Ask HN: Ideas for Hardware Project | Hacker News Reader

Ask HN: Ideas for Hardware Project

13 pointsroschdal8 comments

Please share ideas for useful hardware projects relevant for new hardware startups.

- software-defined IO; see: software-defined radio, software-defined networking, reconfigurable MIDI interfaces, using VR headsets as monitors, etc. even to some extent, eBPF embodies this idea

- hardware that enables probabilistic data structures, whatever that means, for even faster approximate execution of algorithms

- memory that executes instructions in place, avoiding data movement latency

- differentiability is big right now - differentiable/invertible hardware could be pretty dang cool

- state of the art hardware in pretty much any field is pretty much guaranteed to be extremely complex - greater exposure of underlying processes (hopefully in a universal / composable reporting interface) could allow easy + deep/recursive + simple inspection/visibility of hardware as a component of a massive-scale system

- extremely high density + speed + reliability + "hook-ability" + expandable storage media

- formal verification of data IO at a hardware level

- physically unclone-able functions for e.g. encryption, device verification

- photonic computation

I think that there could be a lot of innovation in the field of data diodes. Allowing data to exit a network at speed, while allowing zero infiltration of control, is a very powerful tool for monitoring, and backups.

I believe that a very large array of 4x4 bit lookup tables, clocked like checker boards, could be the key to scaling past the Von Neuman architecture. There are numerous advantages in terms of provability, scalability, security, hardware fault tolerance, and compilation.

The design space for small 5 axis machine tools hasn't been completely explored. Imagine a set of 6 actuators, and a spindle, which can be bolted onto something large that needs to be machined, like the portable milling machines used for key seating, etc. The legs are in pairs, forming the classic 3 point support used to lift flight simulators, etc... and your spindle sits between them. Equip it with a standard R8 or similar collet, and a probe. You could then put a 1-2-3 block on the object to be machined, and probe it to calibrate the axis.

I believe that it is possible to manufacture a custom silicon chip using wet electrochemistry in quantity one. Piezoelectric transducers could be used to move a probe to do etching or deposition of material. Aluminum vias would have to be replaced with copper.

Wrt using a lattice for data flow and exploring nonvon architectures you should check out Dave Ackley [1]

I like the idea of composing Stewart Platforms, they really show the power and versatility of linear actuation.

Could you talk more about using wet electrochemistry for fabrication?

[1] https://www.cs.unm.edu/~ackley/

[2] https://en.wikipedia.org/wiki/Stewart_platform

Electrochemical etching can be used to selectively remove material, and if a very small probe were used, perhaps it's possible to directly etch a channel where required to make a transistor, or to apply a voltage gradient to help cause a layer of silicon dioxide to form where insulator is required.

Sounds like it might be another technique to use with the epitaxial process.

I'm have no idea if it's possible, but I've been thinking about encrypted walkie talkies, whether that means end-to-end encrypted, so there can be multiple "channels" over the same frequency, or simply to add another leveled of protection by using a shared key

Ah, so I think this is certainly quite possible- but the FCC has banned encrypted amateur radio channels, I believe

If we believe the Metaverse is about to be big, then something to support that.

Some sort of input device? Perhaps a feedback device?

Some technology that would be attractive to a larger company, even if it’s not perfected.

Reply on news.ycombinator.com