Brandon Mair, Texas Instruments, “WSP-Wafer Socket Probe for Flip Chip Applications“:
Wafer socket probe (WSP) technology has demonstrated better physical and electrical performance and lower cost of ownership (COO) than traditional vertical probe cards for testing wafer level chip scale packages (WLCSP) at 0.4 mm (400 µm) pitch. These WSP probe heads are typically built using spring probes instead of buckling beam structures used for vertical probe heads.
Currently, Texas Instruments (TI) uses cantilever and vertical probe cards to test flip chip devices at pitches below 400 µm and as small as 150 µm pitch. This experiment was to determine if WSP technology could be used for flip chip devices at these tighter pitches. Initial engineering evaluation and production qualification were performed to compare the performance of the WSP technology versus a plan of record (POR) vertical probe card.
One advantage of WSP is the use of spring probes with a crown tip that self-aligns and only pierces the side of the solder ball compared to the POR flat tip vertical probe which dimples the top of the solder ball often changing the shape of the ball. Operationally TI likes the WSP solution since they are easier to repair on-site and do not require planarization of the probe tips.
A four lot (100 wafer) production qualification test run with approximately 20K touch downs showed better performance for WSP than the vertical probe card in many aspects. However, median contract resistance (Cres) was elevated by 1.2 ohms for the WSP solution over the vertical card but at the same time WSP had a tighter standard deviation than the vertical.
With these promising results, TI has recommended a full scale head-to-head comparison between WSP and vertical probe cards in long term production to verify the consistency of these results.
- What over drive did they use? What is the maximum over drive? They used last touch plus 60 µm over drive. The maximum over drive is a specification value of the spring probe selected.
- In the Cres measurements shown what is being measured? The Cres plots show the maximum value read for the worse pin. Not sure what current setting was used for the measurements.
- What is the lifetime requirement for the probes? On bigger solder bumps they have seen 3 M touchdowns but they are only at 20 or 30K in their experiments. However, they expect lifetimes for the WSP solution in the same range.
- Are they worried about cleaning changing the tip shapes? They are using the same cleaning media for larger tips too so they do not expect issues.
- Which cleaning media? They are using Mipox 6000 with abrasive.
- Did WLCSP have a net increase of 1.2 ohms? This is from all the pins from a failed die. Each device had 720 I/O pins and 9.5 ohms was average on a good die.
- How do you determine when you can replace spring pins or need to rebuild entire head? They can continue to replace single pins almost indefinitely. However when there is catastrophic pin damage or damage to housing you will need to replace head.
Jim Brandes, Multitest, “Wafer-Scale Contactor Development and Deployment“:
Multitest and its sister company Everett Charles Technology (ECT) have a long history (over 30 years) in developing spring probes. In 2009 they introduced the Gemini and Mercury family of probes. Using a Mercury probe, Multitest developed an eight site contactor for a 182-ball WLCSP.
Initial checks using singulated devices in a “hand test” mode verified the electrical performance of the contactor. However, they encountered issues when using the socket with a prober both during check-out and long term production:
- The prober had problems locating the tips of the spring probes since the tips moved more than a traditional probe card (lower positional accuracy). Typically a spring probe “self-aligns” and the X/Y position does not need to be as accurate as a probe. They needed to manually align the probes to the bumps to work around this issue. Design changes and an improved prober alignment algorithm resolved this issue.
- There were customer complaints about planarity which they are working to improve.
- Due to hygroscopy the contactor is not dimensionally stable. The 0.3% growth of their housing material can cause a substantial change in length across a multi-site array and can cause shearing of the solder bumps. The material can be baked to remove the excess water but the customer would prefer to not do this. They are looking to choose a material that is more stable in future versions.
Even with were these issues, the customer was willing to work through them since their Mercury technology had increased yield compared to the customer’s existing probe technology. The specified 500 K touch downs was achieved, in fact, some of the contactors worked for over 1 M touch downs. Based upon resistance and force versus deflection measurements (“FReD” plots) before and after 1 M touch downs Multitest believes there is significant additional life in the contactor. In general, spring probes will achieve longer life when used for wafer level test since the environment is generally gentler than packaged part test.
- Why select a crown tip over a traditional flat tip? The crown tip achieves penetration through debris and oxide on the ball. And oxide is thicker on solder balls than pads. Multitest has found that the tip geometry is more important than tip plating.
- Why did they select a plastic housing over ceramic? This was due to company history. The engineers were originally worried about ceramic from machining to pressing parts (hardware and pins) into the ceramic. Ceramic is also a more expensive material and has longer machining times than plastic. They plan to offer ceramic housings.
- Solder balls are less coplanar than pads – what is the variation range? Each manufacturer specifies a specific range.
- The scrub marks shown appear to be a two point tip. If a four point crown was used wouldn’t the self alignment be better? They used two point crowns. Alignment was as shown.
- What was the cleaning media used? They are also using Mipox yellow sheets.
James Tong, Texas Instruments, “Multi-point Probe Contacts for Flip Chip Wafer Level Probing“:
As the number of total probes increase in typical multi-site configurations, the required force will exceed the mechanical limits of TI’s existing test cells. Currently they are probing flip-chip devices with 7,000 pins using conventional vertical probe cards however some devices are starting to approach 12,000 pins. At the same time, greater electrical performance is required for these new devices.
They evaluated “multi-tip” configurations of FormFactor’s MicroSpring technology. Instead of previous FormFactor designs with a single pyramid or a blade (“Blade Runner” technology) tip, configurations with three triangles, four medium triangles, and four large triangles probe tips were evaluated. Unlike conventional vertical probe cards that use a flat tip (which deforms or flattens the top of the solder ball on the flip-chip device when probing), these multi-tip points pierce the perimeter of the solder ball leaving only small dimples. The extent of deformation using conventional flat tip probes is typically a flattening of the top 1/3 of the solder ball, which requires the solder ball to be reflowed after test to reshape the ball. The minor piercing of the multi-tip probes is acceptable and does not require an additional reflow operation.
Due to the smaller size and contact area of these multi-tips, higher contact pressure can be achieved using lower probe force. Conventional vertical probes have a typical force of 12 gF per probe at the desired over-travel. At the same time the multi-tip probes were only 9 gF per probe at worst case overdrive, resulting in a significant total force reduction.
Data presented included effect of over-travel on contact resistance (Cres) and contact stability. In addition, TI compared actual over-travel versus prober programmed over-travel using clay pucks to measure true probe compression and wafer chuck movement.
In conclusion, TI determined the piercing action of multi-tip contact at lower force enables larger multisite (~12K probes) with lowest cost of ownership (COO).
- How do you determine if you will apply over-travel from last or first touch? You don’t want to over compress the hardware so you need to fully characterize the test cell. This setting needs to be based upon card manufacturer’s recommendation.
- Your colleague just presented a paper about using spring pins but this is the opposite direction – why develop this? TI wanted to prove piercing was the correct method for probing solder bumps. So they went to multiple vendors (NHK & FormFactor). FormFactor was also qualified since they wanted to use FormFactor’s multi-layer ceramic (MLC) solutions for multisite probing to improve COO. In addition, the FormFactor force per probe is lower than TI’s baseline probe technology.
- Looking at four point tip on the FormFactor probe technology, will you get movement in X/Y plane beyond that which is in the scrub direction as a result of the Z axis compression? They turned off micro-force movement to lock in the X/Y motion.
- Since the crown tip “holds” the probe in place (versus typical sliding motion) are they going to do longer testing to see if there is a change to the probe? Yes, that is why they are doing a full production qualification.
Thorsten Teutsch, Pac Tech USA, “A New 3D Laser Bonding Process for Single Spring Attach on 300 mm Probe Cards“:
Pac Tech has developed a series of automated tools to laser bond MEMS probes on substrates from which to build probe cards. The laser system provides precisely controlled localized heat for the solder attach without thermal stress outside of bonding area. Therefore, individual probes can be attached or reworked without affecting neighboring probes.
The tools, which were demonstrated with videos, are:
- Cantilever Sorter – Inspects, then removes the micro-cantilever probes from the MEMS substrate, and finally loads them in waffle trays.
- Solder Jet – Applies a controlled solder shape on the substrate.
- Cantilever Bonder – Picks up the individual probes from the waffle tray, precisely positions the probe, and then uses the laser to bond each probe. After bonding the tool performs an optical inspection and measurement of each probe.
Currently each probe needs to have a 0.5 mm square base for the equipment to handle the probes and to absorb the reflow energy. In addition, the base of each tip needs an alignment structure for the vision system to properly measure its position. Both of these features are required but Pac Tech is working on reducing the size of each.
Using this equipment, Pac Tech has achieved an 80 µm pitch with a process time of 10 seconds per cantilever when run in an “engineering mode”. In a production mode, the post bond base and tip inspection can be skipped saving four second per probe. It is possible to achieve positional accuracy of X/Y +/- 2 µm and Z +/- 4 µm for each probe across a 300 mm diameter substrate.
Also demonstrated was the ability to remove and rework a probe using laser heating. Rework requires probes with solder already attached since the normal process uses the Solder Jet to apply solder to the substrate. In the rework case, they use the Solder Jet to add solder to the base of cantilever before placing the replacement probe.
- Since 80 µm pitch was demonstrated, what is the minimum pitch achievable with solder? The probe is 30 um wide, so 60 µm is currently the minimum pitch. The substrate pad needs to be the same size or slightly wider than the probe.
- How do they load very small probes? This is a very specialized and customized part of the system. This is why Pac Tech developed the whole assembly line – which is the three tools discussed. The Cantilever Sorter removes the springs from the plating substrate where it is tethered in two points. This machine uses a different laser to cut these connectors away. Then the Cantilever Bonder (placement tool) uses vacuum on the hump and on the top of the probe – in a “L” shaped bracket.
- What are the temperature limitations on solder? The tools do not heat the entire substrate so the process can use low temperature substrates with high temperature solder.
- If the probe card is going to test at 150 C will the solder soften? No, since the process can use high temperature solder such as SAC or AuSn which will not soften.
- How to manage multisite? Multisite design is a limitation of the build up process since it is not possible to do fine pitch from all sides since the tools need space to hold the probes.
- After solder process are there any cracking issues? There have issues with ceramic substrates where the metallization (layers of metal used to form pads) was not providing sufficient adhesion for the solder.