CARTS Europe 28 2-23 October Helsinki, Finland Benchmarking of LTCC Circuits up to 4GHz and Comparison with EM Simulation D.E.J. Humphrey, B.Verner, V. Napijalo TDK Electronics Ireland 322 Lake Drive, Citywest Business Campus, Dublin 24, Ireland TEL:- +353 1 41332 FAX:- +353 1 4133295 E-MAIL:- humphrey@tdk.de Abstract This paper assesses the use of Low Temperature Co-fired Ceramic (LTCC) modules in applications where the frequency of operation is higher than 2 GHz. To design at higher frequencies, the dielectric materials used to manufacture the circuits need to be characterised and while some information has been previously published in this area, detailed information is very rare. The work presented here shows the characterisation of some of TDK s LTCC material for frequencies up to 4GHz, using known evaluation techniques. In addition the difference between EM simulation and closed form design equations is discussed where the calculated parameters in this paper are seen to give a better first order design approximation. Introduction Low Temperature Co-fired Ceramic (LTCC) modules are used in many applications. These have generally been for frequencies less than 5 GHz, but with market forces now demanding cheaper integrated modules and at increasingly higher frequencies, the demand for LTCC circuits in high frequency applications has risen. A quick search shows that some modules have been developed for applications in excess of 2 GHz [1] and others using the LTCC as a substrate carrier for frequencies in the 6GHz region [2]. To design at these high frequencies, the materials used to manufacture the LTCC modules need to be characterised and while some information has been previously published in this area [3], there has in general been little activity. The work presented in this paper shows the characterisation of TDK s most common green tape TM for frequencies up to 4GHz, using previously reported evaluation techniques. The aim of this work is not simply to enable the LTCC material to be used a substrate carrier for a MMIC or other type module, but to be used in MIC design where the designed structures contribute to the circuit performance. In addition, as cheaper LTCC technologies use screen printing rather than photolithography, this paper assesses the printing accuracy using a screen printing process and draws conclusions to the applicability of this process for use at higher frequencies. In this way the fabrication cost can be further reduced. Some common RF structures have been designed and the measurements are compared to both linear and electromagnetic simulations. In this way the substrate parameters are confirmed and indeed the difference between circuit simulation and EM simulation at higher frequencies is identified.
Test Structure Design and Printing Ring resonator structures are an effective method of determining the dielectric constant and substrate loss parameters [3]. In this paper both microstrip and stripline ring resonators have been made, as well as visual inspection structures to determine the usability of a screen printing process at high frequencies. A picture of the visible circuit structures are shown in figure 1a while the test structure shown in figure 1b is used to define the line / space printing accuracy of different paste viscosities and their suitability to print the intricate details which are required at higher frequencies. Printing Definition Ring Resonator Filter Fig.1a:- Microstrip test structures Fig 1b:- Printing test structure Many different test structures can be seen in figure 1a where filters and different resonators are indicated. In addition different TRL calibration standards are printed to enable accurate measurement of the printed structures. Since the photograph shows the LTCC top surface, only the microstrip structures can be seen, although internal stripline structures including resonators and filters are also present.
Space Line Fig 2:- Minimum paste definition structure While figure 1b indicates that the line widths and consistency are maintained in the parallel direction of printing i.e. the paste application tool moves along in a direction parallel to the lines, other structures indicate that consistency is also maintained in the perpendicular direction to the paste application. To avoid confusion we define here that a perpendicular line is a line in the direction that is perpendicular to the paste application tool i.e. along the direction of paste application. Results Measured results show that, in general, line widths can be screen-printed to a tolerance of less than 1%. In addition for perpendicular lines, a standard deviation for the line width of about 3um is maintained, compared to 5um for parallel lines. While the results are in general very consistent, larger than normal errors were observed for 1um and 15um lines on one sample. Results for the structure, shown in figure 2, indicate that lines as narrow as 25um can be printed in the perpendicular direction, although this is not recommended for high production yields. Rather, our measurements suggest that in production a 1um line and gap should be maintained, although this could be reduced to 5um in exceptional circumstances with small volumes. While the previous discussion has concentrated on line widths, the gap between them can be intuitively determined using the following equation [4]:- 1 error (%) = ( n1e1 + n2e2 ) (1) 2 where n i is the width of the i th line normalized to the gap size and e i is the percentage error of the i th line. Thus for equal line and gap widths, and with a -1% printing error on both lines, the gap will increase in size by 1% [4]. Ring Resonator Tests Figure 3 shows the measured result for the microstrip ring resonator shown in figure 1a. To avoid the loading of the ring by the feed structures several rings were designed with
different coupling gaps drawn [2]. Here a 2um gap between the line and the ring was found to be enough to excite the ring, without loading it and thus moving the resonance points. -1-2 -3-4 db (S (1,9)) db (S (8,7 )) db (S (6,5)) d B (S (4,3 )) db (S (2,1)) -5-6 -7-8 -9-1 2 4 6 8 1 12 14 16 18 2 22 24 26 28 3 32 34 36 38 4 42 freq, GHz Fig 3:- Ring resonator results These results are summarised in the next section. Results Using Edwards standard equations for dielectric calculation [5, 6], the effective dielectric constant (ε eff ) and loss parameters are calculated. 5.9 5.8 5.7 5.6 Eeff 5.5 5.4 5.3 5.2 5.1 5 1 2 3 4 5 Freq (GHz) db/cm 1.2 1.8.6.4.2 1 2 3 4 5 Freq (GHz) Fig 4:- (a) Variation of E eff up to 4GHz (b) Microstrip loss results The microstrip ring resonator described here was designed for a substrate height of 32um and a material of relative dielectric constant of 7.5. However, when the Eeff value shown in figure 4a, is converted to its equivalent Er value using Linecalc TM [7, 8] a value of 7.55 is calculated. It has been shown previously that the total loss of the material consists of both conductor loss (due to conductor resistivity and surface roughness) and substrate loss [3, 6, 9], and
so it is possible to determine the substrate loss component by considering the loaded Q- Factors. Having measured the surface roughness of.58um and paste conductivity of 2e7 S/m using the test pattern shown in figure 1a, the dielectric loss tangent can be calculated using Linecalc TM [7, 8]. This variation is shown in figure 5a and the combined loss components in figure 5b:- Tand Losses.8.7 Average.6.5 tand.4.3.2.1 1 2 3 4 5 Freq (GHz) 1.2 1.8 db/cm.6.4.2 1 2 3 4 5 Freq (GHz) cond + rough + tand Measured Fig 5:- (a) Calculation of tanδ (b) Individual loss Components Figure 5b shows that at higher frequencies the dielectric loss tangent becomes the principle component of power loss and will become the single most important limiting factor to using cheaper LTCC materials at higher frequencies when conventional transmission line (microstrip) circuit design is used. As a figure of merit, the results here suggest that the cost benefit of the cheaper materials will be negated due to degradation in circuit performance at frequencies > 35 GHz using this material. Electromagnetic Simulation Figure 6 shows an example of an amplifier where the techniques used in this paper can be applied. In addition the importance of electromagnetic simulation (EM) of all circuits at higher frequencies is highlighted. Fig 6:- 24GHz 2-Stage Amplifier This is a 2-stage amplifier circuit designed using microstrip interconnects where the coupled line sections are used to provide dc blocking of the different amplifier stages. The entire circuit can therefore be simulated initially using a circuit simulator and accurate device model. This is often not possible for LTCC circuits which are multilayer in nature and therefore more suited to 3D simulators. Using the previously calculated dielectric constant of 7.55, we can see there is a sizeable frequency shift between the EM and circuit simulation. At low frequencies this change is barely noticeable, but as the
frequency increases a small percentage change introduces a large frequency shift. As always circuit simulations neglect fringing field effects which tend to change the effective phase lengths of the structure. Knowing this can quicken the design cycle, where the initial design can be performed using linear circuit analysis and the design frequency band given an offset to ensure accurate length calculation before electromagnetic simulation. Using this technique to quickly adjust the phase lengths of the relevant matching sections, the measured results are compared to the EM simulation of the final dimensions and initial circuit simulation in figure 7. 2 18 16-5 14-1 Gain (db) 12 1 8-15 -2 Return Loss (db) 6-25 4 2-3 -35 21 22 23 24 25 26 Frequency (GHz) Circuit Simulation E M S im u latio n M easurem ent Fig 7:- Final results for the 2-stage amplifier Good agreement was again obtained between simulation and other measured test structures. By way of example, figure 8 shows the simulation and measurement of a transconductance mixer which was designed using Agilent ADS TM [8] for the 24GHz IMS frequency band. Once again the performance was seen to agree well with the harmonic balance simulation:- Fig 8:- 24 GHz transconductance mixer
The advantage that the LTCC material offers in these designs, is that by further integrating decoupling capacitors and inductors within the ceramic (i.e. components for low frequency), the overall circuit component count is drastically reduced with lower overall board fabrication costs. Thus a lower cost surface mount equivalent circuit can be provided compared to the conventional technology. Conclusions The results presented here show that it is possible to design cheaper high frequency integrated modules using LTCC materials where the material used here is low cost and shows remarkable parameter stability up to 4GHz. The calculated circuit parameters shows good agreement with conventional circuit simulation techniques and although LTCC materials have a higher dielectric loss tangent than materials traditionally used in high frequency design, the results here suggest that this material could still be used in excess of 3GHz. In addition the screen printing technique was found to maintain the consistency of the definition of the printed structures to within 1%, proving that such a method could be used for the integrated structures required by high frequency circuits. Measured results show that screen printing is accurate enough for circuits designed in the 2-3GHz band, although for frequencies in excess of this it is advisable to use a photolithographic approach. However, the loss in microstrip structures at these frequencies becomes even more significant. To compensate for these losses, the advantage of LTCC design means that composite waveguides could be made within the LTCC block to reduce these loss mechanisms for higher frequency circuits and thus their applicability at even higher frequencies is more likely. References [1] Kulke et al., 24GHz Radar Sensor integrates Patch Antenna and Frontend Module in Single Multilayer LTCC Substrate, EMPC, Brugge, June 25. [2] Young et al., Monolithic LTCC SiP transmitter for 6GHz wireless communication terminals, Microwave Symposium Digest, IEEE MTT-S International Volume, pp.115-118, June 25 [3] Kulke et al., RF Benchmark up to 4GHz for various LTCC low loss tapes, IMAPS-Nordic, Stockholm, 22. [4] Humphrey et al., RF Benchmarking of TDK LTCC Circuits up to 4GHz, ICEP Proceedings, 23. [5] Edwards T., Foundations for Microstrip Circuit Design, Wiley, 2 nd Edition, 1992. [6] Kulke et al., Investigation of Ring-Resonators on Multilayer LTCC, MTT-S, Workshop on Ceramic Interconnect Technologies, Phoenix, 21. [7] LINECALC TM, Agilent Technologies, 395 Page Mill Road, Palo Alto, CA9434, USA, 21. [8] Advanced Design System TM, Agilent Technologies, 395 Page Mill Road, Palo Alto, CA9434, USA, 21. [9] El-Bakly A.M., Optimization Study of Stripline Resonator Technique for Dielectric Characterization, Ph.D Dissertation, Virginia Polytechnic Institute and State University, 1999.