What would a quality PnP setup cost?

One alternative to various EMS I have been talking about in another thread, is to do it your self (I know it might be crazy)

Cost for assembly per board might be close to a dollar, including freight

I was just wondering if I bought a PnP machine, like a NeoDen for 4000 USD:

Then I bought a reflow oven, 2500 USD:

I need a paste machine then, 350 USD:

That machine would require a lot of time on my hand

Let’s say it’s 10k USD, for a small setup. What kind of quality could be expected for that setup?

(Completely ignoring that the time I spend could be used on other stuff)

I made a liteplacer and have been running openpnp on that, at a cost of about $2000.

But it’s far from reliable…problems range from feeders (3d printed strip holders), calibration and definitely the computer vision. Oh, and finding documentation for the vision functions. Getting two USB cameras to work required Ubuntu (win 10 and catalina didn’t work).

Only consider something like this if you are comfortable spending time tweaking, browsing the interweb and baby sitting the machine a lot.

I have found a local maker space that have a smallsmt pnp for members to use. It’s a $10k machine and I have seen what it can do. They also have paste printer, reflow and vapor phase ovens, microscopes,etc.
I will get my own smallsmt or other proper pnp some day, but I won’t buy something I haven’t actually tried and tested both hw and sw.

The neoden has a reputation for bad sw from what I can tell, and the feeders are so-so. And a pick and place without reliable feeders is an expensive toy. I would only consider yamaha cs feeders.

Many report of problems picking LEDs, so pay special attention to feeders if you have many LEDs.

Good luck, though! :blush:

1 Like

I have a non-vision Neoden TM-245P. Picked it up on eBay for $600 because it was basically in an e-waste stream and was intercepted by the seller who had no idea if it worked. It turns out it’s fine.

Without vision, the part position error just from the part rattling around in the tape makes it unsuitable for fine-pitched parts. But it is sufficiently reliable for placing 0603, SOT-23, and QFP-44 for functionally acceptable boards.

The biggest limitations are the limited component heights and the “fixed mount” feeder configuration. If you don’t need to swap reels at all, it’d be ok. But wholesale swapouts of parts between incompatible BOMs can eat up an hour or two to prep the machine, as well as discarding about 45 mm worth of parts for each part being loaded.

It’s definitely not as easy-to-use/flexible as the MyData setup that my local CM has! Having vision on the 3V should help with tighter parts placement, but I expect job setup would still be a significant amount of work if you switch out jobs.

For $11k you could buy a whole smt line with neoden 4 with conveyor belt, 5 zone oven, etc.

Getting everything from one place might be good for support, I don’t know.

I have done quite a bit of reading about low cost pnp lately and it seems that many would rather buy “old iron” like Samsung, kayo or other 10-15 year old machines with yamaha feeders than buying new, Chinese made machines with chinglish sw and unknown support. At any rate be prepared to spend time troubleshooting and fiddling.

Quite amazing that one can buy a machine like this for 7k USD:

I have a Chinese CNC, at least 10 times smaller at 3k USD

I guess at first ask yourself what is the requirement or define “quality PnP”. Like super fast and flexible time to market ? low stock and flexible manufacturing. Many different models and Rev of PCB, NDA based stuff you cannot outsource etc. How complex are the boards. Throughput, number of indiv components

I have my own experience with a low cost home setup (CHM-T36v etc). but it has limited feeders (i think like 22x 4mm and 6 more wider). Precision is acceptable for 603 components, yield with 402 i found not acceptable due to high rework %.
If you dont have a lot of Throughput than even paste can become an issue - they (the flux) get bad over time, in uncontrolled areas (hum, temp) they have different behavior. Not a pblm for 603, 804 but smaller components or finer pitch.
Thats why its unused after a larger “basic” job and will sell it soon anyway.

If use case is flexibility those setups suck. If you need to swap reels or change to new jobs on low cost PnP takes hours - even with scripts. I discussed elsewhere and i guessed that costs for getting the workflow right are at about 20k$ worth my time. Setup costs PER JOB are between 200 and 500$ (worth my time).
Best those cheap setup are for when your time is worth little, you have only few different PCBs with little count of different components.

Old PnP aren’t better actually as they are optimised for “single model, place 100.000 boards” - setup is easy as long as they are tied into a CAM system - standalone you’re lost. Than still have the “problem” with stencelling/pasting (continous quality and repeatability) particular with fine pitch and cheap stencils.

If you are interested into learning “Manufacturing” or make this your hobby - than those devices are ok. Given CM turnaround times and cost today it doesn’t justify apart from an ego, learning or individualist element.
My definition/need of “quality PnP”:
I am working with an Essemtec Fox, plus autom pasting, from time to time - That is a 100k$ device.

i have the neoden 4 and i like it fine (once i modded their software/setup and a ulp for eagle) except for the fact it doesn’t have easily swappable reels. at the price its pretty good but i really wish and have even considered making a mod for it to use swappable feeders

the newer neodens are definitely at the price where i would probably go for a used samsung/juki/philips machine, they are for the most part a lot better made and there are always parts for them around.

A lot of good info

I am working on a board that I dream of selling in 10k+ quantity. I can get EMS to do it, but I hate paying for something where I could build up my own setup through investment. Anyway, the boards will be the same, so I would do a large batch at one time (maybe for at full year production), and then put them on the shelf. That way the cost you mention about paste and setup is low. And I can tune the process for highest yield

About quality, I would need a board with high yield. Failures at customers is not an option. So question is really, how much would one need to invest to get comparable quality to Chinese EMS without being an expert and with a setup that is automated so I need to only fill material in one end, and get boards in the other :slight_smile: (granted, it wont be a in-line production line)

To reduce cost to produce as much as possible, inhouse production is often nessesary

Wrt components, smallest is 0402 and largest is QFN32

Actually a Juki can be had for 12k USD:

actually i have a juki and it cost me 3K

@charliex You seem to know a lot about this stuff.

Let’s say speed isn’t the big issue, but 0201 is a hard requirement. What do you like?

The SMT550 is claimed to handle 0201.
It needs electric feeders for 0201 but is prepared for it. 0402 s and larger can use regular CD feeders.

Price $7500 with free shipping if I remember correctly.

Personally I really like this machine, but I’ve not seen or retired if in person. And I don’t know what the support is like. Sw is available in English. There are lots of YouTube videos showing how to solve various issues, check out John Plocher’s.

I’m very tempted to get one - does anyone on here have one and can share experiences?

I am surprised that most discussion is about cost of the PnP. While it is the time which is expensive and can’t be bought.
If a PnP vendors says it can do 201 - than you need to adjust stencilling process. If paste isn’t 100% for those small pitch, than tolerances in PnP make it worse, having a “average” Oven makes it worse. You can mediate by inspecting (1x after paste, 1x after pnp, 1x after reflow) + final inspection & burn in.
People go crazy on 5k vs 10k for a PnP while you waste 50k easily every year because the line-up is so inefficient. What you usually pay for is repeatability, low tolerance.
It’s like this parts-per-hour or “heads” discussion which is so useless if yield is 80% or job-change is 2 days. The time you “rework” or setup a new job you lost for selling and mrketing the product or develop new features.

Back to the original Question: Assuming this is all business (and not hobby) based for a 1-2 man show. Also is for final Production and not low volume protos; than plan in 100k invest + 10-20k/yr (unproductivity). If that doesnt calculate out with profit than get a contract manufacturer.
For the one man band and 10k+ Volume you have to choose being the Design Engineer or the Production Engineer i am afraid.
I also dont let the argument count that you invest into production and than it’s done and you make money - in fact it starts generating cost and use your time by that point. The latter is most precious we engineers have to invest mindful.

1 Like

I mean you see this same thing in almost all electronics tbh - how many times do people complain about some tool that costs $500 and would save them days of work. So hardly just a complaint about PnP discussions :wink:

This is going to be borderline impossible to run without a fair amount of hands-on time, but may be possible. It’s easy to imagine you can tune the machines to work without much oversight, but it’s not true in practice IMHO, especially with low-cost machines. You are going to need to be spending time watching over them, fixing stuff, reworking boards, etc. But talking to techs at larger fabs they seem to do a lot of manual fixing, oversight, etc.

Note I’m coming from the specific place of someone who set up a PnP + other gear for < $10K, so this isn’t theoretical. Our setup is something like this:

stencil holder, CHMT36VA for PNP, cheap inline reflow oven. FWIW I’m working on writing up a lot more details of this, but my coles notes version:

  • You can hit 0603 no problem, 0402 works pretty well, I haven’t tried 0201 but I doubt it would work on this.
  • TQFP-64/0.5mm “mostly work” but with some issues, you need to inspect each one to check if it’s centered and nudge (at least a few seconds per part). Mostly an issue of rotational jitter problem - QFN parts are fine as they are smaller. Smaller TQFP or larger pitch TQFP also work pretty well.
  • Stencil printing quality is a major limitation at 0402 and smaller - without a fully automated printing (that moves the blade a consistent amount etc) it seems to be difficult. Keeping solder paste in “good” condition when you only do a small number of boards is also harder than if you are running boards all day long (like any real fab would be).

You aren’t going to beat a “good” Chinese fab, you might hit an “ok” fab quality (PCBWay/JLCPCB). You are still going to need to inspect boards for issues like tombstoning etc - “good” fabs really just have better inspection processes, now that will be you doing inspection of each board. You won’t have real AOI equipment, although there is some newer ones you can use, it’s not the same.

You can beat cheap fabs on through-hole quality (where they always seem to suffer). You have the advantage of doing partial builds to fit your budget too.

If you are running a small number of products, doing your own assembly isn’t ever going to pay for itself IMHO. Either in terms of cash outlay, or ongoing cost of you running the equipment.

In our case - one driving factor was having this product where there is almost nothing on the boards, but we have a lot of variants (20+ variants), and low quantities sold for many of the variants (like qty 100/year on some):

When starting out it was helpful to be able to do partial runs to keep total costs down, but I’m not convinced it was a smart decision compared to just getting a loan. We also ended up doing it ourselves in a large part because we have the exact same board (say for an STM32F microcontroller as above) with the only difference between the STM32F microcontroller mounted changes. I couldn’t get manufactures to quote me a good price for “total of qty 1000 boards, but 200 each with a different part mounted”. Eventually you get frustrated and just order the gear…

Anyway - despite having the gear sitting here idle 80% of the time, and having a full-time electronics technician that would do the runs, we still mostly outsource board assembly. Some of the medium-cost targets boards are partially outsourced, where a low-cost fab does the SMD & we do the pin mounting (as mentioned - that through-hole issue). Only a few things are handled in-house - mostly those lower cost target boards and some other stuff that we already have a good setup for running or is easy to do (eagle-eyed readers might notice we have a “DIFFPROBE” label on the photo above for example).

That said we have done more complex runs in-house when needed for time reason. Sometimes a handful of parts are long-lead, and we don’t want to wait for everything to be in stock. Or we ran a few hundred of a cheap “ChipWhisperer-Nano” product we wanted to bring to a conference, and there was only a few weeks until the event (i.e., no chance of contract manufacture delivering in time without huge rush fees).

It’s also handy for doing prototypes for placing common parts, although from when we started all this services like JLCPCB have made that much less compelling.

I was almost about to buy a STM460 a year ago, but it was going to be too much hassle moving up the stairs (would have to take apart, even then the weight etc). I had someone give me good feedback on the STM460 that was my basis, so am also curious about the STM550.

EDIT: One additional comment as I realized you mentioned higher quantity - the advice from everyone else who has done this has been to buy a good used machine. A 10 year old used machine will be >> value than a cheap brand-new machine with the same cost as the used machine. Not sure what the technology limit is there, but you see lots of auctions where facilities shut down and the machines come up. If you are located anywhere that such auctions make it possible to check them out in person or freight isn’t too much, that changes the balance a lot of used vs. new. Unfortunately where I am that balance isn’t great (local machines would be rare, and freight is non-negligible). As mentioned the setup of these older machines is non-trivial, so access to someone who has dealt with it before is a must-have.

I have a motivation for doing assembly myself (or at least locally) because I want to stamp “made in Norway” on them rather than “made in China”… My competitors are mostly Swiss…!

Furthermore I cannot risk a Chinese (or any third party) putting wrong or counterfeit parts on the boards because that would violate their homologation or make them out of spec.

Lastly, I don’t want to make huge runs for what is a low volume, high margin product.

So cost isn’t everything. Control is important. Which is why I be assembling myself for now :blush:

I like the Samsung CP45FV depending on which 0201 you mean :slight_smile: it is widely available, reasonably priced and lots of spare parts. Juki’s that are past the 5xx series since the heads on the older machines are ridiculously hard to get and you have to modify them to get more than a fixed rotation.

I have a philips pnp in storage its pretty well made too but we have never run it, it just was a great price, don’t have the model number offhand but it might be a topaz/assembletron

it’s sensible to discuss pricing/budgets, you can spend a lot more on a new pnp that has less features than a new one, you can also possibly spend more time setting up the old one than the new one, but the reverse is also true. a used juki/samsung is a lot higher quality than most of the cheaper current desktop pnp’s

if i can save 5K on one machine, i can spend 5K on a different part of the process that might help me save more money over all. It’d be great to have the budget to buy all the things at the best, but if you’re a small shop or just starting up or its a low yield then its even more important.

so it is a nuanced discussion, and requires research to do the right thing that makes sense for you or your business and all aspects should be considered, we can’t just say don’t consider the new vs old debate, or is it more expensive to save now or later.

Great insight and very nice setup you have. Looks very much like what I was aiming for

I guess you are right that I am not going to save money on doing it myself, so might need to reconsider this. I have a friend that is an expert, so he could help in the setup. Anyway, one compelling thing is the flexibility of the setup and the no markup component sourcing. The EMS I have talked to, almost all of them take a significant markup for purchase components, and for small series the setup cost will be significant

Very interesting discussion!

Sounds like the pro of an in-house set-up is that when the PCBs arrive in the mail you can just say “cool, lemme populate a bunch today!” the con is that when the PCBs arrive in the mail you realize “shit, now I have to spend the rest of the day just to populate a bunch!”…

So I’d ask around a little too - I’ve had very good success with Bittele (https://7pcb.ca/) - you can see several of their boxes in the upper right corner of my earlier photo :wink: They have an online quoting systems now so you can get an idea what it would be without even talking to anyone (a feature I always love).

They do assembly in both China & Canada. Notably they are purchasing parts from normal distributors (Digikey/Mouser/etc) to avoid the issues with sourcing parts via questionable means. The markup is low & well worth it to have them deal with the “making sure the parts are in right format for PnP runs” deal.

They are almost entirely pay upfront is one way they keep margins reasonable - they don’t have to float production costs during the run & don’t have bad debt costs.

FWIW my experience is the quality of the companies can change over time though. So I’m always a bit worried giving recommendations, as I’ve worked with companies that were great & then some critical technician left or they made some business decision and we suddenly had problems.