Manual/Semi-Sutomated SMD line

[andrea.longobardi85]

Hi,

I have started a topic a while ago about buying an  automated SMD line for my business:
https://www.eevblog.com/forum/manufacture/smd-assembly-line-within-30k-budget/

Just to recap we do HW design for external clients and from time to time we build for them prototypes (5-10 units) and small batches (100-500 units).
Unfortunately I have started to realize that with our budget (30k) we either go for Chinese brand SMD line or second hand one but still not top brand or quite old.
In both case I have the feeling I will spend a lot of time babysitting the machines which is something we cannot afford especially for building 5-10 prototypes...

That said I had another plan in my mind and would love to get people feedback on it:
Considering the relatively low volumes and not need for high throughput I was thinking to setup a manual/semi-automated assembly line composed by:

1. manual or semi-automatic paste printer like:
https://www.fritsch-smt.de/en/printing/translate-to-englisch-printall005l
or
https://www.orion-industry.com/solder-paste-printing/sr2700-smd-semi-auto-screen-printer.html

2. manual P&P:
https://www.fritsch-smt.de/en/manual-pick-place/lm901

3. reflow bench oven:
https://www.fritsch-smt.de/en/reflow/batch

This should be something in the 15k range of investment (20k with semi automated paste printer).
And in case things will scale up in the future I can update to a good automated P&P later and get an inline reflow oven.
Also I think the bench oven will be handy anyway as probably I wouldn't turn on the inline reflow oven to bake 5 units...
I wonder if anyone has similar setup...
Is it a bad idea???

[mikeselectricstuff]

I don't think manual P&P systems like this represent good value, compared to a simple manual setup - foot-operated vacuum pen, good lighting /magnification  and some tape-holding fixtures.
I don't know the cost of that Fritsch system,  but I'd imagine you could get a Neoden YY1 for less - however fiddly and annoying the Neoden is, it'll be quicker than any sort of manual/semi-auto  placing. Those semi-manual things are just a dead end.

For stencilling, a simple jig using pins for alignment works fine -I use the Eurocircuits stencil printer, but this is significant overkill - vastly overengineered.
https://www.eurocircuits.com/ec-equipment-ec-stencil-mate/

Even for the 100-500, the cost of a semi-automatic printer is unlikely to be justified - with a good manual setup, printing not a big part of the time so not something to be spending a large part of the budget on.

Even the best setup will have some setup time, so for the 5-10 units requirement, the difference between the cheapest and most expensive system will be disproportional to the time saved. There is no point spending 30K for this.

TBH I'd be inclined to spend 5K now on a cheap Chinese setup and use that as a learning exercise to better understand what you really need.
There is no magic solution to SM assembly and everyone's requirements are different.  Whatever you spend, there will be significant time learning and figuring out the process that works for you.

[andrea.longobardi85]

I don't think manual P&P systems like this represent good value, compared to a simple manual setup - foot-operated vacuum pen, good lighting /magnification  and some tape-holding fixtures.
I don't know the cost of that Fritsch system,  but I'd imagine you could get a Neoden YY1 for less - however fiddly and annoying the Neoden is, it'll be quicker than any sort of manual/semi-auto  placing. Those semi-manual things are just a dead end.

it is around 10k with camera, xyz axis blocking option and other bits...

For stencilling, a simple jig using pins for alignment works fine -I use the Eurocircuits stencil printer, but this is significant overkill - vastly overengineered.
https://www.eurocircuits.com/ec-equipment-ec-stencil-mate/

I have seen it!! doesì it work only with eurocircuit ec-stencil format or with every generic stencil?

Cheers!

[SMTech]

We have a manual p&p, 20+ years ago we built batches of 30 boards at a time on it, complex but fairly low number of BOM lines. Those who diss these fail to appreciate the level of assistance a stable frame and vacuum pickup offers. They are a country mile ahead of trying to use tweezers or a handheld pickup pen. However even if you add all the bells and whistles that the Fritsch offers building something genuinely complex using any kind of manual process is pretty unrealistic. Mike is also quite correct, they cost far too much money and prepping to build is going to take just as long as for a real machine.

We still use our manual p&p, very short strips, loose parts or new component shapes that have not been trained or tested on the machine (and we might only see once) get fitted there instead. We might sometimes also build prototype runs of 1-5 boards as we make a lot of things that are incredibly simple and this will also depend on timing and complexity - is the automatic machine busy, conveniently loaded with most of the right parts etc. I would suggest if you found a manual p&p on eBay for cheap, totally worth with to use the way we do. However you should choose an automatic machine to do as much of the work as possible, an older machine would still be a valid choice for those bigger batches you mention, Neodens less so.

You can argue all day about stencils and printing, it is the most critical process in SMT and the root cause of most failures. Doing it properly requires at the very least a stable fixture and a fully tensioned stencil. Those cheap framed stencils from JLC/PCBWay are a good start, we have a semi automatic printer (Essemtec SP003 aka PBT GO23), we gave up on it as the mechanical side of getting a perfect print was complex and possibly damaged. I would suggest getting some hands on time with the TWS unit, see how easy it is to setup, how it copes with different thicknesses of PCB and how much the cameras help.
You might also need to look at how it holds stencils, does it adjust for different frames, does it use a proprietary system and can you get them made by your preferred supplier etc. 
 Keep in mind 10-15K would buy a 2nd user big brand automatic printer that can do so much more (it can also buy you a Chinese one direct).

I'd steer away from the little box ovens, one of the TWS conveyor ovens would be cheap and effective, or you can go Chinese there are lots of them from the IN6 all the way up to huge.

[mikeselectricstuff]

I don't think manual P&P systems like this represent good value, compared to a simple manual setup - foot-operated vacuum pen, good lighting /magnification  and some tape-holding fixtures.
I don't know the cost of that Fritsch system,  but I'd imagine you could get a Neoden YY1 for less - however fiddly and annoying the Neoden is, it'll be quicker than any sort of manual/semi-auto  placing. Those semi-manual things are just a dead end.

it is around 10k with camera, xyz axis blocking option and other bits...

10K just gets you a minor speed-up of manual placement - still needs someone doing the placing 100% of the time
Totally not worth it - 10K will buy a much better solution, New China or used  big-name

For stencilling, a simple jig using pins for alignment works fine -I use the Eurocircuits stencil printer, but this is significant overkill - vastly overengineered.
https://www.eurocircuits.com/ec-equipment-ec-stencil-mate/

I have seen it!! doesì it work only with eurocircuit ec-stencil format or with every generic stencil?

Cheers!

All you need is an unframed stencil with 2 holes in the stencil and 2 matching ones on the PCB, 30mm spacing stencil to PCB hole - I have a component defined with the corerct sized/spaced holes in the appropriate layers, so just need to place 2 of these at the edge of the PCB
https://www.eurocircuits.com/ec-registration-system/

This works surprisingly well once you get the hole sizes dead on - 0.5mm QFPs on a 250mm PCB are no problem at all. The EC unit does a pretty good job of tensioning

Another benefit of a jig using unframed stencils is they take a lot less storage space!

[mikeselectricstuff]

You can argue all day about stencils and printing, it is the most critical process in SMT and the root cause of most failures.

100% agree, but it's not hard to do well manually with the right fixturing and a bit of practice. IMO auto printing, especially for smaller qtys should be fairly low down the priority list if budget is limited.

A while ago I visited a medium-sized assembly place who did small to medium  runs of some fairly complex boards ( BGAs etc.) who did all their printing manually, they rotated staff and  the same ppl who did the printing did inspection & rework so everyone got good at it.

[Styno]

We have a manual p&p, 20+ years ago we built batches of 30 boards at a time on it, complex but fairly low number of BOM lines. Those who diss these fail to appreciate the level of assistance a stable frame and vacuum pickup offers. They are a country mile ahead of trying to use tweezers or a handheld pickup pen. However even if you add all the bells and whistles that the Fritsch offers building something genuinely complex using any kind of manual process is pretty unrealistic. Mike is also quite correct, they cost far too much money and prepping to build is going to take just as long as for a real machine.

Thank you. Yes, this is my experience building small series up to 100 pcs since ~2001 with an Essemtec manual placer , the stability is a great help and much quicker than setting up feeders for a automatic p&p for short runs.

If interested in a manual p&p, a local trade company in the Netherlands has a secondhand for sale, a Dima Fineplacer FP-600.

[mikeselectricstuff]

Seems the biggest advantage of a manual P&P is stability - I wonder what the scope would be to adapting some kind of existing pantograph/tapping arm type mechanics to get the stability, then a simple finger-operated up/down/rotatable vacuum pickup. Strap a camera on if you need magnification.
 

[SMTech]

Seems the biggest advantage of a manual P&P is stability - I wonder what the scope would be to adapting some kind of existing pantograph/tapping arm type mechanics to get the stability, then a simple finger-operated up/down/rotatable vacuum pickup. Strap a camera on if you need magnification.
 

Given how easy it is to buy precut extrusions, fixings, slides and bearings. A DIY clone of the core features should be easy. There are some cheaper simpler options than the Fritsch if you take the time to Google, all have their limitations. Visualplacer has some interesting concepts with laser pointers and a screen too, it is unclear if anyone has actually made it. On that note Visualplacer or SMT Assistant (forked and plugged on here recently) are useful tools for building things by hand when you can find where D15 is... But then so are the HTML interactive design plugins for Altium and Kicad.

[mikeselectricstuff]

Seems the biggest advantage of a manual P&P is stability - I wonder what the scope would be to adapting some kind of existing pantograph/tapping arm type mechanics to get the stability, then a simple finger-operated up/down/rotatable vacuum pickup. Strap a camera on if you need magnification.
 

Given how easy it is to buy precut extrusions, fixings, slides and bearings. A DIY clone of the core features should be easy. There are some cheaper simpler options than the Fritsch if you take the time to Google, all have their limitations. Visualplacer has some interesting concepts with laser pointers and a screen too, it is unclear if anyone has actually made it. On that note Visualplacer or SMT Assistant (forked and plugged on here recently) are useful tools for building things by hand when you can find where D15 is... But then so are the HTML interactive design plugins for Altium and Kicad.

yes - I was thinking maybe standard ali. extrusion with 3D printed joints/bearing holders etc. Something else for the never-ending list of things to do when I get time...

[SMTech]

You can argue all day about stencils and printing, it is the most critical process in SMT and the root cause of most failures.

100% agree, but it's not hard to do well manually with the right fixturing and a bit of practice. IMO auto printing, especially for smaller qtys should be fairly low down the priority list if budget is limited.

A while ago I visited a medium-sized assembly place who did small to medium  runs of some fairly complex boards ( BGAs etc.) who did all their printing manually, they rotated staff and  the same ppl who did the printing did inspection & rework so everyone got good at it.

I build boards in batches of 5-250 and I still do it on a manual printer, not through choice but because until we set aside the budget for a fully automatic printer the options in between are annoying stopgaps with limitations . A semi-automatic printer is a poorly defined thing, and most of them leave a lot of the tweaking and alignment up to the user and rely on the PCB being in exactly the same place every time, at finer pitches the variability in board fixture means that simply isn't true so you have to interrupt the semi-automatic routine to adjust the alignment. The essemtec semi auto printer has a "go" button. After aligning your first PCB, you fit a board into the fixture, lower the stencil, and press go. The printer then lifts the PCB table up to the stencil and moves the squeegee from front to back or back to front depending on where it is currently parked. Once that stroke is complete it lowers the table, you lift up the stencil and remove your board.. and repeat. That might well have worked just fine once upon a time but for fine pitch I think it needs to raise the PCB to the stencil, stop wait for a human to confirm the alignment still looks good and then do the stroke. It is possible some of the semi auto printers with a screen do work like this such as the Reprint Miniko, the TWS shown. I'm 90% sure the PBT FA23 does work like this because those that used it as an Essemtec SP150 really like that machine and I have seen a demo of the process over whatsapp for its Essemtec predecessor (confusing called the SP-200)

0.5mm and below is the point it gets challenging by eye and the more fine pitch you have and the bigger the board gets, the more diffcult it becomes.

[Jeroen3]

We have stopped doing prototypes ourselves and send them to DeltaProto. They also sell their machine they use to do assisted pnp.
But the service is very convenient, upload files, you pick parts (just like with circuithub) or send parts and you get a clean vapor phase soldered board back.

I did do some boards myself with the ec-stencil-fix during covid and a homemade assembly jig and VisualPlace.
But it's very time consuming. Ordering parts, assembly, solder, reworking... For 15k you can have a lot of boards assembled by a contractor!
Kind of wishes I had a pixel pump when I was doing that.

[jmelson]

I have a Quad QSA30A, made by Samsung, but using Quad's electronic feeders and alignment cameras.  It took me too long to learn all the ins and outs of the software, the manuals were truly AWFUL!  But, the machine is built like a tank and is very accurate.  I routinely place 0.5mm pitch FPGA parts with perfect accuracy.  A P&P does need some babysitting, thinking you can just press a button and walk away is a fantasy.  I do manual paste printing anyway, so I have to be circulating right there, printing, load into P&P, take placed board to oven, then repeat.  On the other hand, a decent production machine is SO FAST that it is hard to keep up with it.  Many of my boards are completed in just a minute or two, so you shouldn't think of hours babysitting the P&P process.
Before the Quad, I had a Philps CSM84, made by Yamaha, and the feeders were too weak to pull the cover tape.  This caused a HUGE number of feeder jams, but the Quad feeders work a WHOLE lot better.
Jon

[MR]

You have to categorize babysitting.

There are a few reasons:
- Pickup issue due to feeder issues
- constant vacuum test issues (eg. nozzle lost due to some epic event, or nozzle changer issues)
- constant optical recognition issues.
- Gantry crash / overdriving endstops (linuxcnc will also stop if this happens).
- etc.

not all issues require babysitting during the pick place process, feeder / pickup issues could result in ignoring feeders and the operator could fix it at the end of the procedure.

[andrea.longobardi85]

You can argue all day about stencils and printing, it is the most critical process in SMT and the root cause of most failures. Doing it properly requires at the very least a stable fixture and a fully tensioned stencil. Those cheap framed stencils from JLC/PCBWay are a good start, we have a semi automatic printer (Essemtec SP003 aka PBT GO23), we gave up on it as the mechanical side of getting a perfect print was complex and possibly damaged. I would suggest getting some hands on time with the TWS unit, see how easy it is to setup, how it copes with different thicknesses of PCB and how much the cameras help.
You might also need to look at how it holds stencils, does it adjust for different frames, does it use a proprietary system and can you get them made by your preferred supplier etc. 
 Keep in mind 10-15K would buy a 2nd user big brand automatic printer that can do so much more (it can also buy you a Chinese one direct).

Indeed I have got a quote for a used DEK 248 around 6k, so lower than the TWS semi automatic printer...

I'd steer away from the little box ovens, one of the TWS conveyor ovens would be cheap and effective, or you can go Chinese there are lots of them from the IN6 all the way up to huge.

Is there a particular reason for it? TWS itself told me that for small proto batch (5 units) makes more sense to use bench oven instead of conveyor ovens...considering warm up time, energy cost etc....

[andrea.longobardi85]

10K just gets you a minor speed-up of manual placement - still needs someone doing the placing 100% of the time
Totally not worth it - 10K will buy a much better solution, New China or used  big-name

Considering the quotes I have got so far, the only P&P in the 10k range is the Neoden 4c (without feeders), and for what I have read it is a nightmare to swap feeders and to use it reliability.
So yes a manual machine needs someone there 100% of the time, but it may take 1/2 days to assemble 5 units...I doubt that setting up the Neoden for a new build will take less, am i wrong?

[mikeselectricstuff]

10K just gets you a minor speed-up of manual placement - still needs someone doing the placing 100% of the time
Totally not worth it - 10K will buy a much better solution, New China or used  big-name

Considering the quotes I have got so far, the only P&P in the 10k range is the Neoden 4c (without feeders), and for what I have read it is a nightmare to swap feeders and to use it reliability.
So yes a manual machine needs someone there 100% of the time, but it may take 1/2 days to assemble 5 units...I doubt that setting up the Neoden for a new build will take less, am i wrong?

That depends a lot on how many BOM lines you have, but even the YY1 shouldn't take "days" to set up, even if you're using all feeders. Unless the setup process is so broken that you can't import your pick/place data and have your positions, rotations etc.basically done with minimal effort.

With any machine, it's worth spending some time on your data transfer process - e.g. making sure all parts in your PCB library have consistent rotations relative to the feeder tape orientation, if possible include fiducial positions (have fids as components). That way you have much less time to spend checking placement orientations etc. 
Also things like having different component definitions for R's.Cs, LEDs etc. in the same package size, as the vision characteristics and heights are often  different, so having them as sperate part definitions in the PCB and P&P machine's part libraries can save a lot of time 

Bear in mind that even a poor-but-still-useful machine is likely to place most of the parts OK by itself, so any manual intervention is going to be for a minority of parts.

[mikeselectricstuff]

not all issues require babysitting during the pick place process, feeder / pickup issues could result in ignoring feeders and the operator could fix it at the end of the procedure.

That assumes that the software can be told to ignore errors, carry on with all parts it can place, and let the operator deal with all issues at the end of the job.
'm told  a lot of the Chinese machines can't do this, and just stop as soon as they hit a problem, making them need more babbysitting than would be the case if the software had been done properly.

[jmelson]

So yes a manual machine needs someone there 100% of the time, but it may take 1/2 days to assemble 5 units...I doubt that setting up the Neoden for a new build will take less, am i wrong?

I can't answer for Neoden, but my old Philips machine could be set up for a simple board with a dozen different parts in an hour.  My newer Quad machine takes maybe two hours since the parts list needs to be synched up with the part library in the machine's vision system.
Then, a simple board can be assembled in about a minute each.
Jon

[BreakingOhmsLaw]

I have an older Fritsch LM901 on my desk, and I absolutely love it. Spare part prices are fair, their service is great. They are highly sought after units, so you can always sell them when you upgrade.

The only drawback is that the dispenser is only time/pressure based, it's not a real volume-controlled dispenser. So when you change paste (or just got it out of the fridge), you need to fiddle with the time settings. You can easily change nozzles to change from 0603 pads to, let's say, the tab of a TO252.
I've built my own camera for it, but ended up using a pair of magnifying glasses made by my optician instead. You just can't beat 3-dimensional sight.

The flipping station seems a good investment if you use it a lot, it's around €600.
https://www.fritsch-smt.de/en/manual-pick-place/features

Keep in mind that you need an air compressor air to run it! (the dispenser option, not the P&P) I use a small silent compressor that sits under my desk:
https://weldinger.de/WELDINGER-Fluesterkompressor-FK-65-pro

[jayx]

This flipping station is very good idea! Anyway as SMTech said, in case of manual P&P, finding particular component position on the PCB is usually quite time consuming.
I think for €15-20k you should be able to find older models 2nd hand Essemtec or Autotronik with ~50 feeders, and perhaps just hand place some components which are too low qty per PCB to bother loading in the machine.
Also I'm fairly sure for €3-5k you can get fully automatic paste printer. I've seen MPM2000 last year sold for €3k.
Also as BreakingOhmsLaw said, don't forget about a compressor. I'd add you may need compressed air dryer too, or at least add cooler between compressor and the tank.

making sure all parts in your PCB library have consistent rotations relative to the feeder tape orientation

Parts in the library should have consistent rotation, but surely not relative to the feeder tape orientation. https://www.eevblog.com/forum/manufacture/pcb-component-zero-orientation/msg5035114/#msg5035114

[mikeselectricstuff]

making sure all parts in your PCB library have consistent rotations relative to the feeder tape orientation

Parts in the library should have consistent rotation, but surely not relative to the feeder tape orientation. https://www.eevblog.com/forum/manufacture/pcb-component-zero-orientation/msg5035114/#msg5035114

Why not?
Why would you want orientation be consistent relative to anything other than taping?
OK different machines may have different conventions, but as long as orientation is consistent relative to taping, that is easily handled by a global constant rotation in the data transfer process. 
The machine itself should deal with orientation changes due to feeder locations, i.e. a part with a specified rotation would always place the same regardless of which side it was fed from, as the machine knows where the feeders are.
The only fly in this ointment is parts that can come in either tape or tubes, or tape & tray.
One way to deal with that is to have seperate part definitions for taped or tubed versions, though you don't necessarily  know which will be used at design time. But that's only going to be a small proportion of parts so not a big issue.

If your process is set up such that you know that all taped parts will be placed correctly with no per-job tweaking, that saves a lot of setup time.

[mikeselectricstuff]

The flipping station seems a good investment if you use it a lot, it's around €600.

Or just put parts on a slightly compliant surface and give it a tap.

[48X24X48X]

Ever since running a P&P machine. I reorient the footprint in own KiCad library following the component orientation in the tape. Quite often that not, the ones on the tray are the same as the tapes in terms of rotation angle. Tubes are used less, so just some manual rotation during setup on the software if required. Whether feeder is located at the front or back of the machine,  that is taken care by the machine software automatically. But most important, is how the center of a component is defined which is based on the visual alignment process. This prompted me to redraw certain irregular parts like connector that are not asymmetrical in terms of it's metal pins by shifting the component centre as seen by the pick and place machine visual alignment process.

[SMTech]

Ever since running a P&P machine. I reorient the footprint in own KiCad library following the component orientation in the tape. Quite often that not, the ones on the tray are the same as the tapes in terms of rotation angle. Tubes are used less, so just some manual rotation during setup on the software if required. Whether feeder is located at the front or back of the machine,  that is taken care by the machine software automatically. But most important, is how the center of a component is defined which is based on the visual alignment process. This prompted me to redraw certain irregular parts like connector that are not asymmetrical in terms of it's metal pins by shifting the component centre as seen by the pick and place machine visual alignment process.

Here's a snippet of the Essemtec process, other modern system will be similar. When importing data you build up a set of rules (on a per client basis in my case) when it sees a specific package it applies a rotation to that part so it matches the internal rotation definition. A brand new design from a new client with lots of new part numbers and shapes could take a while to import and check, an established one or one full of boring parts you have seen before, takes no time at all. Similar rules can replace odd CAD defined package names with ones that match the ones on the machine, combine package and value to match an internal part number.

A defined part number matches a part number against a package definition and compatible carrier (tape/tube/tray), you get to define the orientation in each compatible carrier and a part number can have multiple assigned carriers, handy since many SOIC can easily come in either tube or tape and 2512 resistors cannot decide between 8mm and 4mm pitch for example. The rotation here has NOTHING to do with the placement rotation, it simply tells the machine the orientation it will find the device in, when picked. Both placement and pick rotation are defined relative to the machines defined 0 degrees.

The package definition  contains the physical size of the device, lead type, count, tolerance, package markings (for display only), placement settings including offsets to allow for the mismatch between the leads center & cad center and of course vision settings