Reading time ( words)
Although there are currently no industry standards for specifying placement rate, the authors offer advice on how to determine a machine's realistic cph and how some machines perform closer to their advertised rates than others.
By John Pompea and Jim Gustin
The term placement rate, expressed as components per hour (cph), means different things to different people. To the manager responsible for producing PCBs, it may mean the total number of parts placed over an eight-hour shift, including component replenishment, downtime and set-up time. To the marketing manager of a placement machine manufacturer, cph is the maximum speed the unit can achieve. Typically, this is measured by placing chips on a bare board. It does not include nozzle changes, fiducial recognition or board conveyor time. If the machine features multiple spindles on the head, it is assumed that all nozzles will pick simultaneously from adjacent feeders, move via the shortest path to the board and sequentially place the components side by side.
Neither definition is realistic. The first includes overall shop efficiency, which is unrelated to a machine's capability; the second is irrelevant as it bears no resemblance to a typical production board.
What is needed is an industry-approved method of determining placement rate. Ideally, an independent trade association would provide CAD files and their corresponding component lists for several "standard" boards with each having different component groups. For example, one would have all chips, one all leaded components and a third a mix of the two. Every machine manufacturer could then run the standard boards and publish the placement rates achieved.
Two JobsThe marketing manager's job. Obviously, any machine specification that can influence a sale will get the marketing manager's attention. Since there is no industry standard to refer to, the marketing manager is free to devise a placement-rate test program highly optimized for the company's machine.The purpose of the test program is to make the unit look fast at trade shows and to advertise the highest (but still credible) cph possible.
The production manager's job is to be concerned with how the machines will operate on the production floor in conjunction with his/her people, products and schedule. The production manager must evaluate the technical claims of each machine with respect to the company's needs, and not rely solely on advertised specs or casual observations at trade shows. This is because if one looks closely at a machine's "dry cycle" at a trade show, it is apparent that no components are being picked and placed. The head is simply making a repetitive pattern from the feeders to a bare board. Vacuum sensing, component centering and correction moves are usually not included in dry cycling.
The Derating GameSales catalogues contain specification listings that usually include the heading "Maximum Placement Rate." Often, prospective buyers use this number as an easy way to compare the speeds of the machines. Most buyers know these rates are unrealistic so they derate them, generally applying an equal derating factor to all machines. This is a mistake, and can lead to serious errors. Because no industry standard for maximum placement rate exists, using a common denominator to determine the "real" placement rates of different equipment is inherently inaccurate. Furthermore, because of design characteristics, some machines perform closer to their advertised rates on real production boards than others. A few examples will make this point clearer.
Figure 1. The three-spindle placement head. Fitted to formerly low-speed machines, the multiple spindles have upgraded handling speeds to 18,000+ cph.
The Multispindle HeadThe Cartesian pick-and-place machine was relegated to the low-speed-market segment. Recently, this machine has made a jump in speed, primarily by mounting multiple spindles on the head (Figure 1). With advertised speeds of 18,000 cph and up, it has challenged the existence of the turret-type chipshooter.Achieving such speeds on a production board is predicated on running a job in a highly optimized manner. This means that all high-usage parts must be positioned side by side on the machine so that each time the head goes to the feeders to pick, it can fully load all nozzles on a single stroke. Each time such optimization is unachieved, the machine falls further away from its maximum rate.Most boards are without identical quantities of each component type. The ability of all spindles to pick simultaneously typically ends after a few placement cycles. The multispindle head must then either make several moves over the feeders to pick a full load of parts, or it will run half-empty. In either case, the machine will not run at its maximum rate.In reality, while no multispindle machine will run at its maximum rate in real production, some machines are better than others. Given a random number and mix of components, it is a mathematical certainty that a three-spindle head will place a higher percentage of parts using simultaneous picks than will a six-spindle head. This is not to say that the six-spindle head will run slower on an average board than its three-spindle counterpart. Rather, the latter will run closer to its rated speed, thus its maximum cph should be derated less.
Optimizing. Even if it were possible to optimize a particular board, the time required to reposition and add duplicate feeders to reach maximum speed may not be the proper strategy for every company. For example, if several jobs are run each day, requiring many of the same components, it may be better to leave the parts in dedicated positions for every job. The time lost in repositioning feeders would be greater than that saved by a higher placement rate.
Figure 2. Dual-beam placement with three spindles/head (left) is compared to a single-beam unit with six spindles. If parameters are uniform, the former machine will come closer in production to its rated maximum speed because the single header must make simultaneous component picks on every pick-and-place cycle.
One Head vs. TwoSome multispindle machines have one placement head while others have two (Figure 2). Their speeds cannot be derated by equal amounts. This is because in actual production, the two-headed machine will come closer to its rated maximum speed than the single-head unit. If all motion parameters are equal, a machine with two independent heads, each with three spindles, will exhibit roughly the same maximum placement rate as equipment having six spindles on a single head. Yet, in real production, the two-headed machine will be faster simply because its single-headed relative must make simultaneous picks on every cycle to achieve its rated speed. If it fails, it will lose time making random picks.
The dual-header can run at rated speed without making simultaneous picks. While one head is placing components, the other can be picking from random feeders without loss of time because picking from random feeders consumes about the same time as placing components randomly on the board.
Table 1 compares a dual-beam with a single-beam machine (as in Figure 2 with an equal number of spindles) running the same program. Each row in the table represents an equal increment of time between the two machines with both picking all components simultaneously from adjacent feeders. This is a "best case" scenario, which is typically how the maximum placement rate is achieved (but rarely sustained throughout a real program). Note that both machines finish placing part number 12 at approximately the same time. "Waiting to Place" represents the unused time of a dual-beam machine following a simultaneous pick; the single-beam machine has no such delay.
Table 2 shows the same program without simultaneous picks. This is a "worst case" example, and is what happens at some point on every board. The waiting time in Table 1 is now used by the dual-beam machine to pick randomly without affecting cycle time, demonstrating that when the ability to make simultaneous picks ends, the single-beam machine's speed will drop off more dramatically than that of its dual-beam counterpart.
The Factory VisitHow then is the placement rate determined?
The placement-machine manufacturer must demonstrate his machine's actual run time on a buyer's board. The supplier will want information in advance to gather all the equipment that will be needed. CAD or Gerber data that will permit a program to be created in advance is also needed.
The machine demonstration must duplicate an actual production run as closely as possible. This means a feeder is to be placed on the machine for every component part number on the board. It is okay to use dummy parts and to place on double-sided tape. If the components and stencil can be supplied, a full process demonstration should be requested.
While at the vendor's site, gather facts and impressions. Take a stopwatch and time everything that matters. This includes the time it takes to place all the components on the board, load a tape reel onto a feeder, load a feeder onto the machine, and unload a tape reel from a feeder after the run. Remember, operators must perform all these steps every time a job is changed. Do not estimate any of these times. Many machine manufacturers can time a production run using their software and can display the results. When using this information, know what the software uses as start and stop points.
The Local User VisitThere may be times when it is impossible to go to the factory to conduct an evaluation. The next best action is to visit the site of an existing user. This is far less desirable than a factory visit, because many users are uncertain of the actual placement rates achieved. On the day of the visit, recheck with the user to make sure boards will be run that day.
The natural inclination for most people is to observe an event and estimate the time it takes. The correct way to measure cph is with a stopwatch. Write a list of the exact component types and quantities on the board being watched. Record the time it takes to build the entire board and what is required to do specific sections of the board. During a typical placement program, the machine will place a number of components, change nozzles, place more components, change nozzles again and continue until completion. (Refer to the activity between nozzle changes as "placement strings.") Frequently, there are placement strings where nothing but chip capacitors and resistors are handled. Count the number of components in the largest of these strings and record the exact time required for placement. This will give a fairly good indication of the rate at which the machine will place chips. Do the same thing for strings that contain simple leaded components and fine-pitch parts. Also, record the time needed for tool changes. Do this several times.
These data can be used to estimate how long it would take a candidate machine to build one of the buyer's boards. The board must be analyzed by breaking it up into probable placement strings, separated by nozzle changes. Compute the time required for each string using the data acquired during the user visit, then add in the nozzle change times.
Other ConsiderationsThe methods used to maximize production will vary according to product mix and lot size. A company that runs large production lots will find it advantageous to optimize each job for maximum throughput. For them, the machine with the highest verifiable placement rate will be the machine of choice. Companies that run small lot sizes will focus on how quickly the machine can be set up. For them, ease of feeder loading and mounting is as relevant as the maximum placement rate. These companies may be better off buying a slower machine and spending the savings on extra feeders and feeder carts to facilitate rapid job changeover.
OEMs may have the opportunity to use one common feeder setup for all of their jobs. They may be better off putting two lower speed (and lower cost) machines in line, thereby providing the feeder capacity they need.
Evaluating machine placement rate is essential. Unfortunately, it is not easy to do because there are no industry standards. One cannot rely on sales literature, trade shows or casual observations. Instead, it is essential to time the candidate machines while they build real production boards. The money and frustration saved over years of equipment use will be the reward for the diligence exercised during these important evaluations.
JOHN POMPEA may be contacted at Contact Systems Inc., Miry Brook Road, Danbury, CT 06810; (203) 743-3837; Fax: (203) 790-6322; E-mail: email@example.com; Internet: www.contactsystems.com. JIM GUSTIN may be contacted at Cornerstone Technology, 73 Hat Shop Hill Road, Bridgewater, CT 06752; (860) 350-3029; Fax: (860) 350-3978; Internet: www.cornerstonetechnology.com.