electrical characterization of packages for use with gaas mmic amplifiers rev. v4 application note m542 ? north america tel: 800.366.2266 ? europe tel: +353.21.244.6400 ? india tel: +91.80.4155721 ? china tel: +86.21.2407.1588 1 visit www.macomtech.com for additional data sheets and product information. m/a-com technology solutions inc. and its affiliates reserve the right to make changes to the product(s) or inform ation contained herein without notice. abstract a test methodology will be presented which combines the advantage of on-wafer rf probing with a trl calibration to create a completely de-embeddable, novel ?test fixture? capable of electrically characteriz- ing most any style package or device. this scheme has been used to characterize many of the currently available microwave packages in order to identify ap- propriate packages for our mmic amplifier products which cover frequencies up to 12 ghz. in addition, the technique has been employed to characterize in- jection-molded plastic packages and to evaluate non- probeable mmic's. introduction most package vendors have very little microwave design and characterization capability. their lim- ited characterization efforts typically involve the use of poor fixturing, which obscures the true frequency response of the package. companies specializing in fixturing, while invest ing considerable mechani- cal engineering effort, expend far less on electrical considerations, often producing fixtures inadequate for use at microwave frequencies. consequently, there is very little microwave performance data available from package vendors. therefore, to evaluate a nd identify candidate pack- ages for each of the amplifiers in our mmic ampli- fier product line, specific fixturing had to be devel- oped for each package style considered. a novel fixturing approach was designed and implemented, which not only eliminates the need for expensive, package specific fixtures, but also overcomes the frequency limitations of traditional connectorized, plunger-style fixtures. additionally, a rigorous cali- bration method was developed which allows com- plete fixture de-embedding. this test methodology is applicable to practically any style device. table 1 lists the package styles investigated. through this work, proper electrical characterization of commonly used packages has indicated useful frequency ranges broader than expected by even the package manufacturers. this finding has allowed us to use low-cost pack- ages for frequency applications where our competi- tors typically resort to high-priced custom pack- ages. package description manufacturer 5 lead, ceramic kyocera 6 lead, ceramic kyocera leadless, 6 port, ceramic stratedge 7 lead, ceramic kyocera 8 lead, ceramic kyocera 8 lead, glass mini-systems 8 lead, glass, ground straps mini-systems leadless, 8 port, ceramic oxley leadless, 10 port, ceramic alcoa table 1. summary of packages design approach to eliminate the need for expensive, device spe- cific, traditional fixtures and overcome their fre- quency limitations, an rf probeable ceramic sub- strate was designed as the interface to the device- under-test (dut). figure 1 illustrates this coplanar probe to microstrip transition. it is a 50 ohm line fabricated on 10-mil thick alumina, with an 8-mil pitch, ground-signal-ground (g-s-g) probe pattern at one end. the two ground pads are connected to the substrate backside with 8-mil diameter plated vias. the g-s-g pattern can be probed using com- mercially available microwave probes on a stan- dard microwave probe station. the opposite end of the substrate can be bonded to a test port of the dut. to complete the ?test fixture,? only a thin brass block is required to serve as the mounting surface for the ceramic substrates and the dut. if neces- sary, the brass block could be machined to com- pensate for any difference in height between the substrate and dut test port. to fixture practically any dut, all that is needed is a brass plate and the probeable ceramic substrates. figure 2 shows the configuration used for characterizing our maam71200-h1, a packaged 7-12 ghz gaas mmic low noise amplifier.
electrical characterization of packages for use with gaas mmic amplifiers rev. v4 application note m542 ? north america tel: 800.366.2266 ? europe tel: +353.21.244.6400 ? india tel: +91.80.4155721 ? china tel: +86.21.2407.1588 2 visit www.macomtech.com for additional data sheets and product information. m/a-com technology solutions inc. and its affiliates reserve the right to make changes to the product(s) or inform ation contained herein without notice. figure 1. probeable ceramic substrate figure 2. fixtured maam71200-h1 to de-embed this ?test fixture,? a set of through- reflect-line (trl) standards was employed. a ?zero-length? through, a short, and two delay lines were fabricated. these standards, shown in figure 3, are used with the common trl de-embedding algorithm. this allows any measurement made with the probeable ceramic substrates to be de- embedded to yield data for only the dut with con- necting bonds. bond wires can also be de- embedded by first characterizing and modeling them using this same ?probeable ceramic? tech- nique. for this work, multiple bond wire and ribbon lengths were characterized to generate fully scal- able bond models. figure 3. trl calibration standards to demonstrate the package characterization method, the evaluation of a standard kyocera 8- lead ceramic flat pack will be examined. figure 4 shows how one feedthrough structure in the wall of this package was tested. package leads were cut close to the package body, and the ceramic sub- strates were mounted flush to the package ports. two short 3-mil wide gold ribbons bond the sub- strates to the package. similarly, sealed packages with leads internally terminated with 50 ohm chip resistors were tested to determine the cross-coupling between opposite and adjacent leads. through-lines within sealed packages were also measured. with this data, the true electrical performance of the package was de- termined and models for the feedtrhough and cou- pling were developed. figure 4. fixtured feedthrough this information allows the identification of an ap- propriate package for ex isting mmic products and provides an accurate model for incorporating pack- age effects into future design work.
electrical characterization of packages for use with gaas mmic amplifiers rev. v4 application note m542 ? north america tel: 800.366.2266 ? europe tel: +353.21.244.6400 ? india tel: +91.80.4155721 ? china tel: +86.21.2407.1588 3 visit www.macomtech.com for additional data sheets and product information. m/a-com technology solutions inc. and its affiliates reserve the right to make changes to the product(s) or inform ation contained herein without notice. experimental results the feedthrough walls of each package listed in table 1 have been tested and modeled. this feedthrough data alone largely indicates the useful frequency range of each package. figure 5 shows the frequency response for the feedthrough of the 8-lead ceramic flatpack. this package, previously thought to be useful only at lower frequencies, demonstrates excellent performance well into x- band before resonating. based on this result, we assembled our 2-8 ghz gaas mmic amplifier into this package. the performance of this packaged amplifier, part number maam28000-a1, is shown in figure 6. figure 5. feedth rough frequency response figure 6. maam280 00-a1 performance using the de-embedded feedthrough data, y- parameter extraction followed by a constrained op- timization was performed to derive the feedthrough model shown in figure 7. figure 8 shows the measured versus modeled in- sertion loss and input return loss for this package feedthrough. the model simulates the feedthrough performance closely over the useful frequency range of the package. figure 7. feedthrough model figure 8. measured vs. modeled performance coupling effects between package ports were also measured and modeled. a y-parameter extraction showed that the coupling could be attributed to equivalent capacitance values. in the case of the 8-lead ceramic flatpack, coupling between adjacent ports along one side of the flatpack can be repre- sented by a 0.03 pf capacitance. between alter- nate ports along the same side, the coupling ca- pacitance is nominally 0.003 pf. coupling be- tween internally terminated ports on opposite sides of the flatpack was modeled with a 0.0007 pf ca- pacitor. this coupling model accurately predicts the measured input to output isolation, as illus- trated in figure 9, over the package?s useful fre- quency range. -20 -15 -10 -5 0 5 10 15 20 0246810 gain s11 s22 frequency (ghz)
electrical characterization of packages for use with gaas mmic amplifiers rev. v4 application note m542 ? north america tel: 800.366.2266 ? europe tel: +353.21.244.6400 ? india tel: +91.80.4155721 ? china tel: +86.21.2407.1588 4 visit www.macomtech.com for additional data sheets and product information. m/a-com technology solutions inc. and its affiliates reserve the right to make changes to the product(s) or inform ation contained herein without notice. figure 9. package isolation characterizing the packages in table 1 produced interesting results. the five relatively inexpensive packages (the 5-, 7-, and 8-lead flatpacks) are commonly used for fairly low frequency applica- tions. however, as detailed above, the 8-lead ce- ramic flatpack, supplied by kyocera, exhibits excel- lent performance into x-band. mini-systems? 8- lead glass flatpack also exhibits excellent perform- ance into x-band, and their version with ground straps has similar performance through c-band. the kyocera 5- and 7-lead ceramic flatpacks, often used in switching applications, have higher inser- tion loss and lower return loss, but demonstrate reasonably good performance into x-band and c- band, respectively. the oxley manufactured lead- less 8-port ceramic package has excellent perform- ance through c-band. the remaining three packages shown in table 1 are all advertised for high frequency applications. of these, stratedge?s leadless 6-port ceramic flat- pack exhibits the best performance through 20 ghz. the alcoa 10-port ceramic package also works reasonably well up to 20 ghz. kyocera?s leaded version of the 6-port ceramic package dem- onstrates reasonably good performance to 16 ghz. at least one suitable package was chosen for each of the small signal amplifiers, and one of the power amplifiers, in our gaas mmic amplifier product line. table 2 lists all the packaged amplifiers now offered as standard products. this test method was also used to characterize the lead parasitics of the sop and ssop plastic packages. that data has been incorporated into the design of several new products specifically targeted for high-volume, low-cost, commercial applications. bond wires, bond wire pairs and ribbons have also been characterized with this test method, resulting in scalable, empirically-derived models. in addi- tion, this test methodology is widely employed in our engineering test lab to rf probe mmics which are otherwise not rf probeable. p/n (maam-) function package style 02350-a2 0.2-3.5 ghz ifa 8 lead, ceramic 12000-a1 1-2 ghz lna 8 lead, ceramic 23000-a1 2-3 ghz lna 8 lead, ceramic 37000-A1 3-7 ghz lna 8 lead, ceramic 71200-h1 7-12 ghz lna leadless, 6 port ceramic 28000-a1 2-8 ghz wba 8 lead, ceramic 26100-b1 2-6 ghz pa 7 lead, ceramic table 2. packaged amplifier products a novel fixturing and test methodology has been designed and implemented which allows accurate microwave frequency characterization of virtually any device. this approach has been used to evaluate many of the curre ntly available microwave packages. appropriat e packages have been identi- fied for our gaas mmic amplifiers, resulting in many new standard product s. models for package feedthrough structures, pl astic packages, and bond wires and ribbons have all been developed using this method. acknowledgements written by stephen r. smith and michael t. mur- phy. the authors thank scott mitchell and ted begnoche for testing these devices, brenda mil- inazzo for assembling them and bill fahey for help- ing to prepare this paper. ? 1993 ieee . this paper first appeared in the ieee 1993 mtt-s international microw ave symposium digest and is reprinted here with the permission of ieee . -70 -60 -50 -40 -30 -20 024681012 db (s21) coupmod db (s21) r317co up frequency (ghz)
|
