DIY Space Grade PCBs

3D print circuits using silver ink and ceramic substrates to make high performance, multilayer, fine line < 3mil, PCBs

Similar projects worth following
Low temperature co-fired ceramics (LTCC) is a well established technology for making many of the electronic components we use today. Ceramic capacitors and resistors for instance may be made using LTCC technologies.

The LTCC process of constructing a PCB is remarkably straight forward and cost effective. A flexible sheet made of alumina and glass powders held together with a binder are (1) first drilled and blanked using a CO2 laser, the vias from the drilling process are backfilled (2) with silver ink using either a screen or inkjet printer. The conductor traces are layed down (3). The various layers are then (4) stacked, pressed together (5) using a hydraulic press, and then (6) fired in a furnace. The result is a very hard ceramic substrate, with virtually perfect electrical properties extending well into the microwave, inert to water, high thermal conductivity, and extremely high dielectric strength exceeding many KV/mm with great solderability and wire bonding properties.

The DIY Space Grade PCB is the first of 5 projects that comprise my entry to the 2016 Hackaday Prize. The ultimate goal for this very involved project is to demonstrate my solution to one of the most difficult problems in all of engineering, that of producing a fully functional and field operational Neural Interposer, Gibson's Cyberspace Deck, aka Brain Machine Interface, not surprisingly it also makes a very capable device for Molecular Biology.

Project 1 outlines the ceramic PCB process.

Frequently I find that the problems I like to work on require very complex PCBs with buried and hidden vias and even vias in pads. With vias in pads comes 3 mil lines. Just for fun lets make it a 12 layer board. To find a manufacture of such a board my choices are now reduced from hundreds perhaps thousands of board shops around the world to a countable handful. The cost of a single prototype went from a few $ per square inch to many hundreds of $. I could expect to pay several thousand $ for just a few boards with a lead time of several weeks. Well, one thing is certain, I'm not going to pay a board shop thousands of $ for something I can do myself!

OK, if it's so great why isn't it used more often? The answer is shrinkage. When ceramics are fired there dimensions change. There are ways to keep the shrinkage under control but they are uneconomical and expensive in volume so typically 99.9999% of the time FR4 will do.

At least for now, I'm content to know that the technology of LTCC can provide the boards I need. I don't expect to get to the highest levels of performance immediately, but I know they are obtainable. That's why it's called the art of the hack.

The ability to fabricate circuitry with high tolerance is limited by the precision of the mechanical components of the system, the stiffness of the structures, vibration control, step size of the stepper motors, pitch of the screws, and many 2nd order effects.

To reach the goal of producing multilayer PCBs it's first necessary to place vias in the material, deposit ink within the holes, and lay down as many layers of silver ink as necessary to reach the desired conductivity.

Rather than use multiple machines for these operations I've elected to build one machine. I call it a 3-D fabricator.

My 3-D Fabricator is a big project. It sucks up money, time, and resources like a black hole. I even remodelled a room in the house just to hold it.


Except for the instrument rack that I've had for decades the other major pieces of hardware, like the laser on the left side and the X and Y traverses have all come from EBay. Just below the slotted bed on the left you can see an air exhaust hose and on the right, a water pump and radiator for cooling the laser, the blue container is the water reservoir for the cooling system. Just above the slotted bed is the X axis traverse with a multifunction tool head that has a laser focus assembly, and ink jet printer head, camera and Z axis traverse along with an Arduino and a BeagleBone Black. This assembly is called the short bed. The plywood above that is a mechanical mock-up that duplicates the basic mechanical dimensions of the "long bed" that weighs several hundred pounds.

A closer view of the laser/printer head. The plastic hose supplies a constant source of air or an inert gas for protecting the optics. In the center is the print head assembly and further to the right the camera hidden from view along with the Arduino. The Beaglebone Black is dismounted.

For illustrating scale, here is a stack of "green tape", the ceramic material of interest. The material is from Hereaus. Each sheet costs about a dollar per square inch. The ink comes from ElectronInks.

This is the working end of the laser articulation. It takes 8 mirrors to get the beam down to the work. The assemly to the left is another stepper that controls the Z-axis elevation of the laser so that it tracks the z-position of the main tool head located on the short bed.

This machine is so big that it...

Read more »

  • 1 × Heated Isostatic Press Method for compressing substrates to 3000 PSI@ 70 degrees C
  • 1 × Ceramic Furnace capable of reaching 850 degrees C Dental furnaces seem to work well
  • 1 × Precision Screen printer Registration needs to be about 1 mil or better, with screens of 350 lines/inch
  • 1 × Precision via punch or laser cost is the inverse of via size!
  • 1 × Tape stacking fixtures Alignment of the layers ultimately define the minimum line width and via size

View all 10 components

  • One mans junk is anothers diamond in the rough!

    Chuck Glasser10/16/2016 at 07:13 0 comments

    Just a few days ago I awoke with the thought that I needed to check Ebay for bargins: It didn't take long to find it.

    One of the realities of dealing with all ceramics is that they shrink when fired. That is especially critical when there are many layers with carefully placed traces and vias interconnecting the various layers.

    In general there are three methods to control shrinkage. Each method tries to minimize shrinkage in the X and Y axis and confine material changes to the Z axis.

    Heraeus, HL2000 "zero shrink" tape is made up of uniformly sized grains of material say 8 to 12 microns. This methods gives a shrink factor of much less than 1% in the X & Y whereas the Z may change by %20, whereas one might expect to find 15% in X & Y with ordinary tape in which the particles are whatever came out of the ball mill. The second method is constrained sintering in which the green tape is glued to an alumina substrate and fired. Electro Science Laboratories provides materials for this method. The third method places a weight upon the surface of the tape while it is fired. That is what this pressing furnace is designed for.

    You can probably guess why it was such a great bargin! No matter. An ATMega or Beaglebone Black, display, a few relays and sensors should make it right in no time.

  • There is great rejoicing in the village, Persius has slain the dragon!

    Chuck Glasser10/07/2016 at 19:14 0 comments

    Before your eyes is the weapon that slew the dragon! A little nicked from the hard bones. Four layers of tape with really thick, "too thick", silver ink. The circuit layout used 5 mil design rules. Most of the components are 0201 size.

    From here on, it's practice, practice, practice, and lots of process development.

  • The hard parts

    Chuck Glasser09/23/2016 at 05:58 0 comments

    After the ceramic green tape is blanked and via punched via a laser, printed with inks, the various layers stacked on a fixture it is necessary to laminate the layers together via the isostatic press.

    To laminate the tape, the various stacked layers are placed in a pouch and vacuum sealed. So, I thought to myself, why go to the trouble, "not really that much trouble", of vacuum bagging when I could just manually press the pouch free of most of the air and rely on the extreme pressure of the hydraulics to overcome whatever air is present within the pouch? Wouldn't that be enough? I think the answer is yes. However, there is a problem.

    A few layers of tape were cut out using the template at the lower right,


    and sealed in a bag which was then sealed.

    The pouch is trimmed.

    The pouch is then placed within the isostatic press

    The tape stack is inside the piston cavity along with the hydraulic fluid at 0 PSI and heated to 70 C for a hour or so until everything has had time to get to the pressing temperature.

    Then pressure is applied for around 10 minutes. You can see on the gauge that the pressure is around 4,000 PSI.

    Here is what I found after taking the tape stack out of the press.

    Hydraulic oil in the bag. Dang, and double dang. Note that the tape does not appear flat. That has nothing to do with the failure of the pouch, it just the nature of the details of this step.

    The stack was cleaned with a towel and placed on a quartz substrate. Not even close to being flat!

    The tape stack was fired anyway.

    You'll note that the quartz fused to the tape. That's a surprise! More importantly, note that the tape did not flatten out under the influence of of the firing temperature of 850 C.

    Brittle, fragile, useless junk!

    Oh, what to do?

    Fundamentally, the reason for vacuum bagging the contents is that a bag that has a good seal is visibly different than one that has a leak. Now I know.

    Alumina, Al2O3, otherwise know as aluminum oxide is a fundamental industrial material. Almost as hard as diamond. Very, very refractory! In one of it's many forms it comes in sheets. Very flat sheets. Along with making up 50% of composition of the LTCC green tape.

    Considering the materials extreme hardness, how exactly is it conveniently cut?

    The waterjet works brilliantly! It takes about 55 seconds to cut a 25 mil thick plate with a 2.25" circle.

  • Registration

    Chuck Glasser09/07/2016 at 06:01 0 comments

    So the essential detail in building a system that can produce working circuits with 3 mil design rules is keeping all the various processes registered. That means that if a hole is laser drilled at X = 3.455685, Y = 5.343366, and later come back to via fill that hole, the fill operation is at the same coordinate, and so on.

    That means that any artwork needs to be made in such a way that lines of ink end up the same way. Always located at a precise location corresponding to all the other precise locations. In other words, all the various processes have the same reference point.

    What this means is that even though the ink jet printer may not be used to print silver ink, it serves very well as part of the mask making process for the screen printer.

    To review, there are only 6 operations in making a LTCC circuit.

    Blanking, via filling, circuit printing, stacking, pressing, and firing.

    For blanking either a CO2 or 455nm laser may be used, or a drag knife, and/or a hot drag knife. For minimal residue, a knife seems best. For holes, pulsed laser.

    For vias and lines we'll use a screen.

    The screens will be made by first printing a registered transparency onto the vacuum table. The Ink jet is used to print the pattern.

    A registered screen is then place on top of the transparency and the artwork is exposed from below or more practically, simply removing the transparency and registering it to the screen frame.

    The key is that everything is registered to a set of pins on the CNC table. Tape, artwork, screen printing frame, laser, camera, cutters, etc.

    In this case the 6 holes at each of the hexagon apexes will be used.

  • It's a horse race

    Chuck Glasser09/05/2016 at 04:02 0 comments

    I've just posted a production order form DirtyPCBs for my ADS1299 amplifiers. Pay for the DHL services and you'll have your boards in no time!

    So I need to make some electrodes that mate up to this pattern.

    Seems simple enough. Only a few pads are active (12), but which ones? The other pads are placed to preserve planarity.

  • A new press !

    Chuck Glasser08/01/2016 at 02:45 0 comments

    This project didn't start off with a big huge press. Time to go back to basics. I had started off with this Short Jack, 1/2" of lift rated at 9,250 psi, with a piston 3" in diameter. Actually perfect for my needs.

    When I initially tried to change the hydraulic fitting it only stripped in the socket wrench that was perfectly sized. So, I got frustrated, and lost interest. Better to spend tons of money, and time on a boondoggle.

    With the big press down for some considerable time, and needing to move forward, I took a wrench to the offending fitting and welded it in place. The fitting now came off quite easily, with only a bit more torque that would previously strip the fitting. No tool, the wrench, has ever been so enthusiastically destroyed.

    Even more important, these jacks are really inexpensive! New parts cost, jack, hoses, fittings, hydraulic pump. Around $250. This seriously improves the odds of overall project success!

    Also, of some considerable importance is understanding why it is necessary to use a screen when printing the conductors. The simple answer is that with a screen in place the ink can only go in one direction, Z.

    Here's the new press . Isn't it beautiful!

    Nice and clean. So simple. This is my first, and really positive experience, using perforated metal. Now when oil leaks it will simply drip through onto the drip pan below.

    You'll note that the pressure gauge indicates 3000 psi. Oh happy day!

    There are three valves visible. The silver handle sources the piston to the hydraulic pump via a check valve. The check valve allows the chamber to be pressurized without the pump running. Very important.

    If the silver handle is thrown in the other direction then the chamber releases the fluid to the low pressure side and will flow back into a reservoir where it may be filtered and returned to the pump. One line goes to the bottom of the reservoir and supplies oil, the other is at the top of the tank and provides air. So by pressurizing the reservoir with low pressure air either air or oil may be forced into the piston, or my manually pressing on the piston push hot oil out of the chamber and refill it with air to retrieve the sample.

    In this image we see the heater block running and the system at pressure.

    The industry standard seems to be 3000 psi at 70 C for 10 minutes. That's a big block of material, with the thermal sensor located in the top right corner of the gray aluminium block. As it turns out if the system is pressurized to the working pressure of 3000 psi and the pump shut off, then the check valve will maintain pressure in the line. Thanks to Boyle's law, any subsequent increase in pressure is due to a change in temperature. When the pressure stops going up, in this case by almost 1200 psi, we know the fluid and sample have reached the desired goal.

    Other than a few more sea trials to check for leaks I think we can safely put this one to bed! Isostatic press, done! Most importantly, done safely!

  • Never tell your wife you took a shower in hydraulic oil!

    Chuck Glasser07/08/2016 at 01:55 0 comments


    If you look at the big orange machine and study it a bit you will see that the top hydraulic cylinder applies the clamping force at one point. The image of course being that the force holding the top seal in place is along the centerline of the hydraulic cylinder.

    After taking several showers in hydraulic fluid as the seal gives way I can guarantee that the clamping force is nowhere near the centerline of the cylinder. In fact the present clamping mechanism is a disaster that can NEVER work reliably and SAFELY. So it will have to be rebuilt. Not necessarily much bigger, but certainly a lot smarter.

    I'm thinking a set of 4 linear slides at each corner of the top plate that constrain the movement of the top cylinder cap for movement in ONLY the Z axis.

    There are other issues. When the top cap is lowered into place to form a seal it is under no load. Then, as the pressure is raised, the load that was in place to hold the top and bottom in place is reduced.

    To be truthful, it's a bit scary! Two hydraulic pumps going chuga, chuga, big "I beams" bending. You know, just your ordinary day at the mill.

    So, once the top cap goes on, it needs to be preloaded such that it never moves.

    On the ceramic green tape processing here is what I've found.

    Green tape, the ceramic tape before it is fired comes from the supplier on a mylar carrier film. The tape itself may be supplied in various thickness fro 1 to 10 mils. In my case the tape is 5 mils thick. It has the mechanical properties and tear resistance of wet tissue paper. When the CO2 laser hits the material it chars and fuses to the carrier that may have also burned away, or not. The material itself is far cleaner when a shield gas such as Argon is used.

    If I had my own formulation of binders I would use acrylics as they are know to sublime without residue, but I don't and so I'll have to adapt to the needs of the materials I have.

    After laser cutting the tape is stuck to the carrier film. If the carrier is removed and the tape cut directly on the backing plate, today I used a ceramic kiln support plate 7" on a side square, the tape will be stuck to the support.

    It's looking very muck like a cutting plotter and a via punch should be two of the stations on the large bed CNC machine!!!

    It is essential that the blanking and punching operations not produce a situation that will later induce an operation that will change the dimensional properties of the tape. This is not presently the case. So, while slipping a sharp edge spatula between the tape and carrier may seem like progress, it really isn't.

    Registration of the multiple layers of tape requires a fixture. Experiments suggest that staking the various layers onto a platen press with registration pins may be a good solution as alignment and pre pressing in a platen press may keep the layers together sufficiently for later pressing in the isostatic press.

  • The gods of electrons must be looking after me

    Chuck Glasser07/05/2016 at 03:41 0 comments

    Decades ago I picked up a big drip pan. I have no idea where it came from. It had been stored numerous places over the years, outside in the weeds, in the garage, stacked with lumber. Way to handy to toss, and yet for all the many decades I've had it, it was never used. Until today.

    Apparently, when a hydraulic seal is doing it's job, there is a significant friction force. Hundreds of pounds of force in this case, and when the seal is dry, it's even higher. So, I've got the pneumatic cylinders pushing up trying to fill the back side of the piston with salad oil, and gush, as the piston lifts the cylinder off the lower seal, spilling the contents into the drip pan. And there you have it, years of neglect, vindicated. And that's why you should hoard! Save everything. As long as you remember that you have it, it will be there some day to save it.

    It should come as no surprise that the press is a very dangerous machine. The pneumatics are especially dangerous as they are extremely fast. Throw the switch and WHAM. Things move very, very, fast. Hundreds of pounds of force. Better not have any body parts in the way!

    The leads of the thermocouple were reversed at the controller. Now, when the heaters are powered, the temperature goes up, rather than down, which would be wrong. Got to love those really inexpensive process controllers from China!

    The top cap is now suspended by a set of springs that allow a simple placement of the top cap against the cylinder. A set of 4 linear slides would fix the issue, but that would involve a complete disassembly, flow jet water cutter time, several hundred dollars for the linear bearings, etc.. Perhaps later when all the issues are understood.

    So apparently, despite my frustration yesterday when everything was going the wrong way, everything pneumatic and hydraulic is working exactly to plan.

  • The importance of small animal sacrifices.

    Chuck Glasser07/04/2016 at 02:16 0 comments

    Today, it was time to bring up the big orange isostatic press. I had previously drained all the mineral hydraulic oil from the air driven hydraulic pump and it was now time to refill it with an organic Soybean oil. For now I'm using ordinary Soybean oil intended for salads and fries. No point in paying extra for the same thing that is certified to be unfit for human consumption but great for your tractor. I've determined that if there is a leak under hydraulic pressure and I get injected by a hydraulic fluid, the outcome of the disaster will be better if it is organic oil rather than a mineral oil. As it turns out, the oil is not under any constant use as in a machine that is running 24/7 so if my salad oil goes rancid I can just turn it to soap. It's a lot cheaper as well. In any case, I finally came to the conclusion that I didn't need to run down to Harbor Freight to buy a transfer pump when a piece of hose and a syphon would work just as well. The oxygen bottle was filled as well as the pump.

    In the LTCC process, the various layers of ceramic green tape are compressed with a hydraulic process to force the multiple layers of materials into a uniform and solid body. The best press to use is an isostatic press as it insures that the material will see a steady and uniform compression force in every direction and across the entire surface of the material. The industry standard for the big squeeze is 70 degrees C for 10 minutes. So, I throw the temperature controller switch up, by convention, up is true, and see that the temperature is moving in exactly the wrong direction. Dang, the polarity of the thermocouple is reversed. Got to fix that!

    The hydraulic cylinder has two pneumatic pistons attached whose purpose is to move the hydraulic piston up and down to fill the working volume with hydraulic oil and purge air out of the system. The air cylinders operate under the control a pneumatic single pole, double throw air valves energized by a 24 Volt source. Switch up, hydraulic cylinder goes up, switch down ... hydraulic cylinder goes down. The switch is thrown, and ... nada. The cylinder just sits there, not moving. This can't be. There is an equivalent force of several hundred pounds pressing against the hydraulic piston. Yet, it just sits there. Throw the switch and the machine shudders under the forces, but no movement. Ok, perhaps the piston has sealed itself against the piston wall. I'll force it with an irresistible pressure, say 10,000 psi, the rated pressure of the hydraulic pump. Pump on, nada. No movement, whatsoever.

    Hydraulics is a lot harder than I gave it credit.

    The purpose of a gauge is to report the process parameters. The secondary purpose it to indicate what the hell is going on when nothing works! There needs to be another pressure gauge at the pump, or perhaps just after the check valve, or two... their cheap!

    I was successful in making the piston move once after releasing the clamping force of the heater jacket around the pressure cylinder. But it move once, and then no more.

    As if Murphy wasn't completely satisfied, the hydraulic seal on the top cylinder cap has been chewed up from poor alignment as the top plat is lowered down to form a seal.

    The GFI tripped. OK after a reset.

    Time for a small animal sacrifice. I think I'll make a chicken flat bread pizza.

  • Smoke gets in your eyes

    Chuck Glasser06/28/2016 at 01:17 0 comments

    If you go to a hacker space and use a laser cutter, all the magic behind the curtain goes without notice. To get a visceral understanding of what's involved, take a ceramic plate, fill it with wood chips, say a pack of popsicle sticks, pile them high, set it ablaze, right there in the sink. In a few moments, say thirty, you'll have as much smoke as pours out of a laser in the first few milliseconds. To be accurate, it's not the laser that's smoking, it's the workpiece that's in the focal point of a death ray.

    So, behind the curtain your going to find a big blower and flue system sucking away your general ignorance.

    Wait, there's more! The smoke is different than that you have seen from a flickering campfire late at night. The smoke is more like that from a blow torch, a blast of smoke propelled by a death ray. It's very impressive. Of course, the idea is that the smoke system is effective and the nieve go about their day making their gizmos without a care of the physics involved.

    In this case, I would prefer to remain ignorant. Here is what it takes.

    The exhaust plenum took a huge effort. Someday I'll have to powder coat it black.

    Typically, most cutting is done on a replaceable hexagon grating with sufficient depth that the beam has diverged from it's focus when it reaches the bottom of the cutting chamber. In this system the workpiece is rigidly mounted to the cutting table, so air must be drawn across, rather than down.

    The giant gaping maw of the exhaust system is just visible above the edge of the vacuum hold down.

    Smoke condensation from the underlying MDF used in the vacuum platen.

    Cleaned up with a bit of IPA.

    The cover sheet used to support the ceramic green tape is still intact. Can it be that the power level of the laser was perfect? Doubtful. If the energy was not absorbed, and the MDF beneath it was not cut then it must have been reflected back. Interesting! Do not look into laser with remaining eye!

    You can see a small tear in the material in the lower left corner. Ignore the divot on the right side. Too many things going on for two hands to coordinate.

    The shiny side of the tape. Tear is very visible. The system was out of focus. A great opportunity for a new series of DOEs (Design of Experiments)

    Later this week the MDF will be replaced with a refractory Perlite vacuum plate that will eliminate the smoke and surface contamination.

View all 17 project logs

Enjoy this project?



kelvinA wrote 03/25/2022 at 09:18 point

Wow this project sounds really advanced. Were there any more updates to this?

  Are you sure? yes | no

Chuck Glasser wrote 05/16/2016 at 03:20 point

As you might imagine, there seem to be a limitless number of details to describe and discuss. I'll try not to miss any!

  Are you sure? yes | no

Indigaz wrote 05/15/2016 at 10:02 point

I too am very interested to see this!  Any new way for DIY fab'ing of PCB's is an important addition to the home workshops.

  Are you sure? yes | no

federicoortiz12 wrote 05/05/2016 at 00:54 point

You Sir have my total and undivided attention! Please document as much as you can. This kind of setups mean a giant leap foward for advanced DIYers and hackerspaces.

Thank you for doing this!

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates