Out of curiosity: what is the time saving potential of multiple nozzles?

[ma...@makr.zone]

Hi

Just out of curiosity: does anybody know how much time can be saved by using multiple nozzles?

I sometimes wonder if the potential is not overestimated when seen as a frank cost-benefit assesment.

It is obvious that the nozzle tip changing can be reduced. But the changes required remain constant (max = number of tips), they do not scale with part numbers, so with large PCBs or panels the time saved will not be significant (and only for those large job speed really matters for standalone machines IMHO). 

Another benefit would be to pick parts at similar feeder locations, so the summed travel time is reduced. But feeder/pick co-location is not optimized in OpenPNP (AFAIK), so often the head will make large detours to get all the parts, especially when the feeders are on all the machine's sides. Watching videos, I somethimes got the feeling that the different nozzle tips will often force he planner to mix parts in inefficiant pick location patterns. 

On the other hand, everybody does it, so there must be a large benefit, right?  ;-)

Has anybody measured real world job times with and without multiple nozzles?

Preferrably with large PCBs/panels? (and no LED matrix with all the same parts ;-)

Thanks!

_Mark

[Jason von Nieda]

The *potential* is very high. We are not using much of that potential. For it to really matter for speed, we need to optimize picks, as you've mentioned. This is probably close to impossible to do on the fly. We need static planning, as mentioned in the other thread.

I work with a modern, brand new, high end commercial PnP in my day job now, and it's been a real education seeing how it differs from OpenPnP. All planning is done ahead of time - usually offline.

One other case where it's important is for people who use fixed nozzle tips, or a manual changer. In this case it makes it possible to run a job that requires multiple tips without stopping to change.

Jason

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/bbe4ea01-8d55-4ee4-8cdc-4df0beb997a3%40googlegroups.com.

[Marek T.]

If we can add some stupid planner, that we discussed on some day ago, doing picks N1, N2, N3 just from (the manually made) list one by one - we can optimize the picking process and pick three parts without any x motions. So we can have this planned out of the job ahead and offline :-). This is how my ancient Philipses do the jobs.

[ma...@makr.zone]

It's just funny how everybody is so convinced of the potential and building and ordering machines with more and more nozzles, accepting a firework of added complexity and cost without any hard numbers to back it up (it seems). I'm not even talking about the fact that a head is also getting heavier and heavier, its drag chain stiffer and stiffer and the motion therefore slower and slower (everything else being equal).

I'm not disputing some benefits, I'm just curious as to how large they really are. And I'm not talking about big pipelined industrial machines, but the hobby/prototyping variety predominantly discussed here.  

BTW, I agree a multi-nozzle machine is fascinating and that can be enough reason to want to have one! :)

As to static planning: it does not need to be truly static, it should be iterative instead. Otherwise it loses todays error recovery capability (such as disabling a feeder and failing over to the next one or leaving that one part out).

We'd need a solver that gets better and better the more calculation time you invest, something that adapts to newly introduced disturbances such as unexpected errors, something that you can refine ("cool down") as you get closer to actually doing it:

https://en.wikipedia.org/wiki/Simulated_annealing#/media/File:Travelling_salesman_problem_solved_with_simulated_annealing.gif

https://en.wikipedia.org/wiki/Simulated_annealing

_Mark

[Keto]

How's about a simulator that shows the difference between one, two or

more nozzles using openpnp2 planner logic and tip compatibility map?

Using a tooltip changer is required of course.

The results would show that increasing nozzle count would bring you

allways higher cph, never less.

[Marek T.]

Question how much higher and is it important for the prototype machine...
However I have 3 nozzles and don't imagine to have them less, maybe I'm just used to this. But it seems to me that it's better to have 3-4 nozzles even if instead of the changer only.

[ma...@makr.zone]

I was hoping there were some people on the list that had the same job run with just one nozzle (perhaps due to a technical defect) and then with multiple nozzles, and remember the time difference.

_Mark

[Mike M.]

Marek
Can you run a job with 3 nozzles and record time,
Than disable 2 nozzles and run the same job only with one nozzle and record time.
Indicate how many parts /job ?
Mike

[Marek T.]

Not on the board we assemble on the moment. Within this specific job I'd waste a lot of time for the nozzles changing and result got not representatible at all. But I'll try to remember to check it when do some one-type nozzles board.

[ma...@makr.zone]

That would be very nice. Thanks!

_Mark

[Mike M.]

@Marek - & I will remind you from time to time  :-))

[Marek T.]

Mark do you think that some 50pcs 0805 parts assembling will give you an information usable enough?

Mike - what would I do without you ;).

[Bernd Walter]

I did and this is why I want a changer setup and also this is why I changed
my current fixed setup for two Juki 503.
But I don't have the numbers at hand.
I usually look up in Log File history what the cps had been before a change with
the same board.
My Feeder setup is more or less static, so things are compareable even after
a few months.

For just a few boards it doesn't matter that much.
Setup time is way higher than the benefits of a few minutes run time.
However if you do a small batch of 20-100 boards, then it could mean
saving hours.
If I would build this machine again, then I might do 2 nozzles and changer.
I went for 4 nozzles to have all required tips to run my jobs permanently
mounted.
I'm seriously considering to build a second small machine though.
Small traveling paths, small amount of 8mm automatic feeders, lightweight head.
Just to fast assemble the most common used pasives and then do the rest with
the big machine.

--
B.Walter <be...@bwct.de> http://www.bwct.de
Modbus/TCP Ethernet I/O Baugruppen, ARM basierte FreeBSD Rechner uvm.

[Keto]

I wonder how the IPC9850 throughput would be on our machines?

[Marek T.]

Hi Bernd,

I've 5 big machines but usually use two-three of them connected in-line. First one or two machines do simple 0603-1210 parts aligned by the jaws, fast and with limited precision but good enough for the parts they do. And the last one is Openpnp doing all the rest "complicated" parts. It's best setup in my opinion. So your idea with second small machine seems to me being really good.

[John Plocher]

> without any hard numbers to back it up (it seems). 

On my chmt48 - not using openpnp - a 2- head job is easily twice as fast as a 1-head one with the same design, assuming both heads are the same size.

The pick planning is simple- use the CSV file part order to pick parts; with some optimization there, things could easily be another 20-50% faster by reducing gantry movements.

If I have to change nozzle sizes (the chmt has no auto changer), the 2-head is even faster because there is no high level interactive job resubmission work.

 John

[ma...@makr.zone]

Hi Marek

I'm most interested in real world non-trivial mixed jobs with many parts and different nozzle tip sizes.

Like I said it's just curiosity and I guess everybody has different use cases.

The PCB I'm developing currently has ~100 distinct parts, ~400 in total. Most are 1 each, a few have handfuls, one has 50. I guess I'll use 4 different nozzle tips.

_Mark

[Marek T.]

I don't use autochanger and usually plan the job this way to not need nozzle change. Pity i tuner for changing and calibrations... If need more than 3 nozzles' types for some board I use for this the second or third machine inline. So I can only the rest as proposed, some 100 resistors made with three or only one nozzles (I mean the head!).

[Mike M.]

@Marek - with three (3) nozzels it must be hard to avoid crashing into the Juki nozzle changer ?

Mike

[Marek T.]

Not sure if catch what you mean Mike, what the difference 2 or 3... Anyways, I don't have nozzles changer and no crashes then :-)

[Brynn Rogers]

On Thursday, November 14, 2019 at 1:28:42 PM UTC-6, Jason von Nieda wrote:

The *potential* is very high. We are not using much of that potential. For it to really matter for speed, we need to optimize picks, as you've mentioned. This is probably close to impossible to do on the fly. We need static planning, as mentioned in the other thread.

The optimal planning seems like it should be a straightforward travelling salesman type of problem, a classic computer problem.  It actually sounds like fun :-)  If I only had some time.       

Of course (on some machines) you would also want to sort the parts so they are done in size order to some extant - get the small parts done first , and leave the large and taller (and expensive?) parts til the end

 

I work with a modern, brand new, high end commercial PnP in my day job now, and it's been a real education seeing how it differs from OpenPnP. All planning is done ahead of time - usually offline.

What model machine are you working with now?     I learn stuff every time I take a close look at commercial machines.

 

One other case where it's important is for people who use fixed nozzle tips, or a manual changer. In this case it makes it possible to run a job that requires multiple tips without stopping to change.

Jason

On Thu, Nov 14, 2019 at 12:43 PM ma...@makr.zone <ma...@makr.zone> wrote:

Hi

Just out of curiosity: does anybody know how much time can be saved by using multiple nozzles?

I sometimes wonder if the potential is not overestimated when seen as a frank cost-benefit assesment.

It is obvious that the nozzle tip changing can be reduced. But the changes required remain constant (max = number of tips), they do not scale with part numbers, so with large PCBs or panels the time saved will not be significant (and only for those large job speed really matters for standalone machines IMHO). 

Another benefit would be to pick parts at similar feeder locations, so the summed travel time is reduced. But feeder/pick co-location is not optimized in OpenPNP (AFAIK), so often the head will make large detours to get all the parts, especially when the feeders are on all the machine's sides. Watching videos, I somethimes got the feeling that the different nozzle tips will often force he planner to mix parts in inefficiant pick location patterns. 

On the other hand, everybody does it, so there must be a large benefit, right?  ;-)

Has anybody measured real world job times with and without multiple nozzles?

Preferrably with large PCBs/panels? (and no LED matrix with all the same parts ;-)

Thanks!

_Mark

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.

To unsubscribe from this group and stop receiving emails from it, send an email to ope...@googlegroups.com.

[Trampas Stern]

So with multiple heads a build will be faster, basically it reduces the travel time between picking part and placing part.  If you are doing production builds and time is important consider the cost of machine down time. 

For example if the PnP machine breaks you have zero production.  From this my conclusion was if you are not space constrained then it is often better to have multiple machine. If you have two PnP machines if one is down you still have production capability, and you have doubled your throughput.