21 Key Questions about Strip Testing 1. Is strip testing only good for very high volume products? It depends what you deﬁne as high volume! If you only have enough volume to keep one (1) 4-up gravity handler busy then it is probably more economical to stay with singulated testing. However, if volumes begin to ramp substantially, you will soon reach a volume point where the Capex required to add more singulated cells (another handler + another tester for each test cell) may well exceed the cost to put in a single strip test cell. So even if the strip test cell is only being utilized a small percentage of the time on a particular package, the test cost per unit may well be signiﬁcantly lower than with a series of singulated test cells. Here is a simple example to illustrate the point more clearly. For an 8L SOIC digital product with a 2 second test time in a 6x28 matrix lead frame (192 total parts), it’s well within the capability of a strip test handler to contact all 192 parts in a single plunge. This would require a tester with approximately 1600 digital pins which is well within the capabilities of today’s digital testers. The SH-5000 strip handler, at a 2 second device test time, is capable of testing 10.9 million parts per week, assuming 80% utilization, or just over 125K per hour. A typical 4-up gravity handler might be capable of testing around 850,000 units per week at a UPH of approximately 6,200. With this type of information, we can determine the best way to proceed. For example, if you really do need 10M parts per week, the math will tell you that you will need twelve (12) separate gravity cells (a handler + a tester) to produce that many parts per week. In that case, the Capex required to purchase 12 gravity handlers and 12 low pin count digital testers will far exceed the cost of a single strip handler and a 1600 pin digital tester. If you factor in the additional costs of the gravity cells in terms of ﬂoor space, maintenance, operators and power consumption, the cost advantage of strip test becomes even greater. On the other hand, if you only need 850K parts per week, the economics will clearly favor a single gravity cell. The crossover point where strip test becomes a more cost effective solution is somewhere in between 850K/week and 10M/week because you also need to account for the other process steps that are required for strip testing such as the process for lead isolation. MCT can help you analyze your operation and ﬁnd where the volume trade-off point exists for your particular product. 2. If the economics of strip testing are so compelling, why isn’t everyone doing it? This is a very relevant question and one that deserves a serious look at the reasons. First of all, more and more IDM’s and Subcons are doing strip testing because they have ﬁgured out that it is one of the best solutions available to help them reduce their overall cost of test. But one of the main reasons why some customers still have not adopted strip testing is due in large part to the way that the semiconductor test industry works. At most companies, a marketing team is responsible for forecasting the sales for a new product. If they are un-certain about the demand, the initial forecasts may be low and conservative. Management will want to get those parts assembled and tested for the lowest possible cost. For very low volumes, the best way to approach the problem is probably with standard singulated test using either a gravity, turret, or in some cases a pick-and-place handler with a small (i.e. low cost) tester, meaning one with just enough resources to handle the small degree of parallelism that can be achieved with one of the fore mentioned handlers. The problems start when the product becomes popular and starts to ramp. At that point, everyone is under pressure to ship more parts and to do it quickly. The easiest way to get more output at test is to simply duplicate the test cell that is already in place. That means buying more gravity, turret or PnP handlers and more small testers until there is enough equipment to meet the capacity requirements. It doesn’t take long before a signiﬁcant area of the test ﬂoor is ﬁlled with multiple singulated test cells and it is costing anywhere from 2X to 10X more per unit to test the devices than it would have cost if strip test had been implemented from the beginning! And that doesn’t even count the costs for extra ﬂoor space, extra technicians needed to maintain the equipment, extra operators to run it, extra maintenance costs and extra expenses for all the additional power those multiple cells will consume. MCT can help you analyze where the trade-off occurs so that you achieve the lowest cost of test possible. 3. It has been reported that strip test can result in high yields than singulated testing. How is that possible? Yes, that is true! Let’s talk about leaded packages to keep things simple. In general, for leaded packages, strip testing is done before the leads are formed. This means that there is a large surface for contacting which obviously makes it easier to land the pogo pins onto the lead in a good position. Because of the large surface available to the contactor, it’s easy to make good contact and this results in a very high O/S yield. One of the leading subcons reported at the Known Good Die Conference in October, 2009 that they had seen an average increase in ﬁrst pass yields of 3.2%! This was due simply to the fact that the contacting was better than in their singulated test cells. In singulated testing, the leads are already formed and the device is inserted into a test socket. This is where damage can occur and where O/S yields can suffer because of poor contacting. Yield increases of 2% - 5% are common in strip test operations and that obviously translates into huge cost savings. 21 Key Questions about Strip Testing page 2 4. Is it dangerous to form leads and singulate parts after test? Twelve (12) years ago, when strip test was just getting started, that was one of the primary worries. It has now been proven in production that the trim and form operation is an insigniﬁcant source of part damage. In fact, Amkor, in a paper presented at the Known Good Die Conference in October, 2009, reported that they had tested over 1 Billion parts without a single failure attributed to the trim and form operation. In general, the quality level of a strip test operation will exceed that of a singulated test operation. 5. But what about lead damage. Is that a problem with strip test? In fact, just the opposite is true. More leads are damaged in singulated test due to the operation of inserting parts into a test socket. With strip test, the testing is done with leads un-formed and no side loading occurs since the strips are not mechanically précised into position. The devices go thru the trim and form operation after test. The result is that leads never get jammed into test sockets during test. This eliminates the #1 cause of lead damage on leaded parts. Some companies have seen such small amounts of lead damage when using strip test that they have eliminated 100% lead inspection altogether. Lead inspection, if done at all, is almost always done on a sampling basis when parts are strip tested as opposed to 100% inspection that is generally required for singulated test operations. 6. 4-up and 8-up gravity handler change kits are very expensive, making it impractical to do changeover very often. Is it the same for strip test? No! If we compare change kits for gravity, turret and strip test, there are two important facts to point out. First, kits for gravity and turret handlers are almost always much more expensive than for a strip handler…anywhere from 2X to 4X more expensive in many cases. Second, the time to change a kit on a gravity or turret handler can take many hours. This is not so much due to the number of parts necessary to change as it is to the fact that there are so many subjective adjustments that must be made. And those adjustments must be made by a highly skilled technician. With the MCT strip handler, changing the handler from operating with an SOIC strip to a TSSOP strip, for example, is a simple matter. In general, it would take about 15-20 minutes for a technician to convert the SH-5000 handler from one type of strip to another. The reason is that the MCT kits have been designed for quick change. This means there are three (3) things unique to the MCT change kits. First, all parts are keyed or indexed so that they can only be put in one way…the right way. It’s impossible to put a change kit part into the handler incorrectly. Second, all fasteners are captive. This means that bolts and screws do not fall out onto the ﬂoor and get lost when parts are being changed. They stay captive with the part so they cannot be lost. And third, and most importantly, there are absolutely no subjective adjustments with the MCT strip handlers. You can literally change over from one strip to another in less than 20 minutes. The MCT strip handler is the closest thing there is to a “universal” test handler when compared to gravity, turret or PnP handlers. 7. The best temperature accuracy with most gravity handlers is +/- 2 Deg C. Is that the same for the SH-5000 strip handler? Yes and no. The temperature spec for the standard SH-5000 Strip Test Handler is +/- 2 Deg C. However, for applications requiring better temperature accuracy, MCT offers the HATS (High Accuracy Thermal System) option. HATS consists of a special test chuck with multiple heater zones and real-time microprocessor control of each zone. This enables HATS to hold the temperature of the test chuck very accurately; to better than +/- 0.1 Deg C in most cases. 8 Does HATS allow testing of devices at temperature accuracies better than +/- 2 Deg C? Not exactly. While HATS can control the temperature of the test chuck very accurately (as good as +/- 0.1 Deg C), it takes a bit more to control the temperature of the DUT more precisely. MCT’s thermal management system for the SH-5000 consists of HATS plus TIM (Thermal Interface Manifold) plus TCCF (Thermal Contactor Conditioning Frame). TIM provides heated air to the contactor and TCCF provides a heated frame for the contactor. These two optional sub-systems, TIM and TCCF, combine to make the temperature of the pogo pins in the contactor as close as possible to the set temperature for the test (and the same temperature as the test chuck). The whole idea is to make the thermal differential between the DUT’s and the contactor as small as possible so that heat ﬂow between these two parts is an absolute minimum. MCT has been selling the combination of HATS, TIM and TCCF for almost a year and has extensive experience on how to achieve superior temperature accuracy and stability during strip testing. 9. Tester utilization is reportedly better with strip testing as compared to gravity, turret and Pick-and-Place. Why is this? True. The reason is actually quite simple. Strip test provides a much higher ﬁrst pass yield which helps to eliminate the need for re-test. In addition, since a strip handler handles only strips and not individual devices, a strip handler does not have anywhere near the jam rate of a gravity or turret handler. Jams obviously take time to clear and that seriously reduces the amount of time the tester is available for testing. The combination of superior ﬁrst pass yields and tremendously better jam performance allow a strip handler to provide about 15% (on average) better tester utilization than a gravity or turret test cell. This obviously translates into large cost savings. 10. What is the maximum UPH of a strip handler? Great question and one that is often mis-understood by people not completely familiar with strip testing. The maximum UPH for a strip test cell is really a function of the tester resources available and not a function of the handler per se. The handler’s only limitation is that it is designed to handle one strip at a time, so the theoretical maximum number of devices that can be tested in parallel is equal to the total number of devices on the strip, assuming that the tester resources are available. The maximum UPH that can be achieved is therefore determined by the number of devices tested in parallel and the test time. MCT provides a UPH calculator and can easily estimate the expected UPH as long as we know the degree of parallelism that the tester can support, the test time for the device and the strip layout (number of rows and columns in the strip). 21 Key Questions about Strip Testing page 3 11. Can you give some typical examples of the degree of parallelism you can achieve with strip testing? The answer depends on a couple of factors, the most important being the type of device you are going to be testing and the tester resources available for testing them. For example, with a digital device like a Serial EE or a standard logic part, MCT has tested as many as 384 parts in parallel. This was possible using our dual strip version of the SH5000 which can accommodate two (2) strips at one time (the standard SH-5000 accommodates only one (1) strip at a time). With each strip of SO8’s holding 192 devices, this results in a 384 total parts. It’s is not uncommon for customers who run lots of digital parts to contact an entire strip in one touchdown. For analog or mixed signal parts, the story is slightly different. Here again, the determining factor in the degree of parallelism is the amount of tester resources available. For most analog or mixed signal applications, tester resources will limit the degree of parallelism to something between 8-up and 32-up . For Power Mosfets, parallelism of 4-up in a single insertion is now available with a special tester offered by one of MCT’s customers and in conjunction with an MCT Strip Handler or Film Frame Handler (depending upon the type of package). Since most Mosfet testing today is done at 1-up, this represents a large increase in parallelism and hence capacity for a single test cell. For RF parts, it’s yet another story. Because of tester resources and the limitations of testing RF devices in close proximity to one another, high parallelism may be something between 2-up and possibly 8-up. MCT has experience with RF and one of the advantages of a strip handler over a singlulated test cell is that the strip handler can accommodate a very large load board, making it much easier to handle all of the signal “plumbing” necessary to accommodate multiple RF parts during test. 12. Can the SH-5000 Strip Test Handler accommodate both leaded packages like SOIC and TSSOP as well as leadless packages like QFN, MLP and DFN? Yes, it sure can, but we need a bit of explanation to understand the limitations. The SH-5000 was designed primarily as a strip test handler, meaning that it was intended to handle leaded packages like SOIC, MSOP, TSSOP, QFP, TQFP, PDIP, etc, etc. As you may know, the key requirement of being able to test packages in strip form is that the individual devices must be electrically isolated from one another but still held in place in their lead frame. This is a straightforward and well known process for the traditional leaded packages mentioned above. The process is done using a standard trim and form machine but with a special die set that does the lead isolation cut. This process is often referred to as “punch isolation.” It should be mentioned here that QFN and DFN packages that are pocket molded can also be easily strip tested using the SH-5000. They are treated like other leaded packages in that they would go thru a lead isolation process via a punch process. Leadless packages like QFN and DFN (that are block molded and in a lead frame format) can still be tested on the SH-5000 as long as there is a method for electrically isolating the leads. This can be accomplished by one of several methods that are now available in the industry; 1. A half-saw process where the strip is put through a sawing process that cuts only lead frame but leaves the mold compound intact, therefore electrically isolating the leads. 2. A chemical etch process which again etches away the metal in the lead frame but leave the mold compound intact. 3. A laser isolation process which cuts the metal in the lead frame, but leaves the mold compound intact. Of these three methods, MCT’s experience is that the half-saw process is the most common and usually the least expensive. One word of caution….the half saw process is viable as long as the overall thickness of the lead frame is large enough such that the remaining mold compound (after the metal lead frame is cut) is strong enough to keep the strip held together such that it can withstand the handling process. As a matter of practicality, the thinnest half sawn strips that MCT handles today are on the order of 0.55mm (550 microns) thick. Since the lead frame is typically 0.2mm (200 microns) thick, this leaves 0.35mm (350 microns) of mold compound to hold the strip together. Our experience is that this is enough for the handling process. To run a thinner strip in the SH-5000 would probably require some handling tests to check the ability of the strip to be strong enough to be reliably handled. 13. How can testing of QFN’s and DFN’s be done if there is not a good half-saw, laser or chemical etch process available? If this is the case, the solution lies with MCT’s FH-1200 Film Frame Handler. With the FH-1200, the strip is mounted onto sticky tape on a standard 200mm or 300mm wafer ring. The strip (while mounted on the wafer ring) is then run thru a saw where the individual devices are completely isolated from one another via the saw process. We refer to this as a full saw process since the saw must cut entirely thru both the lead frame metal as well as the mold compound. After the saw process, the strip is cleaned and dried. The individual devices are now completely singulated but held in their appropriate place by the sticky tape. This entire process is referred to as testing on ﬁlm frame. The FH-1200 can accommodate both 200mm and 300mm wafer rings. The handler can then present these rings to the test chuck so that the devices can be tested in parallel, much the same as with the SH-5000. 21 Key Questions about Strip Testing page 4 14. Testing leadless devices on ﬁlm frame sounds interesting, but what if the tape stretches during the sawing or cleaning process such that there is tape distortion. Won’t that make testing of devices in parallel impossible? Yes and No. Some degree of tape distortion is inevitable as a result of the sawing, cleaning or handling that the wafer ring must experience during processing. However, the vision algorithms in the FH-1200 have been speciﬁcally designed to handle moderate amounts of distortion. It may not be possible to test an entire strip of devices at one time, but MCT has already tested 2mm x 3mm DFN parts with parallelism of greater than 140-up. This is possible because of the efﬁciency and robustness of the vision algorithms used to align sections of the strip. 15. For ﬁlm frame testing, if the requirement is to test 140 parts in parallel and each part has 8 leads, doesn’t that translate into over 33 kg of Z-axis force required (assuming 30 grams of force per contactor pin)? Yes, it does. This is still within the range of the FH-1200. In fact, the FH-1200 and the SH-5000 both have three (3) options for the amount of Z-axis force that can be applied. The standard FH or SH machine can go as high as 77 Kgf and both have options to go as high as 194Kgf if necessary. 16. Can leadless packages like QFN’s as well as Wafer Level Chip Scale Packages (WLCSP) be tested using the same handler? You can do both on the FH-1200. The machine handles either 200mm or 300mm wafer rings, so all that would be required is to mount the strip (in the case of a QFN) or a bumped wafer (in the case of a WLCSP device) onto the ﬁlm frame. The FH-1200 can handle and test both types of devices without any changes. It would be best to check with MCT about the feature sizes that can be contacted. One important thing to remember is that the FH-1200 is not a prober, but is designed more for ﬁnal test of leadless packages and bumped wafer level packages. 17. Can the FH-1200 accommodate probe cards when testing WLCSP devices? Yes. The machine can be used with a pogo pin style contactor or with an industry standard vertical or cantilever probe card. MCT tends to favor pogo pin style contactors when the pitch and density of contact points allows it. This is because pogo pin style contactors are typically less expensive than probe cards (especially as compared to vertical probe style cards) and tend to have 2X the life of a probe card. 18. Can the SH-5000 test leaded packages at ambient, hot and cold? Yes it can. Every SH-5000 is tri-temp ready, meaning that it can test parts from -55 Deg C to +160 Deg C. The machine can run ambient and hot out of the box. If you need to run cold as well, all you need to do is purchase MCT’s optional chiller. No changes are required to the handler. All you do is connect two (2) hoses from the chiller to the handler and it is ready to run Cold. In question #12, we discussed testing leadless packages like QFN and DFN on the SH-5000. If the devices can be electrically isolated as described, then they can also be tested at temperature on the SH-5000. 19. Is the SH-5000 ever soak limited when doing temperature testing? Other strip test handlers are soak limited and this is because they use a convection heat transfer mechanism to heat the strips. Since convection uses heated (or cooled) air to heat (or cool) the strips, it can take a considerable amount of time to bring the strips to the correct temperature. In order to avoid being soak limited, handlers that use convection heat transfer generally offer some type of chamber to pre-heat or pre-cool the strips before they enter the handler’s test section. The heat transfer mechanism in the SH-5000 does not use convection. It uses a patented conduction heat transfer mechanism where the strips are actually contacted by heated (or cooled) surfaces in order to bring the strips to the correct temperature. The advantage of conduction heat transfer is that it allows the strips to come to the correct temperature about 8X to 10X faster than a convection system. The SH-5000 uses three (3) soak stations that are internal to the handler. This guarantees that each strip is subjected to hot or cold for at least 3X longer than the test time for one strip plus the index times. The result is that the SH-5000 is never soak limited, even for the fastest test times. 20. Testing at cold temperatures with LN2 is often problematic. Does the SH-5000 use LN2 ? No LN2 is ever used in the SH-5000. MCT achieves its superior thermal performance with the use of a closed cycle chiller. The SH-5000, with the optional chiller can operate from -55 Deg C to +160 Dec C. 21. How do you maintain and manage the data from a strip test cell? The easiest way is to use MCT’s proprietary SmartTrak™ system for Electronic Strip Map Management. SmartTrak™ is a production proven system for managing the strip maps. It includes a rule-based system to guarantee that strips are processed in the correct order and at the correct time. SmartTrak™ also includes a full set of applications programs that allow the user to monitor the strip test cell from his own desk and do various levels of analysis on the data from the strip cell.
© Copyright 2018