NeoDen YY1 Pick And Place Machine With Under $3K Price for Hobbiest/Low vol Usag

[drgerber]

Well, that happened as well. The nozzle gets stuck sometimes and it warns in the interface. But If I think, the drift also occurs in Y direction. if it was only in X direction, I could understand that it is the feeder picking.

Ah, okay. Then the feeders might not explain it if it goes in both directions.

BTW mine is running on 2022.006 with no issues.

[asmi]

It would be possible to design a custom board with advanced stuff and closed-loop system for X and Y axes. Then this device would be really accurate.

This device was obviously designed to a price point. It's always possible to make device better by adding more stuff, but the price tend to get out of control really fast.

[alpelectronics]

Finally I assembled a full panel. It is better now with the new controller. I also did minor adjustments for better accuracy. I would say the biggest problem in this machine is the feeders.

But I'm happier now. Thanks to NeoDen for their support.

Caner,
Alp Electronix

[EEVblog]

Finally I assembled a full panel. It is better now with the new controller. I also did minor adjustments for better accuracy. I would say the biggest problem in this machine is the feeders.

If there weren't many reports of feeder problems, I'd be getting the YY1. They really need to solve this. And I would have expected they would be able to get it right as they have been around a decent amount of time now.

[newto]

If there weren't many reports of feeder problems, I'd be getting the YY1. They really need to solve this. And I would have expected they would be able to get it right as they have been around a decent amount of time now.

Finally had the time to get ours fully set up last week (annual inventory was a giant mess and then had a million other things to do...)

The only persistent feeder issue I've been having is shallow 8mm plastic tape (a bunch of 1206 capacitors). Took a few goes, but I managed to sort them out by moving them from the left side of the machine to the right, and putting them close to the motor. I think where I had them to start with at the front didn't put the right pressure between the wheels, so it would pull the tape forward whenever the motor ran. Seems to be solved now, but I definitely wasted a bunch of caps trying to figure it out.

Other than that, the only issues I've had is needing to turn the wheels by hand to break the friction if the machine has been left a few days, and I've had a minor issue picking from cut tape, but I think that's the fault of my custom holder being too tight and pinching one of the tapes and binding the test point loop in it, and another was an 0201 sized diode that I had trouble taking out with a razor blade after the machine failed on it over and over..

I remember seeing some 3d printed designs that used a gear system triggered by the head to advance tape instead of the head itself, I wonder if a similar design could be made to use the needle drop or dragging motion with a simple clockwork to advance without having the head make contact with the tape...

[seon]

This place in Melbroune sells used PnP machines.
https://www.resurface.com.au/used-equipment/pick-and-place

Resurface is Hawker Richards - You'll need a 3rd mortgage and possibly to sell one of your kidneys to afford one of those machines

I have read disparaging comments about HR on here more than once and I'm not sure why. All of my dealings with them have been very good and Resurface provides a valuable service of hooking people up with used machines (often machines they initially installed and have serviced) at good prices. You really need to be on the mailing list to be able to grab opportunities when they become available though as the website listings are largely out of date.

I wasn't saying anything bad about Resurface or HR - sorry if it came across that way! I was just stating that prices for new machines, and prices most local folks are after for their used machines are exorbitant compared to what you can buy used gear for OS. I'm sure HR and Resurface are awesome folks and great to work with.

Seon
Unexpected Maker

[fourfathom]

Has anyone with the YY1 been able to get into the "System Parameters" screen?  A password is required and it's not accepting whatever I type in what the manual suggested (2022) (?)  I just sent a query to Neoden, but timezones, etc...

So I'm finally trying to get the YY1 up and running, and of course the feeder tape-advance throws the 0603 resistors out of the paper tape (I'm starting with a single sacrificial reel of resistors in feeder #1).  The "Pick Test" cleanly advances the tape and picks the component, but when trying to actually run the mounting process the tape advance pin is being dragged too rapidly (I guess) and it isn't returning to the proper location for the next advance.  I have the "Mount Speed" set for 25%, but that doesn't affect the needle drag speed (I realize that the needle drop speed is not controllable, but at the moment this isn't the problem).

So I was hoping that something in the System Parameters would fix this, and I need a working password.  And please don't tell me how stupid I am for buying this, or Neoden is for feeder design or software update issues.  That was already part of my decision process -- this is my hobby and I am prepared to deal with all that.

Thanks,
Paul

[Selectech]

I have a YY1 and 2022 worked on mine.

[fourfathom]

I must have been doing it wrong. 

The password 2022 works, and after doing the Camera, Head and Needle calibrations, and setting the needle speed to 25% the YY1 is now cleanly picking and placing.  I can probably increase the speeds from 25%, and I need to load in the reels that I'm going to be actually using, so I'm sure there's more excitement to come.  But even if I can only place the 0603 parts and have to hand-place the semiconductors (and I doubt if things will be that bad) this is still a big help.  I hand-edited my KiCad placement file into the Neoden format, and while that was easy enough I'm still going to automate that conversion.

This quite cheered me up!

[fourfathom]

I'm now trying to set up a board using the KiCad .pos file.  I've got it massaged and merged with the YY1 .csv file, and that all seems OK.  My feeders are loaded and calibrated, and the machine camera, head, needle are also calibrated.

Component positioning near (0,0) looks good, but the component placement gets worse as 'X' gets larger.  'Y' is probably off as well, but 'X' is quite noticeable, about 0.5mm error once 'X' gets to 7cm.  How do I calibrate the positioning?  I have tried setting a fiducial and a component location as the manual says, but I can't see that this helps.  I also can't see *how* it could help, other than perhaps correcting for an offset.  Does the YY1 even know where the fiducial and component are supposed to be?  At what point during the placement process does the YY1 look for this component and the fiducial?

Perhaps I don't need to understand how it works, but what is the proper method for using the fiducial and component position?  And for correcting any scaling errors?

No doubt I could manually adjust the position of each part, but that's a lot of work and destroys my design process integrity.  And I shouldn't have to do that anyway.

Thanks!

[fourfathom]

Progress:  I now see (more or less) what the YY1 is doing with the "Edit Fiducial" steps.  The component is the first one in my pick-list, and I guess me setting this is a board XY position calibration.  The fiducial is something that the YY1 looks for every time I start the placement, and is probably used as an automatic XY calibration to compensate for next-board alignment.

I've now placed a test-board (double-sided tape) with about 100 0603 resistors/capacitors, a few LEDS, some transient diodes, and a regulator in a SOT-223 package.  It's looking pretty good.

[ycui@eml.cc]

Technically it uses a PCB origin point for X/Y (bottom right corner of your board) and then one fiducial and one component. That is how I believe they basically triangulate the board for alignment. It's not a perfect system, but it works well enough in my use cases.

That sounds crazy. If you can do one fiducial then you can do two.

Unless your software people have such poor maths skills that they can't figure out how to compensate for board rotation... Wouldn't surprise me.

You will be surprised. To fully correct all 2D linear space distortion, that includes X, Y scaling, rotation, and skew, takes multiple PhD to get correct. I am talking from a semiconductor equipment vendor. Certain big name semiconductor vendor cannot correct 2D distortion in their multi-tens of millions worth reticel inspection machine.

[coppice]

To fully correct all 2D linear space distortion, that includes X, Y scaling, rotation, and skew, takes multiple PhD to get correct. I am talking from a semiconductor equipment vendor. Certain big name semiconductor vendor cannot correct 2D distortion in their multi-tens of millions worth reticel inspection machine.

It takes one reasonably smart person with the right character. Someone who doesn't hand wave away the details like most people do, and really gets into those details. After that person has got it working beautifully, watch out for the hand wavers rotting the software during maintenance.

[SMTech]

To fully correct all 2D linear space distortion, that includes X, Y scaling, rotation, and skew, takes multiple PhD to get correct. I am talking from a semiconductor equipment vendor. Certain big name semiconductor vendor cannot correct 2D distortion in their multi-tens of millions worth reticel inspection machine.

It takes one reasonably smart person with the right character. Someone who doesn't hand wave away the details like most people do, and really gets into those details. After that person has got it working beautifully, watch out for the hand wavers rotting the software during maintenance.

Without going looking, I think its a safe bet that there are perfectly good libraries that can handle rotation calculation and can be applied here if you didn't want to code it yourself. I'd also wager, unless they really screwed it up, this core piece of code barely changes, software updates will typically target UI glitches or improvements and head travel optimization. Equally I think there's a perfectly good argument that for a cheap pick and place machine with limited repeatable accuracy, assuming that the PCB is aligned flat against a fixed rail is a perfectly good expectation. This means one fiducial to locate the board, works, even tho' that may seem naff and not "as good" as others.
 For those other 2+ fiducial machines, how many of them are using two for anything other than PCB rotation? More expensive machines have thermal compensation, linear encoders and high end motor control, they can use this sexy maths but without the hardware support it's wasted effort.
The problems with the YY1 seem to be the same issues from every Chinese manufacturer, there is no iterative improvement on software they have already made, or even hardware. They start over, every time they make a new model, keeping some old mistakes, making new ones and duplicating tonnes of effort. It's only USP is a nozzle changer and while that is a good thing, it's very late to the party and would have been better fitted to an established machine in their lineup.

[nimish]

Technically it uses a PCB origin point for X/Y (bottom right corner of your board) and then one fiducial and one component. That is how I believe they basically triangulate the board for alignment. It's not a perfect system, but it works well enough in my use cases.

That sounds crazy. If you can do one fiducial then you can do two.

Unless your software people have such poor maths skills that they can't figure out how to compensate for board rotation... Wouldn't surprise me.

You will be surprised. To fully correct all 2D linear space distortion, that includes X, Y scaling, rotation, and skew, takes multiple PhD to get correct. I am talking from a semiconductor equipment vendor. Certain big name semiconductor vendor cannot correct 2D distortion in their multi-tens of millions worth reticel inspection machine.

Lens distortion correction at the level of pick and place machine is a solved problem. nVidia has a good one in their VPI. For semiconductor? Yeah I can see why attempting to correct nanometer scale aberrations is hard, but that's not the precision a desktop PnP machine needs.

[eflyersteve]

Yeah, I'm going to have to say to anyone thinking about purchasing one of these to make a hard pass.  I've had mine for months but due to a surgery, I've not really had a chance to use it much other than a few boards.  Having started to make more than a couple boards at a time, it seems like the unit is really skipping steps and had no way of knowing where it is in space until it is power cycled.  Also the cover tape mechanism is horrendous.  As mentioned - if the unit sits for a few days, things stick.  I'm probably going to swap out all 8mm cover tape 'rubbers' for the weaker ones as I can't see a downside to this.

I've been in contact with the manufacturer and it seems they don't understand the issue.  I've explained how I can program locations and cycle back to those locations and they will now be off.  I also still fail to understand how one fiducial alignment can compensate for board edge routing errors.  Each board I do come out with the parts either slightly off or horrendously off.  Note that I have decades of experience with Mydata pick and place machines and like to think that I have a basic understanding of how these machines should work. 

Sitting beside this unit I have a TVM802B that just works.  Doesn't pull too hard on the cover tape, doesn't use steppers for positioning and is much more consistent.  The programming interface is somewhat clunky but not any worse than the YY1.  It just isn't a great looking machine and nothing I would want to show off to a customer.  I do have two Essemtec CSM7000 machines but they don't handle 0402 parts at all and 0603 are difficult.  And the user interface is actually wose than the Chinese models as hard as that may be to believe. 

I've been catching up on this thread and hope that perhaps a new board with updated firmware will help some but I still am not sure how we get around one fiducial for alignment.  This seems like something they could fix in firmware but again, the manufacturer doesn't seem to think it an issue.

If anyone has thoughts or suggestions on how to build any faith in this machine I am all ears.  I'll keep going back through this thread though but wanted to come here to vent. 

[asmi]

So, after reading the thread, it seems that there are two types of people who bought this machine.

1. Those who intend to use it for medium-to-high volume production runs. These are the types who thought they are getting a 30K$ machine for $3K, and are pissed off that it didn't quite turned out to be the case.
2. Those who bought it for a prototype/low volume production runs. These are generally happy with machine, despite it having some shortcomings and some odd design decisions.

[nimish]

So, after reading the thread, it seems that there are two types of people who bought this machine.

1. Those who intend to use it for medium-to-high volume production runs. These are the types who thought they are getting a 30K$ machine for $3K, and are pissed off that it didn't quite turned out to be the case.
2. Those who bought it for a prototype/low volume production runs. These are generally happy with machine, despite it having some shortcomings and some odd design decisions.

Solid, these are very different use cases. I need/want a prototyping machine so having quirks is acceptable. For 3k though, you would hope they had some QC.

[EEVblog]

So, after reading the thread, it seems that there are two types of people who bought this machine.

1. Those who intend to use it for medium-to-high volume production runs. These are the types who thought they are getting a 30K$ machine for $3K, and are pissed off that it didn't quite turned out to be the case.
2. Those who bought it for a prototype/low volume production runs. These are generally happy with machine, despite it having some shortcomings and some odd design decisions.

Solid, these are very different use cases. I need/want a prototyping machine so having quirks is acceptable. For 3k though, you would hope they had some QC.

You would also hope that things would generally evolve as they release new models like this. But basic stuff like not being to peel the tape, or having two fiducial support seems like basic stuff they shuld have nailed years ago. NeoDen isn't exactly a new kid on the block.
I think everyone would be happy with slow and not very flexible for the price, but you kinda still expect decent reliability.

[tboicey]

You would also hope that things would generally evolve as they release new models like this. But basic stuff like not being to peel the tape, or having two fiducial support seems like basic stuff they shuld have nailed years ago. NeoDen isn't exactly a new kid on the block.
I think everyone would be happy with slow and not very flexible for the price, but you kinda still expect decent reliability.

I'm following this thread and occasionally chipping in.

I have a Neoden 3V advanced that has been pretty good over the last few years. Support from Neoden has also been good. They have sent parts freely and also sold me the parts to add a 24mm feeder lane to the back rack. (for placing 20 pin SOICs)

It seems like the YY1 has more modern software and a few additional features, but has some cost cutting measures. Overall it is cheaper by a fair bit, especially for that many feeders, so it still might be a good choice. The 3V in comparison seems older, and a bit boring, but very solid.

Bulk parts picking would be a nice upgrade, if it works well. I throw out a lot of parts like everybody else. Passives aren't worth the time, but some SOT23 chips would be nice to reuse once they fall out of the tape, or if I need so few that loading a feeder seems tedious. With the 3V I cheat a bit though by putting loose chips back in the tape and pulling it back a few slots. This works well enough, but I only seem motivated when the parts start to cost 50 cents or more each.

Nozzle changing would be nice, if it works well. Since I usually require one small and one large nozzle, my machine very rarely uses both nozzles at once. So it will go get a part with the small or big nozzle, take a picture, place it, then do another one. Rather than getting a part with one nozzle, another with the second nozzle, taking both pictures, then placing both components, which overall is faster.

With the nozzle changer, I suspect you can place a lot of passives with two smaller nozzles, then switch to a large one for the few chips at the end one at a time?

The 3V peelers seem to work great. Each peeler is an individual motor. The Neoden web site claims the new ones are an improvement over the 3V but maybe not so sure...

Two fiducials seems like a no-brainer, although I usually butt the top edge of the board up against the immovable rail on the bed. So the amount of rotation/skew that could sneak into the process might be less than what matters when placing 0402 or larger. My 3V can use multiple fiducials but if the board is always squared well the others might add little actual value.

[mikeselectricstuff]

I think the main frustration with the cheap machines is that a lot of the pain is not due to the low build cost but to poor design and poor software, which could have been done much better without increasing the cost.

[thinkfat]

Good software isn't cheap. If it was, there would be more of it. Good software developers don't work on idiot pay.

[coppice]

Good software isn't cheap.

Well, is certainly isn't when it can only be amortised over a small number of units.

[EEVblog]

I think the main frustration with the cheap machines is that a lot of the pain is not due to the low build cost but to poor design and poor software, which could have been done much better without increasing the cost.

Yes, I would have expected refinements of new products like this from a company that's been doing PnP machine for quite a long time now.
Even if the product isn't as fast or feature equiped (extra hardware costs money), you at least expect refinement in terms of reliability and usability.

When I hear reports of basic stuff like tapes jamming etc that's just an automatic "nope, I don't want to have to dick around with that".

[dkonigs]

You would also hope that things would generally evolve as they release new models like this. But basic stuff like not being to peel the tape, or having two fiducial support seems like basic stuff they shuld have nailed years ago. NeoDen isn't exactly a new kid on the block.
I think everyone would be happy with slow and not very flexible for the price, but you kinda still expect decent reliability.

Yes, for something like this, slow is perfectly fine.  But if you spend so much time having to muck around with the machine, work around its shortcomings, and wasting components due to it constantly screwing up...  Suddenly its not worth bothering with anymore.

Striking a balance between "I'll just assemble it by hand" and "I'll just outsource it to someone with a real machine" kinda demands a level of fuss that's below what this machine seems to have.