Many of us may qualify as “makers,” but how about a “maker of machines?” [Danielle Applestone] tells us what kinks to look for whilst embarking on your hardware startup adventure. Co-founder of Other Machine Co, the company that makes a PCB Mill that holds tolerances as tight as a thousandth of an inch, [Danielle] holds degrees in chemistry and materials science from MIT and UT Austin. While she may tell you that the math for running a hardware company is easy, knowing what numbers to crunch and keeping track of them has been part of her key to success. So take 20, and give yourself a moment to take in [Danielle’s] tips from her Hackaday Superconference talk on beating the hurdles ahead in the land of hardware startups.
Small Co; “Big” Data
Like many startups, Other Machine’s encroaching launch date pressured the team to release their product into the wild untested. Denied the luxury of time to find bugs, OtherMachine pushed onward with a clever scheme that enabled them to learn the most from their users: shrewd logging of production manufacturing data for every mill ever sold. Who assembled the mill? Check. What suppliers provided this mills parts? Check. What version of the assembly instructions that were used to assemble this mill? Check. With a detailed cluster of metadata, Other Machine can track down and even predict problems in specific batches of machines based on faults reported from a single user. With so much data at their fingertips, they even developed an in-house app to facilitate the process of logging mill data.
Armed with this data, launching an untested product isn’t an aimless leap into the unknown. It’s a well-timed jump with hooks set deep into the process that enables OtherMachine to learn from their products while they’re in the field.
Customers as Royalty
With a field full of untested machines, OtherMachine’s Kickstarter backers became the first guinea pigs at launch. To complicate matters, the OtherMill morphed in design as they were being sold. Parts were exchanged, modified, or added as the team continued to iterate on the design. To supplement the machine’s untested history and continuous shape-shifting, the team at Other Machine was ready to leap onto the customer’s problem at a moment’s notice. Their biggest asset? The fine-tuned tracking of each machine’s history from their extensive logging infrastructure. Sure, machines may have different bills-of-materials or CAD models, but because Other Machine can tell the story in great detail for each of their machines, they can provide the immediate, personal support for the machines that failed in the wild.
A Community-Tested Triumph
With the clever foresight to log the story of each mill, Other Machine Co. ran–and continues to run–a product-wide experiment to produce a PCB mill that works not just on their bench but also the bench of each of their customers. For potential buyers of the future–fear not! The OtherMill is a crowd-funded and community-tested success, and, for the truly skeptical, [Matt] gives us the inside scoop on just how shiny of a gem this tool really is.
the mill is cool and all, but they should sell the tracking software they came up with
Am I the only one concerned about privacy?
Yep.
From the sound of it, the machine doesn’t phone home – the tracking is of the assembly process and the data is stored by the company, not on the machine. Then when a customer calls in with a problem they can pull up that person’s mill and see exactly what was done to it.
Oh, now I understand Enrico’s comment.
Yeah, the data that is collect is “production data” in the sense that the machines are produced at the factory and that entire process is logged. It is not data collected when the machines are “in production” meaning “in use”.
I’ll work on clarifying that in the body of the post. Thanks!
As best I can tell her company isn’t doing anything that any other manufacturer is probably doing if they offer service and a warranty. I don’t believe the data is available to the public as in that this is how Joe Blow’s mill was manufactured is available to someone outside the company.
My impression of the Othermill is that it’s cool, but not $2200 cool. But then again I haven’t done a detailed comparison of its cost against sending your designs to a fab house. Could somebody weigh in on whether the Othermill can be a cost-effective thing for small production runs?
If you have a pcb mill at home you can have boards in the same day and iterate faster
Yep, it’s exactly this. Failing faster instead of finding out you screwed something up 10 business days later.
And sometimes you get it right the first time and suddenly you’ve gone from idea, to design, to finished working reliable board in the course of a day.
and sometimes people win the lottery. If you have no idea what you are doing and just throw random components at a board it might be viable strategy, for everyone else doing 10 half assed prototypes instead of one thought out and properly made, is just silly
Nice straw man you have there, it would be a shame if something happened to it. Ah, I see you already torched it…
If you iterate faster, then you are not ‘throwing random components at a board.
You do the exact same design effort, only you don’t have to wait 10 days to find out what you did wrong. Perhaps if you spend longer on your design you will catch your errors before they reach the hardware stage. That sounds good in theory, but in practice it turns out the two diverge; only those without the capability of self reflection will insist that they can catch all their errors without testing…
It really depends on what you’re doing. For instance, a PCB mill is *awesome* for getting form factors right, connector positioning, and confirming footprints of things. Sometimes you can’t get all the dimensional information you need from datasheets.
It’s also fantastic for “widgets” – connector adapters, simple microcontroller stuff, etc.
Of course, it doesn’t hurt that I’ve got access to a $25k sub-mil precision mill, but I was definitely looking at Othermill when we went to buy things – unfortunately crowdfunding pre-order wasn’t an option.
Yes, well, one can get two or three other (read: Shenzen-made) cnc mills for that kind of money, which are guaranteed to be able to do everything this mill does, equally well (clue: a thousandth of an inch is 0.0254mm – a tenth of a millimeter is called “backlash” and a quarter of that is called… nothing special, even in the hobby cnc market). Yes, that include PCB production as long as you keep expectations realistic (if you want to do 0.5mm pitch, the nominal resolution of your mill is going to be the least of your problems)
What does “holds tolerances as tight as a thousandth of an inch” even mean? Does it not hold these tolerances sometimes? Are you on their marketing team Vasquez?
Sometimes is correct, tolerances will vary based on technique. If you hog material out using a rusty endmill you bought at a yard sale, you may only hold ten thou. If you have a new endmill and patiently cut progressively finer finishing passes you may get accuracy of up to a thousandth. I have never used an Othermill, but that’s what someone means when they say a CNC mill holds a certain tolerance.
we use the othermill(s) at adafruit for the PCB testers we make and use here! there is a case study – https://othermachine.co/stories/adafruits-ecosystem-of-innovation/
(we also have them as something we stock now too)
I own one and they’re pretty cool. If you need a board now, or want to iterate before sending out then it’s ideal.
10mil traces and double sides boards are reproducible. Have got down to 1mil traces, but at that point the copper
delaminates from the board once you try to solder it.
Is that delamination with an iron, reflow, or both?
I’ve seen the copper delaminate using a hot air station. Haven’t done it with an iron but I design with larger traces since they’ll have more adhesive keeping the copper stuck down.
Soldering iron @ <10mil. Reflow <10mil works. But this is only if you're going less than 10mil, it's due to it being FR1 instead of FR4, adhesive and substrate are a little less tolerant to heat / mechanical stress.
It's recommended not to use FR4 because you don't want that fiberglass in your lungs. You can achieve less than 10mil with other materials, just for PCB's that's the practical limit.
I don't chemical etch since I got the mill, but when I did, 10mil wasn't perfectly reproducible. You're not going to be doing large boards for production runs with it, but where it's a small double sided or single sided prototype or a one-off and you don't want wait a few days, or go through a few iterations, it's ideal.
The bottom line is it accurate enough and reliable enough to be useful, and generally quicker than a chemical etch.
Yeah, they know exactly what build of machine you have.
So if you tell them you have machine #061 they know your machine is crashing into the bed because they hadn’t started shipping machines with sealed switches yet. And that you’re out of warranty. $100 service + $95 shipping
Then, two weeks after you get it back, the spindle will seize and the drive belt will melt into the pulley. Of course their tracking lets them know that your machine was built before they started shipping sealed spindle bearings. $250 service + $80 shipping
Of course you won’t even bother unboxing it after you get it back the second time. You’ll have bought a $690 ball screw water cooled mill off of eBay and been happy.
Sorry, if that was intended as parable of sorts, it doesn’t make sense. Who after spending $525 on the machine above purchase price wouldn’t unpack it and put it back to work, even if an alternative mill had been purchased? One might not send it back for service a third time or never patronize that company again, they sure as hell are going to use it up to the degree possible, if they can’t sell it. Either way it’s going to be unboxed. Then again many comments to Hackaday show a lack of business or even personal finances sense. [shrug]
Hi Eliot,
We definitely make mistakes. After reviewing your case in our support system, I want to accept responsibility here and apologize. We should not have charged you, as our policy is to repair all known problems for free when they cause failures. These known problems rarely cause failures, but yours totally did, and this should not have been on you to cover. Your experience does matter to us, and I’m sorry you had to go through this.
An Othermill sitting in a box is the saddest outcome, but I know you have already spent a bunch of time trying to work with the tool and not gotten much value out of it. If there is still a chance that we could earn back your faith in our machine, I’d like to re-open your support ticket, correct our mistake, and work with you to get you going with your tool. If not, that is okay too, I completely understand, and we won’t contact you. You can also message me directly. I’m danielle at othermachine.co as you might have guessed.
Again, I’m sorry we did not provide the right level of service to you, but hope we can have a chance to correct it.
Sincerely,
Danielle
I do like that response!
I don’t, not really. He’s all ready had to RMA twice and deal with their customer service at a significant portion of the total unit price. If this oversight of customer service was a one off problem then it should have been caught the second time he sent it in for part failure. The fact that he is now getting redress after posting his apparently true story on a highly public forum doesn’t really tell you anything about the company except that the people running it aren’t stupid. Meanwhile this is a Hackaday story lauding their decision to not perform testing and quality assurance before shipping what we now know was a defective product with at least 2 critical design flaws.
Hi Eliot,
We definitely make mistakes. After reviewing your case in our support system, I want to accept responsibility here and apologize. We should not have charged you, as our policy is to repair all known problems for free when they cause failures. These known problems rarely cause failures, but yours totally did, and this should not have been on you to cover. Your experience does matter to us, and I’m sorry you had to go through this.
An Othermill sitting in a box is the saddest outcome, but I know you have already spent a bunch of time trying to work with the tool and not gotten much value from it. If there is still a chance that we could earn back your faith in our machine, I’d like to re-open your support ticket, correct our mistake, and work with you to get you going with your tool. If not, that is okay too, I completely understand if you don’t want to deal with it, and we won’t contact you. You can also message me directly. I’m danielle at othermachine.co as you might have guessed.
Again, I’m sorry we did not provide the right level of service to you, but hope we can have a chance to correct it.
Sincerely,
Danielle
This is very late but I’ve had similar issues.
I just replaced my bearings with F688-ZZ Miniature Flanged Ball Bearing 8x16x5 MF688ZZ RF-1680-ZZ. They were $6.55 each via a reputable online bearing supplier.
I replaced the pulley with an aluminum one. For my early machine its a OD 20mm, Bore 4mm v groove pulley with fixed screw. It was $7 on eBay and you should be able to find the same via that description.
Some careful application of a can of compressed air has fixed the x axis limit switch for now. I can’t figure out how to get it out of the assembly to replace it with a sealed switch.
Definitely an interesting watch/read and glad it’s taken off :)
I seriously considered an Othermill because it really does look like a competent little machine. If they could have managed to get it into the sub 1k range I’d probably have one sitting on my desk this second.
sub $1k would be ideal. J bought a C NC router for $1110 from Craig^ – . It’s custom and very precise. But, getting it to route to 10mils is a challenge. 20krpm and higher is ideal. For somebody for a finished product ready to go, will call the $2200 a investment.
There’s no reason for a CNC circuit board mill in this day and age, if there ever was one to begin with. The problem is with physics, in that it’s extremely hard to produce fine traces with milling bits. These machines are also horribly slow, noisy and produce hard-to-handle dust. The glass fibres in FR-4 wear down the bits quite quickly, and relegating to lesser materials (FR-2) will alleviate this problem, but lead to lots of others when it comes to soldering and rework.
A nice etching setup can be had for a tenth of the price, and it’s just the way to go. You’ll be able to go from CAD to circuit board in about half an hour to an hour, without noise, without dust. The chemicals involved are relatively harmless, so basic chemical lab safety rules will be sufficient to ensure that no damage is caused to you or the environment. Of course, you still need to drill holes by hand, which sucks, but with a drill press, SMT construction and some good-quality drills, this is not much of a problem.
The biggest disadvantage of pcb mills is the precision. With an etching process, you can get 12 mil any time without even trying. 10 mil works flawlessly. If you know your process, 8 mil trace/space can be done reproducibly. And if you have optimized your process parameters and know how to design, 6 mil or less trace/space is possible. With milling, you might *just* be able to do 0.5 mm pitch with a highly optimized process, which is somewhat “jellybean” these days for modern parts. So, just not worth it.
I have to agree with you here. 2200$ buys you quite a few boards from a really professional pcb service, with all the most up to date specs like <0.15mm trace (6mil), multilayers, masks, ENIG plating… delivered on time within a few days. In the end, it probably depends a little bit on what you are trying to do. We do still have a little setup to etch boards for simple designs, but it gets hardly any use these days, because if you calculate the time involved in setting it all up, printing masks, etching, drilling, contacting vias… we use way less valuable engineer-time if we just send out our Altium-pcbdoc to our trusted manufacturer and do something else until the finished boards arrive.
Soldering the prototype on a production run board will be easyer and faster too, because we can just order the laser-stencil with it, spread some paste, place components and reflow. Not to mention that the factory sourced prototype, if it works, can be furter used to verify design for manufacturability and even do some preliminary compliance tests and stuff like that too, because it's already very close to a production run board.
People always bring up the argument that quick in house turnaround boards saves you a lot of time, because the designs they do seem to be always full of errors in the first runs. Why don't you do your research, invest some time to actually validate your circuits before you even start to lay a single track on a pcb? This is usually way cheaper and faster than building a ton of prototypes that don't work and trying to figure out what the heck could be wrong with it…
+ X 1000
If diode pumped fiber lasers ever drop in price, milling PCBs will become obsolete. Until then however…
Probably just me, but the guy introducing Danielle in the video should have rehearsed a bit or give up being an mc… Great once it got going though, keep up the good work Danielle!
We are lucky Chris crawled out of his basement at all! ;o)
The video production was very good quality, but I think they should have spent 50% of the time (or more) showing the slides, and splitting the remaining time between showing the stage, and showing the speaker. Alternatively, show the speaker in one corner overlaying the slide. It takes time to absorb the slide, but watching the speaker doesn’t convey much information (shifts weight between feet, smiles, waves arms, etc).
I really despise the term “maker” or “makers” – just hearing those words make me want to do someone physical harm.
Hobby and hobbyist were perfectly fine and descriptive enough words.
Dale Dougherty can go piss up a rope.
And oh yeah, the article itself. Board Mills… Meh. We have a really expensive “do it all” mill here at work and it’s great for something simple, as in a single layer board… But it’s very rare that we need just a single layer. Then registration of the other side becomes a headache. But the worst thing is fiddling with a bunch of little rivets for vias, or stuffing wire through and soldering both sides etc. A couple of holes, no big deal. But again, the need is always for way more vias than a couple.
I wouldn’t waste my time. There are board houses that are fast and cheap enough to make board mills un-needed.
Vias are quickly done by getting some uncovered wire, soldering at one via, and then threading the other end through a number of vias, then quickly soldering. Real simple and quick, 50 vias can be done in 5-10 mins. There are other solutions too, but that’s the fastest / easiest / cheapest.
Regarding solder mask (or lack of it on the othermill), what you find is that because the space between traces is cnc’d, it’s a little below the surface, so solder tends to behave itself and not bridge.
FYI, I’ve had no problems with my othermill yet, except a few software gotchas which I’ve been provided workarounds for (their support is good, a rarity these days).
There are those who have the othermill, and those that don’t. Those who consider it to be too expensive /
not practical have either not tried it, or are not doing enough volume to make it worthwhile. A $2200 tool that can accurately knock out a prototype in hours, or create a batch of beta’s is a pretty good deal when time is money
Soooo….
How to save money as a startup by pushing your Q&A costs onto customers.
Why should startups be at a disadvantage to all those more established companies that push QA (not Q&A) costs on to customers?
Not to mention that customers pay for QA one way or another.