Oreo Construction: Hiding Your Components Inside The PCB

In recent months, the ability to hide components inside a circuit board has become an item of interest. We could trace this to the burgeoning badgelife movement, where engineers create beautiful works of electronic art. We can also attribute this interest to Bloomberg's Big Hack, where Jordan Robertson and Michael Riley asserted Apple was the target of Chinese spying using components embedded inside a motherboard. The Big Hack story had legs, but so far no evidence of this hack's existence has come to light, and the companies and governments involved have all issued denials that anything like this exists.

That said, embedding components inside a PCB is an interesting topic of discussion, and thanks to the dropping prices of PCB fabrication (this entire project cost $15 for the circuit boards), it's now possible for hobbyists to experiment with the technique.

But first, it's important to define what ‘stuffing components inside a piece of fiberglass’ is actually called. My research keeps coming back to the term ’embedded components’ which is utterly ungooglable, and a truly terrible name because ’embedded’ means something else entirely. You cannot call a PCB fabrication technique ’embedded components’ and expect people to find it on the Internet. For lack of a better term, I’m calling this ‘Oreo construction’, because of my predilection towards ‘stuf’, and because it needs to be called something. We’re all calling it ‘Oreo construction’ now, because the stuf is in the middle. This is how you do it with standard PCB design tools and cheap Chinese board houses.

The immediate inspiration for this build comes from designer2k2 and a flat-pack Christmas ornament. This project used castellated pins and a series of holes to mount SMD parts to the side of a PCB instead of the top or bottom. While soldering electronic components to the side of a PCB is somewhat novel, mounting electronic components to the side of a PCB is nothing new. Lumen Electronic Jewelry is producing a PCB ‘heart’ pin (right) with a capacitor and USB port mounted inside a cutout in the milling layer of a PCB. Likewise, other PCB projects — mostly PCB business cards — have experimented with mounting other components in a cutout in the milling layer. I have seen coin cell battery holders that use PCB cutouts with two ‘tabs’ that capture a battery between fiberglass.

The idea of embedding components within a stack of fiberglass and copper is something we really haven't seen before in the small-scale hobbyist world, but it can be done. Embedded components — there's that ungooglable term again — can be done in very expensive products. The reasons for doing this range from saving physical space, better EMI shielding, and making something more difficult to reverse engineer. This is a technique for military and aerospace components, where price is no object.

Boards for military and aerospace work are one thing, but the past year saw a significant amount of discussion over embedded components, albeit for all the wrong reasons. Bloomberg's Big Hack was a story about Supermicro motherboards shipped to Apple and Amazon that had additional components giving Chinese hackers a back door. This story was widely criticized, Apple and Amazon have fervently denied having found compromised motherboards, and any day now I’m expecting Supermicro to file suit in a libel case. This story did however generate a lot of discussion over how such a hack could happen. The top minds of the Twitterverse believe this could be done by embedding a small microcontroller inside the motherboard's PCB, between the baseboard management controller and its Flash memory. This small microcontroller stuck between a few layers of PCB could in theory change a few bits of the BMC's Flash to give attackers a back door, and Trammel Hudson gave an interesting talk at CCC discussing the theory of this fictional hack's operation. It's within the realm of possibility, but the smart money says this didn't happen with Supermicro motherboards shipped to Amazon or Google. In any event, x-ray inspection or even a flying probe test would reveal any ’embedded component’ was in the PCB.

For this build, I have extended these techniques slightly by mechanically bonding the layers of PCBs together with solder. This was previously done by Voja Antonic and his work in building enclosures out of FR4. His approach was to create a strip of bare copper around the perimeter of each side of the enclosure. By mounting these sides of the enclosure at the correct angle, soldering the two flat planes of PCBs into a three dimensional shape is as simple as running a soldering iron over the exposed copper on the perimeter.

Each PCB in the stackup has exposed copper along the perimeter. By applying solder paste and clamping the boards together they’re read for reflow.

I used Kapton tape as the clamping method since it will have no problem holding up to the heat of the oven. After baking it, sandpaper is all you need to clean up the edges.

The circuit for this build is a guitar pedal. More specifically, it's a slight modification of a Dallas Rangemaster, with the actual schematic borrowed from Fuzz Central (the RangeBlaster). There are several reasons for demonstrating this PCB technique in the form of a guitar pedal, and for using a Rangemaster circuit in particular.

The Rangemaster circuit in particular was chosen because it is a very simple circuit. It's only a single germanium transistor and a handful of resistors and caps. My choice of putting a Rangemaster circuit inside a PCB is driven simply by component count; it is the simplest circuit that does something. As for demonstrating this technique in a guitar pedal, I have far more sinister reasons. The market for guitar pedals makes even less sense than the audiophile market. If you come up with a circuit and coat it in epoxy, you’ve just made a thousand dollar pedal. No, that is not a joke. I am simply capitalizing on the gullibility of consumers with an interesting fabrication process.

The basis of the circuit is exactly what you would expect for a guitar pedal: there is a 3PDT footswitch, a pair of 1/4″ jacks, a 2.1mm DC jack (center negative, because Boss), and a standard PCB mount pot 10k, audio taper. The active part of the circuit is a vintage OC44 transistor in a TO-5 package. These are the only components visible on the finished PCB.

This circuit board was first constructed by laying out the through hole components in logical places, then dropping the surface mount components in places that made sense. Again, this is an exceedingly simple circuit with less than a dozen parts, in the schematic. Once that was done, it was only a matter of copying the PCB to a new file and adding cutouts around the parts. This board was done in Eagle, giving me the ability to add many layers to the board which could then be added to the CAM manager to create the Gerbers.

The real ‘trick’ with this technique is encapsulating components within a PCB stackup. While this can be done with a standard PCB thickness of 1.6mm per layer (three layers are required for complete encapsulation, resulting in a final thickness of 4.8mm), I used 0.6mm thick PCBs for the top and bottom layers. This resulted in a final thickness of 2.8mm. This is thin enough that the assembled piece does not register in your mind as a stack of PCBs. It's thin enough that one could easily believe this is just a normal PCB.

It's easy to create a PCB, and if you know what your board house can do, it's easy to create internal cutouts on a board. There is absolutely nothing new about the previous thousand words. The trick to Oreo construction is mechanically bonding the layers together. This could be done with glues and resins, but taking a page from Voja's work, I decided to use solder to attach one layer of PCB to another. This was done by a copper trace around the perimeter, disconnected from any ground planes or pours.

The assembly process is as simple as populating and soldering the bottom layer board with surface mount components, preferably with lead free solder paste. Then, leaded solder paste is applied to the perimeter traces, the boards are clamped together, and the entire assembly is thrown into the reflow oven. After that, it's simply a matter of populating the through hole components.

There are other ideas I considered to connect these different PCBs together. I could ‘stitch’ them together with vias and through holes, using small bits of wire to both align and mechanically attach each layer together with solder.

While you can embed capacitors, resistors, and microcontrollers inside a stack of PCB, there are limitations. First and foremost, the Rangemaster clone circuit calls for 47 μF capacitors. This value is much too large for small SMD caps, and the (physically) smallest caps I can find with this value are on the order of 10mm thick. Unless you want a circuit board that's half an inch thick, these caps are far too large. The workaround for this problem is to add many caps in parallel.

This leads to another problem. The original circuit used electrolytic capacitors, not small ceramic capacitors. Because I’m using arrays of ceramic caps, the actual capacitance is less than the sum of all the capacitance in the array. MLCC capacitors should be derated when biased (as they are when using them as a bypass cap), and the capacitor I ‘constructed’ does not have the correct value in the circuit. Yes, the capacitance of ceramic capacitors is dependent on their voltage, but you can aaay yolo around this by simply adding even more capacitors.

Additionally, no project that uses this technique will be able to use large parts. If you have a project with a small boost power supply, you probably have a relatively large inductor. Inductors of a sufficient rating for a beefy power supply will be too tall to embed into a single layer of FR4. The same is true for components handling high power, as they’re usually physically large and must dissipate heat, the latter being a problem for a component that is effectively trapped inside a fiberglass box.

Despite the problems, this is an interesting technique of PCB fabrication. Combined with the dropping prices of custom-made PCBs — the boards for this entire project cost less than $15 USD total — I would expect to see many PCB artisans picking up this technique.

This project was just a demonstration of what is possible with Oreo construction, but given the huge advancements in artistic PCBs, this is in no way the limit of what is possible. Given we’re now in the golden age of reverse-mount LEDs, it's possible to encapsulate the driver and the LEDs of a gigantic matrix inside fiberglass. With Oreo construction, an entire PCB could be just a brick of fiberglass when it's off, and a glowing rectangle when it's on.

If you’re wondering what this Oreo construction guitar pedal sounds like, well, it's a Rangemaster treble booster. Brian May's guitar work for Queen would be the most popular example, but Brian May is a little too unique to really get a sense of what this sounds like. A better example would be Tony Iommi of Black Sabbath, something from the first two Zeppelin albums, or Clapton on the Blues Breakers album. That's a lot of sonic territory, but this is a better demo of a Rangemaster than anything I could produce. In any event, but the entire idea behind this was to build the simplest circuit possible inside a PCB, not to do anything fancy. A Rangemaster is one transistor, so that's what I built.