Mechatronic Systems P10

[electroDroid]

I am looking at a small/medium desktop size pick-and-place machine. It needs to be able to dispense solder paste automatically and handle 0402 and 0.5mm pitch components. We will be building enough units we need a fully automatic machine.

I came across the Mechatronic Systems P10. Does anyone have experience with this system? Or do you have any recommendations for similarly capable systems? I haven't been able to find much about this system online.

[SMTech]

No, but lets start by pulling you away from the dispense option. Unless you want to use that for some component gluing the only platform that can truly do print and place in one machine is the Essemtec platform, it can't be done reliably with screw or time/pressure dispensing techniques, which is why Essemtec have like 4 kinds of dispense head and why the dedicated Mycronic solder jetting printer is so very very expensive. You will find basic dispense methods on other machines even inside stencil printers but they are for adding extra paste where a stepped stencil isn't practical, or sometimes PoP assembly (a flux unit is more common here however) or for glue/epoxy/underfill or other niche applications.

You will get better results with a manual stencil printer, plus oh how slow are those machines, ouch seriously buy something else. If you want pseudo European - choose an Autotronik (Manncorp in the USA).

Warning signs to my mind include the low quality of info, the speed, the build in general, no focus on the software capabilities at all and the implications of the Chinese machines on this sector of the market. I would suggest unless they up their game there are about 4-5 brands price at double China pretending to be western and performing at a lower level that are just going to vanish. That would worry me if I was looking to buy one. Chinse software might be lacking but those machines are 5 times quicker, never underestimate how much speed you might need it will probably grow faster than you think.

[jmelson]

I am looking at a small/medium desktop size pick-and-place machine. It needs to be able to dispense solder paste automatically and handle 0402 and 0.5mm pitch components. We will be building enough units we need a fully automatic machine.

I came across the Mechatronic Systems P10. Does anyone have experience with this system? Or do you have any recommendations for similarly capable systems? I haven't been able to find much about this system online.

I do down to 0.5 mm lead pitch QFPs, but no smaller passives than 0603.  I do manual solder paste with a home-made printing frame setup.  It works FINE!  And, it only takes me something like 15 seconds to fixture and print a small board.  So, I don't see solder dispensing as a good way to go.

Jon

[jmelson]

Chinse software might be lacking but those machines are 5 times quicker, never underestimate how much speed you might need it will probably grow faster than you think.

My new(er) machine, a 20 year old Quad QSA30A (built by Samsung) does a real life 4100 parts/hour (heh, they advertise 17K) and I really would not want to go much slower.  Reading a bunch of discussion on this forum, it seems like some of the Chinese machines have some REALLY bad software issues.  I struggled with the software on my Quad also, but once I learned the tricks, it is a truly fantastic machine.  One area you need to consider is error recovery.  My old (1995) Philips machine had decent error recovery, the Quad is much better.  When it finds a jammed feeder (3 retries without a good pickup) it stops placing that part and finishes the board.  A report on the screen shows what parts it had trouble with.  You reach in, clear the jam (usually cover tape that didn't peel) and press start, and it places ALL those skipped parts!  WOW, that's like MAGIC!  That's the most SANE way to deal with a jammed feeder.

I don't think the current Chinese machines do anything this smart.

Jon

[Mangozac]

Chinse software might be lacking but those machines are 5 times quicker, never underestimate how much speed you might need it will probably grow faster than you think.

My new(er) machine, a 20 year old Quad QSA30A (built by Samsung) does a real life 4100 parts/hour (heh, they advertise 17K) and I really would not want to go much slower.  Reading a bunch of discussion on this forum, it seems like some of the Chinese machines have some REALLY bad software issues.  I struggled with the software on my Quad also, but once I learned the tricks, it is a truly fantastic machine.  One area you need to consider is error recovery.  My old (1995) Philips machine had decent error recovery, the Quad is much better.  When it finds a jammed feeder (3 retries without a good pickup) it stops placing that part and finishes the board.  A report on the screen shows what parts it had trouble with.  You reach in, clear the jam (usually cover tape that didn't peel) and press start, and it places ALL those skipped parts!  WOW, that's like MAGIC!  That's the most SANE way to deal with a jammed feeder.

I don't think the current Chinese machines do anything this smart.

Jon

There's a LOT of generalisation in these comments. Don't get me wrong, the mispick handling operation Jon describes is totally awesome and certainly useful but in reality for outfits not doing seriously high volume (let's say <1500 pieces in a run) once everything is configured correctly we don't find pick failures happening often enough to say that it's a feature that would make a big difference for us to have.

Quality, used pick and place machines are fairly prolific in the USA, so if you feel confident enough with maintaining mechatronic systems they're absolutely a good way to go. For the rest of the world though it's just not as simple an option.

Like everything else in China you get the quality that you pay for. Mid-to-high range Chinese pick and place equipment may lack many of the advanced software features but our Kayo machine certainly doesn't have any notable software issues. Also, the Chinese are smart and features are regularly being added (we're doing beta testing of a big update for our 1706 machine at the moment).

If I was setting up a high volume line where uptime and unwavering throughput is absolutely critical I'd certainly be considering a current Yamaha line or something of that level. We're talking literally ten times the price though, so it's really horses for courses.

[48X24X48X]

Chinse software might be lacking but those machines are 5 times quicker, never underestimate how much speed you might need it will probably grow faster than you think.

My new(er) machine, a 20 year old Quad QSA30A (built by Samsung) does a real life 4100 parts/hour (heh, they advertise 17K) and I really would not want to go much slower.  Reading a bunch of discussion on this forum, it seems like some of the Chinese machines have some REALLY bad software issues.  I struggled with the software on my Quad also, but once I learned the tricks, it is a truly fantastic machine.  One area you need to consider is error recovery.  My old (1995) Philips machine had decent error recovery, the Quad is much better.  When it finds a jammed feeder (3 retries without a good pickup) it stops placing that part and finishes the board.  A report on the screen shows what parts it had trouble with.  You reach in, clear the jam (usually cover tape that didn't peel) and press start, and it places ALL those skipped parts!  WOW, that's like MAGIC!  That's the most SANE way to deal with a jammed feeder.

I don't think the current Chinese machines do anything this smart.

Jon

My HWGC machine latest software has that feature. 3 retries (or whatever attempt number which you can set) for pick up and it will buzz you to fix the problem OR you can optionally let it complete the job and fix the problem at the end. Chinese machine software is not that bad... usually all the negative vibes comes from people that don't actually own one.

[Mangozac]

usually all the negative vibes comes from people that don't actually own one.

That's what I've found as well.

[SMTech]

My own impressions of the features missing or odd in the software come off these forums or users/training videos for those machines, they are real and clear shortcomings, the ones that bug me most are on the setup side. I would actually be quite surprised if that particular feature @jmelson mentions were missing on anything but the most basic machines. On our Essemtec that goes one step further, not only can you set global and component level retry attempts and whether to stop or continue when they are encountered but you can also define "only place this component if these other ones are already placed" rules. This means if you run out of 0.1's mid board for instance, while the machine will continue uninterrupted, it won't place other components that might block their placement (i.e tall ones placed very close). At the end you can then load a new reel and complete the assembly, or you could even reload the feeder  while it was running. On our version this is a manual thing, more recent feature matrices suggest its now automatic standard but still is a premium option (Collison avoidance) on some other brands.

As far as I can see from my non-owning one perspective, the cheaper the Chinese machine the simpler the products they are capable of making, which makes sense. Also however the smaller the range of products you would want to make on them just because  everything from loading feeders (especially the drag type) to setting up the jobs is that much slower with considerably more steps. One manufacturer recently introduced or trialed a component library (Kayo?) in their software, that is a massive step forwards towards letting the machine know and recall settings you otherwise input repeatedly by hand. (For example the Neoden 4 seems to need to you set both the pick and place height and measure those manually before typing the measured numbers from one screen into another)

I keep an eye on these things because, a cheap effective machine for certain classes of build would have value. So far I don't think they are quite there, also what's with the piddling feeder counts? I have zero interest in doing in-house tinkering, Essemtec just works, big older machines just work and have multiple sources of support. Uptime matters to us, it might matter less if its something in a lab you run once a week for one or two products you have difficulty out sourcing on terms you find acceptable.

[Mangozac]

I would actually be quite surprised if that particular feature @jmelson mentions were missing on anything but the most basic machines. On our Essemtec that goes one step further, not only can you set global and component level retry attempts and whether to stop or continue when they are encountered but you can also define "only place this component if these other ones are already placed" rules. This means if you run out of 0.1's mid board for instance, while the machine will continue uninterrupted, it won't place other components that might block their placement (i.e tall ones placed very close). At the end you can then load a new reel and complete the assembly, or you could even reload the feeder  while it was running. On our version this is a manual thing, more recent feature matrices suggest its now automatic standard but still is a premium option (Collison avoidance) on some other brands.

As your description indicates though, this kind of logic is actually very complex. The machine needs to have a thorough understanding of the component height relationships to be able to do this and ultimately it's not a market that the low to mid range Chinese machines are trying to cater to. As I said above, it's a really cool feature but I can honestly say that in our moderate volumes (20k boards per year) I don't believe it would make a huge difference. If a reel is about to empty our operator has a new reel set up on a spare feeder. The pick fails, the machine alerts and in 20 seconds the operator has swapped the feeder out and the machine is running again.

One manufacturer recently introduced or trialed a component library (Kayo?) in their software, that is a massive step forwards towards letting the machine know and recall settings you otherwise input repeatedly by hand. (For example the Neoden 4 seems to need to you set both the pick and place height and measure those manually before typing the measured numbers from one screen into another)

Yes, we've been trialing the new Kayo update with this feature. They still have a few bugs to fix on it but overall it's very promising and a big jump in capability. They previously weren't as manual entry as you describe for the Neoden, but still setting up was a lot of work. They realise now that to stay competitive and even simply be appealing to Western companies they need to continue to add these features, since setup is where a lot of the high Western labour rate costs are.

[max666tr]

I have a pick and place P10 machine made by mechatronic systems as well. I tried to operate the machine for BGAs unfortunately the machine is not calibrated. I emailed the manufacturer for a user guidance but asked me around 1000 euros to send me a pdf. Has anybody got experience for working this type of machine. The interface software is MES studio made by the company.

[SMTech]

I have a pick and place P10 machine made by mechatronic systems as well. I tried to operate the machine for BGAs unfortunately the machine is not calibrated. I emailed the manufacturer for a user guidance but asked me around 1000 euros to send me a pdf. Has anybody got experience for working this type of machine. The interface software is MES studio made by the company.

Sadly this kind of fee for even basic access to data, parts or support on a 2nd user machine is not uncommon. The worst example I am aware of is Nordson AOI - lose your software dongle $7k, register your seconds user machine for support and servicing $15k. I think there is a user on here with experience (not good) on these systems...

[MR]

Check if the calibration can be done mechanically (eg. adjusting camera or focus).
You didn't write about which part of the machine needs to be calibrated.
A pick and place machine has numerous parts that can be calibrated.
As a side note single head machines are significantly easier to calibrate than multihead machines.

If missing documentation will cost 1000$ I'd probably go for it before trying other things.
But your machine is old and limited it seems for me it would not worth to spend much time on it, but you need to know yourself.

[max666tr]

The machine is not able to detect the fiducials accurately. I marked them manually but still not placing a component accurately. it seems that the arm is not calibrated or the top camera can not process images correctly. The software is very basic and not possible to calibrate it easily. Top cam can not focus while pick and place. So there are many uncommon issues while operating the machine.

[MR]

Check lightning conditions for the fiducials. Which kind of fiducials? HASL or chemical gold plated, ...?

HASL tends to reflect the light into every kind of direction leaving the camera sensor with optically different sized fiducials.
The original software of our Mechatronika MX80 cannot handle HASL based PCBs because the fiducial always looks different depending on the uneven HASL status of the PAD; as a workaround we wrote our own software for their machine and are happy with it.

Pick and place machines do not mess around with the camera focus - it's usually set statically. If the component is blurry over the staging camera you have to take care about one of those 2 items:
1. component height, the software should usually drive to the staging camera and move Z (minus component height) to always reach the same hardware set focus. So if some components are correct, but your BGA is not okay over the staging camera check the component height.
2. camera lense focus

The actual Z distance to the staging camera is usually a static calibration value.

There's plenty of information about illumination at https://www.vision-doctor.com/

[jmelson]

Check lightning conditions for the fiducials. Which kind of fiducials? HASL or chemical gold plated, ...?

HASL tends to reflect the light into every kind of direction leaving the camera sensor with optically different sized fiducials.
The original software of our Mechatronika MX80 cannot handle HASL based PCBs because the fiducial always looks different depending on the uneven HASL status of the PAD; as a workaround we wrote our own software for their machine and are happy with it.

My Quad QSA30A (made by Samsung, similar to their CP30 but with Quad feeders and alignment cameras) occasionally rejects fiducials on HASL boards.  The fix is rub the fiducial with an INK eraser, that gives it a matte finish and the camera picks that up just fine!
Jon

[max666tr]

Apparently a hidden menu exists for calibration and controlling the arm manually through software but need to be activated by a password. There is another password which opens basic hidden menu foe testing nozzles, etc. Has anybody got any experience with this machine?(P10 Mechatronic Systems.