I recently fabricated a set of SSD and battery brackets. Since almost everything about the Heirloom required a custom solution, there was only the question of how the bracket designs would work. The primary criteria would be that they make use of the peek array if at all possible.
I did find a couple of potentially useful SSD brackets, but sourcing them was all but impossible. The solution in this case is very simple. A majority of the extant brackets seem to rely on the threaded holes in the sides of the drives. But the ones on the bottom made more sense in this case.
So I just milled up some straight bars with mounting holes at the SSD and peek locations. Since there is really no penalty for simple flourishes on the CNC mill, I added a little curved detail at the outboard end of the SSD brackets.
The battery mount was another story. Bunnie designed a case elastomer strap for his Novena enclosure but it is not compatible with the Heirloom layout. The spatial constraints around the battery location are significant and were made all the more so after Bunnie and I came up with the peek-mounted fan/heat sink arrangement that sits right between the SSD and the battery.
I pondered the issue casually for a while and had a sense that there should be a way to use the peek array. It wasn’t until I came in one morning with the battery bracket at the top of the list that I was compelled to find a solution. I had a vague notion that the battery would rest against the inner front wall of the enclosure and so the front mount would have to mount to that same vertical surface above the battery. [screen capture or photo of layout here]
Having set up the milling vice for the 1″ wide aluminum strap used for the SSD brackets, I decided to make use of the setup and designed a system of aluminum clips that would capture the back face and top front face of the battery. An o-ring winding its way between the three would join the three brackets into a complete solution for holding the battery in place.
Bead blasted and ready to anodize
I’ll post the complete assembly when the parts return from the anodizer and the final machining work is done on the base panel.
“That’ll take a day” I thought as I checked outstanding tasks on my Azendoo app. The last couple of weeks have been filled with just such occasions. Frequent days with overfilled schedules falling short of the goal led to this summary assessment. So I make note of what I might have learned and move onto the next task.
Much of the Novena Heirloom work at present is about bringing the various parts together for initial assembly into something looking like a complete laptop. In particular, the LCD panels required a lot of steps to complete them and get them properly hinged to the bottom enclosure.
On revisiting the process for cutting hinge mortises on the CNC mill, I realized there were a number of these prerequisite steps to be carried out before I would be ready to begin with hinge mounting. Among those would be some of the final sanding of the LCD panel assemblies. These sanding operations are carried out by simple manual methods which inevitably lead to tolerance variations greater than those of the corresponding CNC-fabricated aluminum parts. This has led to some significant challenges along the way.
This combination of hand-sanded wood surfaces and cnc-machined metal components is especially interesting in that regard, since there is variation part-to-part with the wood components that has to be accommodated in the machining setups that bring the two materials together. Thus in a small batch production like this, many tasks are essentially a series of one-off processes strung together. (In a larger production, there could be resources to minimize the tolerance variation in the wooden components.)
Fortunately, the vacuum table system I developed to assist with this part-to-part customization has been extremely useful. Minor variations in the wood composite components can be accommodated by applying measured differences to the table adjusters during hinge mounting.
With this installation complete, the value of the custom torque hinge design has been proven. Their four, easily-accessible screw locations make hinge mounting simple.
One of the next tasks will be fitting the lid-latch magnets to the aluminum side plates. When the original face-magnetized designs proved inadequate to fulfill my design goal, I managed to track down a bit of unobtainium in the form of edge-magnetized units in the same form factor. I had been told by several suppliers that these would be a custom only option, but diligence paid off when I found them on the shelf from one of my previous suppliers.
The Novena Heirloom design uses an aluminum, wood-composite and solid wood enclosure assembly for the motherboard and primary electronic components. The wood enclosure panel is joined to the aluminum side panels to form the complete enclosure. Before that assembly task can be completed, the composite and solid panel must be trimmed to a length which corresponds to a nice fit with the system keyboard.
The trimming process must be done carefully to avoid chipping the wood surfaces. Several methods were considered for the task but I finally settled on a simple tablesaw operation. Speed and feed rates are critical to avoid chipping and, since the speed at which the tablesaw runs is fixed, the only available variable is feed rate. The slow feed rate required will be noted in the video below. An additional precaution common with similar situations in conventional woodworking is the taping of the edges to be cut. This provides reinforcement as the blade emerges from the surface of the material and can capture any small chip that may defy my best attempts to avoid chipping. The chip can later be rejoined with the trimmed part.
One of the tasks for which the vacuum table was created was trimming the upper edges of the bottom panel to proper height for mounting the front and rear rails. The rear rail also serves as a mounting location for the hinges. For reasons examined in an earlier post, the rear rail must be mounted slightly proud of the top surfaces of the aluminum side plates. A subsequent operation entails sawing to height to match up the rear rail to the side plates.
The image above shows the side plate loosely inserted into the bottom panel assembly. When the side plates are finally installed, the clamping pressure will align the adjacent wood and aluminum surfaces. The aluminum side plate slightly overlaps the wood composite everywhere except at this top location. The proper fit posting discusses how this fitting is done.
To make this work out properly, the rear rail mounting location must fix the rail in the correct position to allow for the fitting process. Measurements are made to determine where this height is along the back edge of the composite bottom pan. Then the part is set up in the custom vacuum table and trimmed to size with a 72 tooth slitting saw. Downhill cuts are made on both composite surfaces to prevent tearout or splitting of the veneer.
The video below shows the sawing to height operation on the CNC mill. Note the dual passes to create a downhill cut that prevents tearout in the composite surface veneers.
The proper fit issues that effect the Novena Heirloom fabrication process lead to a concern with making minor tweaks for each individual part that is prepared for assembly. Sometimes this kind of thing can be handled by essentially telling a CNC controller that the object is not in plane with the table and just let it do the math to figure out how to compensate. That is typically done for rotational displacements along the Z axis – that’s like spinning a part flat on the table. Each instance could conceivably have a modified parameter in the g-code. But was I was confronted with is a rotational displacement along the Y axis or perhaps the X axis. There are apparently some CNC controllers that can handle this, but mine does not.
My Aciera manual mill can rotate along Y with the table itself. If it were a CNC machine, that would probably suffice to resolve the issue. But CNC mills don’t often have such a feature and mine is no exception. Since I wanted to hold the parts in a vacuum fixture, I resolved to find a solution that would hold the part and simultaneously allow for the subtle X or Y axis tweaks required of the task.
I had earlier made a vacuum table base that uses the CNC machine table T-slots as a plenum. The vacuum pump attaches directly to the mill table as shown above. The sub assembly then bolts down to the table and provides an interface for various insert materials to be used as the working surface. Since my first attempts with it were based upon drawing air through mdf panels, I used one of the previously precut panels for this setup.
I could get slightly better vacuum by sealing off the areas outside of the vac seal. But so far, the system works just fine. I could remake it out of some plastic or aluminum in the future.
As configured above, the setup assumes a panel will be placed on top of this and drawn down to the hard surface by the vacuum while compressing the foam seal around the perimeter. That works quite well but is not applicable to the challenge at hand. So I spent quite a bit of time experimenting with trial and (mainly) error looking for a way to adjust the part hold down assembly on that Y axis. I finally had a minor flash of inspiration and went off the hardware store looking for some relatively thick, closed cell foam tape. Having that in hand, I reconfigured the surface of the insert by removing the 1/8″ round foam cord and replacing it with the newly acquired tape.
Oddly, I have been experimenting with tapes for holding small parts down on the CNC mill and had a particular candidate that was not up to that job but which turned out to be perfect for sealing the intersections of the foam tape. It has no backer, is very tacky and maintains the flexibility of the foam tape at the corners – crucial for what I wanted to achieve.
A pair of index pins can be seen at either side of the tape layout.The table that will hold the part is then fitted over those index pins at three pairs of holes. This is due to the working envelope limitations of my CNC machine. This table version actually came before the final working solution and was intended to be used with the standard foam cord in a flat position on the table in case I found a way to work this out with g-code. That was not in the cards, so it was from here that the thicker foam tape idea sprung.
A few modification later this version of the table materialized. Several features are noteworthy.
The outriggers with thumb screws are used for tweaking the plane of the table.
There are wavy foam seals at either end made by cannibalizing some bottom panel composite material to support the closed cell foam tape. They work nicely to seal the vacuum against the wavy bottoms of the panels.
The closed-cell foam cord along the other two edges are cut in with a horizontally wavy pattern because the point at which they make contact with the bottom panel is at the bottom crest of a wave. The wavy placement of the cord insures that some of the panel contacts the flat surface of the fixture rather than ambigously falling into the cord groove.
Visible but not clear are the red markings indicating the six positions the panel can take on the vacuum insert. (OK, that’s more than several features . . . )
Most things like this are first sourced from within the shop scrap pile and these screws have been lurking for a long time receiving occasional use for projects such as this. Because this is a very short run application, I just used Appleply with threads cut into it for the screws.
The complete arrangement begins with a lower vacuum table assembly that seals at the bottom against the CNC table/plenum chamber and presents a 1/4 inch closed-cell foam surface at the top. On top of that goes this vacuum fixture with seals designed specifically for the panels that have a number of operations needing to be done on them. The key to its operation is the thickness and flexibility of the 1/4″ foam seals. As shown in the video, the foam seal interface provides enough flexibility to allow adjustment of the vacuum fixture across both X and Y planes.
I’ve worked with curved and bent wooden forms as a luthier and camera maker for many years. The beauty of a fair curve is one of my favorite things in the design of wooden objects. Often one curved surface will be joined to an adjacent flat surface as is the case when building a guitar. For many of my camera designs, the same holds true as curved surface joins flat with a proper fit. The fitting process is tricky under the best of circumstances but, once the joint is made, there is an opportunity to fair the intersection between members at the joint by careful sanding or scraping.
I often will actually glue up the joint with one surface rough cut to overlap the other for trimming after the joint is complete. This is pretty standard practice in woodworking and it relies on pre-established datum to insure proper orientation of the completed part. When the glued assembly has cured, the excess material can be trimmed off with a router, a plane, a sanding block or file or any number of other methods that will leave a proper fit.
The aluminum port plate of the Heirloom laptop was trimmed to proper fit by a series of sanding operations one of which is seen here. The port plate has a pair of threaded bosses that align it with the side plate. So the port plate was left slightly long at both ends in anticipation of this fitting operation. Of course the parts are not yet anodized when this procedure occurs. That will happen after final hand finishing and bead blasting for these parts.
A prominent feature of the Novena Heirloom is the wavy bottom surface of the enclosure. The intended benefits of this form are threefold.
It produces a monocoque structure that contributes greatly to the stiffness of the complete assembly.
The gaps resulting when the panel is placed on a desktop are intended to increase airflow and contribute to cooling the computer system.
The visual effect when examining the bottom of the computer is a pleasing surprise.
The side plate design incorporates a curved flange whose profile matches that of the bending form for the wood composite that makes up the wavy bottom panel. The reality of producing such a bent form in a wood/glass fiber/cork composite includes the natural tendency for the material to respond more actively to the environment than metal. Then there is the fact that a composite form coming out of such a process requires quite a bit of cleanup. This includes sanding to remove seam tape and minor amounts of epoxy that makes its way through the pores in the wood to the outer surface. This means that when the composite and aluminum forms are presented to each other for gluing, there will be minor discrepancies in proper fit.
The two components are brought together in a press with a resilient-surface form to remove the effects of those discrepancies and produce the required proper fit. But because of the relative complexity of the mating surfaces, the alignment between the top of the aluminum side plates and the solid wood rails glued to the top edges of the composite bottom panel varies.
Developing fixturing and procedures to achieve proper fit under these conditions took about a week of concentrated thought, trial-and-error and reliance on previous experience. Part of the solution lie in developing a fixture table for the CNC mill that could accommodate tweaking for the side-to-side variations that occur in the bottom-panel-to-side-plate fits for each assembly. (That is dealt with in another post.)
The key difference between this process and the ones described previously is the fact that when the final glue-up occurs, the aluminum side plate will be anodized and the composite panel pre-finished, thus eliminating and prospect of fairing the fit afterwards. And since the fit along the rear edge of the bottom panel assembly also governs the hinge placement, all of the fitting must be done before final assembly of the side plates to the bottom panel.
The image below shows a proper fit being measured to apply to the remainder of the assemblies. A granite surface plate is being used just because it was handy and definitely flat.
This measurement is used to determine height for trimming of the remaining composite panels where the solid rails will attach. (That is also covered in another post.) When that rail-to-composite-panel assembly work has been completed, the fitting procedure shown in the video below will be used to set up the assembly with the necessary rotation on the Y axis to achieve a proper fit between the rail/composite assembly and the aluminum side plates.
The resulting data is used to set up the custom, tilting vacuum table covered in another post.
A circular array of holes for the Heirloom speaker grill holes seemed like a pretty obvious thing when I first considered it. But the truth was a bit more complicated. It turns out that making such an array look right is very challenging. After a few runs at it, I flashed on a recollection of the iconic radio designs of Dieter Rams from the 1960s. Many of his radio and hi-fi designs utilize speaker grill holes with this round pattern.
This German designer, whose body of work is said to heavily influence the designs of Jon Ivey and the Apple design team, had resolved the pattern for round speaker grill holes in a beautiful fashion. His design even appears in some of the promotional material from the time. It is a cleverly conceived quadrant system that produces a pleasing result that appears natural in spite of its very organized layout.
With a modest modification, I used this round hole layout for the Novena Heirloom speaker grill holes. I’ve begun referring to them in my notes and documentation as Dieter holes. They contribute nicely to the slightly vintage look I was after in the Heirloom design.
This video describes finding a method to successfully drill the speaker grill holes in the composite material created for the Heirloom laptop design.
The original notion for switching the Novena Heirloom was to use touch switches below the rear panel with graphics engraved above to indicate their location plus an LED to indicate the switching action.
I briefly tested the idea with an AT42QT1010 breakout board and found it to work pretty well. But when Bunnie put together a PCB design based upon my intended configuration, it proved to be unreliable.
So that idea was shoved onto the back burner for consideration at a later time. I would need a substititute in the form of a mechanical switch.
When the time came to begin resolving the issue, I retrieved a Snaptron sample box I had acquired prior to the Novena project. This seemed like the right application for these simple, mechanical dome switches.
They only require a split pad on the PCB and a small vent hole to vent the switch area under the tape used to position it above the contacts on the board.
The available depth to accomplish implementation of a mechanical switch would be the thickness of the rear panel composite. The PCB would be mounted flush to the bottom and the space above the rear panel would be occupied by the LCD panel in the closed position.
The Snaptron switch domes would occupy maybe a third of this available depth, so a switch actuator would have to reside within the remaining dimension.
The hinges on the Heirloom have a dark bronze chemical finish. I had been thinking about other elements in the design that might pick up on this color and decided to try something in brass that could be finished to a similar color.
I like switches that use a recessed area for an actuator. They are common in space constrained situations and environments where accidental switching would be undesirable.
It occurred to me that I might create a sculpted look for the switch base component by cutting the recesses on the CNC mill using a slightly open spacing between horizontal tool passes.
After these were finished, I made the brass actuators on the lathe. The push button tip projects just to flush with the rear panel surface. The actuators have a projecting pin on the bottom that pushes down on the dome switch to give it just the right audible and tactile feel. They then received a chemical darkening finish followed by a coat of a clear nylonic finish.
Bunnie and I discussed ideas for a CPU cooling system design while he was preparing a trip to China, so he decided to look for a viable fan while he was there. I had a centripetal fan in the shop that seem potentially usable, but it was too large for the available space. Although the small cooling device I had located earlier might have been useful with a typical CPU heatsink, a standard fan would require moving the heat from the CPU to a different location. So we returned to using heat pipe technology. Bunnie designed an evaporator plate that cleverly takes in the CPU and the adjacent IC with a single machined component. A single heat pipe then makes its way over to the fan assembly through a few bends.
A bit of correspondence regarding possible layouts for the fan/heatsink assembly led us to a place between the SSD and the battery which is as close as we could get to the CPU. Bunnie had already tracked down an interesting copper heat sink and was then able to design the condenser plate – also machined from copper but with a fancy plated finish.
We cobbled together a simple setup of the the case filled to the brim and Bunnie began working up a software testing environment to check it out. He then hooked up the fan to the Novena system.
It quickly became apparent that some firmware work was needed to get the fan to operate with proper hysteresis instead of cycling on and off. So Bunnie got his Xobs on the phone and got the CPU cooling system working properly.
The system was then run under extreme conditions to test the effectiveness of the CPU cooling system. It is slightly overkill for this computer so it worked quite nicely in testing.
So now we had a complete system but no way to mount the fan into the case. The fan’s normal mounting points were misplaced for convenient use. I noticed that the fans had a removable metal cover plate on one face and it occurred to me that this might be the key. I could simply make a replacement for this plate that would bind everything together.
Three iterations later I had a CNC machined aluminum mounting plate that uses mounting locations on the peek array below. It provides a place for thermally mounting the heat sink and additional real estate for a potential plenum assembly.
When I first considered this idea, Bunnie and I both commented on the fact that the extant cover had a slight bulge in the area around the fan blades. I examined it to find that there was no immediately obvious reason for the shape as there was plenty of clearance for the blades without it. And since the fan exhibited other examples of possible legacy design details, I concluded it would be reasonable to ignore that feature. This assumption would soon come back to bite me.
The first example of the CPU cooling fan assembly looked to be fine so I sent off the plates to have them anodized in red. But when I had my assistant Darrell begin assembling the remaining fans we discovered that the fan housing castings were inconsistent in how flat a surface they presented to the base plate. It was enough to possibly be an issue with the fan’s efficiency so I set up a lapping operation on the lathe and had Darrell begin to flatten them. As the law of unintended consequences loves to hide away in my shop, we soon discovered that the removal of enough material to properly flatten the fan housings led to binding of some of the fan’s blades against the new bases.
Back to the CNC mill with the fan plates for clearance cutting . . . Fortunately, I had a fitted fixture from previous operations that made it easy to set up for this job.
So now, many months after initially considering how to keep the Heirloom version of the Novena cool, we now have a functional CPU cooling system thanks to the genius of Bunnie and a bit of effort on my part. I still have to asses whether to route the condenser air out of the case. But given the nature of the Novena system, it is actually unlikely that the system will make regular use of this cooling system since the very act of accessing Bunnie’s amazing hardware as intended will allow enough ambient air exchange to passively cool the system. But for any who find themselves using their computer as a conventional laptop, the CPU is protected.
The production Novena laptop as envisioned by designer Bunnie Huang has an open-top layout with plentiful ambient air exchange. This means that a simple, passive heat exchanger mounted to the top of the CPU is adequate to the task of cooling the system. But when Bunnie and I first discussed the Heirloom design, we came to the conclusion that it would be a slightly more conventional clam-shell type layout.
In order to maintain the desired access to the system hardware, a portable Bluetooth keyboard was chosen to incorporate into the design. I integrated the keyboard into the Heirloom design by nestling it into the standard keypad location found in a typical laptop configuration. In this position, the keyboard hovers immediately above the CPU on the system board. The area behind the keyboard is occupied by a rear panel that includes the speakers, system switches and a lid closure sensor. This setup effectively seals up the enclosure so that ambient air is no longer sufficient for CPU cooling.
Initially, the task was left to me to find a way to resolve the cooling issue. Bunnie and I agreed it would be worthwhile to pursue a passive approach and avoid the use of a noisy, watt-hungry and space-hogging fan. I began research in earnest and eventually found my way to a handful of companies specializing in CPU cooling systems. Correspondences with those firms led me to a lengthy round of dialog with engineers over how to manage the passive approach to the problem. This, combined with the quest for proper hinges for the case, became the major issues in the early development of the Heirloom design.
With this passive cooling idea in mind, I designed the speaker mounting arrangement so that there was a potential dual-use aspect to the speaker holes. The speakers are shock-mounted at a small distance below the Dieter hole array so that air flow through the same holes could potentially serve as supply air when the LCD is in the open position.
An early prototype enclosure housing also included an experimental array of slots in the bottom to facilitate this top down air flow approach I came to appreciate from conversations with CPU cooling system engineers.
Many approaches were considered including “finning” the left side plate and connecting it to the CPU with a pair of heat pipes. This mass of aluminum at either side of the Heirloom enclosure seemed the logical candidate to dissipate heat from the system.The short length of attachment area available for the termination of the heat pipes effectively eliminated this idea. The fins may or may not have appealed . . .
The apparently obvious choice of the right side plate, with an uninterrupted surface to accept the heat pipe connection, was seen as too far away from the CPU for an effective solution.
I pushed through several other concepts punctuated by lengthy dialog with engineers including using an off-the-shelf heat sink material running along the back edge of the enclosure. The inability to provide a top to bottom air flow path through this heat sink proved to be its demise.
In the midst of this lengthy round of dialog over heat management using a passive approach, I was introduced to a new product promising to provide an interesting middle ground between passive and active solutions. A company acquired by Aavid, one of the firms whose engineering staff was very helpful, was developing a low-current, diaphragm style pump with a very small footprint. It is intended to be used with a typical on-CPU heat sink by enhancing the flow across it when ambient air is insufficient to maintain desired temperatures. Unfortunately, ongoing production issues have stalled the release of the product so I was never able to acquire one for testing.
My last attempt was an implementation of what one engineer described as a cold plate. This entailed using the peek plate as a mass for pulling heat from the CPU into the peek plate whose perforated surface would then serve to dissipate heat. One engineer indicated that this was a potentially useful solution. Another stated that it wasn’t technically a cold plate since there no fluid flow involved. And it does not address the issue of what to do with accumulated heat in the enclosure or even if that is actually an issue. I think I may still try it out at some point and see how it works.
So with the clock still ticking (yes, I’m old enough to remember that), I decided to consult with Bunnie and see if he might have some alternative thoughts about solutions to the challenge. After a quick update, he suggested that we might look at a fan-based solution. An upcoming trip to China would have fan-shopping added to the list.
Once again, I looked at various options for placement of fans, heat pipes and heat sinks and consulted with Bunnie. With the hinge issue still unresolved, it began to fall on Bunnie to come up with a plan while I wrestled with that.
NEXT: Designing a fan-based cooling system foe the Heirloom.