Upon reading your book "High Speed Digital Design", I was disappointed to find that a "new", and increasingly popular termination technique, using Schottky-diodes to Vcc and GND, is not even mentioned. I've read MANY conflicting reports on the relative merits of various termination methods, including diode clamping.
For example, here's an excerpt from Mai Vu's (Harris Computer Systems) EDN article:
"For memory signals, you should use the clamping diode and the two-thirds rule as your first choice and the series-termination concept second. Clamping diodes typically provide a cleaner signal than series termination, especially when you use them at the first and last loads."
Question: Is there any technical basis in making this conclusion (that diodes provide a "cleaner" signal)?
From another EDN article, written by Calif. Micro Devices (makers of Schottky termination products) employees:
"Schottky-diode termination works well in a multidrop situation in which some of the receivers on the line can also drive the line."
Question: This would seem to be the case for memory architectures in particular, where BOTH the processor AND the RAM chips can drive the bus (on write and read cycles, respectively). How would SERIES termination (a SOURCE termination technique) be applied in this case (where the SOURCE can be at either end of the line)? What benefits would diode clamping provide in this case?
Finally, can you comment on the "termination" function of clamping diodes. As I understand it, they do NOT function as true terminators (since they dissipate no energy), but rather function only to limit the amplitude (by CLAMPING to Vcc or GND plus/minus the diode's forward voltage drop) of the propagating reflections. Reflections only die out due to parasitic resistance/capacitance on the line, not due to any diode function. How does the coupling of the over/undershoot energy to the power and ground planes affect the rest of the system?
Any clarification you can provide would be greatly appreciated.
Thanks for your interest in High-Speed Digital Design.
One equation of interest regarding the theory of diode terminations is the relation between the incident signal amplitude traveling down a transmission line (I), the amplitude of the signal reflected at the end of a line (R), and the amplitude of the signal that exits the line and is apparent at the load (T). The relation is:
T = I + R
In English, the signal at the end of the line is the superposition of whatever came in, plus whatever bounced back toward the driver. This same equation may be re-written to express "R" as a function of "T" and "I":
R = T - I
In this form you might recognize that, if the diode perfectly clamps the output (on a rising edge this means T=Vcc), and if the incoming signal amplitude "I" is exactly full-sized (I=Vcc), then there would be zero reflection. That is, a perfect diode, with a perfect input signal, works fine. You don't need to match the transmission line with a resistor. You only need to make sure that the terminator draws just enough current from the line to prevent overshoot, at which point there will be zero reflections.
Next let's look at the practical implementation of a resistive-terminated system, and then compare that to a diode-terminated system.
With a resistive terminator, the size of the output signal is proportional to the size of the incoming signal--k equals one, therefore (T=kxI). If we size the resistance just right, the "T" and "I" signals will be exactly equal (T=I). With a proper resistive termination this relation holds regardless of the incoming signal amplitude. Furthermore, when T=I, we know that R = 0. In other words, the resistive termination works for all incoming signal amplitudes. It dampens the first incoming signal edge, and then, if anything happens to reflect back to the source and return to the endpoint, the resistive terminator will dampen that, too. On a resistive-terminated line, the residual reflections decay exponentially with time.
With a diode-terminator, the "T" signal is the same regardless of the incoming signal amplitude. If the "I" signal happens to equal "T", you get no reflection. If the "I" signal falls short of "T", you get a noticeable positive reflection. This means the diode termination will dampen the first incoming signal edge (if that edge is full-sized), but after that, whatever signal happens to reflect back to the source and return to the endpoint is never again large enough to activate the clamp. Residual reflections are essentially unterminated. On a diode-terminated line, the first reflection is attenuated, but the residual reflections do not decay exponentially over time (except to the extent that they are slowly absorbed by the driver).
Two other things to keep in mind about diode terminations are that the diode must be very fast compared to the rising edge time to have any effect, and that the diode must clamp to a level offset from Vcc (and Gnd) by the forward voltage drop of the diode.
I am skeptical about the use of diode terminations for fast digital signals, *especially* so if the diodes share common power or ground connection within their package. For more information about problems with the common power rail inside the package, see the section in my book about SIP resistors. The same principles apply.
I hope this explanation has highlighted for you the shortcomings of diode terminations for PCB applications.
In slower applications, like SCSI, diode terminations are great, because (1) the signals are intentionally slowed down to meet radiated emissions requirements, so the diodes are naturally faster than the signals, (2) we can afford to build fancy clamping-voltage generators at Vcc-Vf(diode) and Gnd+Vf(diode), and (3) we can specify receivers with tight V(IH), V(IL) margins that are tolerant of the lingering residual reflections.
Next time I'll talk about the other part of your question, bi-directional connections.
Dr. Howard Johnson