Hébert Labs

PWB Lamination Press

Introduction

The time came, in the ongoing development of my next-generation mass spectrometer, that I had no choice but to expand my capabilities to include multilayer PWB fabrication. The initial circuit design for the control electronics worked fine on a two-layer board because it was built around a specialty chip produced by Cypress Semiconductor. That allowed the board to be small enough that the traces were short compared to the rise time of the signals.

But then Cypress discontinued that chip. I redesigned the circuit to utilize only standard parts, but that made the whole thing just large enough that reflected signals were now causing problems. The control electronics for my mass spectrometer were failing because of impedance discontinuities along certain bus traces. I needed to redesign the board with controlled impedance traces for the data and address busses, and that meant I had to design, and fabricate, a four-layer board.

The obvious first option would be to simply buy a lamination press. LPKF, who made my C60 circuit board milling machine, offers a couple of models along with all the needed supplies and a complete set of "Insert Tab A into Slot B" instructions. Unfortunately their lamination presses include a hefty 5 digit price tag.

I did consider the option, but even the used models I found were many thousands of dollars. I was certain I could build one for significantly less.

But building a press is one thing, and developing a process to use it is quite another. Even as I write this I am still looking for ways to improve my process reliability, and frankly I don't expect that effort will ever be fully complete. Still, getting started was particularly challenging.

How much pressure did the press need to apply? What temperature did it need to reach? Did I need to enclose the process in a vacuum envelope? Where, for that matter, does one buy B-stage (prepreg)? Does anyone sell it in prototype quantities?

Finding these answers was not simply a matter of asking my favorite search engine? There seemed to be precious little information available, on the internet at least, on the physics of laminating multilayer PWBs, and what I did find was, to be kind, often unreliable. Most of the results pointed to fabricators offering to produce multilayer boards, but they weren't at all interested in sharing the details of their lamination processes.

So I turned to LPKF for such information as I could glean. Sadly, that wasn't much. The LPKF customer service technicians were quite expert on using their system, but were less than expert on the "whys" of what they did. For example, when asked how much pressure to apply, they could only give me a gauge pressure from their machine. They had no idea how much actual force the platens would actually apply to the panel stackup.

To be fair, the information is there. It's just that consumer-based search engines aren't prone to list it on the first page of their results, that is unless you carefully choose your keywords. With diligence, and well chosen search terms, I did eventually find the information I needed, including one resource that was particularly invaluable, the LAMAR Group Multilayer Handbook. If you are planning to duplicate my work, to set up your own prototyping PWB lamination press, I strongly urge you to make it part of your preparatory research.

The Press

The press consists of the frame, the pneumatic cylinder/piston and the pressing platens. As a system it must apply and control both temperature and pressure to a PWB panel stackup in a unified process. So while the frame itself is a static fixture, both the pneumatic cylinder/piston and the pressing platens are dynamic subsystems, each with their own control accoutrement.

 

The Frame

 

I began with a fundamental question. How much force do I need to apply to my panel stackup? According to the Lamar Group Multilayer Handbook, the required pressure is a function of the area of the panel being laminated. Using their data table I found that function to be

 

P(a) = 200 + 0.347a

 

where P(a) is the required pressure in psi, and a is the area of the panel being laminated. Per the spreadsheet I used to do the calculations, the regression of this function from the data provided produced a standard error of 4.3 x 10-14 (effectively zero), and an R2 value of 1. In other words, within the limits of the spreadsheet, and certainly within the implicit 3 significant digits, this function is a perfect fit to the data.

Since my panels are 9" x 12", for an area of 108 in2, I would need an applied pressure of 237 psi. Applied to 108 in2 of area, that would amount to 25650 lbs (12.825 tons) of force.

I hope this number gets your attention, especially if you are planning on duplicating my work. These are potentially lethal forces. A poorly constructed frame could result in literally catastrophic failure.

As you can see in the photos above, the force required to laminate a multilayer PWB is real, and dangerous. Rather than trust my welding skills, I fabricated my frame out of a series of angle iron frames bolted together with 5/16"-18 UNC threaded rods (see schematic drawing to the right). Even still, I found it necessary to reinforce the top members with flat iron trusses.

 

The Pneumatic Cylinder/Piston

 

Application of pressure in the lamination press is a dynamic process. As the prepreg resin changes phase states the resistance to the applied force changes, thus necessitating the monitoring and regulation of that force.

The magnitude of the force required to laminate a multilayer PWB would seem to favor a hydraulic system. But regulating a hydraulic system, though possible, is neither inexpensive nor convenient.

Pneumatic forces, on the other hand, are readily monitored and regulated with common and inexpensive technologies. The added process capacity of compressed air also makes the use of pneumatics favorable, and I will discuss this at the appropriate time, but it was the regulation issue that drove my choice.

Pictured to the right is the pneumatic cylinder and piston I had built for this press. The inside diameter of the cylinder is specified at 10.860" (+0.004" -0.000"), with the piston seal provided by a Parker 2-450 O-ring. This provides an effective area of 92.63 in2. Therefore, to deliver a total force of 25,650 lbs would require a cylinder pressure of 277 psi.

Made of steel, the cylinder weighs 39.6 lbs and the piston weighs 38.4 lbs for a combined weight of 78.0 lbs. Though inconvenient to pick up and move, the weight of the piston distributed over the effective area adds less than 0.5 psi to the measured pressure, or less than 0.2% of the total.

To generate and deliver this pneumatic pressure I purchased a Makita AC310H high pressure air compressor. It will produce and regulate up to 375 psi.

 

The Pressing Platens

 

Both the heat and pressure are applied to the panel stackup via a pair of heated platens, and two geometrical properties of those platens are critical. It is critical to the rheology of the prepreg resin that both the pressure and heat are delivered evenly and uniformly over the expanse of the panel. Therefore the platens must be as parallel as possible, and they must be as flat as possible.

Keeping the platens flat ensures maximum contact distributed uniformly over the entire expanse of the panel stackup. If the platens are not flat, they will only contact, and thereby deliver heat and pressure to, the panel stackup at the high points, resulting in poor and uneven lamination.

I achieve flatness by sanding the panels smooth with a fine grit sandpaper on a vibrating sander. This provides a clear visible delineation of contours in the surface topology, and provides for maximum surface contact.

The two platen surfaces must also be parallel during lamination. If they are not, then contact will not be uniform over the expanse of the panel stackup regardless of how flat the platen surfaces are, and two design features of my press contribute mutually to highly parallel platens. The platens are free floating, and they are compressed by a self-leveling pneumatic piston.

Most lamination presses utilize a floating platen compressed against a fixed platen. The floating platen is constrained to move up and down along fixed columns. Since those columns of necessity produce a low effective aspect ratio of diameter to length, the platen will bind if moved very far from the horizontal.

In contrast, my press utilizes a pair of free-floating platens that are fully removable from the frame. Since both are free they can naturally align with each other during pressing. Moreover, they rest on a large diameter pneumatic piston. The high aspect ratio of diameter to length allows the piston significant freedom to self-level.

These platens are constructed from two 1/2" plates of aluminum. I machined channels into the aluminum plates to accept 240 V heating elements, which are held in place with 1/8" sheet steel covers.

Lamination platens with heatin element channelsLamination platens with heating elements in situThe functional part of these elements (i.e. the part that gets red when heated) is contained within the platens. The protrusions get hot to the touch near the platen, but more as a thermal  load than source.

I added grounding brackets to the heating elements for safety, and I electrically insulated the contacts with elastomeric coating and electrical tape. The power to the elements is provided by a solid state relay and a PID controller. Feedback is provided by a K-type thermocouple placed in contact with the panel components. The completed platens are pictured below, both open (showing the pressing surfaces) and in situ (connected to the control electronics).Completed PlatensPlatens in situ with control electronics

To minimize the thermal capacity of the process, I stitched together an insulated thermal envelope that encases the entire press during part of the lamination process. This effectively restricts the thermal mass of the process to the press and the air inside the thermal envelope. Without it, the system would be essentially acting upon all the air in the room. Stated differently, without the thermal envelope the air in the room would act as a giant heat sink, dramatically reducing the efficiency of the process, possibly even to the point of creating unacceptable thermal gradients in the panel stackup.

The Panel Stackup

The panel stackup is the sandwich of materials that go between the pressing platens. It includes the panel components, PTFE (polytetrafluoroethylene, a.k.a. Teflon®) coated pressing plates, a thermal buffer and pressing cushions.

 

The Panel Components

 

In order to produce a four layer board I begin by milling a two-layer core, typically with a ground plane on top and a signal layer on the bottom. I use a core that is 1 oz copper on both sides and 0.024" thick. To this I add 1 oz. copper single sided panels, 0.007" thick and two sheets of 2116 prepreg (for a total of two copper panels and four sheets of prepreg) to each side. This yields a final panel thickness between of approximately 0.062".

Now you might be asking why I don't just laminate two double sided panels together with prepreg between them. If so, pat yourself on the back. That's an excellent question. Nonetheless, there is a very good reason not too. It's just not an immediately obvious reason.

Registration of the multiple layers is accomplished on my LPKF C60 by means of a two-pin registration system. Those two pins define the x-axis of my machine. Now consider a feature that I mill on layer 1 at coordinates (x, y). When I flip the board over, that same feature is now at (-x, y). It's a very effective means of keeping the layers registered.

But what happens when I add other layers. If I machine layers 1 and 2 on one board, and then I machine layers 3 and 4 on another, both will be well registered to themselves, but I will have to maintain the registration between them during lamination, a process wherein the objective is to make the interstitial material move.

Now consider the panel stackup I use. My core is layers 2 and 3, and they are well registered to each other relative to the two registration holes. So far I've machined no features into layers 1 or 4, so I can laminate them to the two sides of the core without concerning myself with registration. All I need to do is punch holes in the stackup materials with enough clearance around the core registration holes that they will still be accessible when I return the board to the C60 for further machining. This way the outer layers are machined relative to the same registration points as the inner layers, and any feature at coordinates (x, y) on layer 2, or (-x, y) on layer 3, will be at the same coordinates on layers 1 and 4 respectively.

So yes, my preferred panel stackup is slightly more expensive, but only slightly. It more than makes up for the added expense with dramatically improved registration, and ease of fabrication.

 

PTFE Pressing Plates

 

Any surface irregularities, creases, textures or even scratches, on whatever you place in contact with the outer layers of your panel will be permanently embossed into the copper, whether or not they're even noticeable on the original surfaces. The first first function of the pressing plates is to provide microscopically smooth surfaces to press against the copper, surfaces that are free of embossable features.

PTFE provides just such a microscopically smooth surface. Still, even something as insignificant as a particle of dust can result in a very noticeable deformation in the finished copper surface. This is why I use lint rollers to clean the surfaces of the pressing plates and the outer layer copper sheets before lamination.

Then there's another reason that the pressing plates are PTFE coated. The PTFE provides a very low coefficient of static friction.

As the stackup temperature increases the copper expands. If it is pressed against something with a high coefficient of static friction it will not be free to slide. The result will be an outer copper surface that is unacceptably textured with small undulations. By using a PTFE pressing sheet, the copper surface remains perfectly smooth. Panel laminated with PTFE pressing sheetPanel laminated without PTFE pressing sheet  

For my PTFE pressing sheets, I use single layer (not double walled) Teflon® coated cookie sheets. I purchase them in packs of two at Wal Mart for $8.00, and trim them to size with a shear in my shop.

 

Thermal Buffer 

 

Crucial to the lamination process is the rheology of the prepreg resin. In order for it to flow evenly it must be heated evenly and uniformly. Toward that end, a thermal buffer is placed between the PTFE pressing sheets and the heated platens. It slows and evenly distributes the heat flow from the platens to the core, thus helping to produce a more uniform temperature distribution and a more uniform resin flow in the prepreg.

This buffer is made of multiple layers of kraft paper. That's right, plain old brown kraft paper. I find that four layers of 24 lb kraft paper between each pressing cushion and PTFE pressing sheet is sufficient to produce good results.

 

Mechanical Pressing Cushions

 

The final component of the panel stackup is a pair of mechanical pressing cushions. I use old cardboard-like underlayment material from my LPKF machine. Like the kraft paper just mentioned, the pressing cushions also help to buffer the heat flow from the platens, but more importantly they add compressibility to the stackup.

In any controlled process, a degree of capacity is needed for smooth regulation. In the case of controlling pressure, that needed capacity is added as compressibility.

Imagine a system that had no compressibility whatsoever. Any infinitesimal change in the applied pressure would be immediately manifest as an impulse of pressure in the system. In this case, imagine the prepreg resin begins to flow. This would result in a drop in pressure, which would trigger a corresponding increase in applied pressure. But without compressibility, that increase would be manifest as an impulse of pressure on the flowing resin causing it to be abruptly expelled. That, in turn, would result in another (this time more abrupt) drop in pressure and a vicious cycle is born. In short, this would be the opposite of effective, or reliably reproducible control. This would be a process out of control.

But with compressibility in the system, instead of being manifest as an impulse, the increase of applied pressure is manifest in a more moderated and controlled way. The process is brought smoothly back to its nominal set point and kept under control.

I should point out that since my press does not use hydraulics, but pneumatics instead (with its own generous supply of inherent compressibility), the importance of the mechanical pressing cushions is probably significantly diminished. Nonetheless, I've not tried operating without them, nor do I intend to.

 

Platen Protection Foil

 

The final and outermost layer of my stackup is a sheet of aluminum foil protecting each of the platens. As the prepreg flows, it can run down onto the platens where it is problematic to remove. Also, sometimes the pressing boards can bond to the platens. Two sheets of aluminum foil are a trivial cost, far more than worth the labor, not to mention wear and tear, that they save.

 

The Lamination Process

 

There is more to achieving a high degree of reliability in the lamination process than simply applying heat and pressure to the stackup. The rheology of the prepreg resin requires the right thermal and pressure profiles in order to flow properly, which is to say to fill all the voids of the circuit topology effectively.

But even before that, two steps are necessary to prepare the stackup for lamination. It is necessary to remove any moisture that may be absorbed into the FR-4 components of the stack, and micro etch the copper surfaces, prior to lamination.

 

Moisture Control

 

Why doesn't water boil at room temperature? In short, it will if the external pressure is low enough. Put a glass of water in a vacuum and it certainly will boil at room temperature. Water boils, not at 100 °C as you might think, but at the temperature where its vapor pressure is greater than any applied external pressure.

Above water's triple point, its vapor pressure increases exponentially with temperature. It happens that at 100 °C the vapor pressure of water is just a little more than 14.7 psi, or one atmosphere of pressure, which is why it boils at 100 °C (at one atmosphere of pressure).

But just as lowering the external pressure below its vapor pressure will induce water to boil at lower temperatures, increasing external pressure will prevent it from doing the same. Of particular interest to the topic is the fact that at 180 °C (the processing temperature of the prepreg in the PWB stack), the vapor pressure of water is approximately 170 - 180 psi. Since the lamination process requires over 200 psi during the whole lamination cycle, moisture trapped in the FR-4 can and will remain in the FR-4 throughout the lamination cycle. 1

Under the right conditions of temperature, pressure and circuit topography, that water can interfere with the resin of the prepreg and weaken its bond strength with the copper layer. But more importantly it will remain trapped, only to cause problems later when the panel is once again subjected to higher temperatures (e.g. during assembly, or soldering). Trapped moisture, unimpeded by lamination pressure, can then cause delamination, or even micro-explosive ruptures in the board (called popcorning).

In my own experience it caused problems during electroplating. Trapped moisture caused lamination failures during hole wall activation (during the pyrolysis stage), creating voids which led to short circuits which formed during electroplating.

It is therefore important that the FR-4 panels undergo baking to remove any absorbed moisture before lamination. I have found that baking panels at 125 °C for 90 minutes is sufficient, provided adequate ventilation is employed to remove the vapor.

 

Micro Etching

 

The second stack preparation step needed to produce a reliable PWB is micro etching of the copper surfaces. Commercial PWB manufacturers call this step an oxide treatment, but the important function of this step is to greatly enhance the bond strength of the prepreg to the copper by greatly increasing the surface area of the copper. It roughs up the surface on a microscopic scale, providing a physical topography that the resin can flow into and mechanically bond with.

Now commercial PWB manufacturers use a variety of exotic and often proprietary chemistries to accomplish this step. They sometimes incorporate the formation of exotic oxide compounds that produce novel and useful reactions during the lamination process. And yes, I imagine they consequently realize efficiencies of five or even six nines (i.e. 0.99999 or 0.999999 efficient). But for the fabrication of a single prototype board, such efficiencies are literally, statistically and empirically meaningless. Let me explain.

Consider my own experience. I needed to produce a PWB with approximately 1000 through holes and vias. Without baking out the moisture from the FR-4 stack layers, and without micro etching the core copper layers, my lamination efficiency was approximately 99% (0.99, or two nines). So out of 1000 through holes and vias, I found an order of magnitude 10 (i.e. ~10, give or take) lamination failures (manifest as short circuits) on each board I made. By baking the FR-4 layers before lamination I was able to raise my efficiency to approximately three nines (0.999, or 99.9% efficient), but that still left me with an order of magnitude 1 (i.e. single digit quantities) shorts per board.

By micro etching the copper core layers before lamination I raised my efficiency to approximately four nines (i.e. 0.9999 or 99.99% efficient) which means I now expect that if I made 100 copies of this board, I would expect to find a lamination failure in about 10 of them (give or take). But I do not produce quantities, only single copy prototypes. So if I did achieve a higher efficiency, I would never know.

And so rather than spending large sums on exotic chemistries, I found an inexpensive alternative that easily achieves the 0.9999 efficiency I need. I use a mixture of hydrogen peroxide (H2O2) and hydrochloric acid (HCl).

More specifically, I use the hydrogen peroxide you can buy at any retail drug store (typically a 3% concentration) and the same muriatic acid you can buy at any retail hardware or building supply store (which is typically a 31% HCl solution). I mix 2 parts 3% H2O2 with 1 part 31% HCl, which yields approximately a 5 to 1 mixture of HCl to H2O2.

The small amount of H2O2 breaks down in solution into H2O + O, providing the free O to act as an oxidizer needed to dissolve Cu in HCl. I'm not a chemist, so I won't try to explain the reactions involved, but once the H2O2 dissolves some Cu into the solution, that Cu reacts to form an intermediate compound which acquires additional O needed directly from the atmosphere. Consequently, the solution actually improves with use, and when in time it does begin to saturate it can be easily revitalized by adding additional HCl.

Before inserting it into the stack, I dip the core in this micro etch solution just long enough for the copper surface to turn reddish brown. This is the oxide layer formation. I then thoroughly rinse the core in deionized water and pat it dry before baking out any absorbed moisture. I use deionized water to prevent residue from forming on the core surfaces during baking. If any residue does remain on the surface after baking I wipe it clean before inserting it into the lamination stack.

 

The Thermal/Pressure Profile

 

Once press and panel stackup are brought together, a process is needed to reliably produce a quality multilayer PWB. The following table shows the process I currently use.

 

period duration or ramp rate nom. temperature nom. pressure
Phase I 5-8 °C∙min-1 amb. - 180 °C 210 psi
Phase II 60 min. 180°C 280 psi
Phase III ~6 hrs. 180°C - <50 °C 280 psi
Phase IV > 16 hrs. <50 °C - amb. 0 psi

The process begins by applying a pressure of 210 psi to the panel stackup. The temperature is then raised by approximately 5 to 8 °C∙min-1, to a temperature of 180 °C. It is during Phase I that the crucial resin flow occurs, specifically in the temperature range of 60 °C to 140 °C. At temperatures above 140 °C the resins begin to cook, or set up.

At Phase II the pressure is raised to 280 psi, and the stackup is held at 180 °C for 1 hour. It is during this time that the resins finish setting up.

During Phase III, the cool down phase, the pressure is maintained at 280 psi and the heating elements are turned off. The thermal envelop is maintained and the entire assembly is allowed to cool over the next 6 hours or more, resulting in a cool down rate of approximately 0.3 °C∙min-1.

Phase IV, the final stage, is the post lamination cure. The stack is removed from the press and left to rest for a minimum of 16 hours before being subjected to any further processing.

 

Summary

 

LPKF sells a benchtop lamination press for well into the five digit price range, though you can occasionally find a used one for under $10,000. Fortunately, these were not my only options. The pneumatic cylinder and piston cost me about $900 to have made, and the high pressure Makita compressor cost about another $700, bringing the total cost of this lamination press to well under $2000.

Yes, it took quite a few failed attempts to finally work out the details of how to efficiently produce a reliable multilayer PWB, but those are difficulties I would have encountered regardless of which press I used. In the end, I would have to conclude that it was well worth the effort.

 



1 Yes, this is an oversimplification of the mechanics involved in vapor migration in the stack during lamination, but it is nonetheless a valid and useful one.

 

A Brief History of Computing Technology

 

What would you consider the advent of computing technology? I suppose if you consider any tool that helps one do calculations, the first computing technology would be the stick, used to scrawl marks in the dust. But I don't think that's what people were thinking of when they coined the phrase, "computing technology."

And if we don't count the stick, then we can't really count the abacus either. It too is simply a mechanical means of keeping track of calculations the user is doing. Likewise for the pencil and pen.

Slide Rule and Vacuum Tubes

I suppose the earliest computing technology, certainly the earliest in common use, was the slide rule. It didn't just help the user keep track of calculations. It gave the user the results.

But the slide rule's days were numbered with the advent of electronics.

Every electronics technician in the world knows about op-amps, but it's surprising how many don't know that they're called operational amplifiers because they were designed to perform mathematical operations (sum, difference, integration and derivative).

In fact, the earliest electronic computers were analog, not digital. But soon relays and binary mathematics gave us the first digital computers.

Did you know that the term "debugging" was supposedly coined by a computer operator who found that a program had stopped running because a moth had been caught in the contacts of an open relay? When asked how he fixed the problem, he said, "I debugged the machine."

Punch Cards and Calculators

In early computers, programs were written on punch cards (see above, left), and had to be fed into the computer in sequence. One trick programmers used was to write on the edge of the stack as a visual indicator that the cards were in the correct order.

It wasn't long before computers were reduced to handheld calculators, though the first ones were very expensive. The one in the center above was my first calculator, purchased in the mid 1970s for about $129. It had the four arithmetic functions, a percent key and a single memory location for keeping track of long calculations. Powered by NiCad batteries, it was the reason students made certain their seats were next to outlets.

My wife's first calculator (pictured above, right), purchased only a couple of years later, had more functions and cost significantly less.

Microprocessors

Today computers are reduced to single integrated circuit chips and are found in virtually every electronic device you can buy. And they're not just in desktop and laptop and tablet computers and smart phones. They're in your cars and your radios and your hair dryers and your vacuum cleaners, and ....

We've come a long way from sticks and slide rules.