Pick and Place with vision.

[electrodacus]

Has anyone seen this or have one. I seen first one today on aliexpress for $5800 and then found this youtube video.
I will realy love one for my Solar BMS project there are many small parts and the worse one is probably a 48pin 0.4mm QFN
Seems realy similar to TM240 and includes dual camera. Based on this video seems excellent.



Edit:
I went to the website mentioned in the above video and it seems is ?28800 = $4635 and that is fantastic price.
Here are some photo from that taobao website

[pickle9000]

Have you seen this one? Vision is shown at the end.

?? ? SY-R46404VZ T15-F30-V4 SMT BGA ??

[rx8pilot]

Looks like a logical and MUCH needed upgrade to that type of machine.

[pickle9000]

This is an ad, I can't tell if these are manufactured by the same company or not.

http://translate.google.ca/translate?hl=en&sl=zh-CN&u=http://item.taobao.com/item.htm%3Fid%3D39350261877&prev=search

[electrodacus]

Still no news about this ? or a similar priced model with vision.
I can still wait a few months before I realy need one. I hope something will come up.
If not I will probably get one of the least expensive ones without vision and place the QFN and other smaller parts by hand.

[mikeselectricstuff]

On the 4-headed one, feeder layout looks rather poor - looks like feeders only on one side so numbers very limited.
Doesn't look like either has auto tool change.
Not obvious if either had a fiucial cam

Not sure I see much value in a small QVGA touchscreen display built in - too fiddly to do anything useful - much better to do it from a PC - maybe a web interface?


Surprised that nobody seems to have thought of using the system used on my 20 year-old versatronics- flying camera, flipping a mirror under the nozzle on it's way from feeder to board - significant placement speed increase with no extra mechanical speed needed.

[electrodacus]

On the 4-headed one, feeder layout looks rather poor - looks like feeders only on one side so numbers very limited.
Doesn't look like either has auto tool change.
Not obvious if either had a fiucial cam

Not sure I see much value in a small QVGA touchscreen display built in - too fiddly to do anything useful - much better to do it from a PC - maybe a web interface?


Surprised that nobody seems to have thought of using the system used on my 20 year-old versatronics- flying camera, flipping a mirror under the nozzle on it's way from feeder to board - significant placement speed increase with no extra mechanical speed needed.

Thanks Mike,

Not interested in the 4 head version well over my budget.  The small version has to heads and you can have a head configured for smaller components and one for larger ones.
The screen is not so important and I'm sure it does not add to cost much. But the fact that it can work independent from an external PC is a big plus for me. I have limited space and energy available.
It has dual cameras one on the bottom for checking component rotation and position and the other one mounted on the side of the head is probably used just for fiducial.
The mirror thing seem like an ingenious way to save time but not sure how it will work. If you flip a mirror under the nozzle the camera will be to far and will see things at an angle also flipping the mirror may not be much faster since this machine is extremely fast and probably the most time is spend for image recognition.
The nozzle is quite close to PCB so the mirror will need to be flipped out somewhere outside the PCB area not sure how exactly that versatronics- flying camera was designed.
Anyway even with bottom camera they can populate a few thousand parts per hour and that is more than sufficient I can not bake PCB's that fast anyway.
I can do a few hundred parts per hour manually so is 10x faster not to mention it will not get bored or tiered

Dacian.   

[jc101]

I tried all sorts of searching with regard to this, couldn't find a huge amount and what I did find was in Chinese which didn't translate.

One place I stumbled on mentioned they had some in the UK, so could be bought without suffering the import duties as it was shipped within the UK.  I thought I had bookmarked the page but didn't, may have another look at some point.  If I find it I might see if I could go and take a look at one...

[mikeselectricstuff]

I tried all sorts of searching with regard to this, couldn't find a huge amount and what I did find was in Chinese which didn't translate.

One place I stumbled on mentioned they had some in the UK, so could be bought without suffering the import duties as it was shipped within the UK.  I thought I had bookmarked the page but didn't, may have another look at some point.  If I find it I might see if I could go and take a look at one...

There appear to be some operations that stock containerloads of popular Chinese ebay kit in the UK for direct shipping - I doubt they are the sort of place you could go look, or have the first clue about what they are selling.

What isn't clear on the 4-head one is what, if any, feeders are included. 

[electrodacus]

I tried all sorts of searching with regard to this, couldn't find a huge amount and what I did find was in Chinese which didn't translate.

One place I stumbled on mentioned they had some in the UK, so could be bought without suffering the import duties as it was shipped within the UK.  I thought I had bookmarked the page but didn't, may have another look at some point.  If I find it I might see if I could go and take a look at one...

Look this seems to be their Aliexpress store sor of eBay in China but page is in English. I used Aliexpress quite a few times.
http://www.aliexpress.com/store/product/2013-New-Tech-PCB-SMT-machine-pick-and-place-machine-Full-Automatic-Chip-Mounter-Upgraded-from/210206_1395653211.html
They sell a lot of other stuff see
http://www.aliexpress.com/store/all-wholesale-products/210206.html
They also sell the CHMT36 without vision that sells for quite some time and now this CHMT48V
It seems they sold two if this new ones and got one positive feedback.

Is about out of my price range and probably needs some shipping and import tax. I still have some months before I will need one maybe there will be something better for me. But from what I can see it looks to be realy nice.

[jc101]

Look this seems to be their Aliexpress store sor of eBay in China but page is in English. I used Aliexpress quite a few times.
http://www.aliexpress.com/store/product/2013-New-Tech-PCB-SMT-machine-pick-and-place-machine-Full-Automatic-Chip-Mounter-Upgraded-from/210206_1395653211.html
They sell a lot of other stuff see
http://www.aliexpress.com/store/all-wholesale-products/210206.html
They also sell the CHMT36 without vision that sells for quite some time and now this CHMT48V
It seems they sold two if this new ones and got one positive feedback.

Is about out of my price range and probably needs some shipping and import tax. I still have some months before I will need one maybe there will be something better for me. But from what I can see it looks to be realy nice.

Thanks for this, I hadn't found that page. 
One of the previous buyers shows as the UK too, which might be of use.

I will do some more reading and might see what other information they have available in english.
Just trying to keep abreast of what is out there, might get to a point later in the year where I need something and that is about the max in terms of physical size I could find a home for.

[alexanderbrevig]

wow!

[Kjelt]

I find it too expensive as a hobby tool and it looks too crappy for a pro tool.
It all depends on the quality of the software and the longevity of the hardware, both should prove themselves.
If you are really making money with this and need to do >50 pcbs a week and depend on it  than a second hand pro machine as Mike has or now rx8pilot probably is a better investment.

[rx8pilot]

I find it too expensive as a hobby tool and it looks too crappy for a pro tool.
It all depends on the quality of the software and the longevity of the hardware, both should prove themselves.
If you are really making money with this and need to do >50 pcbs a week and depend on it  than a second hand pro machine as Mike has or now rx8pilot probably is a better investment.

Agreed. The middle ground is a strange place for Pick and Place that has not been addressed. $5k+ is a lot to put into a hobby for an assembly machine that also requires a slew of other pieces and parts to make it useful. A super low-end random manufacturer is scary for those that are banking a living on the machine to perform when it needs to.

I would be more than will to sacrifice speed for reliability to keep the price in check for low-volume. 1 head, up/down cameras, nozzle changer, flexible feeder system that can support big parts, etc. For cost savings, I like the head indexed feeder concept. The automated feeders on commercial machines are very expensive to get the extra performance. Head indexing is slow, but who cares when you are only doing 150 PCB's per week.

If I was going into the $5-7k PnP business, I would focus on the needs of small businesses that will spend that in a flash if it is not a toy. For me and many other businesses - speed is not the goal. Ease of use, ease of project changeover, and reliability are what is important to me. Not sure what I could accomplish in that price range, but then again I will not likely enter the PnP market ever.

[electrodacus]

I find it too expensive as a hobby tool and it looks too crappy for a pro tool.
It all depends on the quality of the software and the longevity of the hardware, both should prove themselves.
If you are really making money with this and need to do >50 pcbs a week and depend on it  than a second hand pro machine as Mike has or now rx8pilot probably is a better investment.

I have limited space in my home also limited energy (OffGrid) I find that old machines are not a good solution for me. They are large shipping may be extremal expressive then it may need expensive parts hard to find and uses antique software tools and external computer to run.
This are plug and play machines all includes much smaller and extremely fast. The only problem is that at the current volume I can almost do thing manual. But If I try to outsource the assembly I may need to pay same or more as I do for this machine just for 100 units of my new solar BMS plus the risk of fake or inadequate parts and improper soldering.
I will probably end up with the TM220A is quite a bit less and I will do the critical parts few QFN manually.
Longevity should be no problem for such low volume and I can probably easy sell that after I'm done with the job for not much loss. I will probably keep it for future projects.
I realy like this machines they are extremely fast and work in-depended from an external computer they just need a csv file for placement.
Pick and place SMT parts is not a job for humans I done that on the fist Solar BMS and that was no joke plus that had only half the parts this new one has.   

[rx8pilot]

Your strategy is reasonable. I nearly went that direction myself. Some luck went my way and I found a used Quad machine that should be able to out class my needs for the foreseeable future. It is, as you point out, heavy, big, and nearly 2x the cost after it's all said and done. The good news is that it is about 10x more capable but it sounds like that capability would be lost in your situation. At least for now. Hopefully your project will get some traction.

I had considered using TM-240a to place the easy parts and hand place the hard stuff under a microscope like I do now. It would still be an improvement from full manual for sure. Making the transition to production is quite tricky as I have found - balancing time and money.

[Kjelt]

I understand the size problem and maintenance costs are a big issue for non pro start ups.
The problem as rx8pilot also states is that there is not much out there for that class. If you want to buy a brand new car you can,t find anything below 10k at least here in europe and there is a good reason for that.
If some company starts to sell cars for 3k brand new I would be very suspicioys.
The same here , the brand name alone sound all alarmbells. But if there as  you say are already machines out there I would contact that buyer and ask for their experiences and uses.

But If I try to outsource the assembly I may need to pay same or more as I do for this machine just for 100 units of my new solar BMS plus the risk of fake or inadequate parts and improper soldering.

That can,t be right can it? $50 per pcb just for the assembly? Rediculous. I would ask around for other businesses then the startup costs are high i know that.  you can ask how you can lower them for instance make sure the files are compliant with their process.

[pickle9000]

Extreme low quantity and one off's can be a killer. A low cost machine (sub 2000.00 usd) even without vision would make a killing.

[rx8pilot]

That can,t be right can it? $50 per pcb just for the assembly? Rediculous. I would ask around for other businesses then the startup costs are high i know that.  you can ask how you can lower them for instance make sure the files are compliant with their process.

In small qty's that is about right. I got a quote from multiple sources for small batch PCB's in that ball park. The documentation, kitting, setup, programming, testing, shipping divided by a small qty is expensive. That is why I now have a PnP machine. My first 100 sets of PCB's will nearly cover the purchase price. The cost drops very fast as volume goes up - dividing the NRE into smaller parts.

[mikeselectricstuff]

Extreme low quantity and one off's can be a killer. A low cost machine (sub 2000.00 usd) even without vision would make a killing.

The problem is that so far,  the cheap machines have been dead-ends.
What's needed is something that has a low entry cost, but can be expanded over time -feeders, vision, multiple heads etc.

Anything without vision is a toy,  as hardware cost is so low there is little excuse these days for not having it. 

[rx8pilot]

Anything without vision is a toy,  as hardware cost is so low there is little excuse these days for not having it.

Even the software components are open source.

[pickle9000]

Extreme low quantity and one off's can be a killer. A low cost machine (sub 2000.00 usd) even without vision would make a killing.

The problem is that so far,  the cheap machines have been dead-ends.
What's needed is something that has a low entry cost, but can be expanded over time -feeders, vision, multiple heads etc.

Anything without vision is a toy,  as hardware cost is so low there is little excuse these days for not having it.

A modular system would be ideal. I'd still say a basic system (even without vision) but with good placement resolution and cheap feeders is a reasonable place to start.

[electrodacus]

Yes low volume assembly is quite expensive especially when you have a relatively complex design with many parts and also include 0.4mm pitch QFN.
Also each Solar BMS has two PCB's the main one is more complex but there are quite a few parts on the power board (is a metal core PCB).
So yes at least 50 to 70$ to have them assembled and then there is also the risk of soldering quality issues or non genuine parts used.

As for available pick and place there is that guy here on the forum with a $1500 unit with vision but extremely slow based on CNC hardware about 300 parts per hour and that is worse than manual so useless. Then there is the $4000 + shipping for the TM220A or $5000 for the larger TM240A Next is the CHMT36 a clone of TM240A for the price of TM220A and this newer CHMT48V that is even larger (not so important for me) and has vision based on youtube video the best implementation of vision I seen cost 5800 + shipping.
All this P&P have 4000cph with vision to 7000cph without both faster than I relay need.
What is missing is something in the middle $3000 decent speed at least 1000 components placements per hour with vision.
Maybe something will pop up in the next few months before I finalise design and start production.
This CHMT48V is great if it was $3000 I will have invested right now.

[smgvbest]

Has anyone looked at this one.
https://youtu.be/CRSLbo_8nTQ?autoplay=1
opensource, DYI Pick and Place with vision and paste dispensing
Has 1 place head and 1 paste dispensor.

Full site here
https://www.vbesmens.de/en/pick-and-place.html

I've thought about building it myself

[rx8pilot]

As for available pick and place there is that guy here on the forum with a $1500 unit with vision but extremely slow based on CNC hardware about 300 parts per hour and that is worse than manual so useless.

There is no one on earth that can consistently place 300cph for any time at all. You can 'maybe' do a burst of the same component repeated in an easy to memorize geometric pattern. If your board has any complexity at all with a significant component count - you will very much benefit from a 300cph machine. People need breaks, get distracted, place a 1k where the 10k goes, etc.

While you are inspecting, printing another PCB, loading software - the machine would be putting down components. It is FAR from useless. This is coming from someone that has placed nearly 100,000 components manually in the past year. I have all the tricks and TONS of practice. I have great eyesight, tools, and coordination. I have also engineered an organization system for the components and reference materials to speed things up.

I am not saying 300cph is fast - I am saying that it is MUCH faster than what a human can do for any useful length of time.

[mikeselectricstuff]

As for available pick and place there is that guy here on the forum with a $1500 unit with vision but extremely slow based on CNC hardware about 300 parts per hour and that is worse than manual so useless.

There is no one on earth that can consistently place 300cph for any time at all. You can 'maybe' do a burst of the same component repeated in an easy to memorize geometric pattern. If your board has any complexity at all with a significant component count - you will very much benefit from a 300cph machine. People need breaks, get distracted, place a 1k where the 10k goes, etc.

While you are inspecting, printing another PCB, loading software - the machine would be putting down components. It is FAR from useless. This is coming from someone that has placed nearly 100,000 components manually in the past year. I have all the tricks and TONS of practice. I have great eyesight, tools, and coordination. I have also engineered an organization system for the components and reference materials to speed things up.

I am not saying 300cph is fast - I am saying that it is MUCH faster than what a human can do for any useful length of time.

300cph is easily within the same range as manual placement with a decent setup - that's 12 whole seconds per part.
By the time you add setup, a 300cph machine would be of use to a very narrow range of applications.

[electrodacus]

There is no one on earth that can consistently place 300cph for any time at all. You can 'maybe' do a burst of the same component repeated in an easy to memorize geometric pattern. If your board has any complexity at all with a significant component count - you will very much benefit from a 300cph machine. People need breaks, get distracted, place a 1k where the 10k goes, etc.

I actually did around that number 250 to 300cph for quite a few boards and my wife was about as fast as I did after we learned the board of course.
You can look at the end of this video when you can see how I manually placed one of those boards.
Is not something I want to repeat so I ordered a P&P and should soon get here. Is the one without vision that can do about 6000cph with to heads. I will probably reduce that to 2000cph or so since I prefer accuracy over speed the speed is good enough in this case.
Here is the link to the video where I place manual parts. Is a long video look at the end I think last 10 minutes or so  

[rx8pilot]

The only meaningful measurement of CPH is how many components you can put down over the course of a run. Say I want to do 10 of my PCB sets with around 300 components total per PCB set (a few different PCB's that go together). 3000 parts total.

The actual placement time is about 1 hour - 300CPH. But if I look at the whole day - I can only get about half that if I am lucky. The problem is that while I am placing parts, I cannot prep the next PCB, manage the re-flow oven, inspect the previous PCB, manage the parts, answer a phone call, go to the restroom, etc. With a 300cph machine, I can be doing all the other tasks needed and the machine would only stop long enough for me to load the next PCB. That would be a true 300CPH line.

I learned the details when I had my machine shop. We were very focused on how fast the machine could move and never paid much attention to how slow all the other processes were. I was about to drop $500k on a super-fast machine when I realized we can triple the output of our current machines by feeding them faster. For only a few $thousand I dramatically sped up the process by being able to keep our slow machines busy 99% of the time. My only concern was how many parts can I get in a day - NOT how fast the cutter can shred the metal.

In the case of P&P, the concept is similar. I would say that a 300cph machine would at least double the output for someone coming from a fully manual process, maybe even triple. You will be placing parts while you eat lunch, talk on the phone, count parts, day dream, whatever.

[electrodacus]

rx8pilot,  There are so much more things to consider. Yes you are right for your example you need two days of hard work to do those boards but the alternative is not better.
You need to set up the machine that can take a full day easy for a new board then you will need a very long day to be able to finalize the boards the day after setup.
That inexpensive P&P has some other problems except for the speed. I like that guy so I will not want to comment more on those problems. I'm curios what Dave1 and Dave2 will think about this one since they just received the kit.
I paid a bit more than double for a machine that is way superior to that one an place parts 10 to 20x faster and more precise even without vision.
Is now clearing customs so I hope it will be here in about a week and I will give you my firs impression about it if anyone is interested. Is the CHMT28 the smaller brother of the CHMT36. Is almost no difference in price and the parts are exactly the same the 28 is just smaller and has less feeders but I have limited space in my lab and do not realy need the larger size.
The CHMT48V the one with vision I listed in the first post will have been realy great but is quite a bit more and I can not afford at this time and the gain is probably minimal for my use case. 
 

[rx8pilot]

I also got a used machine recently - A Quad IV-C. It's been a few months now and I still have not made a single PCB. There have been a number of things that need to be fixed/replaced in addition to the learning curve. I have alos been too busy to spend too much time on it.

There is no magic solution to P&P, it's a major undertaking no matter how you do it. Even for manual assembly, it takes a long time to get all the parts organized and planned out.

[electrodacus]

I also got a used machine recently - A Quad IV-C. It's been a few months now and I still have not made a single PCB. There have been a number of things that need to be fixed/replaced in addition to the learning curve. I have alos been too busy to spend too much time on it.

There is no magic solution to P&P, it's a major undertaking no matter how you do it. Even for manual assembly, it takes a long time to get all the parts organized and planned out.

I did not got a used machine is a new one low cost based on the popular TM220A and TM240A just with some improvements in my opinion and at a slightly better price.

[charlespax]

The Neoden sales representative in Hua Qiang Bei told me they are launching a vision-enabled machine within the next month or two. I haven't seen it myself, but I'm told it's very professional.

[electrodacus]

The Neoden sales representative in Hua Qiang Bei told me they are launching a vision-enabled machine within the next month or two. I haven't seen it myself, but I'm told it's very professional.

Thanks. I guess it will be similar to the one in the first post.  It will also probably be more expensive than their new version that is also not cheap.
It will not fit in to my budget and I also do not have much need for camera even if I need to place by hand the few small pitch IC's.
The worst part that I have is 48pin  0.4mm pitch QFN and that is the main part there is no alternative.

[electrodacus]

For those interested I got a CHMT28 the smallest version of CHMT48V(the one with vision in original post) and CHMT36.
Look at the second half of this short video for a few clips with the machine in the same day I got it (two days ago).


Everything seems to work as expected but it was just a short test to see that it works. I will have no need for it for quite a few weeks at least.

[jt]

I work for a small business that probably averages three or four quick-turn one-off's a month.  We are constantly fighting assembly quality issues from multiple vendors.  We burn >150k$ on PCBA direct cost a year plus an embarrassing amount of engineering labor troubleshooting issues that turn out to be blatant failures to conform to the IPC class-2 standards we call out. 

It is tempting to bring assembly capability in-house.  However, no one at the company has any experience with P&P and we are worried the learning curve, maintenance, set-up labor and the direct cost of the machines being too much of a burden. 

Does anyone have any experience bringing this capability in-house for a small business?  What kind of support staff is typically needed?

[mikeselectricstuff]

I work for a small business that probably averages three or four quick-turn one-off's a month.  We are constantly fighting assembly quality issues from multiple vendors.  We burn >150k$ on PCBA direct cost a year plus an embarrassing amount of engineering labor troubleshooting issues that turn out to be blatant failures to conform to the IPC class-2 standards we call out. 

It is tempting to bring assembly capability in-house.  However, no one at the company has any experience with P&P and we are worried the learning curve, maintenance, set-up labor and the direct cost of the machines being too much of a burden. 

Does anyone have any experience bringing this capability in-house for a small business?  What kind of support staff is typically needed?

For 1-offs it's all about setup. You need to spend some time to get a streamlined process, so, for example, your PCB libraries match the P&P library (feeder orientation., component height etc), so you know that parts will be the right way round without having to check each one.
Another thing with 1-offs is you want a machine that can fed short tapes without wasting lots of parts on a long leader/trailer - Mydata have short-tape feeders, and you also need the ability to pick from lengths of static tape.

As regards the rest of the process, by far the biggest factor in good quality is getting a good stencil print. Use stainless stencils.
The cheapest decent printer I know of is the Eurocircuits one.
Once you have decent paste print and placement, reflow isn't a big deal - a toaster oven works fine for leaded.  If you really want to use lead-free, and for 1-offs there isn't really a reason to, you may need to look at something a bit more well controlled, maybe vapour-phase.

[jt]

Thanks for the insight!

[ProtoVoltaics]

Hey everyone,

I have actually been working on making a pnp with vision. The unit is almost complete, and we will be selling them. If you are interested I would ask you to visit our Hackaday page https://hackaday.io/ProtoVoltaics and check out our pnp. Here is the latest video on our machine taken yesterday. It was taken with a phone camera, because I forgot my actual camera at home 

https://youtu.be/C7Pi4akL5og

Thanks!