Protomat S103 - LPKF

[LPKF_USA]

My university has an LPKF ProtoMat S63, we've had it for almost a year. Very simlar machine to yours, spindle is a bit slower and maybe some other small differences, but very similar. Here are some of my thoughts on the machine. I'm just a student who somehow manages to get involved with alot of stuff going on in the university, though, so keep that in mind.

We've had some major problems with it, more often than would be desired. We've had what I assume is the same problem happen aswell: when returning a bit to the holder, it misses and crashes the bit into the side of the holder, making a mess of the whole thing. We managed to bend the holder back into form each time it happened, though, and with some help from the local distributor, the problem was fixed and hasn't happened again. I don't recall what the fix was, though, sorry. Not even what was the root cause of that. I can ask another person who was involved with fixing the problem if he recalls it if I see him on skype over the weekend.

Other problems we've had include the spindle dropping the Universal Cutter bit from too high onto the adjustment slope, causing the bit tip to shatter with ~20% chance (that took a while to figure out that that was the problem); drills going too deep; drills not going deep enough; inconsistent universal cutter cutting width across the board and much more.

When the thing works, it is a very nice machine and really effortless to use, though. I've been a pretty active user of it and have milled numerous boards with it (for all kinds of random projects, most not related to studies or any work at all. very random). I gathered a pile of them just now, took a picture and attached it. With the Universal Cutter bit, 8mils accuracy is very much achievable and repeatable. I prefer to use 10-12mils traces just in case, though (we tend to overuse the bits, most of the stuff students here mill are pretty crude and accuracy isn't that important). with the microcutter bit, even better accuracy is probably achievable. It's a really convenient thing, I can walk up to the controller PC with a USB stick with gerbers on it, and in less than 10 minutes the Protomat is milling a board. I can do something else while waiting, flip the board halfway through the process, wait some more and a double sided board is done. Extremely convenient.

We get support from the local distributor and they have helped with the problems. Sometimes the fix has been just to lubricate something and sometimes to increase/decrease the value of some seemingly unrelated parameter in the service menu until the problem goes away. We would have never figured out the solutions to some problems without their help.

Now the machine has been working very nicely for a pretty long time and me and others are finally starting to trust the machine ("unfortunately" I'm also very close to finishing my degree, which means less time with the thing...).

Hi Tuomas,

Those are some great applications! I wanted to let you know we are having a contest where you could win a Raspberry Pi by sharing an application on Twitter. www.lpkfusa.com/myPCB/ for more info

Shane

[chulomex3]

- Single sided boards are straight forward to mill.
- Double sided do tend to be more tricky as the alignment can be difficult when flipping the board
- I did it quickly and tested the circuit.
- I have to make changes so I have not lost a week waiting for the board to come back.
- My machine has manual tool change and I prefer that as I can keep an eye on the machine.

Let's take a serious look at this example ( i will try not to rant and rave... Promise ) ohhh, this is going to be a long one...

1) Single sided boards are indeed 'easy'. Pop in a plate of copperclad and away we go. Setup time is a few minutes. I remember that from my 95S. Drill two registration holes, put the pins in , load the file, let it generate the milling pattern , set home and off we go. Fine

Problems:
- You need to use their board material ( at least with the 95S ) as it has thin copper ( half ounce). They jacked up the prices.
- you need sacrifical material as well , or you drill in the bed of the machine....
- the plates are made for the bed of the machine. Drilling the regholes means you only could use about 3/4 of the available surface to avoid the head running into the registration pins. Everytime you mill a board you have scrap material.

1.  I buy my material direct from Rogers or whoever sells it.
2.  The bed has a replaceable air permeable board material that lasts a long time (about 9 months depending on use).
3.  Machine uses fiducial recognition so no pins needed.

Now, in order to do this particular board : how much extra time did you spend to tweak the layout so it could be made single sided? How much concessions did you have to do in terms of design rules? That also costs money. Answering 'it's a quick proto , the real board will be doublesided' fails you. The goal of a quickturn is to have the real deal fast. If you need to rework it for production you are burning again time and money doing another layout and introducing additional variables in the system. The quickturn needs to be identical to the production.

With camera and fiducial recognition.  Double sided boards are a piece of cake.  No need to limit to single sided boards.

2) double sided tends to be more tricky. Nah, once you got the registration holes drilled the machine software does the rest. But... Doublesided is one step. Metallisation is the problem. Sure you could tweak the layout so you can use stubs of wire or the component pins if thru-hole, but you still could not place vias under a component. And you fall back to the problem that your proto is not identical to your production. Duplication of work , introducing variables, time, money, yaddayadda... And you have to solder all those bridges which costs again time and money... You can buy their plating system (multicontac II) which i also had. Oh boy now we are golden... Ehhh here is how that stuff works:

Prototype and Production are almost never the same.  By the time we're into layout of real production boards we've changed many attributes because of the prototyping.

-

now you need their board material. No escaping. They give you a special substrate , doublesided but each side has two layers of copper. The top layer peels off and acts as a protecting surface for the mill so it doesnt scuff up the layer beneath when drilling and so it wont corrode. You drill regholes , drill all the holes first , and go through the plating process. They use basically the black hole process with a twist. You first go through two nasty chemicals that etch the fiberglass material in the holes. Here is problem 1: there is no agitation. So on small holes air bubbles can get trapped meaning the copper wont stick very well later.... After a quick wash ( darn it now i need running water and a drain and a permit to dump that washwater down the drain....) you get into the black ink. Again no agitation....  That ink has a limited lifespan... And is expensive. You need new chemicals every 6 months, used or not... And getting rid of the old ones is a hassle too. You are not a pcb house that knows what to do. Oh ,but here is -tatadadaaaa- lpkf to the rescue. You can ship it back to us. We send you two big drums , pour everything in there and ship it back. Of course the drums cost you but it makes life easier. You do need a chemical transport though... Good luck finding one. I spent three weeks on the phone with various companies.( this is hazmat)
-what's in it?
-Ehh, dunno. It's from lpkf. Plating solution and some other stuff.
-We need description and chemical names sir for the paperwork.
-Ehh lpkf process is 'secret' they don't release the names of the chemicals.

I use whatever material I find in the lab.  Some stuff around 25 years old and have had no issues.  Again, I also buy direct from Rogers.  I will buy from LPKF for regular copper clad. 1 ounce and .5 ounce.

Where was i ? Ah yes. So we now have black ink. Time to dry the board and peel off the extra copper foils so we have clean , unscuffed copper and holes with dry ink in em. Time to pull out the pocket microscope and inspect each and ever hole.. Dont want any air bubbles in there.. Time keeps on ticking ticking ticking.. into the future....
Plating time ! Woohoo. This has agitation. Oh but wait... How much current do i need ? How long ? Do i need to add a bit of glycol to the plating solution ? What all that muck floating at the surface ? Is my solution well balanced? I dunno , let's wing it and see what happens.... Fast forward an hour or so ( as their plater is grossly underpowered )

Pull it out wash and dry ( permits....for the wastewater). Now, they don't use pulseplating so you do have problems with phosphorisation( or is it sulphurisation.. Can't remember. The deposited copper is of lesser quality than the real copper and it has trapped one of the chemicals elements from  the plating solution. H2s04.. Ah yes sulphurisation it is. That is what pulse plating is about. It strips the sulphur deposit by temporarily reversing current. ( many times a second )

Anyway. We now hopefully ha a board where there is copper in each and every hole .. If there were no air bubbles, if there was ink in every hole... If if if..

Via plating takes about 30min for me since it has to bake in the oven to cure.  I basically squeegee conductive epoxy over the exposed holes (everything else is covered by sticky tape.  Nice and easy)

Milling time. Time keeps on ticking ..... Flip the board ... Time keeps on ticking ... Oh no , it broke a milling bit ! Oh no the software crashed ? Oh no it crashed and milled a line to nowhere cutting through half the traces ... Redo from start.. And you have to use a complete new panel because you need the double foil....

Milling time is usually 10 min to 1 hour for large panels.  Although you are correct on the software crash/bits etc.  Yes it does happen.

In practice: to do this you spend a whole day drilling , washing, inking, drying, inspecting, plating, washing, milling in the hope to have a board. 8 hours.... The shop around the corner, which is a real board shop, can do it in 4... With soldermask silkscreen and flying probe test... Oh wait ! What? Test !!! Darn it, i knew i forgot something. Assuming everything went right and the machine did not mess up , we still need to test our vias! surely we are not going to assemble a board with vias under components without testing it ! So grab your favorite ohmmeter, magnifier and prepare to burn the midnight oil testing each and every via.... Timekeeps on ticking...

The nice thing I end up doing is perfecting my design so when I send it out to the proto-fab house I end up with a close to production prototype board that has all the silk screen/solder mask etc.  Then I continue refining. 

3) i did it quickly and tested the circuit. No you didn't, you spent half a day trying to make a singlesided layout , another half a day futzing around with a machine and assembling a board that was nowhere near the real deal. If you find a mistake you are going to burn another day futzing more with that machine. You self admitted you don't like the toolchanger because then you can keep an eye on the machine. These machines are very addictive. Just like the 3d printers . You assume to shove something in there and let it do its thing while you go off and be productive at something else.. Not! You sit there watching the darn thing. Meanwhile the dollars fly out the window.... And the boss pays ( or if you are a one man shop... You pay.. Through the nose)

I had a 15GHz filter built and tested on the Network Analyzer within 45min of starting the board so it was pretty quick.  (granted nothing went wrong with software/bits etc)

Now, your board has what, 20 parts on it ? You could have grabbed a little demoboard from the cpumaker for 20$ , spiderwebbed some thru hole stuff on it and tested your design equally well. The time it costs you to solder the parts on that pcb is equal to the time soldering the spiderweb. You saved the time on the machine and the time doing the special singlesided layout. So i just shorted all that time and money. If it is a one off : build it on perfboard.

Again, once I got my design down on the s103, I sent it out to the proto-house to get a nicer version.

4) a week ? Any pcb shop out there offer 8 hour or next day turnaround. Live in a high-tech area and you can get 4 hour. 8 hours doing it yourself or 8 hours where someone else does it while you do something else. You surely are not sitting around waiting for the board are you ? There is other productive things to do meanwhile. And in raw cost ... I'm betting the pcb shop will be cheaper, even on an 8 hour turn. Plus: real board with soldermask, silkscreen , full electrical test and surface finish to your liking (hasl, enig, enepig, osp, tinflash, silverflash : you pick)

Making mistakes on my prototype in house boards was great being that I had high confidence that my nicer version of board house boards were going to come in with most errors ironed out.

-i can keep an eye on it. If my boss catches me wasting time staring at that machine he'd first fling it trough the window and then me through the door.

Sounds like your boss needs to work on his management skills.

[chulomex3]

PS  I'm with Free Electron on a lot of points he brought up.  This machine is far from perfect but there were some details that are no longer an issue that I felt compelled to dispel.

[Sabesto]

Just for future reference for others having problems with this machine and finding this thread, I’m posting a few problems and solutions that I’ve encountered with the S103:

System: S103 with automatic tool change and vacuum table

- Drill bits breaking often / wearing down fast:
Clean and lubricate the chuck. We found out that if this is not done regularly it will shoot the bits out of the chuck when calibrating the tool height (which it does whenever changing tool). When hitting the sloped calibration surface it will shatter the tip on the bit, this cost us a lot of them.

- Material moving:
We tape down the material as much as possible, at least on the last layer before it routes out the board. The vacuum table is often not good enough to hold the material when routing out / removing large areas of copper. Also, if you choose to rub out large areas, keep in mind that the head slides along the surface. If the whole head moves over a rubbed out area it can snag when hitting a copper edge, ripping it off the track or moving the material.

- Bits touching the surface where it should not:
The replaceable vacuum boards vary *a lot* in thickness, this is not a huge problem as the head rides on the surface. However, it seems like the play in height difference the head can tolerate is less then the difference in thickness of the vacuum boards. We measure all 4 corners when replacing a vacuum board, take the average and adjust the "vacuum table height" setting. it is unclear how this should be measured as the manual does not mention it, but it seems to be from the aluminium surface the vacuum table is mounted on to the top of the replaceable material.

- Layers out of alignment when using camera to read fiducials:
Not figured this one out yet, even when calibrating the camera head offset every time the machine is booted up its still misaligned.

All in all its a good tool to have, but I would not recommend anyone to buy this with the intention of letting many different people use it. With all the hiccups and bugs it should be operated by one user only.

If anyone from LPKF reads this, I have a lot of tips regarding improvements to the system (just listing a few, please PM me if you are interested in speaking to me) I have spent so much time on this machine, I don’t want to spend more time rewriting this and sending it to LPKF:

- If you haven’t already, the manual requires a lot of work, this would also save you a lot of phone calls.

- Add an emergency stop button to the software!

- When you flip the board and the machine tries to find the fiducials, which it never manages to do automatically unless you lucked out and flipped it perfectly, let the user select if he wants to direct the head manually before having to wait for it to fail the automatic recognition which takes time.

- When calibrating milling width with the camera, let the user manually measure the width. Why? read the next problem

- You should have a picture from the camera with optimal settings as an example to go by when adjusting the image settings. When measuring milling width the machine more then often finds the wrong edges. What looks good for a human does not work at all for the recognition software it seems.

- When routing out boards, the contour router sticks out the exact same depth every time, causing it to be worn out just in that point. This should be fixed so that you could either manually or automatically vary this depth.

- In the "Board production wizard" why does the machine not switch to camera reference when setting up the material area? I still forget to change to camera as reference and not drill before I start the wizard (or at least let the user change this while the wizard is up).

- Let the user pause the machine when using the wizard as it really should be supervised at all times when milling.

- Explain how to remove the dummy bit that is in the chuck when the machine is delivered, the program says no bits are mounted, we where afraid that it would try to pick up a new tool with the dummy bit already in the chuck. Also, add an option to manually release the chuck.

- In general, the machine does a lot of things automatically, like releasing the dummy bit on first tool change, many things like this is not specified in the manual, and because of all the hiccups and bugs I would like more manual control in general. An option to reduce contour routing speed for instance would be nice as it could reduce the risk of moving the material.

- Tell me how on earth the solder paste dispenser is suppose to work (granted, the last time I tried this was one year ago, maybe the software has been fixed). Machine said 0,X mm nozzle was mounted while we only had the 0,Y mm nozzle, I found no way to change this

[Scribe]

Hi guys,

We're looking at purchasing the Protomat S103 for R&D work, aside from its limitations in routing 0.5mm BGAs, I've been generally impressed with what I've seen, we require accuracy (some RF) that I just can't achieve with non-commercial etching and the simple process to multi-layer is also helpful, however reading this forum post has opened my eyes to the fact it may not all be so clear-cut.

A year later, I would be interested in knowing if LPKF has resolved the problems highlighted in this post and what issues are still outstanding.

I would greatly appreciate any feedback you can offer.

[nctnico]

If you are into RF I suggest using a simulation tool like Sonnet. From my experience the simulation results are spot on compared with real world PCBs. And even if you can mill to tight tolerances the production PCB will be less accurate. Making multilayers sounds very clumsy to me. Unless your time is free getting an LPKF seems like a waste of time and money to me.

[G0HZU]

I've been using T-Tech and LPKF machines at home and at work for many years and they can give very impressive results. However, you do need to have training and experience to get the best from them or you will end up confused and bitter like free_electron and you can also easily damage the machine through inexperience/incompetence.

It's true that in 2014 other PCB prototyping options are now far more attractive for the majority of people but for rapid prototyping for RF work these milling machines are still hard to beat. You can also make front panels or laminated tools and even mill flexi cables with them.

Sonnet is very good and I use it a lot at work but it isn't ever going to be a complete solution because you can't easily simulate complex 3D enclosures and screens with it and it can't always model lumped components properly. So the rapid prototype from the T-tech or LPKF machine that you can hold in your hand and test is still very valid here. Especially if you mill a filter and you want to tweak it and have it ready for a presentation at a design review the same day.

These machines aren't for everyone, and the group of people that will get benefit from them is getting smaller every year because of other (mostly non RF) PCB prototyping options that are available at lower and lower prices.

If I thought they were as hopeless as some people here would have us believe than I would have got rid of both of my machines by now and the one at work would never see any use. But in fact, it gets used a LOT at work. Only one member of staff is allowed to use it because if you let untrained people use it then you end up with poor results and possibly a damaged milling machine.

[mazurov]

What is the reason to spend money on these machines? Here is a closeup of a board I just made on a regular non-dedicated mill (I mill aluminum on it, as well as plastic, wood, whatever ) with Bosch Colt palm router as a spindle. SO08 package with 0603 passives. With some difficulty, I can mill a trace between 0603 pads. Milled using eBay-sourced generic bits/FR-4, affixed to the base with a carpet tape; better resolution can be achieved using auto-leveler.

A setup like mine can be put together for less than US$800 - and this thing can do so much more than mill PCBs.

[G0HZU]

What is the reason to spend money on these machines?

In terms of value for money for the home user it's best to buy a used LPKF or T-Tech milling machine rather than a new one.

I bought my first T-Tech 7000S machine over 10yrs ago for about £800 including tools and PCB materials and Nilfisk extraction system and it can mill over a 19" x 13" area.

The second one was even cheaper but was faulty (needed new spindle bearings)

I've milled flexi cables with the T-Tech 7000S on double sided PCB material that is as thin and flexible as paper.