I. Introduction

As the electronics industry transitions to lead free assembly, PCBs will need to endure increased mechanical stresses due to the elevated reflow temperatures of Pb-free solders. This inherent change in process parameters has led some companies to examine the reliability relationship between laminate pre-preg and oxide alternative. It is important to note that historically it was assumed that “rougher isbetter” when it came to adhesive bond promotion. The thought being that the more surface roughness on the copper surfaces the better adhesion will be. In some cases this is true. In many other cases, however, rougher copper is not necessarily a prerequisite for good adhesion and may actually have a deleterious effect on electrical signals.

As PCB signal speeds increase, the signal’s current path switches from the path of least impedance to the path of least inductance. This path of least inductance is what is commonly known as the “skin effect” whereby an AC signal primarily flows nearer to the outside surfaces of a copper conductor rather than through the entire cross section of the conductor. As frequencies increase, the skin effect becomes more pronounced. Therefore it is very important to understand not only the mechanical reliability aspects of a given PCB design but also theelectrical performance criteria it must function under.

It is well known that the roughness of the surface of copper conductors can negatively influence electrical signal propagation at high frequencies. The experiment and results described herein quantify the electrical effects an individual OA may have on signal attenuation as well as an examination of the measurement method itself.

PCBs were fabricated using a standardized test vehicle using two different oxide alternatives. The electrical characteristics were then measured using the Short Pulse Propagation (SPP) test method [1]. The conclusions are based on these characteristics.

II. Adhesion Promotion Using OA:

Figure 1. Typical OA Surface Topography

Lamination adhesion promoters, whether traditional oxide or OA, act as an intermediate layer between the copper surfaces of an inner layer and the b-stage pre-preg resin. Without this intermediate layer, the resin used to bond a multilayer PCB together would not reliably adhere to the bare copper surfaces.

In the case of an OA, an organo-metallic layer is formed through a complex chemical reaction between the copper surface and the process chemistry. This chemical reaction results in a thin, micro-rough surface that greatly increases the surface area footprint of the copper features to which the resin will be bonded during the subsequent lamination process.

Figure 1 is an SEM photomicrograph depicting typical surface topography that an OA will impart on the inner layer copper features. Note the rough texture. This rough texture enhances the mechanical bond between the adhesive resin and the various copper layers in a multi-layered PCB

The topography that is created when copper is exposed to an OA is a very complex series of chemical reactions that results in a uniform, micro-roughened surface that is dark reddish – brown color. This surface is the basic building block that ultimately holds the entire multi-layer PCB together so its importance should not be minimized. It is very important to understand that even though two different OA chemistries may operate at the same etch rate, they may impart very different surface topographies in the copper.

III. Test Vehicle Description:

Enlarge this picture Figure 2. Test Vehicle Cross Section

The test vehicle (TV) has specific design features that allow for the assessment of electrical performance using SPP methods and apparatus. A signal – ground launch structure consisting of 0.006” plated through holes (PTHs) on a 0.020” pitch are utilized to minimize capacitive and inductive discontinuities. Two distinct stripline signal layers are included in the design, allowing two separate core/b-stage building blocks to be assessed.

The cross section of the specific test vehicle used in this analysis can be seen in Figure 2.

IV. Test Vehicle Fabrication:

Figure 3 – SEM Example of Test OA 1

A 4 cell test matrix was developed. Two OAs, each yielding different surface roughness characteristics, were utilized, each with and without an additional organic post dip. As previously indicated, this organic post dip is a product developed to enhance the chemical bond between the OA and tough-to-bond-to resin systems. It was desired to determine if this postdip had any effect on the electrical performanceof the resultant structure.

Figure 4 – SEM Example of Test OA 2

Six (6) panels were run in each cell, each consisting of 6 TVs, for a total of 36 TVs in each cell. Figure 3 and Figure 4 depict the different surface topographies imparted by the two test OAs. Note the distinct difference in the macro and micro surface topography between the samples.

The TVs were sorted based on ability to meet impedance targets, and a minimum of 10 pieces from each cell were provided for the performance assessment.

V. Thermo mechanical Assessment:

Enlarge this picture Table 2. TMA Test Results

Since a PCB not only has to have acceptable electrical characteristics, but also acceptable mechanical reliability characteristics, a subset of all test groups was submitted for T-240, T-260, and T-288 thermal stress testing utilizing a thermomechanical analyzer or TMA per test method IPC-TM-650, Method # 2.4.24.1. The results of the TMA testing are shown below in Table 2. All T-240 and T-260 samples exceeded the 30 minute threshold with no delamination. The T-288 results also performed beyond specification.

VI. Electrical Performance Assessment:

Enlarge this picture Figure 6. Nominal Attenuation Comparison, Resin Rich Layer, S3

The theory behind the Short Pulse Propagation (SPP) technique is fully described in [1]. In summary, the technique allows for the extraction of frequency dependant dielectric constant, εr(f), and effective dielectric loss tan δ(f) through the creation of broadband, fully causal transmission line models based on the measurement results.

Standard TDR equipment is used in conjunction with high performance cables and tight pitch probes to launch a fast rise time pulse through an Impulse Forming Network (IFN). The resultant waveform is then applied to two traces as nearly identical in nature as possible, but of sufficiently different lengths. These measurement results, along with the low frequency characteristics, are then processed using IBM developed software (http://www.alphaworks.ibm.com/tech/gammazandcz2d), resulting in a frequency dependant attenuation which accurately describes theperformance of the structure.

Enlarge this picture Figure 7. Nominal Attenuation Comparison, Resin Poor Layer, S6

The parts are then cross sectioned and average physical attributes of the structure are documented. It shall be noted that the laminate material losses determined through this test methodology are ‘effective losses’, lumping both laminate losses and losses due to skin effects. Therefore, when the Cu roughness is significant, the extracted tan δ(f) will contain both elements. This exercise focuses on the relative impact to overall performance, with all else constant except Cu roughness as a result of the OAprocess.

Finally, nominal design models are created using the previously extracted εr(f) and loss tan δ(f) parameters. These models then result in attenuation characteristics in which the delta issolely driven by the skin effects resulting from the differing Cu roughness imparted by the two OA process chemistries. See Figures 6 and 7 for a comparison of these nominal conditions.

Depending on the structure, one observes an increase in attenuation of between 15% ~ 25%1 – 10GHz range due to use of OA1. This can be considered significant depending on the amount of margin within any given design.

VII. Surface Roughness Analysis:

Enlarge this picture Table 3. Roughness Comparison

Table 3 provides mean data for several industry standard parameters as measured using a Zygo white light interferometer. The Zygo interferometer was employed as a means to quantify the delta in roughness between the sample test groups.

Figure 8. OA1 Cross Section

Note that there is ~ 21% delta in Rmax, while the surface area shows a 37% delta. Thus, for the specification of Cu attributes which contribute to performance, it is then determined that these two characteristics are primary contributors to the impact on performance.

How these roughness attributes manifest themselves as physical characteristics can be observed in cross section as seen in Figures 8 and 9 below.

Figure 9. OA2 Cross Section

One observes that use of OA1 can result in a roughness on a finer scale, dubbed ‘micro’ roughness. This is an indication of an increased interaction of the OA chemistry with the Cu grain boundaries. Conversely, use of OA2 can result in less interaction with the Cu grain structure, effectively imparting a more ‘macro’roughness surface texture.

VIII. Skin Effects:

As previously indicated, skin effects are a well known phenomena. To briefly characterize their expected impact, one can use the following formula:

Skin Depth (δ) = Sqrt [ρ/(fπμ_o = 4π10^-7
(permeability in a vacuum), f in MHz [2]

Through our work cross sectioning and characterizing the Cu foils used in PCB manufacturing, an average resistivity has been determined: ρ, = 2.076 x 10^-8 Ω-m (vs. textbook value of 1.76 x 10^-8 Ω-m) at approximately 25°C.

Therefore, the skin depth (δ) at 1GHz = 2.35 um (0.09 mils). Since approximately 5δ are required to account for ~ 99% of the current flow [2], the bulk of the current flows in the outer 11.75 um (0.45 mils). From the data in Table 3, it is seen that the bulk of the current flow does penetrate the area of the increased roughness.

IX. Conclusion:

The results contained herein quantify the signal attenuation imparted by not only the incoming laminate / Cu foil characteristics, but also by two Cu foil adhesion promotion processes. The twodifferent oxide alternative chemistries used in this assessment resulted in different ‘macro’ and ‘micro’ roughness characteristics. Based on the data presented here, these different roughness characteristics appear to result in distinctive performance levels in otherwise similar PCB designs.

The macro roughness characteristics alone, as dominated by R_max (peak to valley), appear to impart less performance degradation than when combined with added micro roughness characteristics. The latter is perhaps better characterized by the overall surface area. The metrology used to assess the roughness characteristics should be considered so as to ensure adequate, accurate representation is provided.

The PCB fabricator has an opportunity to balance the electrical performance characteristics with the mechanical reliability of the particular design through optimization of the adhesion promotion process.

Furthermore, the OEM should be sensitive to the signal attenuation driven not only by the laminate material properties and the incoming Cu foil characteristics, but also by the output Cu foil roughness, as dictated by the adhesion promotion process

Note: This is a condensed version of the original publication. If you would like a the complete version of this paper please contactblee@macdermid.com