I don’t even know where to begin this blog post. Do I tell you what a pick and place machine is? Do I give you the backstory about how our OoberLights project got to where it is today, and why buying an open-source LumenPNP might be a really good fit for us? Do I tell you that I have no idea what I’m doing?!
We’re going to talk about the OoberLights because producing these boards are what’s driving my decision. Other than that, I don’t know where I am going here. I’m writing this to help me make a decision. I hope that this train my thoughts are currently riding will be helpful for you as well!
NOTE: I just realized that I’ve never written anything about the adorable and hopefully reasonably priced OoberLights Micro boards! I need to correct that soon!
Let’s condense the OoberLights history down to a few paragraphs
It was a cold evening in Scranton, PA sometime around 2006. No. We don’t really have to go back that far, and I have no idea how accurate that year is.
A long time ago, I saw an LED imitation of a Dekatron on hackaday.com. I thought it would be neat to replace the num lock, caps lock, and scroll lock lights on my IBM Model M keyboard with tiny Dekatrons. I figured I could use them as CPU meters, disk usage meters, and I could spin LEDs around to indicate network speed.
Some number of years ago, I told a friend of mine about this. He said, “Why don’t we do it with Neopixels? Why not do concentric rings? Why not make them bigger?!” We ended up with a monstrosity of a board with 90 LEDs and an ESP8266. It was sized to fit in a 5.25” drive bay in your little home server.
Right as we ordered prototypes, the pandemic and its associated supply chain issues hit. The prices of our components went up, and some of our components were almost impossible to buy in any reasonable quantities. So we put the project on hold.
We eventually scaled back to something closer to my original vision. We dropped the outermost concentric ring, skipped the idea of having any brains, and squeezed the LEDs as tightly together as we could. We now have a simple board with 21 LEDs that has about the same footprint as a Cherry MX keycap.
Money is hard
We ordered a a batch of 20 prototype OoberLights Micro boards from PCBWay. The boards were panelized, populated, and soldered for us. We ordered because PCBWay was running a deal. If I remember correctly, we wound up paying the same rate for PCBA per board as you would normally pay for 100 boards. We got a much better deal than the normal rate.
We paid $24 for the unpopulated PCBs and $98 for the LEDs and assembly work. The total for the order after shipping was $129. That’s $6.45 per OoberLights Micro.
Talking about money is hard!
I feel a bit uncomfortable talking about this. We have a product here that we want to sell to you. I’m about to tell you that they may cost us $2.30 per unit. This is going to make you wonder why I set the price in the store to $15. I don’t know if it will be set to $15.00, but whatever the number is, you’ll want to know why I’m putting $12.00 of yours into my pocket when I hand you a $3 doodad!
Not only that, but what if I told you today that I expect the OoberLights Micro boards to sell for $10? What happens next month when I learn that I can’t make that happen and I raise the price to $15? How are you going to feel about that?!
Some of the numbers are extremely concrete, like what we actually paid for the prototypes. Other numbers are more nebulous, like how much we think PCBWay will charge us if we order 1,000 units in a single batch. I am quite comfortable talking about the former, but the latter make me a little nervous!
Back to how difficult money is!
Our ingenious PCB designed has spent time punching numbers into PCBWay. If I remember correctly, he says we get the cost of each fully assembled unit down to $2.36 if we order at least 1,000 units in one go. That’s about $2,500.
We would still need to order one more batch of prototypes. At full price, I expect that would be 20 prototypes for a little over $200.
That’s pretty reasonable. We’d need some sort of packaging. We need to buy shipping materials. Even so, I would think we could list them in the store for $12 or so. That’s not bad!
This is risky. What if our container falls of the boat on the way over from China? What if $2,500 get confiscated at a border crossing? What if nobody buys them, and I find a box with 973 OoberLights Micro boards in my closet in 2035?
If you’re placing a $2,500 order every month, and one goes wrong, that isn’t going to be the end of the world. If our first big batch goes wrong, the whole project is probably in big trouble!
How much will it cost with a LumenPNP?!
The printed circuit boards are cheap. They were only a little over a dollar each when we ordered 20. They’ll get even cheaper as quantity goes up, but for the sake of this post, I will just that they’re going to cost us a buck.
What if I buy reels of WS2812 LEDs? What if I run my own pick and place machine? How much will each OoberLights Micro board cost?
Here’s the lazy and naive answer. It is $2.18 per board. That doesn’t include the cost of the LumenPNP. That doesn’t include labor. That doesn’t include the time and energy it will take me to assemble and learn how to use the LumenPNP. That’s just the cost of components, and I’m just taking a guess at how much the boards will cost from PCBWay.
I don’t have a value to assign to the labor
I have no real idea of how much time I will have to spend in front of the pick and place machine. The first few times running the machine will be rough. Then it will get easier. Then it will eventually become an easy, normal process.
No matter how efficient I get, even if you value my time at minimum wage, I don’t believe it will be possible to match PCBWay’s price per board of $2.36.
And that first batch of boards that we make in house is going to be expensive! We will have to amortize the cost of the $1,145 LumenPNP kit over quite a few batches before it pays for itself.
The LumenPNP would help us minimize our risk
This is another thing that is difficult to assign any sort of dollar figure to. What are the odds of receiving 1,000 dead OoberLights Micro boards in our first batch? Are they going to fall off the boat? What are the chances that we just can’t sell 1,000 boards?!
Here’s what I do know.
We could actually make a single assembled prototype board for about $3.00 with the LumenPNP! How awesome is that?!
We don’t have to buy 1,000 fully assembled boards to get down to a good price point. We could instead make dozens of OoberLights Micros at a time.
The LumenPNP seems to be a lot less risk. If this all winds up being a horrible idea, I would much rather have a LumenPNP and $200 of unused Neopixels on my hands than 1,000 OoberLights boards that nobody wants. I could always sell the LumenPNP and the unused Neopixels to recoup some of our money!
The LumenPNP will make us more flexible
A small batch of unpopulated PCBs from PCBWay takes a week or two to arrive. A small batch of fully assembled OoberLights Micro boards took about six weeks to arrive.
What if we decide to add or remove a ring of LEDs to make a bigger or smaller OoberLights board? If we are assembling the boards in house, we can have the prototypes ready in two weeks, and if they work well, we can just start populating PCBs and have them in our store the same day.
If we don’t have a LumenPNP, we’d have to wait at least 6 weeks just for the prototypes. Then when we see that they work, we’d have to spend another $2,500 to buy the first 1,000 boards. Then we’d have to hope we can actually sell them!
I have been starting a lot of sentences with the word “then.”
Things get even better once we have three different size OoberLights boards. As stock runs low on each part, we can just produce the ones we need. We won’t have to order 1,000 at a time. We can produce just a few dozen at a time.
The LumenPNP comes with its own risks
What if I can’t manage to assemble the kit? What if I am unable to calibrate the thing and get it picking and placing? What if I can’t figure out how to get some sort of reflow oven going?!
The LumenPNP is an overgrown 3D printer. I’ve assembled enough 3D printers that I am confident this won’t be too much of a challenge.
A pick and place has a lot in common with a 3D printer or CNC router. I will be genuinely surprised if I can’t puzzle this thing out, and if I can’t, I’m sure I can find some help!
What about more expensive industrial pick and place machines?
When I mentioned that Stephen says the LumenPNP can place 500 components per hour, a friend of mine pointed me at the Neoden 3V Advanced saying that it was ten times as fast for only eight times the cost!
The Neoden is definitely faster, but not my such a large factor. The Neoden can only do 5,000 components per hour without vision. With vision, the maximum speed is 3,500 components per hours, but the recommended speed is 1,000 components per hour. That’s only twice as fast as the LumenPNP.
Let’s just ignore the fact that I can’t afford an $8,000 machine just for this project, and lets just make the assumption that Stephen’s 500 component per hour number is pushing the LumenPNP as hard as the Neoden 3V Advanced would be pushing itself at 3,500 CPH.
Our OoberLights Micro boards only have 21 components. What if we fit a panel of 50 units in the pick and place at the same time? It would take a little more than two hours for the LumenPNP to populate those boards. The Neoden 3V Advanced might be able to do the same in around 20 minutes.
That seems too fast! If it takes me 10 minutes just to load either machine with fresh LEDs and PCBs between each job, that means the LumenPNP will be operating 92% of the time while the Neoden will be waiting for me to do work at least 33% of the time.
In my garage, it sure seems like it’d be a better value to own two, three, or even four LumenPNP machines. By the time I finish loading one and kicking off the next job, the next machine will be finishing up.
Is the Neoden 3V Advanced the right machine to be comparing to? I have absolutely no idea, but it is an inexpensive commercial machine, so it seems like a reasonable thing to look at!
What if you really do need to produce thousands of OoberLights Micro boards?
We can still farm out the work to PCBWay! Just because we can do the work in house doesn’t mean we have to.
If we start selling 100 or 200 OoberLights Micro boards each month, then it would most definitely be an awesome idea to have PCBWay make us a big batch!
Am I going to order a LumenPNP?
The answer to this question is almost definitely yes. It is Thursday as I am writing this. I am pretty sure I was completely convinced that this was a good idea when Jeremy and I interviewed Stephen Hawes on the Create/Invent podcast on Tuesday. I’ve been trying to find a good reason not to pull the trigger.
I haven’t found one yet, but I decided that I should at least wait until Monday. There’s currently a six week lead time on LumenPNP orders, so it will be two months before I get to post a blog titled I Bought A Pick and Place Machine: I Have No Idea What I’m Doing. That will be fun!
Conclusion
I wrote everything above on Thursday. It is Friday morning now, and I am about to quickly reread this before publishing. The more I talk about the LumenPNP out loud, the more convinced I am that buying one is absolutely the right thing to do.
What do you think? Are you running an open-source pick and place machine? Is your machine the LumenPNP or something else? How is it working out for you? Is it as reliable as my Prusa MK3S? Let me know in the comments, or stop by the Butter, What?! Discord server to chat with me about it!