Can You Trust Pick & Place Machine Ratings?

Phil
Welcome to Board Talk. This is Jim Hall and Phil Zarrow from ITM Consulting, otherwise known as Pick and Place, the Assembly Brothers. And today, our topic is on Pick and Place.

Jim
The topic comes from Q.R. in Phoenix, Arizona. Q.R. asks, "We're setting-up an SMT line. In estimating the capacity for equipment selections, why can't we use the Pick and Place manufacturer's rating for components per hour? They design the machines, don't they?"

Phil
Oh, yes, they do and they're trying to sell you machines. I think the real answer to your question is we don't live in an ideal world. Let's look at it from two major components.

First, the Pick and Place machine manufacturer, is trying to sell a machine. He's got to demonstrate the machine at some specific rate.

Cycle rate is ridiculous, because that would be putting down the same component in the same place in the same location, and we all know that's totally meaningless. So, generally, what they do is they have some rate they designate, and it might qualify if you're doing passives or simple ICs.

Generally, when you go and look at the machine, you'll see a demo board, and the demo board is usually optimize for the machine under ideal conditions. There have been attempts to come up with standardized boards. The IPC has worked on it, but they're not widely used.

Let's face it, from the manufacturer's perspective, before we get a lot of nasty-grams from them, there are no two applications that are the same. So where do you benchmark it? How do you benchmark it?

So, we have what we call de-rating factors. The biggest thing with the de-rating factor is the real world, your environment, and there's a lot of things going on here.

Let's start with your factory. If we're taking an eight hour shift, are you really working eight hours? Well, most likely you're not. Your employees take lunch. Do you give them breaks?

Maybe there are a few slave masters out there that don't in certain parts of the world. What about change-over time? That's probably one of the biggest killers. In addition to change-over time, even if you're running high volume of a single board, there's transfer time, or from the printer to the pick-and-place machine.

Typical conveyor transfer time is about four to six seconds? So, this is all adding up. There's of course, what we call bumbling time, or Dr. Rhine called, "floundering time." 

Jim
We should start with a more generic thing of down time, and down time is when you're attempting to run the machine, you're all set-up, there's a board waiting to go, or there's a board in the machine, but something happens.

Of course, we all know, what is the thing that happens most of the time? The feeders jam, the feeder runs out, the tape doesn't peel-back right. The machine stops, an operator comes over, ideally very quickly, fixes the problem, and pushes the reset button and away you go.

How often does that happen? Once an hour? Once every half hour? Once every 15 minutes? And you start subtracting this significant percentage of the available operating time, and that all takes away from capacity.

Phil
The sad fact of the matter is, if you were to look at overall efficiencies, typical machine through-put rates running 20% to 40% of what was originally specified?

Jim
The actual capacity rated over the time that the machine is theoretically available, if you can get above 25%, that's very good.

I wanted to follow-up on your floundering time. It is when a machine goes down, an operator comes and doesn't know what to do. So, they walk around trying to find somebody who understands the problem, trying to call a maintenance person, trying to call the supervisor, and the bottom line is: capacity is lost.

When you average these out over a significant length of time, what you need to do to estimate your capacity for a new line, you find that you're actually only going to get 25 percent of the machine's capacity.

Phil
So, that's the difference between what you might read on paper and might get in the real world. And that's about it for today's installation of Board Talk. Jim Hall and Phil Zarrow of ITM Consulting saying, whatever you do, don't solder like my brother.

Jim/Phil
Don't solder like my brother. And keep the kids from sticking their hands under the pick and place machine, for crying out loud.

There are many variables the impact the through put performance of a pick and place machine, many outlined above.

Another significant factor is the number of placements per panel the design you are building has. Machines have a ramp to peak throughput curve to overcome board transfer, PEC check, pre-pick inspect, etc. to reach peak throughput. I presented a paper a SMTA International in 2016 with an approach to characterize this. The title of the paper is "Modeling an SMT Line to Improve Throughput".

Line optimization software is a major part of the placement tool set these days. While every PCB is different the equipment suppliers should be able to give you estimated placement times using data files from PCBs you actually run. You would still want to de-rate that based on operating efficiency. Additionally knowing your capacity goals the placement suppliers can specify the equipment configuration you need to meet them.

You should always have the demo run with your component requirements, it's always more practical.

You are on the right track here guys. You allude to, but don't explain in detail one important fact that the IPC should really look at making clear. The test they use allows gang picking by vendors with inline spindle heads. Every engineer I have ever asked says this is where the real world de-rate happens and makes all vendors numbers look disingenuous; whereas only about 5% of a real program allows for gang picking like the vendors use to a high level in testing. That means they are running at their stated IPC rate only 5% of the time. Vendors with rotary heads do not de-rate anywhere near that level, whereas they pick from only one feeder at a time in their testing as well as true production. Example being Europlacer, whose IPC rate is within 15-20% of real world production rates. Greg's comments are also very true; We are that tortoise whose tools helps us beat the hare every time. Thanks for bringing this up for discussion.

The IPC has attempted to shed light on this topic by introducing the IPC-9850 standard a few years ago. What many people don't realize is that the 50+ page standard is more than just a method for leveling the playing field when it comes to placement rates.

The standard goes into defining the methods for determining other performance characteristics such as throughput (not to be confused with placement rate), reliability and accuracy. Most people know only of the placement rate portion of this standard as this is what most pick and place manufacturers publish in their specifications.

Most only know of the placement rate associated with the 0603 placement pattern as this generally yields the highest placement rates. While this pattern varies the spacing and placement orientation of the components in an "attempt" to level the playing field, it is still not going to provide a means for comparing placement machines when it comes to typical SMT assemblies.

There is only one way to know which machine will perform best on any given assembly and that is to have the pick and place manufacturer do a test with actual board in question and then guarantee the placement rate. That said, the fastest rated machine may not perform best overall because with today's lower volumes and more frequent changeovers, in many cases the tortoise beats the hare.