geekhack Marketplace > Interest Checks

[IC] Evolv 75%

(1/137) > >>

Designed By Alpha Studio

Welcome to the Evolv2, and evolution of a crazy idea.
Read the post below OP for information about the design philosophy behind the revision.


The Arrows

Beginning with the layout, the Evolv opts for a 1.5u right shift over the 1.75u shift widespread in blocked 75 layouts.
By doing so, the arrows line up neatly with the right mods, allowing for a continuous .25u bezel.

The Rotary Joystick

In the upper-right corner of the board, utility wins the battle over aesthetics.
No badges or logos; instead, the Evolv features a design evolution to the knob: the rotary joystick.
With users able to assign macros to five unique positions* and rotation, the Evolv packs functionality in a small space.

The Back

A teaser to what's inside the board, the bottom piece of the keyboard is a stunning visual accent.
Created through a process known as generative design, the Evolv can support a truck without breaking, all while looking fantastic.

The Plate

Render pending, the Evolv redesign will still feature a generative plate, however, it will be available as an extra during GB

The Assembly

To top off the design, the Evolv features no visible seams or screws once assembled.

The Specs

Angle: 8 (Half a degree more compared to v1)
Material: 7075 Aluminum + Cast Aluminum
Colors: TBD (Black + Raw and Black + Gold confirmed)
Plate: Full isolation mounted carbon fibre plate, considering adding more options.
PCB: Custom designed PCB: solderable, QMK, VIA, integrated USB-C port.
Weight: ~1.5Kg
Price: TBD, not as high as the first iteration of the Evolv
PCB compatibility, fixed layouts TBD:

Render Gallery



LBaron, this board wouldn't exist without him. From the minor details to the design of the plate, he was always there to provide invaluable feedback
Abec13, for teaching me how to render
Yangsta, for teaching me about vibration
Fropsie, Stef, and my Discord Server, for all the love and support
The Top Clack Community. Love you guys <2

Update History

5/6/20 - Introduced a More Organic Base to v2 Based on Feedback
5/1/20 - Prototype Writeup and Board Redesign
4/9/20 - First Proto Pics Posted
3/24/20 - First Proto Completes Production
3/10/20 - Black + Gold Wins Color Poll and First Proto is Ordered
2/11/20 - Color Poll Went Live
2/6/20 - Generative Plate Writeup Added to IC
2/4/20 - IC Launch

Old Version 1 Renders



*: The five unique positions for the rotary joystick are (N, E, S, W, and pushed down)

Prototype Writeup

Hello all and welcome to the prototype writeup,

First off, I would like to apologize for the delay in posting this writeup. When I received the prototype three weeks ago, I quickly assembled the board and started playing around with it, excited to collect all of my thoughts and to share them with you. However, as I began writing about the issues with my prototype, I couldn't help but open Fusion to try to fix them. As one weekend turned into one week, I grew self-conscious about making this post without proposing the very fixes I was working on. One week turned into two as I took my midterms and had to turn my attention away from the Evolv and onto school. This break was a blessing in disguise: upon returning to the board model, I realized the fixes I was making were hacks to get around fundamental design issues of the board, so I ripped off the band-aid and redesigned the entire thing from the ground up. However, before I talk about the redesign, I must share my thoughts about the original prototype with all of you.

Let's start with the big one: the bottom of the board. I love it; I couldn't have been happier with how it came out; nevertheless, there is room for improvement. First is the side angle of the board, as seen in this photo:

Despite being the visual focal point of the board, a user of the Evolv cannot easily see the generative design of the bottom as the tendrils 'hide' underneath the top case of the keyboard, almost defeating the purpose of the design.

Aesthetics aside, most of my qualms with the bottom came with the manufacturing process. Not only is metal 3d printing exceedingly expensive (as in, I don't think I reasonably could've made the <$1000 mark if I had gone for this manufacturing solution), but I was also quite disappointed with the quality of the print. The images below demonstrate what I mean:

As can be seen in the red circles in the second image, the inside of the printed part shows clear printing defects. Despite the coating (mostly) covering those issues on the outside (look at the bottom of the left slot to see an easily visible defect), a board at this price point should be perfect inside and out. Furthermore, while the manufacturing tolerances I designed around were sufficient for making the parts fit into each other, I was still extremely disappointed by them, with straight lines in CAD being wobbly messes in person. Finally, and I can't quite put my finger on it, but AlSi10Mg sounds odd, seemingly amplifying structure born sound when typing, making the board a pingy mess (more on this when I talk about the integrated plate). Overall, the direction of the bottom was perfect, but it needed some minor revisions to be near the standards I'm looking for in this project.

Next comes the integrated plate, and note, the plate was not generated for this revision of the prototype: it sucked. The big issue with integrated plates, as many of you know, is they are incapable of dissipating the structure born noise caused by pressing a switch. Check out my very scientific diagram below comparing noise propagation through a case with an integrated plate vs. one with a full isolation gasket:

Every time you bottom out a switch, you cause the plate to vibrate at a certain frequency. When you integrate a plate directly into a case, nothing prevents those vibrations from propagating through the rest of the case and turning your keyboard into a resonator. However, when a gasket material is added between the plate and the case, it acts as a vibration dampener, preventing the propagation of structure born noise. With the Evolv, this was a particular issue, as these vibrations cause the 3d printed part to resonate, turning it into a pingy nightmare.

Finally, I personally found the Evolv's outer bezels to be too thin, evoking the look of a Tofu instead of a premium board.

So, what am I doing to fix these issues?

1: The Issue: The generative bottom is not visible enough from the side. The Fix: A complete redesign of the bottom, pushing the tendrils right to the edge of the board.

2: The Issue: Cost and Quality of 3d printing the bottom. The Fix: Investing into molds for casting the base, lowering per unit production cost by a factor of 5.

3: The Issue: Structureborn noise propagation with an integrated plate. The Fix: A full isolation gasket inspired by boards like the J-01 and Kepler.

4: The Issue: Thin top bezels. The Fix: Thicker top bezels, which as a bonus, accommodate the new mounting solution.

5: The Issue: Lack of plate layout options. The Fix: Separate plate with fixed layout options for ISO and ANSI.

The Evolv has truly evolved, and I want to thank all of you for your patience as I've worked through this redesign. Main OP will be updated with the new board design shortly, and following that, I will open a new IC poll with some questions I have for you, the community.

- Nathan <3


Why a blocked 75%?

Taking inspiration from the satisfaction75 and the 7V, the Evolv crams the functionality of a TKL into a significantly more compact space. Furthermore, by separating the F-row, arrows, and column from the alphas, the Evolv not eases your ability to navigate it, but also provides a much needed visual relief to what would otherwise be a block of keys.

You didn't touch on the 1.5u shift

I love blocked arrows; from both an aesthetic and functionality perspective, they make the board better. However, I've always disliked how blocked arrows don't quite line up with the right modifier keys, in turn forcing an awkward gap to exist somewhere. As such, the 1.5u shift and the 13u bottom row are a compromise: aesthetically, the arrows line-up and the .25u bezel is maintained, while functionally, the Evolv can fit the fourth key into what is typically a three key column on blocked 75s.

How about 2u left shift since so many keysets are missing 1.5u shift?

Incorrect stagger is much worse than a wrong legend, in my opinion. Hopefully, making this bold design decision will make other 75% designers follow suit, and at the very least, I know any keyset I release in the future will include the key.


Unfortunately not possible due to the multiple layout options on the PCB.


Underglow: yes.
Top: maybe.

The Plate Design Process

Quick disclaimer since I don't have a better place to put this: The plate is still a work in progress, I had 50+ iterations pre-IC and assume I will have 50+ more throughout IC. On iterations where I think I've made a significant breakthrough, I'll update the plate render to show off my progress.


Because, why not? We've had a lot of innovation in plate mounting over the past few years, but little innovation in regards to the plate itself. The first time I read the Iron165 IC, my jaw hit the floor. "Here were two people pushing plate design so much further than anyone will ever feel. Why? Because they can." I thought to myself. From then on, I wanted to follow in their footsteps. Building off of the work LBaron did for the Iron165, the Evolv pushes his ideas to the extreme to see just where keyboards can go.

What is Generative Design?

Generative design is an evolutionary approach to part creation. The process allows engineers to input design goals and constraints, along with materials and manufacturing methods. Then using AI and the power of the cloud, the algorithm explores all possible combinations of a solution, returning those that best meet the design constraints of the problem.

Wow that was a lot of buzzwords, how do you apply it to keyboards?

The first thing we need to do as designers is to shift our mindset to comply with principles required by generative design. The process generates brackets; therefore, we need to think of the plate not as a general mounting surface, but as a way to connect each switch to the keyboard, boiling it down to the critical essentials: mounting constraints and the 1mm surrounding a switch cut-out.

As you can see, the large green area represents everything constrained by the case, with the floating cut-outs representing where the switches click into.

Next, you need to tell the generative algorithm not to mess with areas you don't want it to mess with namely, the inside of the switch cut-outs and mounting points for stabs.

As mentioned, setting the constraint was a relatively simple task. Since the Evolv's plate is integrated, I simply need to make the outer frame a static element. For top-mount and gasket boards, mounting tabs would be the constrained element.

Configuring loading conditions is a simple yet tedious task. For the Evolv, there are 167 loading cases: 83 (one per switch) at the "correct" force per switch, 83 (one per switch) at 40N to ensure assembly won't destroy the plate, and 1 with all switches depressed at the same time at 2N.

*Correct force per switch: Mods are pressed harder than alphas. Alphas in the center are pressed harder than alphas around the perimeter of the board.

Now comes the annoying bit, and the part that, if possible, I would like to rework. Despite the 100s of emails I exchanged with professors, companies, and researches, there is no way as of right now to generate plates with the design objective of optimizing plane elasticity. I spent a long time trying to write the code myself, but doing so would've been a Ph.D. dissertation. Instead, you need to settle for the next best compromise: minimizing the weight of the plate with a deformation constraint at each switch.

Once loads and constraints are configured, it's a simple matter of creating manufacturing constraints and choosing how many materials to test. Of the 20+ materials I checked for the plate, 7075 consistently returned the best results for both typing consistency and strength.

The verification process

Verification of plate flex was done using no-penetration static simulations. Mock nylon switches were placed into the switch cut-outs and loaded individually to measure the total displacement during each loading condition. The theory behind this is the same as in the Iron165 IC, so I'll summarize the results below.

The current plate of the Evolv has issues with keys on the edge of the alpha cluster (Q, A, Z, }], "', ?/ ). I prefer to be transparent about this as I have no reason to hide that my generation process isn't perfect yet. Again, the Evolv's plate is constantly evolving :p, and you'll see it change throughout the IC. Anyway, those six switches are HUGE outliers in terms of displacement compared to the rest of the alpha cluster. However, if those six switches are removed from calculations, the Evolv averages .15mm of displacement per key switch with a .001mm standard deviation.

For your visual amusement, here are three selected keypresses:

A request from me to you

A part of the reason why I decided to call the plate good enough and post the IC for Evolv was that I started hitting a wall with where to progress from here. Please read through this write-up carefully and see if there are any holes in it or any details I forgot to cover. For the missed details, I will write in a blurb. For the holes, I'll modify my workflow and re-generate the plate.

Best board I've ever seen hands down. Big love :)
My hype for this is beyond anything else

this thing is wild

Hype hype hype


[0] Message Index

[#] Next page

Go to full version