Troubleshooting a circuit is easy, right? All you need is a couple of hands to hold the probes, another hand to twiddle the knobs, a pair of eyes to look at the schematic, another pair to look at the circuit board, and, for fancy work, X-ray vision to see through the board so you know what pads to probe. It’s child’s play!
In the real world, most of us don’t have all the extra parts needed to do the job right, which is where something like CircuitScout would come in mighty handy. [Fangzheng Liu] and [Thomas Juldo]’s design is a little like a small pick-and-place machine, except that instead of placing components, the dual gantries place probes on whatever test points you need to look at. The stepper-controlled gantries move independently over a fixture to hold the PCB in a known position so that the servo-controlled Z-axes can drive the probes down to the right place on the board.
As cool as the hardware is, the real treat is the software. A web-based GUI parses the PCB’s KiCAD files, allowing you to pick a test point on the schematic and have the machine move a probe to the right spot on the board. The video below shows CircuitScout moving probes from a Saleae logic analyzer around, which lets you both control the test setup and see the results without ever looking away from the screen.
CircuitScout seems like a brilliant idea that has a lot of potential both for ad hoc troubleshooting and for more formal production testing. It’s just exactly what we’re looking for in an entry for the Gearing Up round of the 2023 Hackaday Prize.
Did you read the part where it says you can pick out *testpoints* from your schematic to do the probing on? With a good automation toolchain, this turns the boring and time-consuming process of basic function testing into something you hand off to a robot. I would’ve loved this capability on any number of projects over the years.
I’ve had bed-of-nails fixtures built for actual products, that cost $10k a piece. But the upshot was that we had every single board that rolled off the assembly line individually tested. And that whole testing procedure took like 3 days for about a thousand boards. For small-batch boards, a system like this could mean you pay a living wage for whoever is doing the functional test (probably yourself, if you’re a small operation), or that you don’t risk losing all your profit to RMAs, because you couldn’t afford a functional test on every board you ship.
The reality is that either you build a bed-of-nails fixture for every little board you plan on selling, or you build something like this (which is not new, every *professional* EE I’ve met has designs in their head for exactly this).
“Troubleshooting a circuit is easy, right? All you need is a couple of hands to hold the probes, another hand to twiddle the knobs, a pair of eyes to look at the schematic, another pair to look at the circuit board, and, for fancy work, X-ray vision to see through the board so you know what pads to probe. It’s child’s play!”
Mutant octopi.
I remember a YT video of pcbway where they had one big version ,the whole thnig was mounted on a wall and the pcb’s were placed vertically, so they were looking for errors during production. Nice build!
Does it only parse Kicad files? Interesting, but many of us use Altium.
Uh, no. “Many” of us do not. I have literally never seen an Altium file in an open hardware project, or read about it mentioned here on Hackaday.
The community has resoundingly backed KiCad, and failing that, there’s just the hold overs that are still on Eagle.
Stop, stop, he’s already bankrupt from the $30,000 monthly subscription.
So…
Just by looking at the schematic, it knows where to probe the circuit board?
It’s amazing that it doesn’t need to know how the circuit is laid out on the PCB.
Otherwise it seems like an important step or two is missing from the description.
The connection points on a schematic are tied to respective pads for those components on the CAD side. Picking the trace between LED1 and R1 is enough information to locate the probe at the appropriate pad of R1 that goes to LED1 (or the LED1 pad that goes to R1)
So, while it doesn’t know the path of a trace or if there is a probable (probe-able) via or test pad on it, it does know locations of the pads where the parts are placed. The trick then is just knowing where on the pad to probe for the intended schematic circuit. You would want to probe away from the part to prevent damaging the part or the probe.
This is essentially a DIY version of a flying probe. The one we have at work was $160k, has 4 heads on magnetic rails, and contains an array of measurement methods. It also has several power supplies and frequency generators to inject signals/power into the board so you can probe boards while “live”. It is insanely fast and powerful. I like the idea of a DIY version because the fundamentals of a flying probe test machine are worth knowing.
Thank you for taking the time to explain this to me.
Verry cool indeed. For smaller projects and with a logic analyser, i wonder if a more manual setup but with more probes would be faster?
A bit like the tradeoff between CNC/manual work, both have their place.
Maybe you could supplement this with a thin frame for placing manual probes?
Reminds me of Huntron Robotic Prober machine
Never had such a problem! 300 clicks to measure between 2 points
We had a flying probe machine, that eventually got upgraded with the maximum number of probes (more than 4) and double-sided operation. It was a fantastic machine to watch and was great at troubleshooting. It sat prominently in a spot that would allow employees and visiting guests to see that the very expensive machine was used most by the engineer assigned to program it to manage his stock profile and read news on the internet.
After seeing that, I made sure every test bench I had a part in designing couldn’t access the internet without jumping through a whole lot of hoops, and that could only be done by the administrator whose password was well-guarded.