Printed Circuit Board Assembly & PCB Design Forum

SMT electronics assembly manufacturing forum.

SMTnet
»
Electronics Forum
»
YesTech F1 AOI false calls and misses

YesTech F1 AOI false calls and misses

Subscribe to Thread

Our company is having issues with our YesTech F1 AOI machine... - Apr 02, 2012 by vparis

We currently have 2 M1 systems. Our false call rate has nev... - Apr 04, 2012 by Reese

Reese, Thank you so much for your reply. ... - Apr 09, 2012 by vparis

Hey Reese what is your Mis rate? Do you have any boards that... - Apr 16, 2012 by vparis

That's hard to quantify. I deal with escapes when they occu... - Apr 16, 2012 by Reese

Reese, Thanks for your speedy reply. Vic ... - Apr 16, 2012 by vparis

some things to consider, Is there flux on the board. are... - Apr 19, 2012 by iddgmbj

Programming is different, yes, but so is the terminology. ... - Apr 20, 2012 by Reese

vparis

#66089

YesTech F1 AOI false calls and misses | 2 April, 2012

Our company is having issues with our YesTech F1 AOI machine the programs are currently not utilizing this machine correctly. we are have a false call rates of about 5% to as high as 12% of the total number of inspection points. I have started to reprogram some boards and can see our problems and correct them. But I am wondering what are some real world numbers as far as false calls and number of misses. If your company runs this machine or a similar AOI please chime in and let me know abouts where we should be. Thank you in advanced.

reply »

Reese

#66110

YesTech F1 AOI false calls and misses | 4 April, 2012

We currently have 2 M1 systems. Our false call rate has never been has high as you suggest. There is something wrong with your program and/or inspection algorithms, as you probably already know. You should be somewhere between 1000-1500 vcpmo (false-calls per million opportunity). So this would between 0.1-0.l5% of total parts. Any lower and you can run the risk of escapes. You want failures for "questionable" components that an operator can follow up on. If our inspections creep over 0.2% (2000 fcpmo), then there is a problem with an algorithm. Some of your issues can also stem from poor alignment. I don't know what your issues are, but it sounds like your on the right track. Keep in mind, you don't want to eliminate false-calls completely, but have them at a manageable level. From my experience, this is where you want to be.

reply »

vparis

#66129

YesTech F1 AOI false calls and misses | 9 April, 2012

Reese, Thank you so much for your reply.

reply »

vparis

#66162

YesTech F1 AOI false calls and misses | 16 April, 2012

Hey Reese what is your Mis rate? Do you have any boards that are passed that have actual defects?

reply »

Reese

#66163

YesTech F1 AOI false calls and misses | 16 April, 2012

That's hard to quantify. I deal with escapes when they occur, but I haven't kept track of that data. It doesn't happen to often. But when it does, I take it back to the AOI to analyze why it escaped, and change the algorithm to catch it, if I can.

reply »

vparis

#66165

YesTech F1 AOI false calls and misses | 16 April, 2012

Reese, Thanks for your speedy reply. Vic

reply »

iddgmbj

#66188

YesTech F1 AOI false calls and misses | 19 April, 2012

some things to consider,

Is there flux on the board. are there finger prints on the tops of parts. Is the board warped at the V score from reflow Is the board shifting in the machine during inspection.

These are the issues we have had with our mirtec machine causing high false positive rates. Where the problem would only show up during run and not during programming. All relatively controllable.

with machine vision you have parameters that select what is good and what is bad, and you have parameters that help make bad look worse and good look better making the difference between wider.

Things that can make the difference wider are, lighting, inspector alignment, the black and white threshold on binary or BW type inspectors.

Using 1 inspector to look for each thing, rather than multipurpousing a single inspector helps us out a lot.

If you have a resistor with a label. an inspector to find the part but that explicitly ignores the value of the label is best. then a separate inspector just to check the label and 2 solder inspectors looking only at the amount of solder fillet. The mounting inspector can then also be set to a very low acceptance threshold, as it is only used to align the part.

again i have mirtec so programming may be slightly different.

reply »

Reese

#66194

YesTech F1 AOI false calls and misses | 20 April, 2012

Programming is different, yes, but so is the terminology.

///with machine vision you have parameters that select what is good and what is bad, and you have parameters that help make bad look worse and good look better making the difference between wider./// No over-lap between bad and good inspection. Always the end goal. The helps eliminate escapes and false-calls.

///Things that can make the difference wider are, lighting, inspector alignment, the black and white threshold on binary or BW type inspectors.///

By inspectors, do you mean algorithm? Only so much you can do with a binarized image. It's useful for certain things, but not the cure-all.

///Using 1 inspector to look for each thing, rather than multipurpousing a single inspector helps us out a lot.///

Layers and independent templates.

///If you have a resistor with a label. an inspector to find the part but that explicitly ignores the value of the label is best. then a separate inspector just to check the label and 2 solder inspectors looking only at the amount of solder fillet. The mounting inspector can then also be set to a very low acceptance threshold, as it is only used to align the part.///

Local fiducial or part anchor will work here. Why ignore the value of the resistor or use a separate template to inspect? No separate label is needed for YesTech, but then, I rarely use anchors. I only require them if a part tends to float on the pads or shifts and the mount has to be precise.

Reese

reply »