Did someone use CNC as "pick and place" robot?

[emvali]

Hi, emvali and sorry if I offended you in any way. I didn’t mean to.

No problem, as I said, the whole experience still makes me jump like that after relatively long time.

I assume you are talking about used machines and feeders. Are there any brand new ones available that you can buy with 20-30 tape feeders for under $25K?

There are some, one I could think of is this www.mechatronic-systems.com
The prices start from 9,850 EURO, they have the dispenser option as well and they sent me the software demo, light years more advanced than Madell's if judging by the ease of use and user interface. I found about them much later unfortunately.
This is not a recommandation, I have no experience with this company, but everything looks more serious.

Agree. Let’s refer to it as “learning curve”. In case with Madell its pretty steep and their user manual only makes things worse. I had to read it more than once and then a couple of times backwards. Also, I believe they removed the “state of the art” wording from their web site and even added “education and research”

I still believe that Madell’s solution is just a bicycle with a motor compared to the grown up machines but it’s not an excuse for it not to work upon arrival. Minor adjustments and calibration should be expected as with any other precision equipment after shipment, but again, it should work as advertized. If it does not, the manufacturer should rectify it quickly. I understand it’s not exactly what happened with your purchase.

My big claim is not the way the user's manual is edited, they are Chinese people, can't blaim them for that.
What I do blaim is the fact that they sent me the PX3700 user manual when I asked for details, saying that the software is the same, which indeed is.
But the PX3700 has a step motor on Z while the SX has a pneumatic actuator/plunger (this is what I got originally). Also the machine in the manual is shown with Fuji pneumatic feeders while the SX arrived with the lead weight "invention" and feeders that were good for paper tape only. I had to mill myself extra feedres for plastic tapes.

OK, some calibration is always acceptable, but here is what happened:
They suggested to buy the computer for extra $600 claiming that this way they can make all calibration and settings making the machine "plug&play".
I can say that no calibration was made, even the basic "home" initialization resulted in an error. The nozzle changer location was also not calibrated, the head crushed into the nozzle changer for a beginning, later I found that #5-6 were installed out of machine's limits and the same with some "feeders" installed on the right side (by them).
This was just a little about the "learning curve", but here I had to deal with solving problems, not learning. I only try to think what somebody with no repair experience would do - send the machine back (shipping it costed about $1500)? call a technician from Madell (and as you see I am very far away from US)? This is really not something you expect when buying new equipment.
What happened after that is a long story of mafunctions, wrong assembly issues, software bugs and you name it. I "played" with the machine and my nerves for almost 5 months until I decided to leave it and buy a used machine.

Some antivirus packages think that Madell’s executable is infected with a virus. In fact it is just using code obfuscation algorithm that is known to be also used by some malicious programs to prevent reverse engineering. I pointed it out to them quite some time ago. Were there any other viruses found?

Software is not my strongest side, you are probably more experienced with this point. But.... it started when I wanted to copy the manual to a flash disk in order to print it from my computer. As I plugged the disk-on-key my computer's antivirus (ESET NOD32) jumped like crazy. I called Madell - as they request not to install any software on their computer - and they allowed to install an antivirus and remove it after a computer scan.
Now, thinking logically, while scanning the computer the NOD32 found about 12 viruses/trojans and removed them. After that the file exchange between the machine's computer and my computer was clean.
If this was a reverse engineering protection algorithm, how was it so easily removed by the antivirus?

I remember saying it on adafruit forum that Z-stepper losing steps is totally unacceptable. The table can potentially go unlevel if shipped/packed improperly but using sticky tape for the camera mount is simply wrong. I think I now see where you are coming from with the “piece of junk” description. I hope they have improved the design quite a bit since you saw it. Their recent pictures on the web site show rather nice looking machine: http://www.madelltech.com/SX1000.html

The Z stepper was installed on the dispenser in my case, the head worked with a pneumatic actuator. After realizing that as a pnp I will not use this machine (among tons of software problems also it was not possible to place a part at angles other than 90 deg steps, a basic feature that I needed) I tried to justify it's existance as a paste dispenser for small batches where a mask and printer setup would cost and take time when all I need is a prototype for example.
The stepper got quite hot after short use, and after about 100 up/down cycles it started to touch the PCB and if continued it presses it until something bends. Madell didn't believe me at the beginning, later they simulated the problem in their machine and found the driver they used in my machine was causing the problem. I suggested until they find the solution to make for me a software change where the stepper will raise each time up to the optical switch and there it will be reseted for every cycle.
They made this change and temporarly the problem was solved, until they found a driver that works. Of course, making a 400 dots PCB took about 20 minutes, but who's counting.
This is when I found that the frame is not leveled. The dispenser's needle height at one of the corners was too low, almost touching the PCB if zero was set in another corner or center. I should mention that the PCB was quite small and I used a proper support. I disassembled the adjustable slide of the frame and found that in the opposite end they placed a small spacer, a piece of metal ugly grinded and not straight. I made all measurements, made a new spacer and leveled the frame. A bad manufacturing/improvisation and absolutely not a shipping issue.
After a while the new stepper driver arrived, with no instructions how to replace it and of course very different from the old one. I downloaded the driver's manual from the manufacturer's web site, had to calculate the right current limit resistor, find a 5V line which this driver needed and in plain words, had to figure how to replace the old one myself, as they were unable to help. Quite an aftersale service, isn't it?

Months later the saga started again, when they sent me an upgrade "kit" allowing to use the dispenser's Z stepper for pnp, cancelling the pneumatic system. I will not write about this story that took weeks to accomplish.

About the changes of the machine, what I got is still in their web site, the machine with the optional stand. Yes, I see they make blue annodizing - if I ever send you a copy of my first email I sent them you will see that I complaied about the overall finish of the SX, except for a decent annodizing also the aluminum profiles which looked as they were cut with a chainsaw.

Which Philips do you get? What is the smallest part/pitch can it place reliably? What’s your impression with it? Would you buy another one if you have to?

I got an Eclipse 2 for $7000 including about 200 feeders and purchased additional 12 and 16mm for $150-200. I completed the dispenser part with a CAM/ALOT dispenser in almost brand new condition purchased for $1300 which works lovely. A few months ago I also purchased a DEK265 printer, as now I can make also normal batches and not only prototypes.
The Eclipse, maybe old, but works great (exactly as expected from a 10 years old Mercedes) I never needed any service for it so far, parts are available anywhere and it took several hours only to assemble the first PCB.

I work with large parts, 0805 is the smallest among passives but it assembles with no problem odd QFNs. I had the same requirement from the SX, I didn't expected more. A friend of mine travelled all the way to Madell and they were unable to install 0201 as they claimed - he decided to cancel the order, never got his down payment back and he also purchased an used Philips, I think a Topaz.

By the way, which y*group you are referring to? I only remember discussing it here and on adafruit’s forum.

Sorry, my mistake, I thought I saw you in Yah00 groups

[scsi]

There are some, one I could think of is this www.mechatronic-systems.com
The prices start from 9,850 EURO, they have the dispenser option as well and they sent me the software demo, light years more advanced than Madell's if judging by the ease of use and user interface. I found about them much later unfortunately.
This is not a recommandation, I have no experience with this company, but everything looks more serious.

Looks like their low-end machine has only manual tape feeders. What was their price for automatic feeders? I highly doubt that you can buy a new machine with 30 automatic feeders for anything less than $30K today.

[…]Also the machine in the manual is shown with Fuji pneumatic feeders while the SX arrived with the lead weight "invention" and feeders that were good for paper tape only. I had to mill myself extra feedres for plastic tapes.

I guess those Fuji feeders cost close to $1K a piece brand new. Lead weight is well under $1 and works as well or even better because of the vision assist upon pickup.

What happened after that is a long story of mafunctions, wrong assembly issues, software bugs and you name it. I "played" with the machine and my nerves for almost 5 months until I decided to leave it and buy a used machine.

About a year ago I saw one SX machine sold on eBay in Israel. Was that yours?

Software is not my strongest side, you are probably more experienced with this point. But.... it started when I wanted to copy the manual to a flash disk in order to print it from my computer. As I plugged the disk-on-key my computer's antivirus (ESET NOD32) jumped like crazy. I called Madell - as they request not to install any software on their computer - and they allowed to install an antivirus and remove it after a computer scan.
Now, thinking logically, while scanning the computer the NOD32 found about 12 viruses/trojans and removed them. After that the file exchange between the machine's computer and my computer was clean.
If this was a reverse engineering protection algorithm, how was it so easily removed by the antivirus?

This indeed sounds like a virus infestation to me and not a single false positive on the PX3700.exe

[…]The stepper got quite hot after short use, and after about 100 up/down cycles it started to touch the PCB[…]

I’ve heard there were some issues with the Z-axis stepper driver quality. Do you remember if yours was made by US Digital?

[…]I completed the dispenser part with a CAM/ALOT dispenser in almost brand new condition purchased for $1300 which works lovely.

Is that time/pressure or an auger dispenser. If later, I would like to hear more about it.

[…]A friend of mine travelled all the way to Madell and they were unable to install 0201 as they claimed - he decided to cancel the order, never got his down payment back and he also purchased an used Philips, I think a Topaz.

I believe they only recently figured out the 0201’s and had to re-design the feeder brackets for it to work reliably.

Topaz and Eclipse II are nice machines with laser centering and all that. Still there is a market for this kinda thing...

http://4.bp.blogspot.com/__GU_oynHw2...-on-orient.jpg


-scsi

[emvali]

Looks like their low-end machine has only manual tape feeders. What was their price for automatic feeders? I highly doubt that you can buy a new machine with 30 automatic feeders for anything less than $30K today.

I think they can, I have their price list and the automatic feeders, a simple type however, are avilable.

I guess those Fuji feeders cost close to $1K a piece brand new. Lead weight is well under $1 and works as well or even better because of the vision assist upon pickup.

Madell claimed they are about $250 each, but probably not originals. Also an I/O card is needed.

About a year ago I saw one SX machine sold on eBay in Israel. Was that yours?

Yes, actually the guy who bought it got a completely redesigned machine, with software updates and special software changes I requested, completely rebuilt pneumatic system and step control on Z. He made a good deal, this machine would cost >18000 at Madell. But I already purchased the Philips so decided to sell it.

I’ve heard there were some issues with the Z-axis stepper driver quality. Do you remember if yours was made by US Digital?

No, the first one was chinese. Later they sent me one made by Gecko.

Is that time/pressure or an auger dispenser. If later, I would like to hear more about it.

This machine was originally equipped with auger pump, but was taken off and replaced with a plain syringe. I got also a smaller CAM/ALOT for parts, equipped with auger pump so I installed it on the first one (all the system was originally made for it, so it was no problem).

However, the syringe worked fine and the solder paste specialists don't like the auger for several reasons. For most my applications I use the time/pressure option.

I believe they only recently figured out the 0201’s and had to re-design the feeder brackets for it to work reliably.

This is what I am talking about - it is ok for a product which is finalized to get updates from time to time, this is true for any complex machine involving software, DSP, electromechanical and pneumatic components.

But what Madell are doing is they release an unfinished product (in this case claiming to be able to place 0201), assembled by request and then use the customers as guinea pigs for experiments and debugging. As I posted somewhere in the past, usually people should get paid for this, and not pay for being into this experiment.

Anyway, they had a major problem with the SX feeders, partially solved using my suggestions:

The part is advanced by a pneumatic plunger that goes down and pulls the empty part of the tape. This plunger is practically a 1/4" bolt with the head removed then grinded and attached to a pneumatic actuator.
The tape is inserted in the feeder, which is a simple slot with 2 covers. after it passes the first cover the removable tape is pulled back and the fishing weight clipped to it. The uncovered part passes the second cover and goes in the machine. On it's way it passes another aluminum surface, this is where the plunger hits the tape and pulls it. As the distance between the covers is fixed, several small parts are often uncovered and exposed. They use a separate surface for the plunger in order not to cause the exposed parts to jump or flip, this is what they explained. But a pneumatic actuator is strong, the plunger hits the tape seriously and this separation is good up to a certain point. After all, both supporting rails are bolted to the same aluminum frame and the shock passes, specially with small parts such as 0402 or 0201 which are light and because of their size several are exposed.

By the way, any "normal" feeder has adjustment for the pitch and the part is exposed only short time before being picked by the head.

Now to the second problem.

The software (maybe since then it was changed) is using step units, not metric or imperial. The steps are machine steps and when using inch based screws the steps are related to imperial units - if I remember right, 2000 steps = 1".

By the way, Brian Dockerty, the guy who wrote the gerber to dispensing translation software and perhaps knows much more then Madell about step motors and screws added in his software a correction option for these steps, you can measure a certain distance, find the difference between the number entered and the real distance, divide by 2000 and enter the practical value, while the program re-calculates accordingly. A feature Madell is not even aware of.

So, the machine works in imperial steps (a unit of measurement invented by Madell which needs to be learned by the users) while all SMD tapes are metric.
The result is that after a while, with 0805 parts this is about 12 parts, the first of at least 2 exposed parts is very close to the second cover, a place where the camera is not looking for the part.
The part is skipped, then enters under the second cover and later, if it won't jam the tape is crushed by the plunger. This the good part, the worse is that the camera is unable to find the second part at all, the plunger keeps advancing the tape and jamming the machine or errors are displayed.

I made a simple calculation, and found a plunger advancing step closer to the metric division - I don't remember exactly the numbers, but instead 50 steps 53 or 54 are much closer to mm division.

This was a software change initiated by me and adopted by Madell.

Speaking of software, are you able to display gerbers with parts rotated for example at 45 deg? How about placing them?

I assemble round LED boards, where parts are at many angles around the perimeter. Guess what they suggested, as the setup allows only 90 deg rotation (the machine's specs stated full A axis control) - to make pictures of the part in all angles (I have some 30 different angles and polarity matters) then open the same part in the part list several times and in each to "lie" the machine about the angle by providing the various angles snap shots.

The second option suggested was to make a "blind" placement, with no part recognition.
To better explain myself, the angle can be any entered, but the camera recognition only "looks" for 90 deg steps.
I guess they are waiting from some other user to solve the problem and give them the solution.

Believe me, this is still a small part of what I've been through with this "piece of junk" - if you'll hear the story with the way the dispenser worked you will laugh. One thing for sure, I gave them a solution for this too, which saves them lots of money and parts. They supplied the machine with a dispensing controller like the manual stand-alone type which later I discovered (and approved by them) was not necessary, with a small change the pneumatic system can do the job. Ha... time/pressure system, working with a relay that simulates the pedal of the stand alone unit, which activates another relay in this controller, which activates a valve.... excellent timing I may say.

Really, only thinking of the time I invested in this machine instead of working makes me feel like an idiot.

[scsi]

This machine was originally equipped with auger pump, but was taken off and replaced with a plain syringe. I got also a smaller CAM/ALOT for parts, equipped with auger pump so I installed it on the first one (all the system was originally made for it, so it was no problem).

However, the syringe worked fine and the solder paste specialists don't like the auger for several reasons. For most my applications I use the time/pressure option.

Can you tell us a bit more about the auger dispenser from camalot. What were the reasons your soder specialists didn't like it? I imagine cleaning and maintenance would be one big reason. Tweaking feed pressure is another one. What else? What was the smallest dot size were you trying to dispense?

I imagine that for larger dots (0.5mm dia and above) the regular time/pressure systems will work but for smaller deposits (0.3mm and below) the auger is a better way to go.

I'm developing a new kind of auger pump for microdispensing and asking people experienced with augers to provide some feedback here on adafruit: http://forums.adafruit.com/viewtopic.php?f=42&t=16273

Thanks in advance!

scsi

[emvali]

Can you tell us a bit more about the auger dispenser from camalot. What were the reasons your soder specialists didn't like it? I imagine cleaning and maintenance would be one big reason. Tweaking feed pressure is another one. What else? What was the smallest dot size were you trying to dispense?

I imagine that for larger dots (0.5mm dia and above) the regular time/pressure systems will work but for smaller deposits (0.3mm and below) the auger is a better way to go.

Usually I use #18 stainless steel needles, for SO and even TSSOP I need to switch to #20. I think that 0.3mm still can be done, but not less.

#20 is also the smallest recommended for the solder paste we use, which is KOKI 951.

The KOKI distributor explained that the solder paste, which is basically a mixture of flux and small solder spheres, has 2 major things that sould be taken into attention:

The size of the solder spheres is what limits the needle size, as they should flow through it smoothly.
Regarding auger pumps, he claims that there is a probability of "crushing" the spheres, this is what he is afraid of.

Now, about cleaning, this is not such a problem:
The CAM/ALOT has a purging function. All you have to do is take away the solder paste syringe and install one with a special cleaning grease, hit "purge" button and this is it.
The removal of the syringe has no effect on the setup and it is always recommended to put it in a refrigerator anyway.
By the way, working at room temperature for long time should also be avoided as the flux becomes too liquid and the solder spheres accumulate on the bottom

[scsi]

Thanks for the info!

Usually I use #18 stainless steel needles, for SO and even TSSOP I need to switch to #20. I think that 0.3mm still can be done, but not less.

#20 is also the smallest recommended for the solder paste we use, which is KOKI 951.

Trying to acheive 0.3mm dots with #20 needle is more than pushing it. Essentially the rule of thumb is that the smallest dot diameter cannot be smaller than 1.5 x ID of the needle. In case of 20 gage we have an ID of 0.6mm, therefore the nominal smallest dot that can be maintained reliably has to be 0.6 * 1.5 = 0.9mm

For 0.3mm dots you'd want to use #25 or even smaller needles and at that point you have to use much finer solder powder (Type-5).

I'm currently testing my auger with #25 and 27 needles and it works quite fine with the proper paste. Here's a quick video showing 0.2mm dots dispensing: [nomedia="http://www.youtube.com/watch?v=HoT4s2wY17k"]YouTube- Microdispensing testing with Type 5 Solder Paste 0.2mm dots dispensing[/nomedia]

[...]The size of the solder spheres is what limits the needle size, as they should flow through it smoothly.

Yes, they usually recommend to choose the powder size so that the largest sphere size will be less than seven times smaller than the ID of the needle. With type-5 paste it's 25um for the largest shere. Otherwise the plugging will occur.

[...]Regarding auger pumps, he claims that there is a probability of "crushing" the spheres, this is what he is afraid of.

Correct. The alloy is very soft and the spheres can easily get crushed/flattened by the auger if the gap between the screw and the shell (tube) is not sufficient. This commonly happens with augers that are not designed for solder paste dispensing. In my case I had to widen the gap (make the auger smaller) to almost 0.2mm for this same reason.

-scsi

[rayscott]

I am not sure of the exact materials to move around, but you could build your own, load and unload mechanics. I did this with a PLC integration to automatically load and unload Chrysler timing gears for a fanuc mill. The integration worked great. It all depends on your budget for the project. If you do the work in macro programming, it could be done inside Mach3. Not considering the costs of labor, the project could surely be done on a hobby budget.

Ray Scott

[bootstrap]

I just posted new thread about a super-cheap way to build a pick-and-place device that can place BGA, QFN and other "hidden contact" components with ~1 mil precision.

Since you guys have been thinking about pick-and-place issues for a long time, please read the message and make comments. Thanks!