-
EEVblog #588 - How To Do PCB Production Testing
Posted by
EEVblog
on 09 Mar, 2014 22:46
-
Dave shows how the assemblers will production test the µCurrent PCB panels using his new test jigs. And he will does a trial run on 50 µCurrent PCB's to determine the average testing time. What will take the longest - PCB testing, or wrapping, or packaging?
! Private video
-
#1 Reply
Posted by
krivx
on 10 Mar, 2014 00:11
-
So 2 failures out of 50? Is that typical for a product like this?
-
#2 Reply
Posted by
kony
on 10 Mar, 2014 00:27
-
Wouldn't be from an perspective of ergonomy more convinient just slap in small micro on the test jig and do test indications via buzzer ?
I'm sure i'd be pissed in just an hour by bending my back every single time to check LEDs, not to mention how easy is to overlook indication of bad board once worker slides in stereotype and gets bored.
Short beep for in specs board and some longer alarm signal for failed is making much more sense for me
-
#3 Reply
Posted by
TVman
on 10 Mar, 2014 01:01
-
So 2 failures out of 50? Is that typical for a product like this?
That seems average for something like A uCurrent.
that would be 100 bad out of 1000.
At least he tested 50 of the uCurrents
-
#4 Reply
Posted by
EEVblog
on 10 Mar, 2014 01:20
-
So 2 failures out of 50? Is that typical for a product like this?
When you are still trying to iron out bugs in the assembly process, yes, totally expected.
-
#5 Reply
Posted by
EEVblog
on 10 Mar, 2014 01:21
-
Both of those two faults could so easily be intermittent. I would be surprised if you don't get those faults slipping through the test phase.
That's a risk with any product. That is why you identify these and tweak the production process.
-
#6 Reply
Posted by
EEVblog
on 10 Mar, 2014 01:22
-
Short beep for in specs board and some longer alarm signal for failed is making much more sense for me
Ever worked in production test? Beeps are annoying!
-
-
Short beep for in specs board and some longer alarm signal for failed is making much more sense for me
Ever worked in production test? Beeps are annoying!
And the production floor can be fairly noisy.
-
#8 Reply
Posted by
BravoV
on 10 Mar, 2014 01:50
-
Talking about loose nut at the banana connector, this reminds me the older ucurrent that I bought, its was out of alignment on the spacing at both input and output banana connectors like below photo.
Dave, I guess this time your contractor also aware of the spacing when fastening them right ?
-
-
Hi
2 bad on 50 look bit high for such simple board even on first batch
if you have >50 0402 part on each PCB that may ok
but it look more like 0805 on that board and ok the shunt may pose a bit of challenge
but normally that spotted and inspected manually before delivery to customer
-
#10 Reply
Posted by
calexanian
on 10 Mar, 2014 03:52
-
Wow. most of you are way too critical on them. For an assembly house to not really have the test rig yet and be assembling blind on the first runs only having two failures, and those being simple to correct is not bad! Especially with mixed boards like this. This is why you always have an in house re inspection and in most products cases a monitored burn in. Real production problems are things like bad plating and down the road stress failures, Via separation, Thermal shock that leads to latent failure, material incompatibility (Some plastics do not play well with each other!) Solder mask pealing or under oxidation, board washdown or defluxing issues (My current problems) BOM translation failures on a massive scale (Why oh Why don't passive manufactures just put in huge numbers the parts values on the spools. I write them in big sharpie marker numbers on them) The best feeling on earth. 1000 boards with 47K resistors in place of 4.7K and the darn thing still passes test, but eventually will fail..... Honestly I would rather find a few random bugs and mistakes and have to troubleshoot them to get a good feeling of how things are going to shape up and figure out how that particular manufacturer does things.
In other words, Stop complaining! You put your pants on one leg at a time too! Unless you somehow drop into them while they are being held open, which would be pretty cool... Testicular torsion might be an issue though if you happen to be male.
Looks good Dave! Cleaning a small portion of my bench in anticipation of its arrival! Trying to decide which meter I will use it with first...
-
#11 Reply
Posted by
EEVblog
on 10 Mar, 2014 04:06
-
Dave, I guess this time your contractor also aware of the spacing when fastening them right ?
No need, the new one is engineered to b a precise spacing. No way to mount out of alignment,
-
#12 Reply
Posted by
SeanB
on 10 Mar, 2014 05:07
-
Whoever gets 261 will know it has been inspected, reworked to reach spec and checked twice. One poor solder on the fist run is pretty good, and better the nut was not tight enough than they strip them. Brass is pretty soft if you have enough torque.
-
#13 Reply
Posted by
Pillager
on 10 Mar, 2014 07:31
-
Brass is pretty soft if you have enough torque.
Agreed. But there are specs for the connectors, and hopefully values for recommended and maximum torque. With the right tools, that shouldn't be a problem.
-
#14 Reply
Posted by
Pillager
on 10 Mar, 2014 07:50
-
Wouldn't be from an perspective of ergonomy more convinient just slap in small micro on the test jig and do test indications via buzzer ?
I'm sure i'd be pissed in just an hour by bending my back every single time to check LEDs, not to mention how easy is to overlook indication of bad board once worker slides in stereotype and gets bored.
Short beep for in specs board and some longer alarm signal for failed is making much more sense for me
When doing repetitive work like this, workers usually listen to music, either on a radio or an ipod or some such device. And that worker won't be sitting in a room all by her-/himself, but at a workspace next to other workers performing similar procedures on other products.
If everyone has a buzzer, aside from the noise level, soon you wouldn't know if it was your buzzer going off, or that of your neighbour. Then you would have to recheck that board and that takes time, which in turn costs money.
So the visual indicators are really the only pratical way to go on a factory floor or test room. At home, when no one else is around, it might be ok, if you can stand the constant beeping.
At as far as bending forward to see the LEDs: The workbenches used in production/testing are adjustable, and there will be a custom or adjustable jig for holding the panel in place and at the right angle to minimize physical stress for the worker. Not just because its nice for the worker, but because it increases productivity
-
#15 Reply
Posted by
Gecko
on 10 Mar, 2014 10:32
-
@Dave: Will you make a Video of the production and testing at your manufacturer?
And can you make an overview over the costs in the end? I think it would be interesting how much such a fairly small size project costs from front to back.
I guess a lot of experience goes into it and knowledge from the uCurrent 1, but a video on how to estimate such costs could also be interesting.
Thanks for another interesting video :-)
-
-
Hi Dave.
I noticed that you did not (yet) have a "from" address on the packages going to Aust Post. Remembering the hassles with the pcb rulers (I received an empty envelope with no return address), I hope you will be putting your return address this time... please!
Cheers,
Ross
-
#17 Reply
Posted by
EEVblog
on 10 Mar, 2014 13:08
-
I noticed that you did not (yet) have a "from" address on the packages going to Aust Post.
Yes, there is a return address on them. Where did you see otherwise?
-
-
The first failure examined on the video was the SMD shunt, which was not soldered properly - Dave pointed it out on the video.
The second failed unit was a not tightened binding post/jack, which was also explained on the video.
However, the video shows that the second unit's shunt resistor isn't soldered too well either. It looks to me like it could use a touch up as well.
I know absolutely nothing about production SMD boards, but would one consider a solder paste stencil or application problem even if most units worked, but the joints were marginal?
-
#19 Reply
Posted by
essbdub
on 10 Mar, 2014 13:55
-
To be fair to the assembler, it appears as though the spring washers may bind up slightly on the locating portion of the post before they contact the board. Applying a bit more torque opens up the washer slightly, allowing it to contact the board properly. Might be worth pointing out to an external assembler...
-
-
Hi calexanian
i run myself a assembly house for 10 year
and personally 4% failure on 0805 / 0603 part board whit less that 100 part on each PCB was bit over expected rate
even if assemble "blind" 0805 / 0603 was a 20 year old assembly process so if parameter control was right you have no defect
and normally today shop have automated optical inspection system ,and operator manually inspect critical part
but many shop skip or do it fast for final inspection since it very labor intensive ,so cost money ..
Qoute"
bad plating , Via separation, Thermal shock that leads to latent failure, material incompatibility , Solder mask pealing or under oxidation, board washdown or defluxing issues"
all this was trouble that not normal and was clear sing that PCB fab not follow any of the IPC standard and quality control
as example final rinse water must have Resistivity over 1 to 5 M?·cm (cleaning solvent and flux was quite very conductive)
so easy to see if still have ionic contamination present on PCB whit a simple TDS meter reading of < 1PPM
Excavatoree
on video solder volume look good ,so it likely that it caused by ,wrong pressure on pick place head for push part on the past
and / or bad thermal profile of the reflow oven , for small batch it may bit harder to control the reflow ,put to many cold board on it may change to much it temperature ,so heat become non uniform and one side of the solder may melt before and surface tension twist part so one pad lift a bit and not solder (if part was push well on the part it stick a bit so it help)
that a huge issue on smaller part like 0402 that not have big mass so part flip (Tombstoning)
normally the guy at inspection find that part whit uneven pad or thermal mass and closely inspect it on every board
p.s. i not tell Dave PCB fail rate was really bad , ONLY a BIT over expected rate for that board ..
Best regard
-
-
Personally I love the testing rigs that Dave created ti test the products. I really like the episodes that showed the creation of the panel edge connector, and the pass/fail rig. I would love to see more about the design of the test rigs (the other two that haven't been covered yet).
When those rigs are created I can see they are basically a similar form factor as the uCurrents themselves (for obvious reasons), does that mean you had a whole pannel of those different rigs manufactured as well? Are they hand assembled or assembled at the assembly house?
For some really big eevBlog fanboy, getting their hands on these limited production run devices would probably be pretty sweet! (Future kickstarter idea)
-
#22 Reply
Posted by
lewis
on 10 Mar, 2014 20:50
-
i run myself a assembly house for 10 year
Dude, PLEASE get rid of that music on your website. It plays every time I click a link.
-
#23 Reply
Posted by
Simon
on 10 Mar, 2014 21:35
-
The sockets should have a torque on their datasheet and this needs passing to the assembler. Where i work we do mechanical assembly of radiators and everything has to have a torque set on it be it the standard for that bolt size or specific because it's something else. We also expect our electronics subcontractors to tighten panel mount switches etc up to a specific torque to ensure they are well tight (military kit) but not brocken.
-
-
Hi lewis
yup totally agree ,have request it to my webmaster for some time ,still un-fixed
same for got static banner for all page after main one ..
will go kick some ass now ..
Fix was Done ...