EEVblog #552 - DFM Automated PCB Panel Testing

[EEVblog]

Dave demonstrates how to add testing capabilities to your PCB panel for easier production testing. Using an example of his new uCurrent design, and also a commercial product designed to be automatically ATE tested.
Essentially Part 3 of the PCB Design For Manufacture series.
PCB DFM Part1:
PCB DFM Part2:
uCurrent Test Jig:
Testing uCurrents:

! Private video

[poorchava]

"this video is private"

[JoannaK]

Video #552 is Private.. 

[jancumps]

I could watch it.

[daqq]

Thanks for the video Dave.

Question: Why not use pogo pin mated with test points instead of the connector on the panel? It's cheaper per panel (exactly 0.00 per panel), does the same job. You can then make a quick, small test connector with pogo pins that mates with it. Or some such.

[qoole]

According to your circuit diagram your -IIN (J2) is directly linked to your -V output (J4) so that means you only really need 5 test connectors as the ground is common. (Shame it can't be shared with the battery as then you'd only need 4 connectors) So if you were to sacrifice one tiny piece of top-side real-estate with a tiny via in the corner or if you chance it and try to make two traces fit you could make an ENTIRELY automated test rig. Failing that the other up side is it would mean the person doing the testing would only need to plug one banana plug into the +IIN connector.

[sivalley]

I had a bit of a derp and didn't wait the whole video through before posting a comment so I'll copy-pasta it here now:
Silly passing thought looking at your schematic and routing limitations and I thought I'd toss in my two cents:

Ideally it looks like you could knock down your test connections to 5 points and with a little clever cover make it not destroy the top surface continuity.

Mandatory traces on the bottom side for input positive, output positive, battery positive, and battery negative leaving no trace pairs along the ear to short out.

The trick for I/O negative; since they are tied together and looks like they are your flood plane on the top side would be to flood it "out" across all 10 units and viola!  You can test all ten simultaneously after all.  Of course this might be an issue when it comes to the routing since it might leave a nasty copper edge.  I have no idea if your manufacturer can run a 90 degree bevel for the routing cut to leave a cleaner edge if you flood the whole top surface.

Final connection could be accomplished with a 24 pin (20 I/O positive pairs, 1 I/O common negative, 2 battery +/-, 1 NC key pin) IDC header since you obviously don't want to parallel all those sense inputs or device outputs. ?

[mikeselectricstuff]

Thanks for the video Dave.

Question: Why not use pogo pin mated with test points instead of the connector on the panel? It's cheaper per panel (exactly 0.00 per panel), does the same job. You can then make a quick, small test connector with pogo pins that mates with it. Or some such.

An Edge connector would be a more universal solution - you can use pogo pins or an edge connector - as you probably don't need a full-size testbed, a clip-on edge con on a cable is probably the better solution.

[EEVblog]

Question: Why not use pogo pin mated with test points instead of the connector on the panel? It's cheaper per panel (exactly 0.00 per panel), does the same job. You can then make a quick, small test connector with pogo pins that mates with it. Or some such.

Yes, I may not actually use a connector, it was just the most convenient footprint to put down in a hurry.

[Isaac]

Hi, Dave,

I really enjoy the DFM episodes and I'm sure others do too. Why don't you make a series of videos about DFM, with not only the DFM itself, but with related stuff as well like industry stories, more visits to manufacturing houses and etc. I'm sure some manufacturing houses would love the advertising in australia and in other places since your vlog is seen by lots of Hackers, Makers and students and professional engeneers out there.

[peter.mitchell]

I'm curious of this, as the routing isn't precise, and it is just for test connections, what are the downsides of using a very wide track that will inevitably be slightly routed into, but at least never be fully cut (and of course, the 2 track variant too). What are the chances that the routing tool will tear the copper? I understand on things with metal housings this may be an issue because of exposed edge copper, and on the top layer it would be visible, but since this is the bottom layer?

If you're not sure of my explanation, this image may help; https://dl.dropboxusercontent.com/u/677635/explanation%20device.bmp

[cidcorp]

I was going to ask about using pogo pins as well.  As it stands would this require you to solder a single row header on each panel?  Not something I'd want to do on each panel.  At least with the edge connector contacts there is no additional step - simply slide in and test.

If you are limited to 4 contacts coming from the board you could make the power connection simple bare contacts on the bottom of each board, giving you full use of the 4 corner traces.  Then you apply power to each board one at a time marking pass or fail.

Also with peters idea, assuming the copper 'munching' won't damage the routing bit, you could get 8 traces out of each board, 2 per corner and not have to worry about anything.

This is an awesome series of video - keep them up.

Chris

[peter.mitchell]

Also with peters idea, assuming the copper 'munching' won't damage the routing bit, you could get 8 traces out of each board, 2 per corner and not have to worry about anything.

I'm not worried about the routing bit, copper is soft compared to the harsh abrasive that is fibre glass. I'm more about whether the routing bit will tear the thin layer of copper from the board.

[owiecc]

Very nice silkscreen on that first board (3:00). There is a matrix of pads and some external ones have nice labels attached (GND, 1V2, GND, 1V8). How did you do these labels? Plugin for Altium or a custom component?

[TheEPROM9]

Dam right about the YouTube comments system.

Rather clever solutions to what is a complex problem. What automated logic are you going to put on the test jig PCB, thus how automated will it be?

[cthree]

Dave, have you considered using compression type board-to-board connectors like this one from AVX? http://www.digikey.ca/product-detail/en/009258008004062/478-5494-6-ND/2024974

It would require nothing but a simple footprint on the DUT and could be activated by simply pressing the panel down on the test jig. SAMTEC makes some which allow for as much as 10mm between boards. I would think 4 pins to align the panel with the holes in the corners of the panel and them simply press it down. Pogo pins seem like a lot more work and would require a more complicated jig.

[free_electron]

Samtec has pogo clusters. Samtec pcp series.

All you need on the board is landing pads.

The carrier card would have a pogo cluster and off you go.
You could make a big carrier that can hold the entire panel and test all 10 at the same time. Hoom up a 34970 and the keithley current source, make a small testprogram in vb controlling both machines over a usb to rs232 link and bingo. No flicking switches, no nothing.
Youd need a relay card and a multiplexer card.

[mcinque]

Very nice idea!! thx for sharing!

[ion]

... I may only be manufacturing a couple of thousand of these things...

So you won't be crowd funding these?

[manu]

Hello,

I'm a big fan of test jigs and I would suggest a standard .1" right angle connector rather than, as other subscribers said, pogo pins which are difficult to mount mecanically on a board (in my opinion).
You can make the right angle connector removable from your panel if you misalign alternatively each of the pin on the panel, so as JP3 on the chipkit uno32 board (this connector is used to plug a PICkit2 or PICkit3).

Misalignment allows the connector to be held in place (temporarily, during the test session) and making quite good contact without soldering it.

If you also do that on your test jig, you can have a replacable connector (without soldering!) you can swap it with a new one after some uses.

It's a little thing but it prevents you to solder a connector each time you have to test a panel.

[arildj78]

At 28:36 we can see the test traces coming into the board. Is there a reason to bring the trace a few millimeters into the board, before bringing it out towards the edge again?

Arild

[sergey]

Cool video, thanks for it!

The question tho. Top layer is connected to "virtual" ground which is also connected to "-" banana jacks. Now. why not to use 4 bottom corners to trace positive power, current input, voltage output and route ground connection from the top layer? You wouldn't need vias for this and you might add connections to all 4 corners on the top level just for symmetrical look.

Or you think it also breaks look-and-feel?

[steve_w]

Thanks Dave,

Most informative, love your work.

regards

Steve_W

[nitro2k01]

Cool video, thanks for it!

The question tho. Top layer is connected to "virtual" ground which is also connected to "-" banana jacks. Now. why not to use 4 bottom corners to trace positive power, current input, voltage output and route ground connection from the top layer? You wouldn't need vias for this and you might add connections to all 4 corners on the top level just for symmetrical look.

Or you think it also breaks look-and-feel?

I thought about this too. I think the big reason is that you would ideally need 10 separate test current srouces per panel.  A different option would be to use a single current source in parallel for all the boards and rely on the parallel resistance of the shunts to divide the current equally, but that's a bit crusty and just asking for trouble.

[JerTheSmartAlek]

I really want to see a quick little demonstration of your test system, perhaps also with a quick use of the old test method for comparison, maybe as a little benchmarking? I dunno, I just really love it when we see you doing board design.

[kaindub]

Dave, you have connected all the negative output terminals together. Is this correct? The negative output terminal is your virtual ground derived via an op amp from the battery. There may be a small potential difference between the 10different circuits, which will show up as an offset on the output. Since your current is a precision measures, won,t the offsets upset your tests?
Robert

[SeanB]

Simple multipoint test jig is to use a unmilled panel and populate the current points with banana plugs that fit the test panel holes, along with a relay bank or switch to select the unit under test for current source. That way you place it on a bed of pins, plug in the test connector and start testing using the switch to select the device under test for current. when finished testing simply pull off the test jig. Using an unmilled panel and cutting unnedded traces saves spinning a test board specially for this, you only need one unmilled panel.

Sure there will be a market for those panels unmilled as wall art. Might be a good funding project there for excess boards. I would take one or two for $20 each.

[rs20]

Hey, awesome video, I'm really looking forward to seeing this working in the flesh.

One little potential bug spotted by a guy on YouTube though -- you've got all the virtual grounds commoned, which means that if (for example) one board's virtual ground op amp output pin was not soldered down (or basically any other problem with the virtual ground section of the circuit, really), that board would still pass the test because its virtual ground would be overruled to a correct value by the other nine circuits. Once the boards are broken out, it'd become dud.

Probably a very pedantic corner case, but kinda worth being aware of?

[EEVblog]

Probably a very pedantic corner case, but kinda worth being aware of?

Yes, I realised this potential issue myself after I sent the board away, which is always the case.

[rs20]

Probably a very pedantic corner case, but kinda worth being aware of?

Yes, I realised this potential issue myself after I sent the board away, which is always the case.

Bloody Murphy!

[centon1]

Could someone please provide a link or IC number to Dave's 'window detector' reference so I may understand it better.

Thanks.

[RobbieC]

For the test jig interconnect, this may also be an option:

https://www.phoenixcontact.com/online/portal/us?1dmy&urile=wcm:path:/usen/web/main/products/subcategory_pages/pcb_terminal_blocks_and_pcb_plug-in_connectors_p-11/495ba453-0024-4c4a-bf61-897a8daa650e/495ba453-0024-4c4a-bf61-897a8daa650e

I was speaking with our rep yesterday morning and he was showing me some various solutions they may have for one of the test jugs I'm currently working on.

Unfortunately I think the pad size requirement may be a bit too long for your needs. If the connector life and cost is good enough I may use it to reduce our interconnect replacement costs.

That being said, those right angle headers are cheap enough for your test volumes and if you've soldered as many as I have, they're pretty quick!

Cheers!
-Rob

[TheWelly888]

Ah! You made me feel dirty when you used heavy pixellation on that potential nerd porn!

 

Actually, I understand that particular board is commercially sensitive so fair dos.

Informative video as usual, Dave! 

Yeah G+ really felches indeed!

[nitro2k01]

Could someone please provide a link or IC number to Dave's 'window detector' reference so I may understand it better.

Thanks.

You may be interested in watching this:

[centon1]

Thanks Nitro...

I misunderstood the context of the 'window detector' reference in the PCB video but it is 'clear' now.

[mfeinstein]

Hey Dave,

Any reasons for doing this routing with the trace? I mean, why didn't you just continued it straight until the board edge and then turned it 45º down?

[jolshefsky]

 I just wanted to chime in and say this video is extra helpful for me right now. I'm working on redesigning a board for our product and wanted to add test functionality. Our production is incredibly low (specialized industry equipment; <100 per year) but it's a pain to debug a board without adequate test connections. When I say I'm redesigning it, I mean literally right now — to be completed in a month or so. So it was extra helpful to reconsider a card-edge break-out of test pins, and the suggestion to add power and current test/limiting to the test bench.

Excellent work, and thanks!

(By the way, your worry about the virtual ground may be alleviated some in the case that only one board would be powered at a time. I'm not sure what happens to the output pin of an unpowered op-amp, but I believe it floats, so with only one board powered at a time, you could actually test every board. Maybe once verifying the virtual ground on the most-likely-test to cause an error, power them all and run through quick.)

[majbthrd]

When using discrete pogo pins, what works for me to ensure they are mechanically aligned is to use through-hole variants of the pogo pins and order at least one more PCB than I plan to build.

The PCB will likely have some sort of mechanical holes for alignment to the DUT.  I use said mechanical holes to clamp two PCBs together (one PCB is the one being built; the second is the unpopulated spare).  Before tightening the screws between the PCBs, I add the pogo pins.  The through-holes on the unpopulated PCB act to align the pogo pins.  I tighten the screws, making sure the pogo pins have seated properly.  Then, I solder them in.  The PCB sandwich can then be taken apart, and the pogo pins should all be nicely aligned.


There was seemingly another feature of the test fixture (with PCI card-edge connector) that Dave showed but didn't mention.  He explained that the PCI connectors had adapter PCBs to 0.1" connectors.

Surely, a benefit of using the adapter PCBs (0.1" or otherwise) is that the PCI connectors can be swapped out (and the insertion counter reset) without having to scrap the entire test fixture PCB?

[EEVblog]

There was seemingly another feature of the test fixture (with PCI card-edge connector) that Dave showed but didn't mention.  He explained that the PCI connectors had adapter PCBs to 0.1" connectors.
Surely, a benefit of using the adapter PCBs (0.1" or otherwise) is that the PCI connectors can be swapped out (and the insertion counter reset) without having to scrap the entire test fixture PCB?

Yes, that was the idea, I should have explained that. Obviously if you know your test connector has finite life then you desing it to be easy and cheap to replace. In this case a simple adapter board you swap out in a few minutes. "Connector saver" gender changers also work a treat in those situations where you can use them.