If anything, the most complicated problem in lithography is not EUV at all. The hardest problem is the overlay mismatch. ASML had spent decades minimizing overlay mismatch. A new player in semiconductor may claim to master EUV, or X-Ray or contact printing, but everyone must solve the overlay mismatch problem. That is no more easier than the EUV part.
with modern Chinese manufacturing, you get 2-layer PCB with via and soldermask gorgeously made for $4 in 24 hours (actually samples can even be free in China). why bother make PCB at home with all the nasty chemicals or CNC ?
1. It's a hobby in your own right. The nasty is not that bad, back in the day the annoying thing was how fiddly it was but the community thought of better ways like etching with a sponge. You can get better boards made online but that applies to all hobbies.
2. In most of US/EU you get them for $5 in up to _two weeks_, or in two days for $5+$2X or more which adds up quicklyand can be more than the rest of the project. Even next day shipping loses out badly to making it yourself in ~30 minutes, turns into two days delay unless you're lucky to have an idea in the morning and time to work on it in the next afternoon, a lot of the time i see people bodging things i'd make another board for because it's just easier.
It might take 24 hours to make the board but shipping is at least another week of wait time. The dream is for a 3d printer like setup that sits on the desk and cranks out fully assembled multilayer boards
For sure. I tried to setup a collaboration environment for a Customer years ago using WebDAV over SSL in lieu of Dropbox. Everything worked great (authenticating to Active Directory, NTFS ACLs, IP address restrictions in IIS policy where necessary, auditing access in Windows security log and IIS logs, no client to install), but the Windows client experience was hideously slow. People hated it for that and it got no traction.
$3 ADCs are not cheap ADCs. Cheap ADCs are the one embedded inside cheap MCU. These days, TI sell cheap ADC/DAC that is artificially limited to 10/12-bit, but with linearity at more than 16-bit. The cost of producing bad ones versus good ones are not much different. They go through the same semiconductor process. So long as the fundamental design is sound, cheap ADC don’t perform any worse.
The opamp back in the days were pretty terrible. NE5532 was the king of audio opamp for decades until the early 2000s.
Modern Class D are built on advanced semiconductor processes (they are considered legacy node in the eye of Hacker News's primary audience. They are at least a lot better than the early days in terms of performance in analog domain.) When an IC company spend a lot of R&D money to develop Class D amp, they for sure exhausted what they can do before they tape out. That results in the superbe performance of modern Class D amplifier.
There is still oppertunities in getting analog Class AB type of amplifier working better, such as adding motional feedback control sensor-less or with sensor. KEF recently released a motional feedback soundbar with back-EMF voltage as sensor. It sure improve the sound quality for a soundbar. Although physics is physics, one cannot make a 1 inch speaker sounds like a subwoofer, but motional feedback sure can make 10 speakder sounds like a 15 inch subwoofer.
Sound reproduction is not just a flat frequency response. Perfect reproduction of phase information generates wider 3D sound stage, without the need of DSP to fake it.
problem with Lumafield is their pricing. Last time I checked with them. They don't sell the CT machine. It was a lease/subscription at yearly cost of $75,000. It was not justified for what we are doing.
But a conventional CT scanner costs $300k-$1m plus $50-100k per year for software licenses and maintenance contract. Lumafield’s basically giving you the entire system for just the annual software/support cost on an older-style machine. My company is considering one and it’s a more attractive model once you start comparing realistic costs.
I wonder if ultrasound techniques might provide an alternative. If it takes 10+ hours to do a CT inspection, as someone pointed out elsewhere (if I understood correctly), then that's a lot of DSP time.
For that matter, jeez, how long does it take to just whip out a Dremel tool and take the battery apart for inspection? I must have misunderstood that comment.
Scan time depends on material composition in the object you're scanning and your requirements for resolution. You can scan a dense steel object overnight to capture micron-level detail, or you can scan a plastic object in a few seconds to search for a known issue like a crack.
Battery scans are very fast; the scans in the report took less than a second. Total cycle time on a Triton CT scanner is under 5 seconds when you account for part handling.
ITX-Llama is good for young kids to learn computer, programming, etc. My first mathematical application was MathCAD 6 on Windows 3.1 with Win32s add-on, and that worked fine for me, and can work fine for young kids. My most used programming language in DOS was Turbo Pascal, and it worked fine for me and can work fine for young kids.
I see nothing good in limited game option. It was inspiring to play game and then want to also program some game. Or to crack that game. DOS games are easier to penetrate with ArtMoney and IDA than modern garbage with closed source servers. So it makes sense which games are playable on device, not playability of games as is.
Parents may spend some time with kids in network DOS games, discuss how it was to experience it first. And that Raspberry Pi, what is it good for? It does not have DOS legacy, it does not have Amiga legacy. It is disconnected from all great stuff of the past, and does not have own charm. Its games are either emulators or some open source games ported to ARM. An architecture that is new on desktops and who knows how long will it be. There is Windows ARM, but it does not run on Raspberry Pi. There is Mac OS ARM, but also won't run. Proprietary ARM is pushed to be replaced by RISC-V, so should we tie ourselves to ARM if it is doomed?
I mostly see Raspberry Pi in hardware. MT-32Pi for Roland sound emulation in ITX-Llama. PiStorm for fast CPU emulation in MiniMig. Not primary CPU for desktop. I don't know what to discuss with kids in Rasperry Pi. No past and no future.
I have at different times put out windows, mac & rpi as "lounge computer" used by kids (5, 7, 9 and their friends) and their main complaint with rpi is DRM issues, that's pretty much it.
Kids who have used multiple devices understand they can physically look different and the things you click on are in different places. Also updates can move stuff around. Things are just randomly different sometimes for fundamentally un-knowable reasons.
The operating system primarily exists to provide a button that launches a web browser; computers as a skill is ability to open the things you want despite bad ui/ux.
A browser is a thing that if you type in "crazygames" lets you play games. The other important buttons / magic words are "netflix" / "disney" / "youtube". Spotify also a core request.
Median child (<12) doesn't know what an operating system is, advanced kids will know about "apple" being different from "windows" but usually won't have a coherent mental model of why that would be. Engineer's kids know what Linux is.
Its been missing touch input for over 6 years now, still cannot believe their dev teams do not have professional pride to fix such an embaressing flaw.
