Tuesday, February 15, 2000
Low cost bench top memory module test handlers has been around since 1990. As of today, module handlers are mostly using the conventional contactor fingers at the test site. These 1.5 inches long contactor fingers (gold plated beryllium copper) together with the tester transition board causes impedance distortion and signal degradation during high frequency testing.
The next generation memory modules like PC133 DIMM, Rambus RIMM, and DDR DIMM will all be running at higher frequency. The high frequencies render the existing memory module handlers unusable. There is an urgent need for a low cost and reliable memory module handler that will handle high frequency testing at 133Mhz, at 266Mhz and even at 800Mhz.
Back in early 1999, CST did a survey with the memory module manufacturing industry. Memory module manufacturing managers were asked to rank the importance of features on their handlers. The information opened our eyes.
* “Simple-to-use” ranked at the top of the list. Manufacturers have hard time hiring qualify technicians. As a result, most of the setup and usage of the memory module handler falls onto the line operator who has only a high school diploma. They basically wanted no button, no learning curve, no setup and no adjustment when they change between the many types of module everyday.
* They want Reliability and no maintenance. Their major concern is downtime that causes them to miss schedule and lose money. They want the machine to be a workhorse with durable and reliable parts.
* They want us to take away the conventional test contactor (beryllium copper fingers) not only because it causes high frequency measurement degradation, but also for the high cost of replacement.
* Manufacturers also want the fastest handler for efficiency. They just want most “bang for the buck” to justify their investment.
* Static charge, scratch, and stress on the module under test are also high concerns on their list.
* Surprisingly, cost of the handler is not of major issue in comparison to throughput and reliability. Manufacturers seem to be looking for productivity as the number one factor.
MEMORY MODULE HANDLER COMPETITIONS
Currently, there are several memory module handlers on the market. Survey of users unveiled several unfilled areas:
* Users are tired of the difficult procedures in changing over from single side to double sided memory module and vice versa. They want a fast and easy way of changing over.
* Most handlers on the market cannot test module that has capacitors or resistors on the reserve side. Problems also arise when testing module with uneven bottom side.
* Handlers have problem when testing thin profile modules, it causes it to drop unevenly and fool the platform sensor.
* When the output-drop is too high, it damages the other modules already in the bin.
* Users found that sometimes contactor misalignment or mismatch with tester adapter causes intermitting and fall out in the testing. Besides, contactor pins are too easy to break and too expensive to replace.
* Due to the tight pin-to-pin pitch on the Small Outline DIMM module (SODIMM), there has been no sure handling solution for them.
Apparently, there is still a wide open replacement market for an automatic module handler that can overcome all these problems.
The second problem is Intellectual Properties. Since memory module handler has been around for the last 10 years, one has to wonder if any patent was filed on some of the technologies. A careful study uncovered a US patent specifically on memory module handler. After detail study of the documents, our legal counsel had given his advice and the green lights to go ahead.
RESEARCHING AND TESTING OF NEW IDEAS
In order to overcome the conventional contactor, our engineers looked at using a direct test socket plug-in at the horizontal position. It does not only solve the contactor problem, it also clears the test ambiguity between manual test and automatic handler test since the same socket is being used. However, the anticipated question is “how often do I have to change socket?” The engineers knew that they had to proof it.
The commonly available Yamaichi 168pin heavy-duty DIMM socket is used. The manufacturer specification allows 10,000 manual insertion cycles. A special fixture was built to horizontally punch the memory module into the socket and then withdrew it gently. The insertion continued until the module failed the test. The result was a surprised 200,000 cycles. Similar result was also obtained with the 144pin SODIMM sockets. The engineers repeated these tests and the result was consistent. Their explanation was that the Yamaichi socket would last a whole lot longer when the insertion force is regulated and no jerking was applied like the case of hand insertion. With the result on record, the engineers concluded that the socket would work for more than 100 days in the normal production environment. Together with the socket receptacle to allow quick replacement, it would be a much more cost-effective method when compared to the conventional contactor.
In order to fulfill the reliability requirement of the new handler, the engineers were also told to “over design” all the critical parts. All parts in contact with the fiberglass PCB were constructed with hardened material such as stainless steel, plated brass and nylon. All moving joints were on bearings for long lasting. Vigorous tests were also applied to gauge wear and tear of the parts.
Module Manufacturers also had their question on difference between a conveyor belt system and a direct drop output system. A complex analysis based on engineering data was performed. The conclusion was that a properly designed conveyor system out performs a direct drop system by the margin of 3 to 2. While the direct drop system required seconds of mechanical settling time, the conveyor system can output modules faster and safer.
The size of future modules was also a major design criteria. A thorough discussion with the JEDEC memory module committee (the standard setting body) and the socket manufacturers concluded that modules will not likely be more than 5.25 inches long and 1.7 inches height in the next 5 years.
The ROBOFLEX AUTOMATIC DIMM HANDLER
With all the considerations settled, the result was the birth of the new RoboFlex Handler.
The RoboFlex Handler has a vertical input tray hosting the modules to be tested. The modules are then dropped one by one onto holding rails. A pusher-system grabbed hold of the memory module by the notches on its two sides and slides the module horizontally into the test socket. The test socket is an integral part of the memory module tester that mounts on the back of the handler. At the completion of the test, the module is pulled back out of the socket and drops down onto the conveyor belt. The “good” modules are then transported to the right hand exit by the conveyor belt. For the “bad” modules, the conveyor belt reverse direction and exit them through the left hand of the handler.
Parts reduction And Performance Enhancement
Having a working handler prototype is not enough. Refinements must be made to get a great product. The engineers looked at different angles:
* Reduce the number of mechanical parts to make it simple to produce and easy to understand. While the conventional memory module sigulator requires 2 sets (4pieces) of air cylinder to work, the engineers were able to reduce it to 2. Thus saving space and increased reliability.
* Since a microprocessor controller is embedded into the system, there must be functions that can better utilize its intelligence. The engineers were able to control the motor in the reverse direction for output exit. This enhancement practically eliminated a mechanical output director and gain additional reliability.
* On a conveyor belt system, problem arises when the module falls on top of the belt clip. The belt clip is the little piece on the belt use to push the module along the travelling rail by the edge. When the module accidentally falls on top of the clip, the module does not travel properly and thus cause jamming. The engineers were able to use the microprocessor intelligence to locate the position of the belt clip and were able to always align the belt clip out of the way of the falling module.
* Unanticipated events always happen with a mechanical device. Either it is a module loaded up-side-down or a module that has irregular shape can cause a jam and an automatic timeout of the machine. In the old time, the operator had to shut down the system, troubleshoot, clear the problem and then restart the machine. Instead, the RoboFlex handler has a one button reset. Under any condition, the reset button causes the belt to sweep forward and backward to clear the channel. Once the channel is cleared, it resumes normal operation sequence. This is a smart way of saving operator frustrations.
* For safety purpose, an emergency button is included. At the push of the emergency button, all operations will stop and all air cylinders will return to the “release” position avoiding any possible injury.
UPGRADES AND OPTIONS
Since users want minimum setup time when changing from one size module to the other, the new RoboFlex handler is built with tolerance and flexibility. The standard 168pin handler can accommodate both single side and double-sided modules without any adjustment. It can also work with any module height from 0.8 inches to 1.75 inches. Since the future DDR module, Rambus module and PC133 modules are in identical length, it can also test all of them without any modifications.
Changing the setup to test the smaller SODIMM modules are made easy through the easily swap “pusher module”. The “pusher module” includes the input tray, the sliding rail, and the singulator all preset to the small size module. The entire change over can be done in less than a minute.
PROTECTING THE IDEAS
We recognize good idea when we have one. This is why a patent was filed covering a good number of claims to protect them from being copied. We anticipate that will give the inventor the proper credit in the long run.
By: Cecil Ho
Copyright © 2000 CST, Inc. All Rights Reserved