Eric Bogatin's Blog: What I Learned This Month

Check out our next public classes: Essential Principles of Signal Integrity and Advanced Signal Integrity Design, Oct 11-14 in Hillsboro, OR.

Check out our next No Myths Allowed Webinar, “Stack-up Design for Differential Pairs”, presented free on Sept  16, 1 pm EDT.

The design process is a creative process and intuition is the most important skill you rely on first. Once you have a design, then you apply your analysis skills to evaluate the cost-performance tradeoffs.

Wikipedia defines intuition as the “ability to sense or know immediately without reasoning”. I like to think of intuition as what you take away after you have done the reasoning- what has become ingrained in your understanding and you trust to use every day. It becomes your “common sense” of the right way of doing things.

The better your intuition or common sense about signal integrity, the better your first pass design will be. This translates to a shorter and lower cost design process.

Because signal integrity is sometimes “anti-intuitive”, we sometimes have to “re-calibrate” our intuition for these analog electromagnetic effects important in interconnects. The Essential Principles class I teach, is really about building a strong foundation for your common sense about signal integrity. Here in an abbreviated list, are 13 of the most important principles to strengthen your common sense when dealing with signal integrity:

1.    The fastest way of fixing a problem is by first identifying its root cause and then applying the Youngman Principle.

2.    All interconnects are transmission lines, no matter how long or how short they are.

3.    All signals are dynamic and constantly move along the transmission line at the speed of light in the dielectric, roughly 6 inches/nsec.

4.    Signals sees an instantaneous impedance each step they takes along a transmission line.

5.    The return current is exactly coincident with the signal current, flowing in the opposite direction, in the return conductor.

6.    Reflections occur whenever the instantaneous impedance changes.

7.    Don’t confuse the distributed cross talk between transmission lines which rarely extends beyond adjacent lines, with ground bounce cross talk which can extend to many adjacent transmission lines.

8.    Ground bounce occurs due to the dI/dt of the return current passing through the total inductance of the return path.

9.    The differential impedance of a differential pair can be just as well controlled for a tightly coupled as loosely coupled pair.

10.    A real capacitor behaves like a series combination of ideal RLC elements even up to the GHz bandwidth.

11.    Always try to place power and ground planes on adjacent layers with thin dielectric between them.

12.    Use SPICE to simulate the parallel resonances when multiple capacitors and the power and ground planes are connected in parallel.

13.    Assign return path layers and signal routing in the stack up based on the ability to provide a return via whenever a signal via changes layers.

If you want to learn more about common sense signal integrity, take our class, Essential Principles of Signal Integrity, attend one of our webinars or visit our web site, www.beTheSignal.com.

Check out our next public classes: Essential Principles of Signal Integrity and Advanced Signal Integrity Design, Oct 11-14 in Hillsboro, OR.

Check out our next No Myths Allowed Webinar, “Stack-up Design for Differential Pairs”, presented free on Sept  16, 1 pm EDT.

A lossy transmission line screws up digital signals. A clean bit stream going into an interconnect can be jumbled and distorted by the time it comes out. This is all due to the frequency dependence of the line parameters: the resistance, capacitance, inductance and conductance per length. The spectrum of the input waveform is distorted by the interconnect.

To work around this problem, you do what you can afford in the interconnect design and materials selection. In addition, a revolutionary solution is to perform signal processing on the signal’s spectral content. This is implemented with pre- or de-emphasis and equalization. These techniques condition the spectrum of the signal to compensate for the spectral degradation by the interconnect. (These and similar topics are covered in our new class, Multi Giga Bit Design (MGBD))

It has long been known that there is one special waveform that is transmitted through an interconnect undistorted: a sine wave. Send a sine wave into a lossy interconnect and you get an identical sine wave coming out, though with an amplitude and phase change.

Sine waves are not distorted when transmitted through a lossy line.  But sine waves are not the only waveform that can propagate undistorted on a lossy line.

In 2002, Dr. Robert Flake, a professor in the Department of Electrical and Computer Engineering at Univ of Texas, Austin, realized there was another waveform that would propagate through a frequency dependent, lossy interconnect undistorted: an increasing exponential wave. He calls these waveforms, Speedy Delivery (SD). I’ve had the opportunity to meet with Prof Flake and discuss this intriguing effect.

Of course, no wave is going to increase exponentially forever, so it will always be truncated at some voltage level. This truncated exponential wave, when traveling through a lossy transmission line, will maintain its exponential shape. It is a property of the differential equation that describes the lossy line and the derivative properties of the exponential function.

Prof Flake generates his SD, truncated exponential pulses using an arbitrary waveform generator (AWG). He has sent pulses through 18,000 foot, cheap, twisted pair lossy transmission lines and found an exponential edge coming out, though attenuated.

The figure to the left is an example of a truncated exponential pulse as it enters a 200 m long RG58 cable (red), and then overlaid on the measured waveform as it comes out (black). The exponential shape is perfectly preserved, just shifted in time, with the non exponential part of the waveform attenuated and distorted.

Since the shape is preserved, the time delay of the edge through the interconnect can be very accurately determined. The fundamental time delay accuracy is related to the jitter in the AWG. Prof Flake has already demonstrated 2 psec accuracy in a time delay measurement over a 200 m lossy line.

This is an incredible observation. While the obvious applications are for accurate TDR measurements with lossy interconnects, such as long cable spools, heater tapes and power lines,  Prof Flake suggests this unique feature of exponential waveforms not being distorted by lossy interconnects may be useful in on-die applications or even lossy backplanes. It potentially could become as useful in overcoming the degradation from losses, in the future, as pre-emphasis and equalization are today.

Check out our next public classes: Essential Principles of Signal Integrity and Advanced Signal Integrity Design, Oct 11-14 in Hillsboro, OR.

Check out our next No Myths Allowed Webinar, “Link Analysis with Return Path Discontinuities”, presented free on July 7, 1 pm EDT.

The pop quiz this month was, “When two adjacent signal lines transition from one signal layer, through a pair of planes, to another signal layer, the return current flows between the cavity formed by the planes. The impact of the return path discontinuity is strongest on which S-parameter term.”

At 28%, the consensus was S11, followed by SDD11 at 23%. This is surprising, as neither answer is correct. This topic is covered in detail in the SI-Insight for June, 2009, released this month, “Ground bounce in Vias.” We also touch on this topic in this month’s webinar, NMA-820, “Link Analysis with Return Path Discontinuities,” which, if you missed the free live event on July 7, can be viewed from the recording on the web site.

Which quality of the pair of lines is most affected by the return path discontinuity? The impedance discontinuity, which S11 is most sensitive to, is minor. And, as we show in the paper and the webinar, the differential signal is mostly immune to this return path discontinuity.

In fact, the biggest impact is on the single ended cross talk between the two lines. The correct answer to this month’s pop quiz is S31, the near end cross talk. Especially when the adjacent lines are far apart, the edge coupled cross talk can be very small, but the cross talk between the signal vias that pass through the cavity can be very large and long range.

The figure to the left, Figure 19 taken from the SI-Insights report, shows the measured near end cross talk between two signal lines, far apart for three design cases. The red measured response is with no signal vias, just two 50 Ohm lines about 15 lines widths apart. The green response is the measured near end crosstalk between these two lines routed from the top layer to the bottom layer, going through vias, with adjacent return vias.

The blue trace is the measured response of similar via transitions, but without the return vias. The coupling between the signal lines from noise injected in the cavity is more than 20 dB higher than with either no vias or vias with adjacent return vias.

Feel up for the challenge of a new pop quiz? Check out the new pop quiz on the web site.