Insertion loss and return loss: never heard of them? Meet the new universal characterization metrics.

HOW WOULD YOU describe the high-speed performance of an interconnect (backplane, connector, package or even a test socket) in a way that yields a good first-order estimate of how the interconnect might distort high-speed signals, or whether it will even be suitable for an application? Short of using a complex circuit model and running a simulation, what would you do?

The answer lies in the insertion loss and return loss of the interconnect. These two characteristics are becoming the de facto industry standards to describe the high-speed behavior of any interconnect. The return loss provides information about the impedance match of the interconnect to a 50[OMEGA] system, while the insertion loss provides information about the signal quality of the transmitted signal and the bandwidth of the interconnect. The bandwidth, of course, is a rough indication of the highest data rate that can be transmitted through the interconnect.

Both terms relate to how a sine wave voltage signal interacts with the interconnect. If we know how sine waves, over a wide frequency range, interact with the interconnect, we can also determine how time domain signals will interact with the interconnect. Both entities can be measured or simulated for a great many types of interconnects.

When we send a signal through an intcerconnect--i.e., the complete channel path through an IC package, daughter card, connector, backplane, connector, daughter card, package and back to a receiver chip--any impedance discontinuity along the way will cause some of the signal to reflect back to the source. The return loss is a measure of the ratio of the reflected voltage amplitude of each sine wave to the amplitude that is incident to the front of the interconnect.

The signal quality transmitted through the interconnect and picked up at the receiver will be affected by the impedance discontinuities and by the series resistance losses and the losses from the dissipation factor of the dielectric materials. Unfortunately, due to fundamental properties of materials, higher frequencies will generally be attenuated more than low frequencies. The insertion loss is a direct measure of the ratio of the transmitted amplitude of each frequency component to the incident amplitude at the front of the interconnect.

We usually measure insertion loss and return loss in units of dB. This is a great source of confusion. A small amount of reflected amplitude, an indication of a good impedance match, would be a large, negative number in dB. An exceptionally well matched interconnect would have a return loss of -40 dB. If there is a 50[OMEGA] impedance mismatch somewhere in the system, out of 50[OMEGA], the reflected signal amplitude would be about 5%. This corresponds to a return loss in dB of -25 dB. A marginally acceptable return loss, especially at high frequencies, is typically about -15 dB. An open, a really bad interconnect, would be 0 dB.

A great interconnect is transparent to signals. When we insert the interconnect into the signal path, we should not lose any signal. The ratio of the amplitude of the signal at the end of the interconnect to the incident signal should be 1. In dB, a ratio of the amplitudes of 1 corresponds to 0 dB. As attenuation increases, the value of the insertion loss, in dB gets to be a larger, negative number.

As a rough rule of thumb, we usually use the frequency when the insertion loss is down to -3 dB as a measure of the bandwidth of the interconnect. This corresponds to a received amplitude that is 70% of the incident amplitude. With SerDes chips, it is possible to use an interconnect with a total insertion loss in the complete path as low as -15 dB. The frequency at which these values are reached is a rough measure of the highest frequency that the interconnect will transmit, and a rough estimate of the highest acceptable data rate. In FIGURE 1, we see the measured return loss and insertion loss of a 12" length microstrip trace in FR-4, with SMA connectors on each end. At low frequency, the return loss is about--40 dB, a very good impedance match to 50[OMEGA]. But, above about 2 GHz, the return loss is typically above -15 dB, indicating a poor impedance match.

[FIGURE 1 OMITTED]

The insertion loss at low frequency is about 0 dB, meaning that the entire signal is getting through. However, the -3 dB point for the insertion loss is about 2 GHz. Using a SerDes chip set, a usable bandwidth for this 12" section of FR-4 with SMA connectors might be as high as 5 GHz.

Return and insertion loss can be used to characterize an interconnect and provide an immediate estimate of the performance of the interconnect. It can also be used to &fine a performance specification. Expect these two terms to become the primary standard for describing interconnect performance.

REFERENCES

The principle covered in this column is reviewed in detail in Signal Integrity--Simplified (Prentice Hall) and online at www.BogEnt.com.

ERIC BOGATIN is the CTO of IDI, a high-volume manufacturer of high performance interconnects. He can be reached at [e-mail].