Quad QSA-30A will arrive tomorrow

[MR]

jmelson, are the pneumatic switches installed in the head or are they in the base station of the machine?
Our pneumatic switches are currently in the head, but I think I'll migrate them into the base station.

[jmelson]

jmelson, are the pneumatic switches installed in the head or are they in the base station of the machine?
Our pneumatic switches are currently in the head, but I think I'll migrate them into the base station.

There are 6 solenoid valves, 3 vacuum generators and 3 vacuum sensors in the head.  I think the idea is to keep the tubes from vacuum generator to nozzle as short as possible so the vacuum comes on and stops quickly.  The 6 valves are for vacuum on/off and blow on/off for each of the 3 nozzles.

Jon

[MR]

Every nozzle has its own vacuum generator? Does anyone know is that a common thing?
I'm sharing one vac generator for 2 nozzles at the moment, the solenoids are controlled with 24v.

Oh by the way could you make a photo of those modules?

[jmelson]

Every nozzle has its own vacuum generator? Does anyone know is that a common thing?
I'm sharing one vac generator for 2 nozzles at the moment, the solenoids are controlled with 24v.

My Philips CSM84 (made by Yamaha) did exactly the same thing.

Oh by the way could you make a photo of those modules?

Umm, kind of hard to do, they are buried in a big jumble of hoses and wires.  I'll see what I can do.

Jon

[jmelson]

Further updates.  I finally figured out how the machine is set up, the corner that the board locates to is the front-right corner, so all the coordinates grow positive from there.  (I think the prior owners changed something in the machine setup to do this.)  Anyway, now that I finally understand the coordinate system, I was able to program a board with board dimensions, two fiducials and a component, and get it to place a part.  The placment wasn't horribly bad, but consistently off to one side.  So, I figured out that I could apply spotting dye to a nozzle, drive that down onto the feeder rail so that it left a ring of dye, note coordinates and then drive the camera directly over the spot and note the coordinates.  I then used the difference of those numbers to adjust the camera and nozzle offsets, and ran the placement program again.  This time, the part was placed EXACTLY dead center on the component pads!

Now, I need to make my CAD -> P&P converter program do all these manipulations on the pick and place file to get the numbers the way the machine wants them.  I'm getting real close.

Jon

[jmelson]

So, the ASCII import doesn't work.  It reads XY rotation part number, etc into a screen, but when it gets imported to the "step program" all rotations are zero.

CAD import is a different function, it DOES bring in the rotation.  But, I get a bunch of error messages.  The "optimize" button reassigns parts to different steps, and to nozzles and heads and feeder locations.  But, it seems to scramble and duplicate some of the parts.

This is just getting to be such a mess -- 2 whole months so far and still not a single board placed.

Jon

[Reckless]

Why not hand teach the board?  Sounds quicker. 

I heard Quads are good beginner machines but personally I prefer Samsungs.  Since Quad is no longer in business I wouldn't recommend them even though they use Samsung hardware.  I believe Samsungs are the best starter machines, good hardware and simple software. Easier to find support, Quads seem like everyone is out to screw you and since they didn't design the hardware I don't feel as confident in their product.   

[jmelson]

Why not hand teach the board?  Sounds quicker. 

The accuracy and the correctness of teaching a board with 800 components sounds really iffy!  I've had my Philips CSM84 for 13 years, and have NEVER taught a board.  The VERY FIRST board I did, I hand-edited the CAD system's placement file into the format needed by the Philips, downloaded it and ran the board.
Then, I wrote a small C program to convert the file format.  This is what I'm working on now for the QSA.  I have over 15 standard boards that I make, and having to hand-teach all of the placements would be an insane horror.

Actually, I could just enter all the numbers from a printout, but it just seems incredibly STUPID to have a bunch of computers and have to HAND-ENTER data from one to the other.  This is just crazy that this stuff doesn't work.  They have this incredibly "good looking" software to import ASCII files with the placement data, but it seems to be horribly buggy.  I'm seriously thinking that if this is as good as it gets, I may just scrap this machine.  This will be a GIANT step backwards if this import function just can't be made to work.

I heard Quads are good beginner machines but personally I prefer Samsungs.  Since Quad is no longer in business I wouldn't recommend them even though they use Samsung hardware.  I believe Samsungs are the best starter machines, good hardware and simple software. Easier to find support, Quads seem like everyone is out to screw you and since they didn't design the hardware I don't feel as confident in their product.

Well, the Quad QSA-30A is actually a pretty standard Samsung CP30 modified to take Quad feeders instead of Samsung.  Instead of a rack of air cylinders to advance the all-mechanical Samsung feeders, the QSA-30 supplies power to the Quad feeders, and when the nozzle drops down and raises up with the part, the Quad electronic feeder advances the tape to the next part.  Otherwise, it is VERY similar to a Samsung, and actually uses exactly the same software.  There is an option in the setup file that tells it what machine it is running on.

Jon

[SMTech]

It has been suggested before, but just how different is the file format for a Samsung CP40 to a CP30? PCBSynergy http://members.iinet.net.au/~sarason/ supports that format and maybe even the author could add your quad as a supported format if you gave him a simple of a simple taught PCB.
I've never been near one but is it possible the import functions don't work correctly until they have a library of parts (so like a broken relational database) were there any files left on the machine from its previous life you can snoop at, or even files they think they deleted but didn't...?

[Reckless]

Why not hand teach the board?  Sounds quicker. 

The accuracy and the correctness of teaching a board with 800 components sounds really iffy!  I've had my Philips CSM84 for 13 years, and have NEVER taught a board.  The VERY FIRST board I did, I hand-edited the CAD system's placement file into the format needed by the Philips, downloaded it and ran the board.
Then, I wrote a small C program to convert the file format.  This is what I'm working on now for the QSA.  I have over 15 standard boards that I make, and having to hand-teach all of the placements would be an insane horror.

Actually, I could just enter all the numbers from a printout, but it just seems incredibly STUPID to have a bunch of computers and have to HAND-ENTER data from one to the other.  This is just crazy that this stuff doesn't work.  They have this incredibly "good looking" software to import ASCII files with the placement data, but it seems to be horribly buggy.  I'm seriously thinking that if this is as good as it gets, I may just scrap this machine.  This will be a GIANT step backwards if this import function just can't be made to work.

I heard Quads are good beginner machines but personally I prefer Samsungs.  Since Quad is no longer in business I wouldn't recommend them even though they use Samsung hardware.  I believe Samsungs are the best starter machines, good hardware and simple software. Easier to find support, Quads seem like everyone is out to screw you and since they didn't design the hardware I don't feel as confident in their product.

Well, the Quad QSA-30A is actually a pretty standard Samsung CP30 modified to take Quad feeders instead of Samsung.  Instead of a rack of air cylinders to advance the all-mechanical Samsung feeders, the QSA-30 supplies power to the Quad feeders, and when the nozzle drops down and raises up with the part, the Quad electronic feeder advances the tape to the next part.  Otherwise, it is VERY similar to a Samsung, and actually uses exactly the same software.  There is an option in the setup file that tells it what machine it is running on.

Jon

800 components is not too bad.  We have hand taught 3000. I do agree pcbsynergy is a great tool.  I have used for a Juki in the past. 
Nice to know Quad uses Samsung software.  I thought they had their own custom software.  Personally I prefer mechanical feeders as they are cheaper and easier to maintain. 

[jmelson]

It has been suggested before, but just how different is the file format for a Samsung CP40 to a CP30? PCBSynergy http://members.iinet.net.au/~sarason/ supports that format and maybe even the author could add your quad as a supported format if you gave him a simple of a simple taught PCB.
I've never been near one but is it possible the import functions don't work correctly until they have a library of parts (so like a broken relational database) were there any files left on the machine from its previous life you can snoop at, or even files they think they deleted but didn't...?

I think the  format should be identical.  But, the horrible issue is the import program in the QSA-30 has some real problems.  ASCII import throws away all the rotation info.  CAD import handles that properly, but you have to manually enter the fiducial coordinates.  That isn't so bad, just two points.
I have set up the library parts for this first board, that was quite easy.

There are THOUSANDS of files on the machine, but they are pretty inscrutable.  All binary, although the export feature will pull some of the info out and make it human-readable.

I'm still having trouble figuring out how to deal with the automatic nozzle changer.  I can program it so it can put the nozzles in the changer and then get them back out, when I do it manually.  But, when you try to run a program, if the nozzles are in the changer, it gives one message, if the nozzles are on the head and the system KNOWS that, it gives another message.

Jon

[jmelson]

800 components is not too bad.  We have hand taught 3000. I do agree pcbsynergy is a great tool.  I have used for a Juki in the past. 
Nice to know Quad uses Samsung software.  I thought they had their own custom software.  Personally I prefer mechanical feeders as they are cheaper and easier to maintain.

Yes, I was pretty comfortable with my Philips (Yamaha) feeders, and could thread them in under a minute.  The Samsung mechanical feeders were quite similar.
The Quad electronic feeders are quite a bit different, and I'm still getting used to loading them and getting them set up right.

I did have a lot of jams, failure to peel the cover tape and such stuff with the Philips feeders.  I'm hoping the Quad feeders do better in that regard.

Jon

[jmelson]

OK, more progress!  I still find the CAD -> P&P import process quite a bit more cumbersome that I think it ought to be.  But, I have now loaded a data file with designator, X/Y/R and part type into the machine, and with a LOT of hand fiddling, got it loaded in.  They have an import function that brings in the raw data only.  Then, there's an "optimizer" that rescrambles everything to try to maximize assembly rate.  It reassigns tape feeders and nozzles to the parts found in the imported data.  There are a bunch of options in the optimizer, and I'm pretty sure I'm not making the right settings there.  But, by hand, mostly, I got it all assigned.  I don't know if the machine is scrambling the file, if the CPU has bad memory, or it is just me clicking on the wrong things that is causing the scrambling.

But, through all that, and a number of other issues, I did manage to get 14 parts put onto double-stick tape on a test board.  One of the QuadAlign cameras has some debris in it that causes alignment issues, I'm going to have to pull that out again and try to clean it.  And, now, I have to get some more nozzles.

But, I'm getting really close, here!

Jon

[SMdude]

Nice work, it is very satisfying to get these machines to work and pump panels out faster than the reflow oven can bake them, especially with little to no help!
Next thing I need to figure out is if I can copy feeder info from one job to the next. That will save a whole heap of manual setting up when setting up a new job.

[jmelson]

Nice work, it is very satisfying to get these machines to work and pump panels out faster than the reflow oven can bake them, especially with little to no help!
Next thing I need to figure out is if I can copy feeder info from one job to the next. That will save a whole heap of manual setting up when setting up a new job.

Yes, I am a bit disappointed that this is taking so MUCH fiddling.  On my Philips, I could compose the complete placement file on my desktop computer and just download it to the machine.  On the Quad, it seems I will have to do a lot of hand-editing of the placement file (step program) after the optimizer has scrambled things.  I only have the 3 nozzles that came with the machine, I have so far not found out how to tell the optimizer that that is all I have.  Maybe as I gain more familiarity, I will get it down to a routine.

I still have some kind of dirt on one of the QuadAlign cameras, and the 3rd head that was extensively messed with by the previous owner still have a very "gritty" ball spline.  I may have to swap out that whole head to take care of those issues.

But, what it assembled a few days ago looked to be SO MUCH more accurate than my old Philips that I'm getting pretty excited about doing some more challenging boards.  Anyway, I'm going to try an actual production run this weekend with solder paste and reflow, and see how it goes.

Jon

[jmelson]

Despite some lingering issues, I populated the back side of 20 boards today.  There were just 2 part types there, an 0805 capacitor and an SOD-323 diode.
I did have a few mis-picks where the cover tape on the 0.1 uF 0805 caps didn't peel well enough.  This specific part (from a variety of manufacturers) has ALWAYS given me fits on my old machine with the cover tape not peeling and causing the feeder to get stuck.  But, it seems the Quad feeders did better with it.  I may need some new rollers with more grip on them.

The fiducial camera seems to be blinking, I think the connector board on the back of the camera has another failing connection.  I'm going to swap out that board shortly.  Now, I have to learn how to program and teach the SOIC parts on the vibratory feeder, so I can do the front side of the board.

Jon

[jmelson]

So, over the weekend I did a bunch of the front sides of those boards until I ran out of parts.  I had a major fit trying to get the QuadAlign vision system to accept the SOIC-16 parts, although it accepted the SOIC-8 parts with the factory stock settings.  I still don't understand what was going on there, but I just kept fiddling with parameters until it accepted the parts.

My old Philips CSM84 was certainly fast enough for me, this machine, running at 50% speed, is about twice as fast as the Philips, even when the Philips didn't have to go use the mechanical alignment station.

I still don't have the automatic nozzle changer working, it is clearly a software setup issue.  Every time I start the system, I have to reset all the ANC settings and hole positions, and then it forgets them the next time.

The fiducial camera flaked out, but by swapping boards with a spare, I was able to get it running.  Hopefully, I have gotten past the issue with boards going bad.  I need to get a few more feeders and make some repairs on the ones I have, to have enough for the next boards I will be doing.

As for teaching boards, I won't have to do that.  I have now developed a flow from the CAD system, through a C program I wrote, into the CAD import function of the QSA-30, but then NOT using their built-in "optimizer" which totally scrambles the placement data.  It takes some additional hand editing of the placement data to assign parts to feeders, etc. but it works.  I have to manually enter the board dimensions and fiducial location and settings, but then all the placement data is brought in from the CAD system.  (correctly!)

I'm still getting used to the Quad feeders.  At first, I was thinking they were going to take a LOT longer to thread and set up, but it now looks like it only takes a little longer than simple mechanical feeders.  And, I had a LOT of trouble with my old Philips/Yamaha feeders not being able to peel the cover tape reliably.  Once threaded, the Quad feeders seem to have enough pull to peel the tape reliably.  This might actually be a BIG improvement that I was not expecting.  On the old Philips, I had to constantly be leaning in there to mess with cover tapes and unjam them.

I still have to make a waffle tray holder or find an affordable one.  And, I still need to calibrate the QuadAlign cameras.  Placement is already better than my Philips most of the time, but I know that I have knocked these out of alignment when I disassembled things.

Just keeping the list updated on progress.

Jon

[Mangozac]

Keep the updates coming - they're good to read! Your progress thus far is great

[jmelson]

So, I got the front side of 20 boards done, and am now moving on to the second board to assemble.  This one has a bunch of different parts
(crystals, opto-couplers, quad flat packs, big capacitors and inductors, etc.) so I had to program the shapes of these for the alignment system.
Then, I had problems getting the system to recognize the fiducials.  Not sure what is going on there, but it probably needs some parameters to be tweaked.

Then, I have to buy some more 8mm feeders and build a holder for the waffle tray.  These seem to be unobtanium.

Jon

[jmelson]

Wow, it just keeps getting crazier.  I was having a LOT of difficulty getting fiducials to be recognized reliably.  I did get 20 boards done, and started programming the next board.  I just couldn't get it to recognize fiducials better than about 5% of the time.  I tinkered with parameters, cleaned the fiducials, changed the camera aperture, tweaked the focus, nothing seemed to help.  I finally pulled the DSP board that does the vision processing, and found seven bad solder joints on one of the FPGAs!  I couldn't believe it.  I fixed them, and now the machine recognizes fiducials well off-center and puts the recognition crosshairs perfectly on center, 100% of the time.

This is pretty crazy, as the machine had apparently been that way since it was built.

Well, I'm taking small steps, but things are getting fixed a little at a time.

Jon

[Mangozac]

That's crazy! Surely it would have been working properly for the initial stages of its life? I'm guessing the poor solder joints were OK until the machine got trucked across the country to you.

There's a really high level of satisfaction watching a machine recognise fiducials

[SMdude]

That would have been frustrating!
I have had some problems, but that was all due to HASL and the blob left on the fiducial.
Solderwick and some fine sandpaper fixed that for me. Note to self, all boards to be ENIG in the future!

[jmelson]

That's crazy! Surely it would have been working properly for the initial stages of its life? I'm guessing the poor solder joints were OK until the machine got trucked across the country to you.

There's a really high level of satisfaction watching a machine recognise fiducials

Yeah, crazy is right!  No, the only way it would have worked is for the lead to make contact with the solder lump on the pad.  My experience is that can work fine for a functional test right after reflow, but after a couple weeks a little oxide develops and the contact goes flaky.  I've seen this lots of times on the equipment I manufacture.  Now, there were a few leads that still had weak solder attachment, but I could tell, looking at it under a microscope, that at least five of them NEVER had any attachment to the pad.

This machine may be a Frankenstein combination of parts from several machines, so they might have put in a board that was in sombody's used/unknown condition stack, and it worked well enough that they moved on to other issues.  The big killer was that they swapped an entire head from an incompatible machine version and didn't know why it woudln't home.  It didn't take me too long to figure out they'd put a 2-phase stepper motor where a 5-phase motor needed to be.

I'm somewhat concerned that two (other) boards went bad AFTER I got the machine.  A VME interface board stopped allowing the machine to go to the "ready" state, and the fiducial camera had two failures.  I'm guessing that after the last ill-fated "repairs" on the machine in 2014, it was pushed into an un-conditioned space in Austin TX and absorbed a LOT of moisture over 6 years.  (I know when it was last used or attempted to be used from the file time stamps on the disk.)  I hope these delayed failures settle down!

I'm making parts for a waffle tray holder to go on the back feeder rail.  I need that for the next board to be made.

Jon

[jmelson]

I made the back side of 26 boards today.  I had a few small issues, one board needed to have the fiducials cleaned with a pencil eraser, and one of the feeders was not consistently pulling the cover tape, but that cleared up after a while, I guess the roller was a little slick at first.

Also, I was placing 2 0603 parts per board without the right nozzle, so some of them were going on a little tilted.  Otherwise, it just hummed along.

I have mostly completed a waffle tray holder, picture attached.

I'm waiting for some parts from Digi-Key to do the front side of the board.

Jon

[jmelson]

I did a dozen front sides of these boards today.  I learned a bunch about how to set up a waffle tray with several different parts on it, and also set up the vision profiles for about a dozen different parts (1117 regulator, oxicap, inductor, TQFP44, quartz crystal, 4-pin opto-coupler, etc.)  I am waiting on a loan of a calibration nozzle to calibrate the QuadAlign cameras.  Until I get that done the placement is not as good as this machine should be able to do, but it is definitely coming along.

I got some experience with using the partial assembly feature of the Quad/Samsung to put on the missed parts ONLY.  That was pretty cool!

Jon