Author Topic: Mechatronika MX80 Pick and Place Machine (warning)  (Read 1399 times)

0 Members and 1 Guest are viewing this topic.

Offline MR

  • Contributor
  • Posts: 24
  • Country: tw
Mechatronika MX80 Pick and Place Machine (warning)
« on: September 24, 2019, 05:36:04 am »
This is just a small note about the Mechatronika Sp.J.  Pick and Place units (from Poland). This company is owned by Lesław Gajda (61 years) and Dariusz Pieczerak (65 years).

They're bad... really bad. They are probably a dying company not really caring about their customers and their products - and just trying to rip off the last few customers who are willing to throw money after them. There's no explanation why they do not even comment obvious issues of their machines. If Mechatronika ever changes their mind we're certainly open to change this review.

Be warned.

This applies to:
Mechatronika MX80 and many application related issues apply to MX70, M80, M70, M10V

they all share the same pick up head design, which is nothing but rubbish; and Mechatronika does not even seem to configure them properly.
80% of the critics apply to the software, around 20% to erroneous mechanical design.

This is a review about following product:
http://www.mechatronika.com.pl/content/view/31/4/lang,en/

One information ahead, we have reverse engineered the RS232 protocol and wrote our own software and are able to use the machine properly. But now let's get back to what they deliver:

- the optical recognition is based on trained picture patterns, this system fails when components reflect light and are not very close to the same trained pattern. This is a very very critical bug (Mechatronika doesn't care). This caused many component drops. (at least 5% with our units, mainly 0402)

part of a Mechatronika Email:
Quote
- using the Don't care area option please mask areas as on attached picture to prevent of the presence of part of nozzle in the captured image.

see attached picture "0402R template 2" (Mechatronika attached this picture and sent it to us back then)

attached picture component_10uF_crosshair (1).png is what we sometimes get from some components (those were usually dropped by Mechatronika's Software because of unpredictable reflections). The crosshair analytics was done by our software already, our own software is pretty much solid with it [without training and without a template picture] It usually only happens with some of our capacitors.

- The company does not properly calibrate their machines after servicing them (they charged us for that, why upgrade motors for an expensive fee and deliver it with a speed which is even slower than the old model?)
compare:
*
with:
* https://streamable.com/t8cfl (take picking up the component from the right side as reference, the left side is non optimized programming)
The M80 is stepper based, once upgraded to the MX80 it uses Servo motors.
This is a very critical issue, you ship the unit to them pay a lot shipping and service cost and they don't even care to do the setup properly for the money you pay
I know what I'm writing about since I set everything up by myself including the servos, this is not acceptable.

Mechatronika upgraded the motors in our setup, the default setup can only get <1K5 CPH it can all be adjusted - but in my opinion the customer should not even touch the motor settings by himself, especially since it is not documented.

official Placement rate
3200 - 4000 cph


- a big disadvantage is that all different looking components need reference pictures

- by default the component rotation allows only 190 degree (hardware limitation) leaving only space for a 10° fix at to side (the 10° limitation to one side affects all components which are placed 0° or 180°)
(using a rasp this can be extended to 270°, probably even 360° if you change the connector for the vacuum, the limitation is a non documented register in the eeprom.. easy to find). A 190 degree rotation can reach any position, there's just a limitation on the offset correction.
- Component rotation stepper motor is loosing steps (usually customers don't notice that because the rotation is relative after each pickup, while the controlling itself is done using absolute hw degrees (4096 for 180 degree))
- we have had awful many component drops due to the optical recognition and wrong servo settings
- there we come to the Mechatronika MX80, they shipped it with misconfigured servo settings; The portal was shaking/oscillating terribly sometimes (due to a misconfigured PID controller). Mechatronika did not bother to fix that they just left us alone with it (well we fixed it since they're using Delta motors from Taiwan and Delta has a convenient Windows Tool for adjusting the controllers).
- the feeder system is error prone, the used tape might get stuck because the mechanical construction is not good enough for directing the emptied tape outside of the machine
- there are EMC problems with the pick and place head, touching it might reset the top head controller and defaults to the machine factory settings (the application normally reprograms the stepper motor configuration). Guess Mechatronika owners don't really pay attention to it .. when programming our own application we had a close look at everything.
- cameras are CVBS based (interlaced...)
- there's an FPGA in the machine which is able to freeze the picture (full frame), and converts it to CVBS. Back then we've caught the Windows application capturing the picture too early thus returning an interlaced picture which caused a re-run of the optical recognition or dump of the component.

- it takes 5 mouse clicks to place a single selected component, try to place 10 single components - 50 mouseclicks (multiselect is not available), this is necessary eg. when a feeder is jammed and ignoring the part (and trying to fix it at the end).
- pick and place needles are overpriced, the material cost is below 1 EUR, they're sold 100 EUR each (it's cheaper to buy a small used lathe and make some by yourself).

- the machine stops when problems come up, eg jammed feeders, it's absolutely necessary to do baby sitting. - no component tracking option (eg. if a component did not get placed it will not be marked as not placed in the software). In our rewritten application it has shown that component tracking is seriously mandatory to prevent that the machine has to stop all the time.

- No multipanel setup is supported
- We have done a machine upgrade (and paid way too much money for that), it also included a camera upgrade (attached you can see what kind of camera we got for that, those are 10$ chinese cameras but Mechatronika got it even worse)

- Mechatronica was only responsive as long as they could see that they could get money from us for nonsense service and upgrades (technically nothing improved after servicing our machine, the new motors even came with a misconfigured servo controller as mentioned)

On the pro side:
- if you can get it cheap, the RS232 protocol is easy to understand so it's possible to add support for it to OpenPNP (however we did our own application for it... if you want to do a good one count 4 months+ in for everything including component management).
An older Version of Portmon for windows is able to capture all the RS232 communication.

Our pick and place software has around 30K lines C++, 9K Lines webscripting for the component/project management also online feeder setup, even an online preview of the PCB including all the designators. We went through many problems when doing our application, and know the limitations of the machine (eg. components are captured too small, different lenses could increase the entire pick and place quality).

We are using OpenCV, and various algorithms to nail down the components - until now the machine dropped almost no component (compared with before 5% at least, various youtube videos also have a lot components lying around inside the machine).
Again a direct picture pattern approach for detecting components is nonsense and will cause problems.
If we wouldn't have been able to figure out the control commands we would have sold the machine immediately, we have fixed all the issues with our own application - and the machine is good with good software, but garbage with what it's delivered.


To sum up, Mechatronika (the manufacturer) doesn't care and if you have delicate components (slightly reflecting capacitors are already delicate for their system) which cannot be handled by the pattern recognition you're screwed.

At any time we're open to talk with Mechatronika (up until now), but they're refusing

Hope this helps when deciding to buy such a device.

Some items also apply to Mechatronika M10V, M70, M80, MX70.
M70 and M80 use Stepper Motors, while the MX Series seem to use Delta Servos (400 & 200W Delta ASDA-A2)
« Last Edit: December 10, 2019, 11:51:27 pm by MR »
 
The following users thanked this post: Fire Doger

Offline SMTech

  • Frequent Contributor
  • **
  • Posts: 333
  • Country: gb
Re: Mechatronika MX80 Pick and Place Machine
« Reply #1 on: September 24, 2019, 08:33:40 am »
Interesting, the bit where you write on your own software particularly. I think most people would probably scrap the machine, and send in the lawyers "fit for purpose" tests would seem to fail if things are this bad. Many of your issues are inexcusable, sadly I bet the payment terms on this machine were somewhat different to what we see on more expensive ones.

However - "Optical Recognition based on patterns" I'm no expert but how in basic terms else does image recognition work? On a Versatronics a much cruder method is used where it uses pretty natty edge detection algorithm to pick out its maximum dimensions and that's all it uses. Like many other machines Essemtec uses Cognex and that vision uses lots of algorithms, some of which are almost certainly pattern based, but they are used to spot bent leads, missing balls as well as dimension and offset data.

Nozzles - You want shocking try €400 for an Essemtec nozzle which is a metal tip inserted into a machined piece of Delrin. Unlike Fuji/Yamaha etc these are not widely copied and the fine pitch ones are sufficiently small and complex there's not really any way you can copy them cost effectively (the bigger ones are totally feasible). The Versatronics nozzles were purely machined Delrin, I don't think we or anyone who still has one has found a machine shop that would make them for much less than £40 in quantity despite the low material cost and relative simplicity of design, a service dept has to make profit on top of that. Plenty of people don't want the expense mess or space a lathe, training, risk assessment etc required in a real business if machining is not their area.

Its fairly reasonable for a jammed feeder to stop the machine, sadly when you bring down the cost of machine its the feeders that are likely to suffer first, its clearly much trickier to make good ones than most people want to recognize.

From what I have seen of these machines the software is pretty clunky too which means the arrival of the many Chinese options should really kill them off. All that is really needed to finish off the job is proper local support as currently that is limited.
 

Offline Styno

  • Regular Contributor
  • *
  • Posts: 81
  • Country: nl
  • Gently fused by gravity
Re: Mechatronika MX80 Pick and Place Machine
« Reply #2 on: September 24, 2019, 11:21:13 am »
Like the Mechatronika the simple mechanical feeders on our Dima Optimat suffers from the "spent tape curl-up inside the feeder" syndrome, in additiion to the tape sometimes getting caught by an edge somewhere. It requires regular checking on new reels until the first bit of the tape has properly exited the feeder after which they run smoothly.

Some of our second-hand feeders came with the very sophisticated hack of a section of small cable duct taped inside the feeder with the purpose of catching the spent tape early on and to guide it out of the feeder. We intend to copy this to all of our feeders. Perhaps this would work with the Mechatronika's as well?
 

Offline SMTech

  • Frequent Contributor
  • **
  • Posts: 333
  • Country: gb
Re: Mechatronika MX80 Pick and Place Machine
« Reply #3 on: September 24, 2019, 11:38:07 am »
Like the Mechatronika the simple mechanical feeders on our Dima Optimat suffers from the "spent tape curl-up inside the feeder" syndrome, in additiion to the tape sometimes getting caught by an edge somewhere. It requires regular checking on new reels until the first bit of the tape has properly exited the feeder after which they run smoothly.

Some of our second-hand feeders came with the very sophisticated hack of a section of small cable duct taped inside the feeder with the purpose of catching the spent tape early on and to guide it out of the feeder. We intend to copy this to all of our feeders. Perhaps this would work with the Mechatronika's as well?

This is something of a risk with Essemtec CLM type feeders as well although primarily its plastic tape with pockets that is most likely to have an issue, luckily its really quite rare. With the CLM feeder tho' I would say the reason it can be a problem is because they are feeders in banks of 10 so the path of the spent tape is well inside the feeder housing and hard to get to, older feeders had nothing in that area between lanes so tapes could tangle with each other.

If you are working with a lot of short tapes there is an increased risk the front of the tape will hit small ledges and catch, so its best to keep a decent leader on your tape or even splice one on. However splicing may or may not cause issues with some feeder designs. ps...
 

Offline MR

  • Contributor
  • Posts: 24
  • Country: tw
Re: Mechatronika MX80 Pick and Place Machine
« Reply #4 on: September 24, 2019, 01:19:12 pm »
We have put together items where we can attack Mechatronika the legal way as well (but we're also a small business and time is an essential matter for us...), they do not meet certain required standards in Europe. Since Mechatronika is more or less a "partnership under the Civil Code" the owners are liable with their personal money it's going to become interesting.

Quote
However - "Optical Recognition based on patterns" I'm no expert but how in basic terms else does image recognition work?

analyze pictures, check rectangles, etc. it works flawlessly if you do it correctly. I only apply certain groups to the components eg. [0402/0603/etc] the identifier of those components is one side must be longer than another, and a certain area must be met to qualify as 0402 or 0603 etc, the pick and place list also has information about which component it is.
[QFN] .. component must be a square (with some tolerance) ... TSSOP8 component must be longer again and meet a certain area.
It works wonderful.

If you apply direct picture patterns that's never going to work perfectly on a quick run, and imagine how long a setup will require if you put more than 60 components in... some components can be grouped in the Mechatronika application yes but if the shape is a little different or if there are unavoidable reflections it just doesn't work anymore. Back then we've started to copy groups SOT323 SOT323-COMPONENTNAMEA SOT323-COMPONENTNAMEB etc. all not necessary... now it's just SOT323 and needs to meet a triangle shape which needs to meet a certain criteria (the optical recognition checks against the final rotation of the component and does some calculation again).
The key-skill required for doing all that is high-school trigonometry.

Direct template pattern based systems need reference pictures of the particular component. They're nonsense. A serious Pick and Place company might have rules about the landing patterns or about the shape of a components but not a direct reference picture of it - it's too error prone.
Check the attached picture... that's what we're doing with our application now. You can also see some dead pixels in the picture (that shows up quality).
So far it's 99.9% reliable I would say I did not run into any issues anymore.

I have stopped using their application a few months ago because when I tried to pick and place a prototype their software complained 7 times when trying to place 10 components... well, sure you can go on and try to reconfigure the whole thing until it will meet their requirement but it's just not economically and not needed if an application is done the right way.

My point was always the mess that happened to us was always the fault of their application - a proper application will:
a) be careful with customer components
b) not let the operator run into the knife

If you're the victim of that machine (especially the Mechatronika Software) all you can do is to resell the machine and hope that the new owner won't kill you for fooling them. I have seen a few users selling that machine quickly and they seem to be disappointed that they couldn't get the entire process work properly - and it's not even their fault it's the fault of the crap software. Even getting a training from Mechatronika in workarounds is not acceptable.

the last thing they've told me is you want a DIY machine... surely that's why I have bought their expensive unit and loose 4 months for our own products. And I've clearly told them where the problems are.
Finally we're still happy with it because it fits 100% into our process with our application, it now eats the Altium pick and place files directly without a stupid importer (like which they have in their application).
Our application shares a global component database in our company, so no need to duplicate entries anymore.

Regarding the nozzles, they are really not difficult to make by yourself and you can easily get the material for free from the scrapyard of a company which works with metal.. all you need is a small lathe, 20mm soft steel stick and some needles from china.
The nozzle changer supports 8 nozzles, all of them are easy to make.
The material cost including buying a small lathe is around 600 EUR (we've bought an old used desktop EMCO lathe which can do the job).
I know the cheap Juki nozzles (clones) from China, they're ages ahead and I probably would not try to make them myself.

For myself it was an interesting dive into mechanics, since the Mechatronika machines are very easy to understand and on a very low entry level.
The Delta servo motors will just shut down if you configure them wrong using the Delta controller - they're very safe to touch (the automatic emergency shut down of the servos happened after the full upgrade/service from Mechatronica from time to time ... ). After reconfiguring them by ourself the machine is also much quieter. Basically they messed up the de-acceleration settings of the machine.
As a normal operator (only) you'll think how you've fucked up that part of course... we've just decided to have a closer look at all the parts and figured out more or less everything.
« Last Edit: September 24, 2019, 02:31:05 pm by MR »
 

Offline MR

  • Contributor
  • Posts: 24
  • Country: tw
Re: Mechatronika MX80 Pick and Place Machine
« Reply #5 on: September 24, 2019, 01:48:02 pm »
Like the Mechatronika the simple mechanical feeders on our Dima Optimat suffers from the "spent tape curl-up inside the feeder" syndrome, in additiion to the tape sometimes getting caught by an edge somewhere. It requires regular checking on new reels until the first bit of the tape has properly exited the feeder after which they run smoothly.

Some of our second-hand feeders came with the very sophisticated hack of a section of small cable duct taped inside the feeder with the purpose of catching the spent tape early on and to guide it out of the feeder. We intend to copy this to all of our feeders. Perhaps this would work with the Mechatronika's as well?

I think so! I will have a closer look at the feeders next week, we need to catch up with pick and placing our products now since we're already overdue with some products...
Thanks for the hint.
 

Offline wilhe_jo

  • Contributor
  • Posts: 41
  • Country: at
Re: Mechatronika MX80 Pick and Place Machine (warning)
« Reply #6 on: September 24, 2019, 03:14:10 pm »
It seems you're not alone with this impressions.
There's a quite long thread over on the mikrocontroller.net-forum.

https://www.mikrocontroller.net/topic/377363 (english translation: https://translate.google.com/translate?sl=de&tl=en&u=https%3A%2F%2Fwww.mikrocontroller.net%2Ftopic%2F377363)

73
 

Offline MR

  • Contributor
  • Posts: 24
  • Country: tw
Re: Mechatronika MX80 Pick and Place Machine (warning)
« Reply #7 on: September 24, 2019, 03:18:41 pm »
That's our post, cheers from Austria btw.
I have seen that someone from our area (in Austria) has also bought a Mechatronika machine, they have sold it quite quickly and went for Autotronik some years ago.
Block-Feeders are not a K.O. criteria, aside of that we could also put other feeders into it since we know the communication protocol.
Someone already did that earlier in a youtube video.

I guess there are not so many customers of those more expensive machines out there and the lower end machines people just deal with the crap software that company is delivering.
Eg. for us it was impossible to reliably place an OnSemi TSSOP8 with their software (other TSSOP8 components were ok), we have spent hours trying to configure the software.
In our software we have no problem with it, and all we do is put the component into a category TSSOP8 (without any optical training).

Mechatronika is aware that we are posting the reviews online, we have told them. They don't care (the owners are supposed to be around 60 years old).

The machine was bought to increase productivity, effectively it decreased our productivity initially with their software. We are also (and are still) open to talk with Mechatronika but they don't bother and just said we focus on existing customers (one nozzle 100 EUR ... there you see where the money comes from).
« Last Edit: September 24, 2019, 03:37:42 pm by MR »
 

Online Kjelt

  • Super Contributor
  • ***
  • Posts: 5767
  • Country: nl
Re: Mechatronika MX80 Pick and Place Machine (warning)
« Reply #8 on: September 25, 2019, 10:31:32 am »
Well that explains why a dutch distributor dropped them a few years back  :)
 

Offline MR

  • Contributor
  • Posts: 24
  • Country: tw
Re: Mechatronika MX80 Pick and Place Machine (warning)
« Reply #9 on: September 27, 2019, 12:22:29 am »
We will keep testing and extending our pick and place application until December afterwards we can offer an upgrade kit for their machines.


So far we support in our application:
- Web based Stock/Component management, including Barcode/and Printer and an android app for setting up feeders
- The components are stored in a mysql database, our buyer is entering the components directly, and the pick and place machine is sharing the same data
- Components are all optically analysed (no template pictures anymore as mentioned)
- Multiple-Multiplanels are supported
- The pick and place list can be edited like a normal excel table
- Easy 2D AOI
- There's an OpenGL view of the pick and place which can zoom in/out the captured areas of the Pick and Place table
- Components are drawn onto the camera screen and virtual pick and place table view (that helps for verifying that the components will be placed at the correct position upfront)
- Fiducial check can be set up as separate task (Lock PCB, afterwards coordinates are recalculated and components can be placed eg. one by one - or using multiselect many at one time)
- The PNP head can be jogged by clicking onto the camera screen
- Runs with Linux dual screen (one screen for the camera, the other screen for statistical purpose)
- The system works with Fiducials and anonymous markers (I always use Fiducials, anonymous markers are just in case no Fiducials are available on a board - that needs 2 markers and 2 components for the setup)
- Multiselect is available for all tables (Mechatronika does not support that... nowhere).
- Several items like re-programming the component position/pickup height can be done directly on the video screen using hotkeys.

There are still a few more items which we want to implement:
eg. an error list, if a feeder is jammed or reel is empty it should be disabled while the entire panel should still go on with pick and placing, once everything is done the operator can go through the error list and fix the open issues.

Now we just need to get rid of the single pnp head and install 4x it's on our list.

That's the result of several months of work so far and we're very happy with it now, even though that's not what we have bought the machine for to write our own application.... but as usual we've learned a lot and luckily not all is lost.
 

Offline IconicPCB

  • Super Contributor
  • ***
  • Posts: 1331
  • Country: au
Re: Mechatronika MX80 Pick and Place Machine nonsensnsical allegations
« Reply #10 on: September 27, 2019, 01:31:19 am »
Seems to me You bought a secondhand machine then decided to  remake the design  in your own image while complaining how it does not work,  could have never worked, amplifying the allegations by posting same information on multiple boards and hoping to hear others scream "... the sky is falling..."

If I were a shrink i might be qualified to make claims of behavior befitting a sociopath  ( i am sixty five not sixty ) but seeing how i am not... i will not.

All i can say is I have an M10V and i do not see any of the issues you are enumerating.

In any case I look forward to your input on github.
 
The following users thanked this post: rx8pilot

Offline MR

  • Contributor
  • Posts: 24
  • Country: tw
Re: Mechatronika MX80 Pick and Place Machine (warning)
« Reply #11 on: September 27, 2019, 03:43:30 am »
The M10V does not have flying cameras, I think they cause big issues, I'm only using the stationary camera for a few components, I cannot say too much about them.
Many issues also come from the Mechatronika feeders (jammed feeders, or the feeder is not strong enough to advance a tape .. and it's enough if that happens to one out of 60 feeder positions), however we also shipped them to Mechatronika for servicing them and they charged us over 2K Eur for the service. So absolutely no excuse for having it somewhat work. Either it works perfectly or it's shit.

We bought a second hand machine yes, but shipped it to Mechatronika for a full service, they have replaced the motors and cameras (look at the first post and the camera which was inside!), and they shipped it back with misconfigured the servo controllers. We have replaced the top camera with a 10$ CVBS camera afterwards which had a slightly better quality.
If you have a close look at the entire machine there's not much technology behind it, personally I don't see any value in the machine anymore since I dived into it so deep, all standard components.

There's not even an option to throw off a component with air pressure and it's possible to have 0402 components stick to the needle and the software telling the operator to remove it manually, because they saved a few bucks on a pressure switch on the PNP head. I don't know why we had so many issues with Windows back then but it's all gone with our application now.

We have faced big issues with the optical recognition and the flying cameras, we still have some logs around and it's just awful how many standard components were dropped due to misdetection. And as mentioned we don't need to take reference pictures of standard components anymore - so no drops from the optical recognition side.

On-Semi TSSOP - I have spent at least an hour trying to set up a pattern back then - and it did not work properly out of 10 components at least 6-7 drops. Again now - no issue anymore.
I have paid 40K Eur for all that (including service and upgrade from Mechatronika) and believe me for that price I did not intend to write my own application - but we had no other choice because Mechatronika's support more or less did not exist. I would expect that if you pay like 10K for such a pick and place machine to receive an awful software.

Additionally to the main issues I don't want to take reference pictures of standard components, this is really extra work which can be completely eliminated. Only lunatic would want that from customers for standard parts, sot323, 0402, etc.

----
the detailed log about the misconfigured servo drive was:
- we received the MX80 back from Mechatronika with new motors and new cameras
- we have started to pick and place a new project, after a few minutes the portal started to shake/oscillate heavily
- since we urgently needed those PCBs we tried to figure out by ourself what was wrong. First we aimed at mechanical issues, tightened a screw which seemed to be loose to us.
This temporarily fixed the servo drive because it added more tension to the belt which is connected with the gantry (we have told Mechatronika about that), because of that the motor required more power to move the portal and that might have matched the PID controller settings better. After all we know about it now Mechatronika should have immediately said don't touch that - sorry but we need remote access to a computer which is directly connected to the servo controllers via USB to fix this.

So with that setup we were running for around a half year, the vibrating portal happened again randomly but not that often at least.
Early this summer we decided to have a closer look at it again and figured out:
* adding more tension to the driving belt of the gantry almost damaged it
* Delta provides a windows tool for doing the Servo setup, the ramps were misconfigured, it took us 2 days to figure out how to set up the controller properly (yes we've never done that before). The gantry and servo drive are the heart of that machine, once you have that under control you can say it's your device already. No more oscillating portal anymore.

So Mechatronika has made a terrible mistake and instead of fixing it they just ran away, absolutely unacceptable.
« Last Edit: September 29, 2019, 05:34:33 pm by MR »
 

Offline MR

  • Contributor
  • Posts: 24
  • Country: tw
Re: Mechatronika MX80 Pick and Place Machine (warning)
« Reply #12 on: September 27, 2019, 07:06:57 am »
"manual" QFN placing, after changing the Tray (with our software):

https://streamable.com/muepw

regular run (also with our software):

https://streamable.com/t8cfl

in the meanwhile we have fixed the sequencial X/Y drive (as it was done with Windows before and run X/Y at the same time, the sequential drive can be seen in the Video), so now the machine is even faster than with the Mechatronika Software.
Since we grabbed the RS232 commands via portmon for winxp we have implemented the same problems first of course.

One screen is for the video the other one (not seen in this video) for status reports.

This is a very exceptional situation where a customer writes his own application for such a pick and place machine because the Manufacturer is unable to provide proper support and unable to provide a proper application.
« Last Edit: September 28, 2019, 09:42:15 pm by MR »
 

Offline mairo

  • Supporter
  • ****
  • Posts: 141
  • Country: au
Re: Mechatronika MX80 Pick and Place Machine (warning)
« Reply #13 on: September 30, 2019, 09:58:57 pm »
I have Mechatronika gear and it has been working for many years + have never had any issues with correspondence.   
 

Offline MR

  • Contributor
  • Posts: 24
  • Country: tw
Re: Mechatronika MX80 Pick and Place Machine (warning)
« Reply #14 on: October 01, 2019, 01:38:54 am »
- What are you placing?
Our side:
* 0402
* all feeders occupied
* QFNs
- Which machine do you use?
- Is it stepper or servo based?

I think they do not have the servo based machine under proper control (or they just made a mistake with our machine before shipping it out to us). I have also read in the other forum that the stepper based system seems to be more reliable.
Especially those problems added a whole chain of issues and reworking need to our projects and has shown us the critical limitations.

But I guess you would also run into issues if you would use HASL based PCBs because the illumination of the PCBs is a design issue of the machine. eg. ENIG PCBs did not cause issues with the fiducial detection. There's no good way for verifying if the fiducial detection worked out or not with Windows - we have implemented the option "Lock PCB" into our application - it will just try to match the fiducials and draws the designators onto the camera screen so you know where the component is supposed to end up. It also features the option "snap to blob" which centers the detected fiducial automatically (in the training process).


fiducial_a -> default PCB illumination
fiducial_b -> PCB illumination using the shielding.

The shadow of HASL-based PCBs is not predictable, attached is just one example. The fiducial might also end up having a dark spot on the other side. Depending how the process turned out for the particular panel.

For 0402 such an offset can be an issue already - and we had that. The solution was to wrap around the shielding as you can see in the video:
https://streamable.com/t8cfl
I wouldn't use that machine without it anymore.

Just adding more power to the LED ring cannot fix this problem because otherwise the entire PCB will be too bright.

 
Mentioned this -> got no reply from Mechatronika back then.

Since the machine was used -- but shipped to Mechatronika for a full service, I also attach a picture of the feeder did they change this in a newer design? It would be better if every stripe would have its own channel out of the machine.
If a reel is completely full and the friction between the slot and reel is too big the feeder won't advance the tape, the workaround is to unwind the reel a little bit.
A proper solution would be to remove the friction.

I really wonder you never saw such an issue? Because to me that seems to be impossible.

You see the stripe leaving the machine at the wrong place, this gap should not even exist -> design issue.
At the right side of the feeder you can see the bottom of the reel holder, the reel is just lying on the other side so forwarding the tape will cause friction on that plate, and in some cases it's just too much friction so it won't advance the tape (recently a 0.1uF reel). This can also happen if the reel is almost used up because the diameter of the tape on the reel is getting smaller (so it needs to pull harder). Again they serviced the feeders and charged us for it, so no excuse for not having it work properly.
« Last Edit: October 01, 2019, 03:33:02 pm by MR »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf