Improved blades for RF testing at wafer level B. Tunaboylu An improved design for a blade utilised in a probe card is presented. The comparison is made with a current blade design and prototype test cards were manufactured for the study. Measurement results on S-parameters are reported on test cards with the new design and new blade material. Introduction: Ceramic blade probes consist of wire probes attached to a ceramic blade. Different design variations exist to meet various wafer probing requirements and semiconductor test environments. Blade card technology is preferred for probe counts typically less than 100 probes at a minimum bond pad pitch of 100 mm and is preferred for low leakage, high frequency and high temperature applications [1]. Fig. 1 shows a cross-section of a blade card. Each probe is mounted on a separate blade made of ceramic which is typically alumina as the standard. These ‘blade probes’ are then individually soldered onto lands – special wide metallised patterns on the top of the PCB (printed circuit board). Depending on probing requirements, device size and shape, the number of bond pads, signal characteristics, probe card design parameters will vary. Probe material, typically tungsten, tungsten–rhenium, beryllium copper or Paliney 7 TM , will vary depend- ing on test signal characteristics, contact resistance requirements, current carrying requirements, and bond pad material. Probe diameter and beam length will vary depending on contact force and current carrying require- ments. Probe parameters must be optimised for the design and sensitive pad structures [1, 2]. The PCB and tip depth will vary depending on the type of prober interface being used and probing temperature. blade PCB top surface stiffener probe tip probe tip depth PCB thickness blade metalisation blade beam length probe tip length Fig. 1 Blade card cross-section Standard blade probes are used in applications not requiring a con- trolled impedance environment. Radial microstrip blade probes are used in applications that require a controlled impedance environment where the signal path connects directly to the PCB, while microstrip blade probes connect the signal path directly to a coaxial cable or other transmission line. Full metal/metallised ceramic blades are pri- marily used for ground shield connections. Those designs that currently exist have bandwidth limitations for some high-speed applications. We have proposed and patented a new approach [3] where a coplanar wave- guide (CPW) structure on the blade and a ground ring attachment on the probe tip section improve the bandwidth of the signal and provide a more flexible pad layout. Simulation results were also provided, predict- ing significant improvements [3]. There are many factors affecting band- width in a probe card such as connections to the PCB (mounting points), PCB line discontinuities, probe design, configuration and materials. The proposed method reduces the return path by implementing a ground loop close to the probe tip. It also shows a coplanar waveguide design on the blade structure. This Letter presents measurement results on CPW blades made of a new material with a ground loop structure attached to probes of a test probe card. Concept design, prototype and measurement method: Current blade design with an existing PCB is shown in Fig. 2a. The main body is made of 96% alumina and the probes were made of tungsten– rhenium. The base PCB material was FR-4. The 50 V microstripline characteristic impedance is achieved by 15 mil (381 mm) trace width on alumina with a dielectric constant 1 of 9.5. Each blade is placed on the PCB in circular form as shown in the Figure. Signal pin and ground pin probes have different electrical layout. The current blade was replaced with a CPW in an effort to reduce crosstalk between adja- cent blades. Also the CPW is arranged to mate easily with a PCB. The existing ground loop passes through the PCB structure so that the elec- trical fields and magnetic fields are radiated over the board. The new blade board uses a secondary U-shape extender to create a shorter ground loop, as shown in Fig. 2b. So the injected signal passes through the integrated circuit and from the ground probe shorted to the U extender. The new structure has two advantages over the old con- figuration: reduced ground loop length and less radiated energy. As a result, the new structure achieves higher bandwidth and less impedance variations. Another change was the use of a low-loss microwave material, Rogers RO4000 with a 1 of 3.38 and dissipation factor of 0.0021, for the blade and the PCB structure as well. The measurement of S-parameters was carried out using an HP8722D network analyser, on a Gigatest GTL3030 probe station using Model 40A picoprobes for contacting probe tips. Time-domain reflectometry (TDR) measure- ments were also done to observe impedance variations in the same setup. a b c current blade Fig. 2 Images showing blade card with current blades on PCB (Fig. 2a), new ground loop concept and PCB assembly (Fig. 2b), and test setup for microwave measurements (Fig. 2c) S 21 (dB) against frequency –1dB bandwidth 0.18 GHz –3dB bandwidth 0.39 GHz Fig. 3 Measured insertion loss for current blade Results: The insertion loss measurements indicate a bandwidth value of 21 dB bandwidth of 0.18 GHz and 23 dB bandwidth of 0.39 GHz for the current blade design on a test card, as shown in Fig. 3. In this case the PCB design was not optimised and the mounting structure and connec- tions cause reduction in bandwidth performance of the card. Based on TDR pulse responses, the propagation delay and the rise time (RT) measured were 78.25 and 974.34 ps, respectively. The calculated 23 dB bandwidth based on the equation 0.35/RT was 0.36 GHz for the current blade structure. The insertion loss testing made on the new blade structure and the test card showed a significant performance improvement with a 21 dB ELECTRONICS LETTERS 8th December 2011 Vol. 47 No. 25