“Nickel is probably the most mysterious metal widely used in electronics,” is the comment from Yuriy Shlepnev of Simberian Software, my electromagnetics guru, and Scott McMorrow, of Teraspeed Consulting, my high speed measurement guru. They write this in a paper they presented at the 2011 IEEE EMC Symposium.
In their paper, “Nickel Characterization for Interconnect Analysis,” they report on measuring the insertion loss from electroless nickel immersion gold (ENIG) plated copper microstrips and building a simple model based on the electrical properties of the nickel plating to describe what they measure. You can download a pdf copy of the slides from Yuriy’s site, but must register as a member to Yuriy’s site to download a pdf copy of the written paper. Both are free.
It’s important to note that the skin depth of gold is about 2.4u at 1 GHz and scales lower with the square root of frequency. At 3 GHz, it is 1.4u. If the gold plating is at least 2u, little current flows in the underlying Nickel and it plays little role on the electrical performance of the interconnect.
However if only immersion gold is used, its typical thickness may be only on the order of 0.05u, in which case, current flows in the nickel layer, even above 100 GHz and the nickel’s conductivity and permeability affect signal propagation.
Predicting the impact on the signal propagation from the ENIG layer depends on these Ni material properties. This was the purpose of their paper, to extract the Nickel properties.
Yuriy and Scott did an excellent job of reviewing what has been reported in the literature about Nickel. They show that historically, a wide range of values for the permeability of nickel have been reported, based on trying to fit measured data.
They say that this is in part due to the dependence on the nickel properties from the phosphorous concentration and the specifics of the plating process. This means measuring one microstrip sample is no guarantee that all ENIG plated microstrips will behave this way.
Even taking the process variation into account, there is still an anomalous behavior in the losses and propagation delay of microstrips with nickel plating which cannot be explained using a simple value of permeability or resistivity for the nickel.
|
|
|
From Shlepnev and McMorrow’s paper, comparing the measured insertion loss between a stripline transmission line with no ENIG and a microstrip transmission line with ENIG plating. An anomaly also shows up in the group delay of the insertion loss.
|
However, by applying a model, first introduce by Landau and Lifshits in 1935, they are able to match the measured performance pretty well. This assumes a resonance in the magnetization between layers in the nickel plating. This model for the nickel properties has just five parameters that fully describe it.
For the samples of ENIG plated microstrip they measured, they find a good match for values of these terms as:
|
Low frequency permeability
|
6
|
|
High frequency permeability
|
2
|
|
Resonant frequency, fo
|
2.5 GHz
|
|
Conductivity
|
1.55e7 S/m, 27% of copper
|
|
Damping term
|
0.22 x f0
|
Using these values for the Nickel, and the actual cross section of the microstrip, Yuriy and Scott are able to accurately predict the insertion loss and group delay from the currents flowing in the nickel layer.
|
|
|
From Shlepnev and McMorrow’s paper, comparing the measured insertion loss and group delay with the calculated values based on the Landau and Liftshits’ model for Nickel’s complex permeability, matching the anomalous behavior. The large noise in the measured value at high frequency is due to the non identical launches used in the fixture to the lines.
|
The insertion loss is increased by the nickel plating. This is due to the lower conductivity of the nickel and by the higher permeability which decreases the skin depth, further increasing the resistance. The anomalous dip at 2.5 GHz is due to the higher losses in the nickel at the resonance.
The group delay is a little bit higher at low frequency, where the permeability is about 6, and flattens out above the resonance where the permeability is only about 2. This translates to the perception of a slightly higher effective dielectric constant which slows the signal down and increases the group delay.
Of course, the impact on the group delay is only due to the small fraction of the internal magnetic fields inside the bulk nickel which sees this higher permeability. The nickel’s effect on group velocity decreases quickly above the resonance at 2.5 GHz.
This paper ties up effects which have puzzled anyone who has looked at ENIG microstrips and tried to figure out how to think about the impact of the Nickel.
The nickel will increase the loss over pure copper or thicker gold, and will increase the effective dielectric constant.
