When talking about electronics today, people want it smaller, they want it faster and they want it cheaper. They want it to work right, from day one and keep performing for years to follow. I remember the standard used to be running manual functional tests that sometimes took hours or even days to complete. That’s not cheap, that’s not fast and having dedicated test points are starting to take up valuable PCB real estate. There are many times we have to solder a wire to some infinitesimally small point on the board to troubleshoot a signal or inject a voltage when diagnostics are required.
One of the ways we have come up with to deal with these challenges is investment into a flying probe machine, the SPEA 4060. I want to be clear before I get into this, I do not believe this machine to be a complete solution. I believe it is best applied as a complementary solution to a reduced functional test. The main takeaway from a unit passing flying probe is that all of the components it has access to are correct, oriented properly and the nodal signatures match the learned model, i.e., this board looks the same as the last board we tested, as it did for the first unit we programmed in. I say complementary because there are some issues a functional test will not catch and vice versa for the flying probe. Now you could build all that functionality into either, making each one a total complete solution, but it is a “low-hanging fruit” problem. There are some things that are just better suited as a functional test and some that are better performed on the flying probe. You could pound in a nail with a wrench, but why not use the right tool for the job?
Why is the flying probe better at some of these functions? For one thing, it has four needle points that are capable of taking readings off 0201 components. Even with the smallest handheld probes I have that is a difficult task. These probes make four wire kelvin measurements possible, in fractions of a second, allowing for more accurate measurements on smaller value components. And then we can not only control the “x” and “y” coordinates, but also the “z”. This lets us put probe anywhere in 3D space on the board, meaning we can probe inside of connectors. Each of the probes has a fixed angle: -16, -5, 5, and 16. So they are not moving straight down, but allow us to choose the right angle in the right situation. You would not want to probe down on an unsoldered contact and, by mechanically pushing the pin to the board, force the test to pass. Coming in at an angle makes that easier to avoid.
The flying probe is also fast. Not as fast as a “bed-of-nails” in-circuit test (ICT) fixture, but much faster than a human and typically faster than most functional tests. Several thousand tests can be performed in a few minutes. When I first started evaluating these machines, I saw several that used servo x-y “screws” that moved the probe heads around. I still see this technology on many 3D printers and other industrial machines, and it does work just fine. But the SPEA machine is “air-bearing”, meaning it uses magnets on the frames and the head arms to hold everything in position until air is used to push them apart. This is more like air hockey as these metal bars glide around. The arms are held together by magnetics, pushed apart by pneumatics. It is quiet and fast. The needles are very sharp and they are moving pretty fast so one of the features we use by default is called “soft-landing”. Once the instruction to place a probe is given, the machine places that probe there as fast as it possibly can. This leads to witness marks. They are not really that noticeable, but not desirable either. With soft-landing, it again tries to move the probe as fast as it can, until final descent onto the target. Then it slows its movement and gently strikes the point. This means fewer witness marks and is much safer for the smaller targets. Think about it, there is not much solder there on an 0201 and any type of witness mark will expand the solder out.
Because it can run thousands of tests per minute, it cuts down on time and that means smaller cost. The only non-recurring engineering (NRE) costs are associated with programming the device with the specific board to test and that software is guided. There are no fixtures that need fabricated or upgraded. The software is smart enough to calculate where adjacent shorts are possible and run the equations for parallel resistance and capacitance. We do need an engineer setting things up and giving the final okay on the tests that the software cannot resolve, but I would say that is typically less than 10%. That is less development time than functional or bed-of-nails, no external hardware than either of those, and the actual run rate per board is somewhere in between functional and bed-of-nails.
So what does flying probe actually do? I know before I used one I associated flying probe with circuit board manufacturing, i.e., we are just checking the bare board for shorts and opens. Yes, the machine does do that by default. It even cuts down on time by utilizing an algorithm to determine where those shorts are most likely and does not check every master net to every other master net. Only those that logically close are checked against each other. It also checks all passives like resistors, capacitors and inductors. It checks the low-level semiconductors like diodes, MOSFETs, crystals, operational amplifiers, etc. It runs a test called junction scan or JSCAN. This test checks for the electrostatic discharge (ESD) protection diodes in an integrated circuit (IC). The thinking is, if we can read that protection diode on one board, and not on another, then either that second board has a bad component or is unsoldered. Another test used for checking pins that may not have ESD diodes is called electro scan or ESCAN. This utilizes one of the two antenna probes in the SPEA machine and a normal probe. The normal probe makes contact with a master net connected to, for example, the ball of a ball grid array (BGA) that typically cannot be probed because it is under a part. The antenna probe goes on top of that same BGA and then the normal probe injects a radio frequency signal onto the master net. This will cause the die of the BGA to radiate the signal and be picked up by the antenna, as long as the ball is soldered to the joint.
I’ve had several people ask me why do need flying probe? Or, functional test makes this obsolete, right? Yes, we’re testing all the integrity of the build, just like the functional test would, but let me explain a scenario. You have 10K pull-ups on your digital pins. What if the wrong reel of parts was used or if there was a bill-of-material (BOM) error? What if those 10Ks were actually 1K? A functional test might pass and the unit would work fine for a while, but the extra stress on the component may cause a premature failure in its lifespan. Or, if the design is intended to operate on battery, it will draw more current leading to reduced battery life. Flying probe would have caught that.
Another point to bring up is that that scenario would be very difficult to diagnose. I have troubleshot the same board for days or even weeks trying to find that one thread that would unravel the problem. The flying probe may not specifically tell us exactly what’s wrong in every case, but it can narrow the scope of the diagnostics down considerably.
The flying probe ends up testing a lot of the printed circuit board (PCB). After that step is complete, we know that everything we have access to is correct and there are no blatant shorts, especially shorts on power rails to ground. The only part we are really missing at this point is to power up and make sure the entire design is capable of doing what it needs to do out in the field. That is why I think both flying probe and functional are needed. The flying probe checks each discrete component; the functional test checks the total of the design. To learn more about how ACDi’s testing capabilities can help ensure the quality of your electronics product, contact us today. In electronics manufacturing, having the right partner could make the difference between a brilliant success and a costly failure.
Principal Test Engineer