Problems with indexing and large parts

Started by phonoplug, May 11, 2012, 01:31:28 PM

Previous topic - Next topic

phonoplug

I think I posted about this a while back, but am now clearer about what the problem(s) are.

Large parts that need to be visioned in two halves of four quarters. These can only be fed from vib feeders without modification of the lane feeders. And you can't feed them from passive feeders.

Problem is that the software doesn't 'index' correctly when vision is going to be in 2 or 4 shots.

-With passive feeders, it will always go back to the same 'pocket' to collect the next part, even though it knows damn well its just used the part from that pocket. I see no way around this.
-With lane feeders, it will index one time less than its set to, because the first index command normally sent immediately after the part is picked, doesn't get sent. The software also only allows for a maximum of  I think its 3 indexes, or possibly its 4 (even though you can enter any number in the 'indexes' box without getting a warning message). Parts large enough to be imaged in 2 or 4 shots will always need more than 1 index, so you can find that the machine cannot be set to index enough times foe the part. You can accept that it will 'mis-pick' several times for each part: ie two times for every 3 index part, but this is frustrating and very time consuming, specially if you are placing a lot of these parts on the board. The other option is to intercept the signal from the PC to the feeder, recognise the command to index once for that particular lane and send additional index commands. I have essentially done this with a new feeder controller board and a special build of the firmware on it for the larger feeders.
-With vib feeders, it works because the parts feed as soon as one is taken, without any interraction from the PC.

There is a way round this (ref my other post), by setting the 'centre override' option for that package to 'Force fixed', as it will then only vision with a single image. However, on the majority of parts thats no good as there is no detail in the middle of the large part to be recognised (eg a PLCC44), and if it can be made to vision ok, the placement will be inaccurate at best, getting worse with increasing part size.

Another issue is that if the part is visioned in 2 or 4 shots, and then fails analysis, the number of retries is ignored and the machine will forever keep trying until you stop it. If you are using a lane feeder, no index command is ever sent, so it won't index the feeder at all meaning you still have to stop the machine, wind the tape manually then start it going again.

Am I unusual in using the machine to place larger parts? Its mainly connectors, D-PAKs on tape, SO-24W on tape etc. Some of them can be taken off tape and put into tubes, but typically, especially with D-Paks, there are a lot on a panel (90 per panel on the job the other day), so having tubes of 75 is a bit impractical! Not to mention annoying a time consuming slotting them into tubes one by one!

Mike

Quote from: phonoplug on May 11, 2012, 01:31:28 PM
I think I posted about this a while back, but am now clearer about what the problem(s) are.

Large parts that need to be visioned in two halves of four quarters. These can only be fed from vib feeders without modification of the lane feeders. And you can't feed them from passive feeders.

Problem is that the software doesn't 'index' correctly when vision is going to be in 2 or 4 shots.

-With passive feeders, it will always go back to the same 'pocket' to collect the next part, even though it knows damn well its just used the part from that pocket. I see no way around this.
I think this may just be a case of specifying the feeder as a matrix tray not a passive - I'm not sure what the difference is between a passive feeder and a matrix - I think it may be whether it replaces rejects back in the tray or the reject bin.
I've done 1mm pitch SO24's from a static tape which I think were big enough to  need double or quad vision.  Most QFPs from trays also do multiple vision, even 32 pin ones.
Quote
-With lane feeders, it will index one time less than its set to, because the first index command normally sent immediately after the part is picked, doesn't get sent.
Is it not that the second index is missed? - it does one index after picking and a second when visioned, which can be too quick if vision is on the fly.
This can sometimes be worked around by forcing fixed vision to add a delay - locating parts far enough away from the camera if necessary.
Quote
The software also only allows for a maximum of  I think its 3 indexes, or possibly its 4 (even though you can enter any number in the 'indexes' box without getting a warning message). Parts large enough to be imaged in 2 or 4 shots will always need more than 1 index, so you can find that the machine cannot be set to index enough times foe the part. You can accept that it will 'mis-pick' several times for each part: ie two times for every 3 index part, but this is frustrating and very time consuming, specially if you are placing a lot of these parts on the board. The other option is to intercept the signal from the PC to the feeder, recognise the command to index once for that particular lane and send additional index commands. I have essentially done this with a new feeder controller board and a special build of the firmware on it for the larger feeders.
-With vib feeders, it works because the parts feed as soon as one is taken, without any interraction from the PC.
It should also be fairly easy to do some PC software to manipulate indexes on standard feeders, by feeding the serial data into and out of an extra COM port. It's on the list of things I intend to do at some point...
However you may need to do things like specifying slow movement to make sure there is time for the indexes before the next pick.
Quote
especially with D-Paks, there are a lot on a panel (90 per panel on the job the other day), so having tubes of 75 is a bit impractical!
David C told me a while ago that it is possible to have multiple tubes of the same part - I think it involved some sort of fiddly feeder-copy process.
Of course you could set it up so it thinks there are, say, 4 different parts which are actually all the same part loaded into 4 vib lanes.


phonoplug

Passive feeders: It picks from the same pocket even after successfully placing the previous part (and therefore it knew the pocket is empty).

Lane feeders: Its the first index missing. No index as soon as the part is lifted from the tape. Means a vision failure never indexes the feeder. I'm aware that the original feeders don't handle two index commands too close together - obviously nobody thought about buffering the data in the feeder controller as it ignores any data received until its completely finished the index movement. I fixed this with the new controller.

You can do multiple lanes of the same part by using the 'copy' function when allocating feeders. Gets a bit more tricky when the same parts are on more than one feeder though. I guess you load it into the first feeder, then remove that feeder from the job. The part then re-appears in the list, so you can then load it into another feeder. Then load the first feeder back into the job again too.

Mike

QuotePassive feeders: It picks from the same pocket even after successfully placing the previous part (and therefore it knew the pocket is empty).
Not very useful, but a 1-row matrix tray works OK so surely this is not a problem?

QuoteLane feeders: Its the first index missing. No index as soon as the part is lifted from the tape. Means a vision failure never indexes the feeder. I'm aware that the original feeders don't handle two index commands too close together - obviously nobody thought about buffering the data in the feeder controller as it ignores any data received until its completely finished the index movement. I fixed this with the new controller.
My observation for a double-index is it indexes once after the pick and again after vision - this makes sense as it means a tape that's one index out will fail the first time and then pick correctly and index twice from then on -the only problem being when the post-vision index happens too quickly. I've not tried more than 2 indexes so maybe it's different.

phonoplug

I have the same problem with a single row passive 'tray'. I came across this a long time ago when I was sticking a strip of SM HC49 type crystal tape to the working area to place these crystals. I didn't know the machine so well then, but its exactly the same issue. The crystal needs to be imaged in 2 shots as its long (24mm tape), so the machine only ever picks from the one pocket whether it placed the part successfully or not.

Your observation about multi-index parts is correct *if* its visioned in one image. It its 2 or 4 image, the first index is missing and it will only send a single index command as its putting the part on the board (ie after imaging has been processed AND passes), regardless of number of indexes setting, as long as its set to 2 or more.

Aub

This seems to be part of my problem, is the only way round it seriously to use vib feeders?

Mike

I have definitely used multi-vision parts from both trays and normal feeders.
Remember the vision type must be set to "Automatic" for multi-vision to work.

phonoplug

Aubrey - The problems with passive feeders and trying to repetedly pick from the same pocket even though it knows its empty is fixed by using the larger 1.7MB 'place.dll' file.

Its the same issue that causes the lane feeders not to index correctly, and vib feeder lanes to retry indefinitely despite the settings.

These all happen only with parts that are too large to vision in a single shot.

Aub

do you have a copy of the larger place.dll file you could send me? I have partley got round the problem by putting the SOIC 24 chips in a vib feeder. however I am not getting consistant placement when using the VIB feeder. I placed  6 SOIC 24  pin from lane 1, then tried to place 6 SOIC 18 pin from lane 0 of vib feeder. I have tried and checked all settings and don't seem to be getting nearer. The camera looks at the chip, and no matter what setting I have I don't seem to get a perfect image when it shows the outline screen. I am beginning to wonder if I have a machine problem.
It will sometimes place 1 Ic, and them put the next 5 in the bin. it picks them goes to almost the correct position, but then drops them in bin. Any Idea's would be much appreciated.  Regards Aubrey. P.S. my email is aubrey@oldham.demon.co.uk

Gopher

When part fails the visual check, the rejection happens just before placement. If its working sometimes but not others I would watch the placement numbers at the top of the screen that tell you the measurements that have been made by the camera. Here you should see X and Y dimensions it has determined the part to have as well as how far off-centre the pick was, if any of these parameters falls outside what you have set the part will be rejected. With a Vib feeder you are more likely to get failures as a part may not have fully reached the pick position when the pick is attempted, making it more likely the offset will be outside the accepted range. MFOV images tend to look a little fuzzier at the seams, is that machine coping OK with all the smaller parts? If these are your first few jobs on the machine, did you go though any of the calibration procedures documented on this forum first?