Issue link: https://iconnect007.uberflip.com/i/1305670
40 SMT007 MAGAZINE I NOVEMBER 2020 be taken. Are you starting to see that sort of interaction show up in customer installations? Webb: Yes. We have employed something along those lines for some time with our Sherlock algorithm. What the Sherlock algorithm does is analyze what has been tested so far. If you're troubleshooting ICT failures, for example, and the board has gone through AOI and ICT, but hasn't yet gone through functional test, if there are unknown failures that are encountered, the Sherlock algorithm will give you likely places to look based on what has been tested and what hasn't been tested yet. That's a little bit about what you mentioned about using AI. It's using some information that you know to make some educated guesses about what the failure could be. If there isn't already some- thing in the database that says, "This partic- ular problem is fixed by this," it can lead the technician in the right direction based on what has not been tested yet. Horner: There are actually two challenges. One, if you don't have any homogeneous equip- ment in your line, where they're communicat- ing to each other, a seamless flow might not always happen as we would like because there are variances in the protocols talking to each other and might not be fluent. Each of them is going to have their own neutral talking lan- guages. Two, if you have two pieces of equip- ment that are different makes or one is very old, you would still have the same communi- cation issues. When you start looking at the tools like Aster, Will is going to be looking at it from a point of view of test. TestWay and QUAD are going to be giving feedback to the manufacturing line. The value of this software is that you don't have to have a homogeneous line or have that synergy in your manufacturing line. You can now say, "I have a Keysight 3070 and a Goe- pel boundary scan giving us feedback on the line," and they can give that feedback to the command tower or the process people, which will allow them to tweak as they go along. I think that's a hidden value we see as a con- tract engineering service. The contract manu- the CM, and say, "Build and test this for me." A CM will have a more difficult time identi- fying how to pay for the tool. But if you start looking at reducing the gaps in coverage, you reduce your faults going out the door that are going to come back. You look at eliminating the redundancies and can calculate your ROI right there. The tool still pays for itself. There are two different ways of looking at it, but it has been done by both. We've had a lot of cus- tomers of both variants. Johnson: That sets up the next step, which is once you have that kind of information and a facility using your tool to set that up and get- ting the test results back with great coverage, that is just waiting to be used upstream in real- time as process optimization. Webb: Correct. We have another part of our toolset called our QUAD tool. The easiest way to understand what it can do for you is auto- mation of the repair loop. As your boards are flowing through your production test, all of the defect information and everything is being col- lected in a centralized database. When a board fails ICT, for instance, the troubleshooter can pull up that failure ticket from the database. They can click on the failure, and that failure information will be shown visually to them on our viewer so that they can see on the layout where the part is and where it is on the sche- matics. They can also access any prior failure information. If the past five times that C10 has failed, it has been the result of another component on the board U5, they'll instantly see that infor- mation brought up in front of them. And you can see how that would help the troubleshoot- ing loop and the repair loop and help the tech- nician achieve a much quicker result in figur- ing out failures. Johnson: A number of companies are using CFX, pulling the big data, looking to bring AI, automated problem solving, or issue response to the process. Now that you have statistical data, that begs for some change in the manu- facturing process to be identified and steps to