generuso

I was thinking earlier how the comments or annotations could be implemented in the right way, to fix the problems with the Youtube's comments system. It is certainly not trivial, and one can argue whether this is even desirable. But it would be cool to be able to reference parts of different videos and other resources more fluently.

For example, Kurzgesagt channel prepares their mini-lecture videos very carefully, and they often have a document on their site, referenced in the description under the video, and which annotates the video, and justifies the choices they made in how to present things. The document gives quotes and references to academic sources. If one could navigate and update narratives like that, sort of like a wiki, with one-click references to the relevant snippets of the video, maybe that would work. But even for their videos, the comments section in Youtube is largely useless.

World wide web was unleashed by putting together in a seamless way a few things which were already common before it. One click hypertext navigation was already a standard thing for the technical manuals locally on PC. But making it work on-line, instead of typing several strings in command line made all the difference. Maybe today we could tweak the commenting system to make the comments much more convenient to navigate than they are today. Keep similar comments from multiplying unnecessarily. Allow finding and linking relevant information. I do not have a thought out solution, but there is definitely some room for innovation here, if one approached it imaginatively.

It is rather stunning that PCI signalling works despite this style of wiring, but the proof is in the pudding! Dealing with the wire inductance and the interference between the signals could easily be the hardest aspect of this implementation. Probably much harder than designing the logic itself. IIRC, the absolutely minimal working PCI hardware and software could actually be rather rudimentary, if one did not need implement the entire PCI specification in a compliant way.

There used to be small IBM PC compatible "palmtop" computers capable of running for many days on a pair of AA batteries. HP 200LX was pretty cool. I think it had 80C186 running at 10 MHz or thereabouts, with 1MB of RAM, monochrome graphic LCD and PCMCIA cards for storage. Norton Commander and Turbo C run on it without any problems.

Sam Zeloof's new company is a serious business. He went from studying electrical engineering at CMU and doing semiconductor fabrication as a hobby, to running a well funded team to produce equipment for mini-fabs.

It is actually a great example which shows the difference between cobbling something together in a garage, to produce a one-off demonstration, and having a mini-fab which is capable of delivering the results reliably. The difference in the required effort is many orders of magnitude.

The "RAM in a shed" guy started with a Ph.D. in semiconductor technology, and was running a company before he made the improvised lab at home and started to make videos about it. He uses the lab to test some of the steps for the specialty devices which he invents.

The history of precision machinery is a steady ladder of using old machines to produce new more accurate components, and then new more accurate machines.

But simply making something smaller is not synonymous with it automatically becoming more accurate or capable of achieving higher resolution. If anything, very high precision machines are typically larger in size than the less accurate ones.

The guy with the shed is a founder and a CTO of a semiconductor company, and uses his home lab for business.

Modern satellites, in particular Starlink, use optical transceivers that are identical to the ones used for the fiber optic links, with an addition of all the required parts to make them work across free space. So a link between a pair of satellites is very comparable to a single pair of optic fibers.

The cables can go the entire distance between the data centers. The satellites probably connect to their closest neighbors, and any long distance transmission would have to make multiple hops, and share the bandwidth with everything else in the system going through the same nodes.

Since there are hundreds of undersea cables, with a dozen fiber pairs or so each, the cable system has several orders of magnitude greater throughput compared to the current constellation of satellites.

The cable contains not only the fibers, but also optical amplifiers to compensate for the losses. Even for 24 fibers, the amplifier is a substantial and expensive piece of equipment. It also requires electric power, which is not trivial to provide along the entire span between continents.

A very good 10 MHz ovenized crystal oscillator, Hewlett Packard 10811D, ages by up to a few Hertz per year, and has a mechanical capacitor for trimming the frequency by up to 10 Hz, and an electronic frequency adjustment by 1 Hz using a varactor. I am sure the varactor does not improve jitter, but considering that the full range of adjustment is 0.1 ppm, it also should not add that much jitter, assuming everything is well designed.

So we are talking about a reasonably high stability crystal to begin with, and a very narrow adjustment range. In an atomic clock, the feedback loop uses the electronic frequency adjustment to more or less completely remove the aging. This requires a very tiny and a rather slow acting feedback.

In the atomic clock, the output of the crystal is used as a reference for a microwave sweep generator, which then scans the spectrum of atomic transitions. The absorption peak in a cesium clock is something like a kiloherz wide, but with a good signal to noise ratio and with a lot of averaging, one can measure the position of the peak to a very tiny fraction of its width. Comparing the measured and the expected positions reveals the deviation of the reference frequency from its design value, and that is what generates the tuning feedback for the crystal. I do not know off the top of my head how slow the feedback is exactly -- it is something that one could look up in the Hewlett-Packard service manual, but I am sure it is slow enough to be irrelevant for the cycle-per-cycle jitter.

There are different types of atomic clocks, but in most common types, the output comes from a crystal oscillator, or more generally frequency synthesizer, which is then slaved to some spectral feature in the "physics package". That is to say, the phase noise is as good as that of the crystal in the unit, but the longer term frequency stability is much improved by the slaving.

An exception is an active hydrogen maser, which directly outputs the frequency of atomic transition. It has very good phase noise, but is a rare beast, which is only used where it is absolutely necessary.

The cone of Cherenkov radiation produces arcs of "bright pixels" in the detector.

There are very good illustrations of this if you search for "Super-Kamiokande Cherenkov ring." That's a Japanese neutrino telescope with a bunch of huge photomultipliers lining the walls of a giant underground cistern filled with water. Ice Cube has similar light detectors, probably from the same Japanese manufacturer, that were lowered on strings into wells melted into ice.