Notices by 🇳🇴 Thor — backup account (thorthenorseman@octodon.social), page 101

@KnivesOnly Several of the article comments are pointing out that the ion channel issue has been rediscovered many times over the decades, but it keeps being forgotten again because they can't seem to find the *cause*.

I'm not sure what to make of ME...

@bamfic @SuperFloppies Modern CPU pretty much *is* RF design. The only type of analog circuit where things get this ridiculous is when designing RF amps, mixers and filters for microwave (>1GHz) RF circuits, such as those found in WiFi radios and satellite dish LNAs. At these frequencies, lines behave more like whips than rods...

@SuperFloppies With the clock in the center of the die, the delay would actually be a more comfortable 1/12 of a clock cycle. Still not anywhere near the spacious edge timing tolerances one would ideally want, but well, it seems to be sufficient.

@SuperFloppies If there is excess line, one simply makes folds or zig zags. Doing this on a chip die is possible too, of course, but that doesn't help you much if a trigger signal on one side of the die depends on an instant (within the same half-cycle) response from the other side of the die. Digital logic depends heavily on such signals. With clock signals on CPU dies, one tries to have the common source in the middle of the die, like with the droplet in the pond.

@SuperFloppies It's simpler when a single signal needs to travel to equidistant points in only one direction. Think of a droplet making a ripple in a pond. If you stretched 16 wires of equal length in a star pattern emanating from the droplet, the ripple would reach each end simultaneously. To get the same effect in a CPU chip package, one simply makes 16 equally long clock lines emanating from a single clock generator.

@SuperFloppies That's called bouncing. Here are two diagrams showing them side by side. Bouncing is basically rapid switching. Ringing is a form of oscillation, not dissimilar to the effect one hears when clapping in certain rooms and it "sings" a tone back at you.

@SuperFloppies It looks like modern CPU dies are about 13.45mm on each side. That's 1/6 of a clock cycle at 4 GHz, so I'd say they're really pushing it. If I were engineering a digital circuit, I'm not sure I'd be comfortable with a 1/6 cycle delay. That's 1/3 of the half-cycle being used to ramp the signal up or down. I guess it works so long as less than 1/4 of the cycle is used for switching.

@SuperFloppies Fast digital circuits start hitting problems one might describe as analog. Having sufficiently crisp switching edges is one challenge. Make them too crisp and you might induce ringing (little ripples on the edges). Another challenge is the speed of light. At 4 GHz, a half a clock cycle fits across 37 mm of wire. Having components too far apart can induce significant synchronisation issues, thus the shrinking of chip sizes to get higher clock rates.

@SuperFloppies A gate doesn't need to have exactly 0V or 2.5/3.3/5V at its input to see a 0 or a 1, and such errors don't propagate to the outputs. You could say that every gate in a digital circuit doubles as a signal cleaner. Timing errors *can* propagate, which is why digital circuits can get unstable if the switching rate, as determined by the clock rate, gets too high. Switching generates heat—the faster the hotter—thus the need for cooling.

@SuperFloppies There is an obvious reason for the reliability of digital circuits. As you correctly point out, digital circuits are built with analog components. The trick to reliability lies in switching thresholds. If a gate in a digital circuit has soft switching edges or is off by a half a volt, it doesn't matter. Once the signal passes through the next gate, it will get straightened out again.

@shamecrystal Digital circuits are far more straight forward, but therefore also less exciting to design.

@shamecrystal The analog circuit designer's job is to manipulate these chaotic electromagnetic eddies and currents in useful ways. It's very unlike software programming because getting approximate answers is unavoidable. High precision is possible, but usually requires a lot of compensation circuitry and manual calibration, and that calibration will only be valid at the temperature you made it at, because many components are temperature sensitive.

Analog circuit design is an exercise in compromises. You can only ever get a circuit to do approximately what you want, and every component you use has non-ideal properties. A diode is non-linear if there is <0.6V across it. Components have approximate values and can only dissipate so much power. Everything has thermal noise and parasitic resistance and capacitance. Getting it all to work is equal parts art and science.

I've been designing and tuning the peak detector of the 2-source 5.1 channel summing box I'm going to build. The detection envelope can be seen in green, the horizontal lines indicate the threshold voltages for 0, -6, -12 and -18 dB, and the square pulses are the LED on-off control signals for those thresholds. The waves are curved at the top because I'm inducing hysteresis into the comparators to prevent from from switching with the ripples on the detection envelope.

Simulation of circuit for 5.1 summing box, with two channels active. The green line a probe on the level meter. It tracks the loudest channel at any given time. I didn't originally plan on having a level meter, but I decided that the user should at least be able to check if any of the channels are in danger of clipping, since this thing is USB-powered and doesn't have a lot of headroom. Do consumer sound cards generally stay below 1.41Vp-p on their line outputs at max volume?

@tootapp Glad to see that development is continuing. It's coming along nicely. But, here are two deal breaking issues:

- Instance switcher: It's one of those neat but impractical ideas. It takes considerable effort to use it correctly.

- No top tap: Nowhere brings me to the very top of a feed instantly.

These two issues absolutely MUST be fixed before this can be my main client. They are what I'm looking for with every TestFlight update.

@tootapp Glad to see that development is continuing. It's coming along nicely. But, here are some two breakers:

- Instance switcher: It's one of those neat but impractical ideas. It takes considerable effort to use it correctly.

- No top tap: Nowhere brings me to the very top of a feed instantly.

These two features absolutely MUST be fixed before this can be my main client. I'm hoping for them with every TestFlight update.

@idesofmerch Ah, yes, same story here. Favourite book until someone said "I prefer Terry Pratchett for that kind of thing."