Hacker News new | past | comments | ask | show | jobs | submit | tomooot's comments login

I would imagine if the design/assembly information was broadly available (internally) in the past, there's probably one or several "digital twin" emulations of the craft, or at the very least specific subsystems of it's computing resources. There must be some kind of analog/simulation of it's software just for proving "bugfixes" before upload, like the coms error and subsequent setting of the "solar system record" for "furthest distance remote code update" earlier this year.

There isn’t a simulator or digital twin for voyager. It has a bespoke processor made with 74* style logic. One guy will puts together a command and they will have a review where the other engineers will try and independently verify it. Then they copy and paste the command somewhere to “run it”. It happened, fairly recently, that the command had a typo that was caught in review, but the “wrong” pre-review command was used and the attitude became off by so much that they lost contact. It was only by cranking up the power at Goldstone that they got a command through. This fundamentally changed their understanding of the largest angle for which they could still communicate with the spacecraft. They just hadn’t wanted to try larger angles before because it was too risky.

That's the great thing about such a simple design, you can actually sit down with pen and paper and verify operations.

Forget about a digital twin, they don’t even have an assembler for the CPUs. It’s all hex values.

Most racing bicycles are right up against the limit of the minimum weight allowed. Sometimes they even add ballast weights to stay over the limit, and fit them as low as possible as lower weight = better cornering performance due to moving the bike's center of gravity down/fore/aft.


Tiny aerodynamic gains from lack of cabling. But also allows more flexibility (sometimes literally, as anisotropic properties can be engineered into the layup) in the design of the composite material frames, as you don't need to worry about leaving room for smooth cable runs and the stress risers caused by holes/ports used to route them internally.


$100 is plausible though, and if you had some outdated mining hardware (still likely to be single use ASIC hashing setups rather than computers) along with a lack of morals/ethics, then I can see it happening for sure.


One example that springs to mind is the documentation site for WLED, an incredibly powerful microcontroller firmware for controlling addressable LEDs:

https://kno.wled.ge/


As far as I understood the explanation, the incoming ("external") 50Mhz clock signal is a core requirement of the spec: all of those workarounds are just what is required to meet that spec, and be able to TX/RX using the protocol at all.


The squeeze is mostly within the segmentation of VRAM between products, it's basically a commodity and this week the spot price for 8GB of GDDR6 has varied from $1.30 to $3.50 [1].

Yet to get a card with 8GB more than one with comparable logical performance, you'd be looking at hundreds (or thousands in the case of "machine learning" cards) of dollars.

[1] https://www.dramexchange.com/


I've converted a £10 trash tier quadcopter to run off of 30G wire. Only 5 meters range, and the weight and momentum of the wire cause extremely difficult to control oscillations, but it still flies (badly)! The transmission losses over such thin wires are massive but the way around that is to supply a higher voltage than is required, and include a linear regulator inside the quad to burn off excess voltage when current draw is low.

I specced the wire purely based on the weight reduction I got from stripping out the cameras, battery, and then replacing the main body with a 3D printed skeleton, as the original battery contributed more rigidity than I had realised.


Sounds like you were supplying power. That's unnecessary in this case. Just leave the battery in.

The question is if you can reliably push video down a 5 mile pair of copper wires and send control inputs back up. And the answer is yes, even for an unshielded twisted pair. With the right encoding and enough voltage, you can get GBits down that wire.


I think modding communities are a reasonably good analog to this. Up until someone realized you could change the alpha channel of wall textures to make them transparent, the way to get a custom skin in Counter Strike Source (and 1.6), was to drop a file into your install directory.

The game would check for files named the same as it's default resources, read them in for use instead of the original. But then people started using the aforementioned transparent walls in competitive matches, and so a new variable was introduced to force the use of the defaults.

The next game (CS:GO) provided skins through a marketplace, including the use of loot-crate mechanics, the prices of in demand items sky rocketed, they are now used as currency for hackers and online gambling, and the online skin gambling sites have been caught advertising fraudulently through streamers. "rare" skins can sell for tens or hundreds of THOUSANDS of dollars.

In short, a great feature got exploited, commercialized, more exploited, and inspired a great amount of profiteering and sketchy business practices while ruining community aspects of a whole genre of entertainment (lots of copycats).

What I'm trying to say, is I think those paradigms most likely would have been co-opted by third parties in the name of greed and profit, destroying the communal and humanistic aspects of them in short order.


Neural nets using individual tubes as nodes? Although the current trend seems to be quantizing down to a minimal amount of bits to process more in parallel, in an analogue system you could have a near "continuous" range of values.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: