Design007 Magazine


Issue link:

Contents of this Issue


Page 43 of 119

44 DESIGN007 MAGAZINE I OCTOBER 2019 We know that in the signal integrity world, reflections are usually bad. In clock networks, reflection glitches may cause multiple and false clock triggering. In medium-speed digi- tal signaling, reflections will reduce noise mar- gin, and in high-speed serializer/deserializer (SerDes) signaling, reflections increase jitter and create vertical eye closure. Reflections happen along an interconnect at any point where the impedance environment around the electromagnetic wave changes. Fig- ure 1 illustrates this with a simple example us- ing a uniform stretch of transmission line with Z 01 characteristic impedance between Z 0 refer- ence impedance connections. The formulas shown in Figure 1 for the G voltage reflection coefficient are generic and express the complex ratio of reflected and in- cident waves. We can apply the formula to steady-state impedances—something we could measure with a vector network analyzer—or How Much Signal Do We Lose Due to Reflections? to transient impedances, which would be the case when we use time-domain reflectometry. In general, the impedances that go into the for- mula—and as a result, the voltage reflection coefficient itself as well—are complex num- bers with magnitude and phase or real and imaginary parts. Another generic characteristic is that the di- rection of the arrow at the end of the red line has a significance. In the nominator of the volt- age reflection formula, the first term is the im- pedance the wave will enter into by crossing the boundary, and the second term is the im- pedance the wave is coming from. This means that if we calculate the voltage reflection coef- ficient at the same boundary but going the op- posite direction, the sign of the voltage reflec- tion coefficient will change while the magni- tude stays the same. As a simple example, let's assume that we have a lossless transmission line with a Z 01 = 45-ohm characteristic impedance and look at it between Z 0 = 50-ohm reference impedances. This repre- sents the lower bound of a ±10% impedance tolerance for a 50-ohm trace. With all impedances being real numbers in this simple example, the voltage reflection coefficient is also real with a value of G 1 = -1/19 and G 2 = 1/19, or approximately ±5%. Based on the 5% reflection magni- tude, we may expect that 95% of the launched signal will continue after the reflection. To test this assump- tion, we can do a very simple sim- ulation. Figure 2 shows the circuit Quiet power by Istvan Novak, SAMTEC Figure 1: Definition of voltage reflection coefficient.

Articles in this issue

Archives of this issue

view archives of Design007 Magazine - Design007-Oct2019