収束ゾーンでのライフ: Life in the Convergence Zone

Fuck, I hate the new generation of blog comment spambots.

See – some years ago, there was an XKCD comic about training spambots to make more and more accurate and relevant comments that ended with “What will you do when spammers train their bots to make automated constructive and helpful comments?” and “MISSION. FUCKING. ACCOMPLISHED.”

it was really pretty funny at the time

But now with LLMs it’s real easy to make spambots make reasonable comments that are actually talking about your blog posts. They’re on topic, they’re cogent, they’re positive – of course – and the ad is in the URL attached to the commenter name, thus far universally shilling one or another sort of AI tool.

Turns out, on topic and cogent botposts are still just noise of emptiness, since, after all…

…the ad is the only reason they’re there…

…and you can’t not know that…

…so…

…it’s all clockwork and empty, and…

…sorry, Randall. Turns out it’s not Mission. Fucking. Accomplished. It’s more Mission. Unfortunately. Accomplished, and the punchline this time is that the mission itself kinda sucked.

I do think that as this ramps up – which it absolutely will, I mean, it’s incredibly obvious that it will, I don’t even know how you write spam filters against this – Federated comments from bot-disallowing instances will be the only thing keeping blog comments usable at all. Fortunately for me, that’s where most comments on this blog come from these days, so… maybe ActivityPub will save this, too. You never know.

But holy shit, once those spam-blockers stop mattering, with LLM-generated boring-but-cogent comments start taking over, there’ll be no way to have even remotely reasonable blog comment sections on their own.

Not ones with actual people, anyway.

I was so excited about what’s now called LLMs in undergrad, too. The first time my incredibly primitive toy version talked back to me was like being struck by lightning in the best possible way, walking around in the lab shrieking “IT WORKS! IT’S ALIIIIIIIIIVE!” Then when I did my project demo, the head of the department was so disturbed by it that he literally left the room. Just walked out. I was expecting a bunch of Q&A and chat about the models and all the commentary I’d written about the kind of data you’d actually need vs. the bullshit 450-ish word database with made-up numbers I was using and talking about how to connect actual meaning to all this word probability chewing – you know, the actual hard part that the LLM people just decided to leave out – but insead he just freaked and left.

And now, well, here we are. Instead of incredibly cool game engines and entirely new computer control systems, it’s all just… “what can we fuck up today?”

Y’know, I kinda liked Dr. M – the guy who walked out – even if I did make fun of him at times.

Because, well, honestly…

…maybe he was the smart one about all this after all.

Posted via Solarbird{y|z|yz}, Collected.

I never did write up a final followup report on the clothes dryer heat recovery system, did I? A writeup on Version 3.2, which I promised over a year ago when writing up Version 3.1.

Well: it works! It works really well, honestly. The house is more comfortable with a little more retained humidity in the winter but never gets at at all damp – if anything, what we need is more humidity reclamation, not less.

We also successfully reclaim a noticeable amount of heat, same notation. I don’t have any reason to change my numbers, so it even works out to a net operational profit! It even saves money, on top of everything else.

It really does all just work. We got to use it from November until May’s heat wave, even longer than we did last year, thanks to last winter’s particular weather. The very minor strengthening was all it needed, though honestly I’m not sure it needed even that.

The way to “close” it for summer pressure testing turned out to be really simple: a thin sheet of cooking silicone cut to the right size, slipped inside the door to cover the filter intake, held in place by the friction of the door itself and air pressure. It worked perfectly; we had no sign of leakage, the rigid air ducts I used to connect it to the dryer and outside vent never built up abnormal amounts of lint, in short: the pressure testing over summer showed no issues.

Once actively in use this past winter I cleaned it weekly, but I could’ve easily cleaned it every second or third week, based on the amount of lint build-up on the inside of the filter. The furnace filter is still good and can still be used again for another year, no question; I think the charcoal should be replaced more than once a winter, since we did get more laundry smell over time.

Which gets me to why I’m probably not really likely to reinstall it this coming autumn.

Over winter, one of my housemates switched back to scented laundry supplies. They did so for specific reasons which are pretty reasonable, honestly. I couldn’t even criticise it if I wanted to. But… even with the charcoal filter layer, those scents started becoming more and more of a presence. Most of the time just in the laundry room, but when their laundry was drying it’d creep out to other parts of the house, and even when it wasn’t their laundry, you’d still get some of the smell just from accumulated build-up inside the vents.

And I don’t like those scents. I use unscented laundry supplies for reasons. It’s not nasty or something, but it’s artificial and perfumey and I just really dislike it. On clothing, the amount left behind is hardly noticeable – but in the air, it really kinda is.

So in the end: this is a solved problem. Version 3.2 works and works well. I have no meaningful new notes or plans version a version 3.3, much less a version 4, because… it’s done! It works! It saves money and energy and helps keep the house from getting too dry in the winter and it’s low maintenance and safe! It’s genuinely quite nice!

Unless if you have people using scented laundry supplies, and you don’t like the scents. Then… it’s not as nice.

You never know. Maybe I’ll reinstall it anyway. Try changing the charcoal layer every month or something nuts like that. Particularly if it starts getting really really dry, come January.

We’ll see.

eta: This post includes a bunch of photos showing the design. This is of Version 3, but the only differences between versions 3 and version 3.2 are more bolts tying everything together better. Refinements, not replacements. Air comes in the right side, mostly goes out the front panel, excess pressure goes out the left and out of the building, and that’s also the safety release in case somehow everyone forgets to clean the system for several months.

Posted via Solarbird{y|z|yz}, Collected.

So y'have a problem, right? and it's an electrical problem.

And you're like, "okay, sure, an electrical problem, I have some ideas, maybe it's partly a design problem, I can work this out," so you start taking the device apart, and you figure out it's definitely related to one particular subsystem, so that's good.

Then you poke at that subsystem for a while, it's very simple, but also, everything is okay? You still don't like the design, but otherwise, it's fine. So you go backwards up the chain, past the control system, up to the power system, and you poke at that for a while figuring that out as best you can, looking for voltage drops and not finding anything.

Now since this is a power issue you try throwing in a little power buffer into the DC side, but that doesn't help at all, which if you're honest about it makes sense at this point because you're not seeing any kind of power issue on that side anyway. Nothing 'til you get to the AC side.

And you figure out how to measure the problem on that side and eventually figure out exactly what's triggering everything and you're kinda getting worried and trying to figure out how big an electrical problem you have, when you go...

wait

...and you think about physics.

And you pull out the right calculator and you plug in all your numbers as best as you can estimate, and your estimates are pretty good since you know all this kit pretty well, and you hit "calculate..."

...and out comes exactly the number you saw in the wall, and you think...

...godt dammit...

...because you don't have an electrical problem...

...you have a physics problem.

And you go fuck, I can't fix fucking physics.

I mean, really, what I have is a design problem. or strictly speaking, what I have is two design problems, in both cases of something I didn't design and can't re-design.

To wit:

1: PWM heating on a 3D printer build plate is kinda goddamn nasty, but also,

2: Dimmable LED lights should not act like drama queens when seeing a cyclic 1.4v drop across a circuit, particularly when that drop only gets you down to 120 volts, @PHILIPS_LIGHTS.

I am just saying.

CREE doesn't act like that. They're fine.

But seriously pwm heating is stupid and going to create problems in residential environments and it sure would be NICE if it DIDN'T but it does.

Anyway...

...guess it's time to buy new lightblubs.

goddammit physics.

I replaced the hot-end fan on my 3D printer today, like I'd hoped to do. It's not gone silent like the PSU went, but it is better, and so I have an addition to my HEPA-filter comparison scale:

1: Original - printer much louder than HEPA filter on high

2: Replaced PSU - printer much quieter than HEPA filter oh high

3: Replaced hot-end fan - printer mostly inaudible over HEPA filter on high

Mostly today has been that occasional "okay I need to reorganise around my new processes" day, mostly on the 3D printing/electronics bench. It's something you need to do every so often when you've been adding stuff to an area for a few months and each additional bit you've done has been just that little bit ad-hoc, so you need to go okay, let's get this properly sorted.

And there are still a couple of questions to answer, but basically, now I have.

My longest-run print so far, 1 day, 7 hours, 49 minutes, 51 seconds. Basically one and a third days. It's nothing special, just an organiser I made to fit into a specific spot on the wall in my workspace of specific size, but I need that, so yay, it worked.


Ironically, it's with the most troublesome, annoying filament I have. (eSun matte ultra black.) I love how this filament looks the... one time in five times or so it actually prints. Even then it will lift off the bed in some corners.

(Literally the only 100% flawless print I've ever had with it - no flecking, no corner lifts, no separations - was the temperature tower I printed when I got it. Absolutely beautiful. Everything after that has been a nightmare. No other filament I've worked with has behaved even fractionally this badly.)

Version one of this was the first object I managed to print without bizarre spiky flecks sticking out of flat surfaces. I posted about that a while ago, the best solution I've found is "print objects with flat sides rotated 45 degrees."

Anyway, with the little bit I have left I'm going to test my 45° theory with an object that recreates the fleck bullshit 100% of the time, and also try my "maybe a raft with supports going up to the intended object hovering in space will work" theory at the same time, since I don't have a lot left of it anyway.

The chest part of the bench isn't going to be storage. It's a tool, in and of itself.

Combined with a heating pad, it's a warming chamber for the five or six months of the year when temperature down there is too low for too many materials. I can't get it hot with a heading pad, but I can get it more than warm enough to make glues (et al) work, and it's big.

(Though if I do buy myself that toaster oven, I may end up storing that in there. So there will probably be some storage, just not too much, because I don't want to have to unpack it to the ground to use it.)

My other accomplishment today was measuring the old garden chest that I'm using as half my workspace desktop and realising I could rotate it and not lose meaningful surface space - like, only a cm or so - and have the top lid function as a door that opens down, in the front, so I could use the inside of the chest for storage.

I'm going to put a supporting bar inside the chest to strengthen up the new top a bit more - I mean, it's the same thickness as the old top, but with one side not really supported. (Even if the other three sides are actually supported better. I'll just add a crossbeam, it'll be fine.)

Anyway this gives me a bunch more storage and I have some scrap lumber that'll work a treat for the job so I should be able to pop that into place tomorrow and it'll be good. ^_^

ANKER's 3D printer kickstarter - which they're running solely for publicity, they don't need this money - is going around with a bunch of paid videos (LTT is at least upfront about it, but I still don't like it) and I have to say... I'm not real impressed.

Given what's available now at what costs are now, I was really disappointed with that combination of apparent bed size and price. I mean, no, it's not an unreasonably small bed, but I saw the size of that base and was all, "You have my attention" and then I saw the print area and was, "...and you just lost it."

Plus, they are making some really inappropriate speed claims. Not because I don't think they can move the hot end that fast, but because that's how PLA works. (Or any of the other filaments.) They seem to be talking about just linear movement speed and that doesn't matter if your material can't cool that quickly.

I mean, I'm sure it's faster at defaults than my Ender 3 V2 but that's not hard, and at well over twice the price, I expect better.

And all those are big issues, but there's a bigger one, and that's the whole sales pitch.

They want a plug-in-and-forget 3D printer, and they're basically advertising this as one, focusing really hard on ease of assembly (great, I'm for it) and the included auto-leveller (great, those are common, I don't have one and don't need it, but they're good), and they're spending a LOT of time on what they call their AI-driven spaghetti/failed print detection.

And that's fine, but they're making it sound like those are the unsolved problems, when those are in fact the already solved problems.

And a big one is filament. Filaments are not consistent enough for this kind of approach to eliminate all the variables! Particularly not in an open / uncased environment, like this printer! If it had an enclosure, I'd be a bit a bit more copacetic about it, but not entirely, and it doesn't.

Does it have some sort of environment temperature sensor to make temperature adjustments based on ambient temperature? Not that they've mentioned, and I don't see one.

What happens when someone throws some HT-PLA+ at this thing, just to use a filament type I like to use. It can't tell that it's not standard PLA, but the temperature profile is going to be completely different, and not just for this type in general, but by the maker! Each company's HTPLA and PLA+ and so on - they're all different, and there's no way to communicate that to the printer from the filament itself, or the spool.

And changing filaments between jobs, too - it's using standard i3-style printer print heads, and those _need_ to be cleared between filament types. They just do. PLA to PLA generally isn't too bad, but changing colours is not just like popping in a new inkjet cartridge. I have some low-temperature black matte PLA and you need to get that head _squeeky_ clean between prints or you are going to learn new meanings of pain if you switch to something different! Particularly something that operates in different temperature ranges.

And that's just one example.

Here's another: nozzle types. I mean, I guess they can have some sort of hardened-steel plated copper nozzle and that's fine, and that will last a while and tolerate abrasive filaments like wood-infused PLA, but... that's at best a semi-solved problem. Changing those out isn't too difficult, but with that kind of hot end - exactly the same kind I have - you need to get up close and intimate with 200°C metal. (That's... 390° in Freedom units.)

Not to mention the adventures in cleaning it out. By going direct-drive they've eliminated the Bowden tube issues, but... well, we'll see.

Which is all a way of saying "great, they've solved assembly and they're including a levelling sensor with the kit." And they've got a spaghetti detector to stop you from wasting too much filament, that's nice. But that doesn't make it print-and-forget, not by a damn sight. There are all these other very serious issues that they haven't touched and which are not solved here. And in my experience, they're the bigger problems.

But they're kinda selling it as an "all problems solved, print and walk away" solution.

And it's really not.

So I gave in and admitted I need a drill press - even a small one would do - so last week I used some rewards points and ordered a bench unit, and it's showing up tomorrow, and today I'm trying to solve a build problem with Prototype 3 when I think, "Really, I could solve this with a drill press OH WAIT" :D So tomorrow is now exciting because DRILL PRESS ARRIVAL! so yep small excite :D

Right, so, one month in and here's what we've got on prototype 2a at this point.

1: As per the one-week update, it still reclaims a lot more heat/humidity than prototype one. Numbers are within what I'm going to call margin of error, with gains between 1.6°C and 2.1°C depending on who is doing what laundry - with 5ish points of gained humidity. As before, we never exceeded bad limits, though with one particularly long day of laundry from a housemate (which went into the dryer extra damp) I did have to turn a fan to accelerate dispersion of temperature and humidity into the rest of the house.

2: The second fabric screening material I decided to try - cut from one of my dryer bags - appears to be very effective. Weekly cleaning is optional. Biweekly cleaning is asking for reduced effectiveness. Monthly cleaning is probably out.

3: But on another(?) plus side, you don't need a vacuum to clear the lint! You can do it by hand, just like with a dryer lint trap. (I used a vacuum anyway after determining this, for neatness.)

4: As before, hinged access to the filter (lets you vacuum lint off the fabric screening material) is a necessary feature for convenience.

5: Repeat for my own notes basically: a filter this much larger needs central-axis support.

6: Annealed PLA is much more humidity/temperature stable than basic/raw printed PLA. It's staying flat even in a configuration without much reinforcement in that direction. I'm suspecting

7: MERV 13/MPR 1900 continues to be more than good enough for dust control. There's no sign of dryer lint or dust anywhere.

8: You really don't want to use scented dryer sheets with this kind of setup. I didn't expect to encounter this, but a housemate grabbed some because they were out of unscented and we all learned that was bad. (They didn't like it either and got more of the proper kind.)

But if you like scented dryer sheets, you definitely won't want to do this unless you really like that scent.

Items 7 and 1 combine to tell me that we can expect some real durability from a MPR1900/MERV 13 filter, particularly if combined (as here) with a charcoal prefilter and a lint-trapping fabric on top of that. I'm really not seeing any sign of occlusion - and I suspect the charcoal prefilter part is pretty optional, given the lack of buildup I'm seeing in the charcoal. I do think it's getting moderately saturated scent-wise, but most of that is probably that accidental two-week introduction of HEAVY HEAVY SCENT MACHINE dryer sheets.

We're heading into March, so this series of winter testing is probably about done. We'll have a couple more weekends, I imagine, before the humidity additions become counter-productive, but that's most likely all we've got.

However, from this month of testing, I do feel confident enough to move to a prototype 3, which will not be mostly made out of cardboard (lol I hope anyway) and which will need to include some sort of shutoff door. The primary filter isn't showing any signs of clogging after a month; output is being maintained pretty well, and we aren't getting leaking.

I suppose the next question is whether something like this is worth the effort. It's not lengthening the time required to dry clothing, so there's not a downside other than initial construction and operation. It does gain us measurable heat and humidity gain in winter, both as per goals. So it does work.

But does that intermittent (three days a week) additional comfort / heating cost reduction pencil out against cost of materials, including annual filter replacement?

Honestly, I don't know, and don't really have the training to know or the facilities to find out. Obviously, this is no way to heat a house, it's not close to enough and is not exactly what you'd call efficient as a primary system. But in this case, where it's a matter of recovering otherwise-lost heat and humidity resources, it's kind of infinitely efficient. It successfully retains heat that would be otherwise lost to the outside, and in our climate, it does help reduce the dryness in the building, both in amounts measurable (and not subtle) in our graphs.

Ignoring construction cost - is that worth an annual $20 or so for the filter? Is the reduction in carbon output from heating balanced by the carbon cost of making that filter? So far, I don't know.

But furnace filters aren't hard to make. These aren't high-energy-use items.

So maybe.

ETA: Of course there are extant cost of operations numbers for clothes dryers! How did I not realise that?

Based on those average numbers, the cost of heat for our dryer is around $72 for four months, give our typical number of loads. That produces dry clothing, waste heat, waste humidity. Based on surface area of exhaust tube (4" diameter) vs. filter (169" inches) about 92% of that is being recovered, let's round that down to 90%. That's $64.80.

To break even with $20, we need 30.9% of that to be heat recovery to break even on cost of one filter.

I don't know how to calculate that here, but it seems a lot to ask that hot-air-on-tumbled-clothing have a continual 69.1% drying efficiency. I have no idea what the actual number would be, but that seems high. Like, probably a lot high. Particularly given the lower-than-I-expected relative humidity of the recovered air.

And what all that means, if I'm right about drying efficiency, is that, in fact, we're coming out ahead. Maybe by... a lot.

If that efficiency is around 50/50, then we're getting $32.40 worth of de facto electric heating out of this (discounting filter cost for a moment) which translates to about 295kWh at our rates, which translates to roughly a 1500W space heater's worth of electric running a bit under 11 hours a week...

...which gets us very close to back to our number of loads of laundry.

Huh.

So... yeah.

This might actually pencil out!