The Comparison of Sheet Resistances of Screen‐ and Stencil‐Printed Resistors
Abstract
The deposition of the precise thick‐film buried resistors (with better tolerance) is one of the factors that can increase the integration of the low‐temperature co‐fired ceramic devices further. The thick‐film buried resistors cannot be trimmed after sintering or the trimming is very difficult in the contradiction to the surface resistors. The possibility of utilizing simple stencil printing in the fabrication of thick‐film resistors is presented in this paper. Moreover, the comparison of the variability of the sheet resistances of the screen‐ and stencil‐printed buried resistors was performed using Taguchi Design of the Experiment methodology.
Introduction
Thick‐film technology is a well‐developed and cheap technique of a microhybrid multilayer circuit fabrication.1, 2 The application of this technology in combination with low‐temperature co‐fired ceramics (LTCC)3-5 enables a significant increase in the electrical interconnection density and the reliability of the whole device. Moreover, the size of the device can be decreased using embedded (buried) thick‐film passive components inside the multilayer substrates. This solution permits the saving of a lot of space on the device surface. However, it drastically complicates the trimming of the resistors.6, 7 The problem starts to be really important when it is taken into account that the standard commercial thick‐film resistors have a sheet resistance tolerance in the range of ±20%. Moreover, any technological process inaccuracies can increase this value further. Therefore, beside the average value of a sheet resistance, the authors very often give information about the deviation of this value. The topic from a commercial point of view is very interesting. Because high deviation of the sheet resistance excludes such components from more precise applications and permits the use of them only as, for example, pull‐up resistors. Moreover, a high variation in the sheet resistance can lead to inaccurate conclusions when, for example, the influence of the resistor composition on its deviation or stability is investigated. The higher deviation, the easier it is to prove some statistical theses, but it does not mean that these conclusions are reliable and correct.
The thick film can be deposited using several different techniques. The most common are the following: screen printing, laser pattering,8, 9 photo‐imaging layers,10-12 and ink jet printing.13, 14 Screen printing is the best developed technique of a thick‐film deposition. Moreover, thanks to modern fine metal fabrics with the wire diameters smaller than 15 μm, very fine patterns can be fabricated.15 This technique is very useful especially when low costs and precision of the deposition is required. The standard wideness of the screen‐printed conductive paths can be narrower than 100 μm. Additional fabric treatments enable attaining the resolution of even 30 μm.16 Screen printing has also one another big advantage; it permits the deposition of layers with various electrical properties.17, 18 Therefore, screen printing still enables the lowest cost of the deposition and the highest stability and reliability of the thick‐film pastes.
The comparison of the mean values and the deviation of sheet resistances of stencil‐ and screen‐printed resistors were carried out in this paper. The goal of this work was to analyze which technique permits achieving of lower sheet resistance variations. The investigation was carried out using Design of the Experiment methodology.19-22 The achieved results will be utilized in the future in the fabrication of the precise ceramic temperature controllers and sensors. Moreover, the results are very promising for the further development of thick‐film buried components.
Experiment
The Taguchi Design of the Experiment (DoE) methodology was used to estimate the influences of the printing squeegee pressure, printing velocity, printing snap‐off (only screens), and lamination pressure on the sheet resistance and its deviation of stencil‐ and screen‐deposited thick‐film buried resistors. Stencil printing is carried out with direct contact of a stencil and a substrate; hence, snap‐off cannot be investigated for the stencil deposition. The volatility levels of the investigated inputs are given in Table 1. The values given in Table 1 are in a typical range of the volatility used for the thick‐film process deposition and lamination at Wroclaw University of Technology (Poland). Hence, it was useless to investigate wider or narrower range of the input volatility.
| Input | Acronym | Levels of volatility | ||
|---|---|---|---|---|
| 1 | 2 | 3 | ||
| Squeegee pressure [MPa] | A | 1 | 3 | 5 |
| Printing velocity [mm/s] | B | 5 | 42.5 | 80 |
| Snap‐off [mm] (only screen) | C | 0.5 | 1 | 1.5 |
| Lamination pressure [MPa] | D | 5 | 10 | 15 |
A significant analysis of four inputs on three levels of volatility requires at least an L9 Taguchi matrix. Therefore, it was necessary to investigate only nine combinations of the aforementioned input values. The cardinality of resistors for each of the analyzed combinations of inputs was around 160, divided into four measuring runs. The theoretical thickness of the wet thick film was set to approximately the same level for both stencil‐ and screen‐printed resistors by choosing proper properties of both screen and stencil. The stainless stencil was laser cut and its thickness was fixed to 40 μm. The screen had a fabric angle of 22.5°, 325 mesh, 24 μm wire diameter and 10‐μm‐thick emulsion above the fabric. The choice of 40‐μm stencil was affected by the necessity of ensuring approximately the same wet thickness of screen‐ and stencil‐printed resistors. The stencil supplier enabled only to choose between 40‐, 60‐, and 80‐μm‐thick stencils. Hence, the thinness stencil had to be purchased, otherwise, the thickness of resistor would be too high and the resistance would be far below of designed 100 Ω. The resistor dimensions were fixed to 1 × 1 mm in fired state. The stencil‐printed resistors were deposited on LTCC green tape before the conductor paste deposition. The draft of screen‐printed terminations and stencil‐printed resistor configuration is presented in Fig. 1a. The screen‐printed resistors were deposited after the conductive film deposition. The draft of screen‐printed terminations and screen‐printed resistor configuration is presented in Fig. 1b. Stencil‐printed resistors have to be deposited first because of the lack of the sealing between the stencil and LTCC, what in many cases can limit using this technique. The problem will be more precisely describe later in the text. The sealant in the screen printing is basically an emulsion.
A standard CF021 DuPont resistance paste for buried components, DP6146 (Pd/Ag) conductor paste, and DP951 green tape were used for the sheet resistances comparison. The laminated resistors were fired in a box furnace at the profile given in Table 2.
| Parameter | Value |
|---|---|
| Rising time to debinding temperature | 3 h |
| Debinding temperature | 450°C |
| Dwell time at debinding temperature | 1 h |
| Rising time to peak firing temperature | 1 h |
| Peak firing temperature | 875°C |
| Dwell time at peak firing temperature | 20 min |
Results and Discussion
The average sheet resistance measured for the screen‐ and stencil‐printed resistors is given in Tables 3 and 4, respectively (where: set – number of volatility combination; A, B, C, D – acronyms explained in Table 1; Ra – average sheet resistance calculated based on around 160 measurements; σ – standard deviation of sheet resistance measurements; V = σ/Ra – variability coefficient). Moreover, both tables contain information about result deviation. It can be easily observed that the sheet resistance of the screen‐printed components is much more dependent on the deposition process conditions than the sheet resistance of the stencil‐printed components. Additionally, deviation of the measurements is also lower for the stencil‐printed components. The value of the sheet resistance was expected to be equal to 100 [Ω/□] for both screen‐ and stencil‐printed components according to the paste supplier. The comparison of the total average sheet resistance (calculated as an average of all average sheet resistances), standard deviation, and variability coefficient of screen‐ and stencil‐printed resistors is given in Table 5. As it can be seen, the stencil printing process enables the achieving of the expected sheet resistance equal to 100 Ω/□, even before optimization. So, all investigated combinations of inputs permit, with good accuracy, obtaining the expected value of the sheet resistance. Moreover, the relative variation in the sheet resistance of the stencil‐printed resistors calculated as the variability coefficient is extremely low, especially for such a big range of investigated inputs, and is approximately equal to 9%. In comparison with these results, the total average sheet resistance and its variability coefficient for the screen‐printed resistors are equal to 169 Ω/□ and 48%, respectively. So, on this stage of the analysis, it is obvious that the stencil‐printed components are far less sensitive to the process parameters fluctuations in comparison with the screen‐printed components.
| Set | Input level of volatility | Ra [Ω/□] | σ [Ω/□] | V [%] | |||
|---|---|---|---|---|---|---|---|
| A | B | C | D | ||||
| 1 | 1 | 1 | 1 | 1 | 164 | 49.40 | 30.18 |
| 2 | 1 | 2 | 2 | 2 | 98 | 17.93 | 18.38 |
| 3 | 1 | 3 | 3 | 3 | 63 | 32.77 | 52.18 |
| 4 | 2 | 1 | 2 | 3 | 233 | 41.48 | 17.78 |
| 5 | 2 | 2 | 3 | 1 | 114 | 13.88 | 12.20 |
| 6 | 2 | 3 | 1 | 2 | 140 | 14.06 | 10.04 |
| 7 | 3 | 1 | 3 | 2 | 228 | 31.62 | 13.87 |
| 8 | 3 | 2 | 1 | 3 | 157 | 34.75 | 22.11 |
| 9 | 3 | 3 | 2 | 1 | 329 | 47.98 | 14.60 |
| Set | Input level of volatility | Ra [Ω/□] | σ [Ω/□] | V [%] | ||
|---|---|---|---|---|---|---|
| A | B | D | ||||
| 1 | 1 | 1 | 1 | 91 | 5.25 | 5.78 |
| 2 | 1 | 2 | 2 | 90 | 12.75 | 14.17 |
| 3 | 1 | 3 | 3 | 85 | 10.51 | 12.38 |
| 4 | 2 | 1 | 2 | 108 | 4.98 | 4.61 |
| 5 | 2 | 2 | 3 | 100 | 6.12 | 6.10 |
| 6 | 2 | 3 | 1 | 105 | 5.35 | 5.09 |
| 7 | 3 | 1 | 3 | 107 | 5.01 | 4.69 |
| 8 | 3 | 2 | 1 | 109 | 5.25 | 4.82 |
| 9 | 3 | 3 | 2 | 106 | 5.34 | 5.04 |
| Stencil/screen | Ra [Ω/□] | σ [Ω/□] | V [%] |
|---|---|---|---|
| Stencil | 100 | 9.08 | 9.08 |
| Screen | 169 | 81.79 | 48.30 |
However, these initial conclusions can also be confirmed using more sophisticated statistical analysis. The main effects analysis was conducted for both the stencil‐ and screen‐printed components using proposed by the Taguchi concept of the Design of the Experiment. The value of the sheet resistance was expected to be equal to 100 [Ω/□] for both the screen‐ and stencil‐printed components. The main effects analysis calculations enable the prediction of the optimal process conditions and can be carried out for three different cases: Lower is better, higher is better and nominal is the best.19 The nominal is the best case that has to be chosen if an experimenter is looking for the process parameters which enable achieving the expected value of the sheet resistance. The obtained results of the main effects analysis are presented in Fig. 2 (where: A, B, C, D – acronyms explained in Table 1; 1, 2, 3 – levels of volatility explained in Table 1). As a response in this investigation was used SN ratio (signal to noise). Hence, the optimal conditions for a particular input will be achieved for the highest value of SN ratio. The Taguchi permits the conducting of the separate analysis for each of the inputs. Therefore, the optimal process conditions of each of the inputs can be easily found. However, the main effects analysis shall be conducted in conjunction with the analysis of variance (ANOVA). This solution permits the elimination of the insignificant factors and the estimation of the sheet resistance for the optimal process conditions. If any of the inputs are insignificant, then analyses of the trends for these insignificant inputs are useless, because these trends are heavily affected by noise. The trends presented in Fig. 2 will be additionally explained a little bit later in the text.
The main result of the analysis of variance is a percentage impact of the each of the investigate inputs on the sheet resistance. The snap‐off of the screen printing process is insignificant for a given range of volatility according to the ANOVA; hence, analysis of the main effects trend for this input (see Fig. 2a) is useless. Of course, this conclusion is only true for the used range of the volatility of the other inputs, for example, if the range of volatility of the printing velocity would be much smaller, then the snap‐off probably would play a more significant role. The most significant inputs in the screen printing of resistors are printing velocity, printing pressure, and lamination pressure, as it is shown in Table 6. Therefore, the trends for these three inputs can be analyzed in Fig. 2a. If a higher printing pressure was used, the resistor film would be thinner. If the printing velocity increases, the film thickness decreases. The deposited film thickness increases if the lamination pressure is decreased. The thinner the film thickness, the higher the sheet resistance can be obtained. Therefore, there are some optimal process conditions for which the expected sheet resistance value can be achieved. The most optimal results will be obtained according to the main effects analysis (see Fig. 2a) for printing pressure (pP) equal to 1 MPa, printing velocity (v) equal to 42.5 mm/s, and lamination pressure (pL) equal to 10 MPa. The comparison of the expected value, estimated value using ANOVA, and real measured average value of sheet resistance for the screen printing optimal conditions is presented in Table 7. Moreover, extra information about the standard deviation and the variability coefficient of the conducted measurements are given in this table. It can be seen that the variability is three times lower in comparison with Table 5. However, this value is still high and is equal to 16%.
| Input | Screen printing | Stencil printing |
|---|---|---|
| Printing pressure (P1) | 36.43% | 67% |
| Printing velocity (P2) | 47.94% | Insignificant |
| Snap‐off (P3) | Insignificant | Not investigated |
| Lamination pressure (P4) | 7.11% | Insignificant |
| Other impacts | 8.52% | 33% |
| Input | Screen printing | Stencil printing |
|---|---|---|
| Expected [Ω/□] | 100 | 100 |
| Estimated [Ω/□] | 107 | 106 |
| Real measurement Ra [Ω/□] | 89 | 88 |
| Real measurement σ [Ω/□] | 14.23 | 5.65 |
| Real measurement V [%] | 15.99 | 6.43 |
Printing pressure is a significant input only in the resistor stencil deposition process according to the ANOVA. Hence, the investigation of only this input trend makes sense in stencil printing (see Fig. 2b). Taking into account that the stencil is not covered by any sealant, the sealing between the stencil and green LTCC can be achieved only if enough high printing pressure was used. So, in this particular situation, the stencil has to be in direct contact with a soft elastic green LTCC and the squeegee pressure has to ensure that the thick‐film paste does not spread between the stencil and green LTCC. Increasing the pressure further will press the stencil deeper into the green tape and deform the tape. Therefore, going above a particular squeegee pressure, the film thickness decreases and the sheet resistance increases. The lamination pressure and printing velocity are insignificant, and investigation of their trends by curves in Fig. 2b is useless according to the ANOVA. The printing velocity is insignificant because in stencil printing, the paste does not need to have any rheological behavior – it can have the same density during the whole process. It can be surprising that the lamination pressure does not affect the sheet resistance. However, it has to be bear in mind that this result says only that in comparison with the printing pressure, the lamination pressure is insignificant from a statistical point of view. The optimal process conditions for stencil printing were found using main effects analysis after the exclusion of the insignificant inputs. The optimal condition is as follows: printing pressure (pP) equal to 3 MPa. The comparison of the expected value, estimated value using ANOVA, and real measured average value of the sheet resistance for the optimal stencil printing conditions is shown in Table 7. Moreover, extra information about the standard deviation and the variability coefficient of the conducted measurements are given in this table. It can be seen that the variability of the stencil‐printed components is around 6% and it is almost three times lower in comparison with the optimized screen‐printed resistors. Beside this, the average sheet resistances of both stencil‐ and screen‐printed resistors are approximately the same and are equal to 88 and 89 Ω/□, respectively. The expected, estimated, and real measured sheet resistances for both deposition methods are very similar. Therefore, the proper conduction of the Taguchi and ANOVA analyses can be confirmed. Moreover, a very good estimation of printing optimal conditions can be confirmed as well.
At the end of the experiment, screen‐ and stencil‐printed resistors’ cross sections were prepared and investigated. The cross sections were done using laser scratching and breaking techniques. The screen‐ and stencil‐printed exemplary structures are presented in Fig. 3a and b, respectively. The delaminations that are visible in Fig. 3a are caused mainly by co‐firing process. The defect was present for both screen‐ and stencil‐printed resistors, and this delamination was not bigger than 100 μm. So, it was on accepted level for such investigations. The existence of this defect can be probably decreased by increasing firing dwell time of debinding or by increasing of air flow during this stage of co‐firing. Another solution is the utilization of exhaust channels that permit to drive off resistor organic combustion products easier.23 If higher integration would be needed, one of these solutions has to be applied during the buried resistors’ fabrication. This would increase the reliability.
It has to be concluded that in the applications where the deposition of resistors before conductive paths is possible, stencil printing guarantees much lower variability of the sheet resistances. Moreover, the stencil process is much more resistance to the process condition fluctuations and does not depend on paste rheology but only on paste density. Hence, less complex and cheaper pastes for stencil printing process can be utilized. However, stencil printing using bare stainless steel stencils can be carried out only on green LTCC tapes. Such tapes play a significant role in the sealing of a stencil to a substrate. For the screen printing process, an emulsion plays the same role. Therefore, a stencil shall be covered by an additional polymer layer if such stencil shall be used for the deposition of pastes on hard substrates.
Conclusions
Stencil printing can be utilized for the deposition of thick‐film resistors on green LTCC tapes. This technique permits the reduction in the variability of sheet resistances even three times in comparison with the standard screen printing process. The thick‐film thickness depends mostly on stencil thickness and printing pressure. Therefore, the stability of only these two parameters has to be provided. The choice of the substrates on which resistors can be deposited using the bare stainless steel stencil printing is limited only to soft elastic substrates. It is caused by the fact that a steel stencil does not have any sealing between a substrate and stencil. Hence, the deposition is possible only on soft elastic substrates which will act as a sealant and will prevent the spreading of the paste between the stencil and substrate. The sealant in screen printing is basically an emulsion. If the deposition of the resistors on the hard materials is needed, the stencil has to be covered by some sort of an emulsion to provide sealing between a substrate and stencil. Such stencils are also commercially available. The utilization of stencil printing is especially interesting in the applications where resistor trimming is impossible, for example, in the buried resistor configuration.
Acknowledgment
The project was financed by The National Center for Science (Poland) awarded based on decision number DEC 2011/03/N/ST7/00205.
