Moving Toward the Transparent Connector
By Dr. Eric Bogatin
www.BeTheSignal.com
February 2007
The higher the bit rate, the more problems connectors introduce. This article describes two technology options that may offer building blocks that can shift the connector design paradigm from pin-in-box to pad-to-pad and approach an electrically transparent connector.
While a uniform cross-section interconnect maintains a uniform impedance to a signal, any change in cross-section, such as at an interface, introduces an impedance discontinuity and potential reflection noise.
In addition, the less the signal-return path geometry looks like a coax, the more ground bounce voltage is introduced. It’s generally the ground bounce from the connector of a daughtercard that drives common currents that cause EMC failures.
This suggests two important design guidelines for the electrical design of connectors: use a uniform cross-section that matches the target impedance and keep returns completely surrounding signals. A third design guideline, which can also help make the connector electrically transparent, should be the mantra of every interface designer: Shorter is better.
When the return path is not coaxial, there is some total inductance in the return path. The dI/dt through the return path generates ground bounce noise, which is directly proportional to the length of the connector.
The shorter an interface is compared to about 1/10th a wavelength of the highest frequency component, the more transparent it looks. For example, in a polymer dielectric, 1/10th of a wavelength at 10 GHz is about 1.5 mm or 60 mils. At 40 GHz, less than 0.4 mm or 15 mils is short.
The down side of a short interface is the decrease in available compliance when the two uniform transmission lines on opposite ends of the interface are brought together. There is a delicate tradeoff between the mechanical and electrical design.
Two available technology options are potential building blocks to fabricate the next generation of connectors. They are in the class of connectors called interposers, which act as thin, compressible sheets that conduct current only through the sheet, and not laterally, in the plane of the sheet.
When placed between an array of pads, the sheets connect the top pads to the bottom pads. If the pads and their board-level via stack-ups are optimized, the interface can be designed to approach the performance of a transparent connector.
The PariPoser Contactor System from Paricon (http://www.paricon-tech.com/) was originally developed at AT&T Bell Labs more than 10 years ago, and was acquired by Paricon in 1998. The conducting elements are columns of silver-plated nickel balls embedded in a silicone rubber matrix. Four balls make up each column. Figure 1 shows a cross-section of the sheet. The top portion of the figure shows the columns of four balls. The lower-left portion is an illustration of ball columns under compression load. The lower-right portion is a photograph of columns on top of an array of pads on a 1-mm pitch.
Figure 1. PariPoser Interposer Examples
The compliance comes from the elastic properties of the silicone. When compressed, the balls roll over each other, maintaining electrical contact between the top and bottom pads. The interposer acts as an array of very thin columns of wires.
One implementation of this building block in a board-level connector system uses a flex circuit transmission line to connect two daughtercards (refer to Figure 2). “The insertion loss of this connector system was remarkable,” said Roger Weiss, President of Paricon. “It doesn’t matter if you define the connector bandwidth as the -1 dB or -3 dB, the insertion loss of this interface is almost invisible, even up to 30 GHz.”
Figure 2. Board-to-Board Flex Connector Using a PariPoser Sheet Clamped Between Interfaces
The ultimate transparent connector could be constructed by taking an array of semi-rigid micro coax cables, cutting them in half, and reconnecting them with a layer of PariPoser between them. The signal pins connect to signal pins, and the return shields connect to return shields. Weiss says this design may be the future of ultra-high bandwidth, high-density, and low-noise connectors.
A second technology solution for approaching a transparent connector is offered by Particle Interconnect Research and Development Inc. (http://www.pitek.us). Diamond dust particles are occluded in nickel plating and then overplated with gold. The plating can be applied to circuit board pads, flex pads, and even contactor points.
When a gold plated, diamond coated pad is pressed against another pad, it “creates an oxide penetrating, non-wiping, gas-tight connection that allows electrical connection in presence of adhesive, oil, and dirt,” Larry DeFrancesco, president of Particle Interconnect Research and Development Inc. said.
In principle, the connection provided by a diamond particle interconnect pad, “can be as electrically transparent as a solder ball connection,” DeFrancesco goes on to say, “but can be mated and de-mated a million times.”
Figure 3 shows one implementation of an interposer design leveraging diamond particle interconnects. Diamond coated pads extend on the ends of cantilevered conductors, part of a flex strip. The compliance comes partly from the penetration of the sharp tip into the mating pad and partly from the cantilever of the lead. The left side of Figure 3 shows an array of cantilevered strips, the middle shows a close-up view of a single cantilever, and the right side shows a close-up view of the diamond particles embedded in the nickel plating.
Figure 3. Diamond Particle Interconnect Coated Pads in an Interposer
Neither of these technologies are a complete solution by themselves, but they offer possible building blocks for the ultimate bandwidth-transparent connector.
This and other signal integrity topics are covered in Eric’s public classes and online lectures, available from his website, www.BeTheSignal.com. Send your signal integrity technical questions to DoctorIsIn@BeTheSignal.com
Reader Q & A
If you have a question for the SI Doctor, please send it to DoctorIsIn@bethesignal.com.
Question from Tim in Ann Arbor, MI: You’ve mentioned thru-hole vias and stubs and their effects on high-speed serial links, but I don't recall your mentioning blind vias. Are these a better choice than thru-hole for controlling signal effects? Are there any downsides to using blind vias in place of thru-hole vias?
Answer: Blind vias that start on one signal layer and terminate on the final signal layer can be designed with minimal stubs. If the signal transition is from a layer near the top of the board, a back-drilled via, with the bottom stub drilled out, could have just as good performance as a buried via, in which case, it is a matter of cost.
Question from Frank in Santa Cruz, CA: I just got my hands on a 2D field solver and it’s giving me negative capacitance values for the coupling capacitance. Is this an error or numerical noise? If not, what does a negative capacitance mean?
Answer: The negative capacitance in the 2D field solver is perfectly correct, and due to the definition of capacitance used in the field solver. In circuit models, capacitance is defined from a circuit element whose current through it is proportional to the dV/dt across it. The higher the capacitance, the higher the current through it for the same dV/dt across it. In the world of circuit elements, capacitance is always positive.
In a field solver, capacitance, and the capacitance matrix, has a different definition. To distinguish these two different kinds of capacitance, we refer to the capacitance used in a circuit model as the SPICE capacitance, and the capacitance used in a field solver as the Maxwell capacitance.
When calculating the capacitance between any two conductors with a field solver, all the conductors except one are tied to the ground reference point. A voltage is placed on the one floating conductor. Then, the charge on each conductor is calculated. The capacitance between any two conductors is the ratio of the charge on the second one to the voltage on the first one.
Since the voltage on the first conductor is positive, it induces negative charges on all adjacent conductors. With this definition, since the charge is always negative, the Maxwell capacitance is always negative. To convert it to the SPICE capacitance, just change the sign.
Bio: Eric Bogatin is president of Bogatin Enterprises. Many of his papers are available on his website, www.BeTheSignal.com. He is the author of Signal Integrity - Simplified , published by Prentice Hall.
|
Literature
> 
