Author Topic: Pick and Place Assembly Line - How bad could it possibly be?  (Read 1753 times)

0 Members and 1 Guest are viewing this topic.

Offline matthewpang

  • Contributor
  • Posts: 25
  • Country: 00
Pick and Place Assembly Line - How bad could it possibly be?
« on: September 04, 2021, 09:09:03 pm »

Initial Post - How bad could it possibly be?
Sep 4 2021
---


Hey All,

Hokaaaaay - so after another late night of sorting components into a component tray (see attachment - designed to be shaped like the board) only to accidentally drop my phone on said component tray and sending EVERYTHING flying, hand sorting everything AGAIN, building 10 boards, discovering that there was an error and spending way too much time testing and reworking the damn things to figure out which component was messed up - I finally lost my patience and in the wee hours of the morning mostly rage/impulse bought a SMT Line.

I am now somewhat poorer (on nice refurb equipment from reputable resellers) as follows (in USD):

Purchased Equipment
1) UIC GSM PNP Machine ~18k
2) Feeders, accessories, feeder carts, etc for the PNP machine ~17k
3) Ekra X4 Screen Printer ~9k
4) Essemtec Reflow Oven ~5k
5) Conveyors (normal and inspection) ~3k

Outstanding orders (pending refurb)
1) AOI Machine ~ 10k-15k
2) Xray ~ 15-30k

Infrastructure & Rigging
1) New outlets and transformers with 3 phase AC ~5k
2) HVAC modifications for oven ~3k
3) Rigging (men with trucks, forklifts, etc to move and level each machine (multiple tons) ~5k

Have I ever used a PNP machine? Nope.
Have I ever worked on an SMT line? Nope.
Is this the smartest thing to do? Questionable.

Will we probably get away with it? *Yep - but the question is how much curse-cry-screaming is involved*

Haven't seen anybody do this in the open. So - for all your schadenfreude amusement (and as a cautionary tale should anybody want to follow in our footsteps)

Here goes nothing...

Come along for the ride,
M


Update 1 - "F***"
Sep 6 2021
---

Hello friends,

Some updates.

GSM Pick and Place Machine

The machine was refurbished by an excellent company out of Illinois. When it was ready we were invited over for training and acceptance. Due to the short notice, we had to send one of our software people (he is pretty mechanically inclined though). He got a tour of the facility and then set up a video conference so we could all inspect the machine and get training information. Everything went really well except the fact we discovered an error was made and the machine was only fitted with one head instead of the two ordered. F***.

Quick overview - the machines can be fitted with several different kinds of heads optimized for different tasks. We specified that it should be fitted with a second head that would help with high insertion force components and odd sized components. Something got missed and we were charged for the work but it wasn't done. The vendor agreed to fix the problem but told us that it basically involved pulling the whole machine apart and adding almost an entire second motion system - NOT FUN.

They offered us the choice between a refund of our money and a bunch of free stuff (extra accessories, a feeder cart, and some other useful bits) or taking the machine apart again and refitting the second motion system. In the end we decided to take the inducements  - the offer was rather nice AND we are in a low volume application where we can easily have somebody hand fit the funny components in line (more on this later).


Essemtec Oven

We purchased the oven from another reseller/refurbisher here in Minneapolis - the oven is working well but it's (optional) control computer had a broken serial port. We had the refurbisher replace the computer and everything is working well. The oven currently operates on a sensor basis instead of SMEMA ( more on this later) - we explored having the oven converted to support SMEMA but manufacturers are usually grumpy about helping resellers/refurbishers for obvious reasons.

After some grumbling from the manufacturer - they actually agreed to support the SMEMA upgrade but as it turns out it was rather expensive and didn't offer enough benefit - I decided against it.

We were lucky to find a 208V - 3 phase oven that meets our needs (its smaller and only accepts up to 12" PCBs) , but we don't have a need for high throughput and we really really didn't want to have to run a high current 480 circuit just yet (inconvenient and expensive - probably would cost more than the equipment) 

SMEMA and Automation

So the SMEMA interface is used between various pieces of equipment on the SMT line to send commands/signals. Our layout is as follows:

Conveyor -> Screen Printer -> Conveyor -> PNP -> Inspection Workbench Conveyor -> Oven -> Workbench.

Conveyors are useful for several purposes - one is obviously to move the boards, but also - conveyors are useful to match the speed between faster/slower operations [PNP (fast) versus oven (slow)]. Finally - some conveyors come with magnifiers and work surfaces which enable an operator to "pause" a board, inspect/work on it and then send it along.

As it turns out - you can buy conveyors with either one zone or several zones and sometimes they're fixed speed and sometimes they're variable speed.Multiple zones offer better buffering and variable speed can help with this as well. Also there is the question of width to consider. Finally one should consider if the conveyor speaks SMEMA so that it can signal to upstream and downstream equipment as needed.

We were able to cut good deals for conveyors (they're generally not that expensive only 1-2k each for nicer used ones) AND we have the floor space so we decided to go for the full monty and have conveyors throughout except at the end of the line where there is a little slide that will put boards onto a nice workbench. We considered an output conveyor as well - but this would have required retrofitting the oven for SMEMA - something that was too much hassle and expense to be justified. 

Importantly - we were able to source a conveyor that could "pause" the line and had a work surface and magnifier for an operator to sit at, inspect the work, and insert any odd components - this was crucial to work around having one fewer head in our PNP machine due to the error. A nice effect of only having one head in the machine is that we were able to fit a nice tool changer in it - this gives us the ability to have more nozzles (such as specialized ones for MELF parts) in the machine that it can automatically swap to without needed to fiddle with it.


Utilities, Rigging, Leveling

So as it turns out, our PNP machine is really really compressed air hungry (11CFM @ 90+PSI). For reference our Hurco vertical machining center takes significantly less than half of that. The reason for this is that the PNP uses a compressed air vacuum generator (venturi) - which is just stupid amounts of inefficient.  It is so hungry that it would significantly overload our current Powerex Scroll compressor, especially if other equipment is running.

Another scroll compressor costs about 8k (when you factor in the tank, dryer, compressor, delivery, electrical wiring and pneumatic plumbing). I have RFQs out to various compressor vendors - and even some of them that offer "Compressed Air As A Service" (you pay a fixed rate per hour of runtime with a 40 hr/month minimum) plus a one time delivery/installation fee. Some calculations suggest that our breakeven point would be only 3 years so we decided on purchasing a new compressor instead.

Simultaneously (to possibly avert the need to buy and plumb in another compressor) - we are looking at retrofitting the solenoids and venturi generator with a nice vacuum pump and appropriate vacuum solenoids. This should be SIGNIFICANTLY more efficient, quieter and also save us a lot of hassle - also it would be a hell of a lot cheaper both in terms of cost and real estate. [UPDATE : This option was discarded because millisecond timing is required and we don't want to mess with that right now]

We are having to get new venting put in for the oven to exhaust to the outside, as well as new breakers and outlets for all the equipment. This is proving to be a chicken and egg nightmare since all the equipment needs to be spotted in and aligned before you know exactly where to run things, but doing building work on top of equipment is generally a horrible idea. We are looking at some flexible hosing and wiring options to get around this.

Also, work has to be done on the floor - each machine is pretty heavy (with the PNP being quite a few tons), we currently have vinyl floors which will need to be pulled up and then the adhesive scraped clean to give the machine's feet a flat place to lay.

The machines also need to be leveled (if they are not perfectly level - the frames can warp ever so slightly - affecting the linear axes which is not great). The leveling is significantly less bad than for CNC machines (which require many feet to be epoxied to the floor and then carefully placed either in tension or compression to perfectly level the machine) - however it is still a significant effort. Luckily we were able to convince a tech to come out and do this for us.

To add just a little bit more insult, Starrett is currently out of stock of the precision level we needed (that would fit on the PNP's smaller linear rails) - so we had to spring for the NIST traceable version of the same (more $$$) in order to get it in time - another small and irritating expense. But maybe it will be useful for other things.

If you've ever tried to get stuff you bought online delivered on a precise time and date - you can imagine the headache this is causing (multiplied by about 7 different companies, most of which involve heavy machinery, all of whom need to do their work in a specific order of operations - F**********k.

-M




« Last Edit: September 07, 2021, 06:13:05 pm by matthewpang »
entropicengineering.com
 

Offline matthewpang

  • Contributor
  • Posts: 25
  • Country: 00
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #1 on: September 04, 2021, 09:11:52 pm »
Reserved
entropicengineering.com
 

Offline matthewpang

  • Contributor
  • Posts: 25
  • Country: 00
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #2 on: September 04, 2021, 09:18:10 pm »
reserved
entropicengineering.com
 

Offline matthewpang

  • Contributor
  • Posts: 25
  • Country: 00
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #3 on: September 04, 2021, 09:20:27 pm »
reserved
entropicengineering.com
 

Offline matthewpang

  • Contributor
  • Posts: 25
  • Country: 00
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #4 on: September 04, 2021, 09:29:26 pm »
reserved
entropicengineering.com
 

Offline Cerebus

  • Super Contributor
  • ***
  • Posts: 9125
  • Country: gb
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #5 on: September 04, 2021, 09:38:01 pm »
"Pick and Place Assembly Line - How bad could it possibly be?"

This bad, but with reels of components?



Sorry, couldn't resist.  :)
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 
The following users thanked this post: ANTALIFE

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2281
  • Country: us
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #6 on: September 04, 2021, 11:31:22 pm »
Well, I bought a used Quad/Samsung QSA30A P&P machine, after reaching the end of capabilties of the Philips/Yamaha CSM84 that I had used for 13 years.
(Not accurate enough, it used mechanical centering jaws to align parts.)  The QSA30A has side vision cameras riding on on all 3 heads for centering alignment.
Well, the machine was trashed by previous owner's attempts to repair it, and it had a serious electronic problem from when it was built.  The software has a LOT of odd issues, but the manual it absolute CRAP!  They don't explain how anything works, it is just a guide through the menus.  With huge help from the guy who was the trainer in the factory, I managed to get it up and running.  $500 bid price for the machine, $1200 for shipping, several thousand $ for repair parts and a bunch for feeders, etc.  Total of about $7500 to get the machine fully operating the way it should.  I still had an issue with the last board I did, it has some REALLY low-profile MSOP chips that it wasn't planting into the paste.  I didn't understand that I had all the settings available to fix it, just didn't understand what they were.

I really think with better understanding of how to set up all the vision parameters, I should be able to get very good results with it.
I had the rest of my line already working OK, manual stencil printing frame and toaster oven with thermocouple ramp and soak controller.

Jon
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2281
  • Country: us
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #7 on: September 04, 2021, 11:50:43 pm »

Have I ever used a PNP machine? Nope.
Have I ever worked on an SMT line? Nope.
Is this the smartest thing to do? Questionable.

When I got my Philips CSM84, it was from a working line, just weeks before I got it.  I powered it up, and went to work learning the setup.  It was a really simple machine, no vision, just air cylinders for Z, so no height programming.  The manuals were not the greatest, but quite understandable.  I had never even touched a P&P macine, but was making boards in 2 weeks!

 I had no idea how bad it could get.  The QSA30A manuals were astonishingly bad.  Probably due to cultural reasons, they CAN'T tell you how to do anything, that would be condescending.  The coordinate system is really odd.  My Philips had the reference corner at bottom left.  Due to the conveyor stop being at bottom right, you needed to supply an offset for the width of the board.  The QSA30A has the reference at the corner where the stop is, (bottom right) so all coordinates are increasing toward the left.  But, the machine's X coordinates increase to the right.  Completely CRAZY, and nowhere in the manual does it explain this!  But, once I had figured out the damn crazy way this was all set up, I could easily write a little C program that parsed my CAD placement file and put out the coordinates as the machine needed it.

Well, the machine arrived generally in working condition, but then things started breaking.  After getting into the disk drive, it was apparent that the machine had not been powered up in at least 6 years, and likely was left sitting in unconditioned space in Austin, TX.  So, a humid soak.  Fortunately, this Quad model is based on a Samsung model popular in the far East, so spare boards are easily available.  But, it took me several months to get it running, and the frustration I experienced trying to understand how to set up a job might have had something to do with the stroke I experienced a few months ago!

Jon
 

Offline Mangozac

  • Frequent Contributor
  • **
  • Posts: 347
  • Country: 00
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #8 on: September 05, 2021, 11:03:34 pm »
Congratulations! I don't think you'll regret it.

Why the x-ray machine?
 

Offline Cerebus

  • Super Contributor
  • ***
  • Posts: 9125
  • Country: gb
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #9 on: September 05, 2021, 11:52:13 pm »
Congratulations! I don't think you'll regret it.

Why the x-ray machine?

"What's happening under that BGA package?" <- That's why X-ray.
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 

Offline Mangozac

  • Frequent Contributor
  • **
  • Posts: 347
  • Country: 00
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #10 on: September 06, 2021, 12:30:32 am »
"What's happening under that BGA package?" <- That's why X-ray.
I know what the x-ray is used for; I was wondering why he felt the need to add it in at this point. If they've been assembling boards by hand successfully with no x-ray it's not exactly a requirement yet. I know plenty of people running SMT lines and doing BGAs without x-ray.
 

Offline matthewpang

  • Contributor
  • Posts: 25
  • Country: 00
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #11 on: September 06, 2021, 07:49:21 pm »
Initial post updated :)

"What's happening under that BGA package?" <- That's why X-ray.
I know what the x-ray is used for; I was wondering why he felt the need to add it in at this point. If they've been assembling boards by hand successfully with no x-ray it's not exactly a requirement yet. I know plenty of people running SMT lines and doing BGAs without x-ray.

Question of what the time is worth right - in our case we're a consulting engineering company and our clients are billed hourly. With a lot of the latest parts only available in BGA plus the increase of WLCSP and even RF Modules with pads that cant be inspected (MDBT50Q and other NRF52 modules are a popular example) PLUS high layer count PCBS (lots of thermal mass) - many hours are spent debugging a sufficiently complex board only to discover one (or more) of the un-inspectable parts has an issue.

Also assuming after a day or more of hunting - you find a problematic BGA - all you can do is to rework it and pray it worked right (which isn't necessarily a good bet).

X-ray rapidly becomes justified (saving ourselves the frustration, and our clients a lot of $$$) .

For example - we recently had a board for a customer that passed 100% of its functional tests. However - there was an unexplained increase of 400uA in current draw of just some of it's sleep states. Having enough dynamic measurement range to go from ten microamps in deep sleep all the way up to amps in wake is really hard. We only caught this after quite some substantial debugging (there was a partial short defect under the BGA of one of the ICs that had a nontrivial impedance and only showed up when said IC turned on some of it's peripherals. While the uC and Joulescope et al try to solve this problem - it is *very* hard to do this if you also need higher bandwidth/sampling speed and are using a device that will get upset from the range switching.  This is in fact so hard to do that the test gear company in question wont let us say how we took the measurement as it's in an upcoming product.

Many many days of work and a lot of expensive equipment would have been saved if we could have just x-rayed the board because the moment we did it showed up. By the way - reworking this board would have caused the issue to go away, never to be found until it showed up in some small % of units in mass production (there was a small inefficiency in the component footprint and solder mask that increased the probability that this would happen)

Well, the machine arrived generally in working condition, but then things started breaking.  After getting into the disk drive, it was apparent that the machine had not been powered up in at least 6 years, and likely was left sitting in unconditioned space in Austin, TX.  So, a humid soak. 
Jon

Yeah definitely hear you there, we purchased all our machines via reputable refurbishers for this reason. Definitely paid more - but it's worth it for the time and better peace of mind.
« Last Edit: September 06, 2021, 09:22:38 pm by matthewpang »
entropicengineering.com
 

Offline Reckless

  • Frequent Contributor
  • **
  • Posts: 303
  • Country: us
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #12 on: September 10, 2021, 12:22:28 am »
If you need any GSM parts let me know I have a truck load based in chicago along with a UIC GSM for sale and a GSM GENESIS LIGHTNING dual head dual gantry machine for sale.  Tons of gold, black, green feeders for sale to go with genesis.  I love my Essemtec ovens and have extra one for sale in case anybody wants to follow in OP's footsteps.   Essemtecs are the best IMHO (small, compact and precise).  Ive had one Ekra semi automatic it was nice but haven't played with any of their automatic machines.  My tech likes his Ekra E5. 

 

Offline salihkanber

  • Regular Contributor
  • *
  • Posts: 80
  • Country: us
  • Hello all.. It's nice to be here..
    • Dynamic, Versatile, Innovative Start-Up
Re: Pick and Place Assembly Line - How bad could it possibly be?
« Reply #13 on: October 26, 2021, 08:23:52 pm »
Wauv thats a bit of a long text :)
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf