characteristic impedance becomes seriously important when your trace length gets above a tenth of a wavelength. Below that you can mostly get away with treating your trace like a capacitor.

50 ohm single ended is a standard mostly used by the radio world. Digital signals that are fast enough to need impedance matching usually use differential signaling.

Answer 2

I was wondering what the origin of this 50ohm impedance is.

This is a historical accident (as so many things are), and as usual in electronics represents a compromise. You can start here. Early coaxial cables (which were rigid copper tubing) used a range of impedances from about 40 ohms (for high power) to 93 ohms. The US (primarily through military) settled on a compromise of 50 ohms, and Europe picked 60, although they later changed their standard to 75. Even so, flexible coax used to be made to quite a variety of impedances within the original range. And 93 ohms has remained the standard for TV antennas in the US.

Are single ended digital signals on ICs designed per this spec? It wouldn't make much sense to create 50ohm traces on a PCB if the driver and receiver ICs were not designed with these characteristic impedances so as to impedance match.

In the 60s, 70s and 80s, 50 ohm logic drivers were few and far between, due to the need to source 3 volts or so into 50 ohms. This results in a source current of 60 mA, and this was not easily done. In the TTL world, only a few ICs, such as the 74128, were rated for 50 ohms. Much more common was the use of twisted pair with a nominal impedance of 132 ohms. Load termination was a 220/330 ohm voltage divider connected between 5 volts and ground, with the signal source driven by an open-collector to ground, which would provide a 3 volt 1 and a 0.4 volt or so 0.

Or is this a non-issue for low frequencies?

Yup. TTL logic had rise times on the order of 10 nsec. At distances of a foot or two, which is a very large pcb, there is very little problem with ringing. So trace impedance matching simply wasn't a big deal for most logic systems.

The exception to this was the use of ECL systems, such as the old Cray supercomuputer used. This had faster edges, so trace lengths were an issue. Fortunately, ECL used higher currents to get its speed, AND it was well-suited for differential signalling.

Of course, those pesky IC designers insisted on making chips that got faster and faster, so trace impedance became an issue. Fortunately, the same technology which produced higher speeds also used lower voltage swings,and at the same time MOS let output transistors provide more current than the old bipolar technology, so the issue became more manageable.

As Peter Green has answered, though, the use of differential signalling for high speed signals also became more widespread, since this allows easier rejection of noise spikes induced by all the high-speed edges running around digital systems.

Answer 3

Another perspective is if the rise time is < the propagation delay of the path then ringing will become apparent without a matched impedance. The frequency of the ringing lowers with longer lengths.

Even a scope probe is unmatched with a 20cm inductive ground lead, so false capture ringing occurs on fast transients with ringing > 20MHz or pulses with < 50ns rise time.

The origin of the value 50 Ohms and 75 Ohms and 300 Ohms and 600 Ohm standards has to do with the distributed RLC and gap spacings of conductors and insulation.