prova_modena

I'm going to push back on this statement, as I spend my day-to-day more immersed in the non-techy car world than in the HN tech world. I see a lot of frustration with new tech features among car reviewers and enthusiasts. You can read almost any car-related social media forums and see ire directed at stuff like complicated touchscreen-driven interfaces and flaky emergency braking systems. I also know many older folks who simply ignore and avoid all these tech features that now come standard on luxury vehicles. Anecdotal evidence, I know, but I don't think there's widespread acceptance of this stuff as you imply.

You're right, car dealers/manufacturers do use options packages to increase profit margins by bundling less popular high margin features with more popular ones. The number of brands that let you spec out a new car using a truly a-la-carte selection of options has shrunk, and is now mostly confined to the ultra high end brands (which make huge margins on everything anyway).

However, it's not a new phenomenon, manufacturers and dealers have done this with "luxury" or "sport" features for a long time. This has usually meant adding stuff like alloy wheels, different interior trims, better speakers etc. The difference now is the huge amount of complexity that modern "smart" features bring.

I recently decided to refresh my statistics knowledge after about 10 years of not touching it. My last experience with stats was during my (non-math/stats) undergraduate degree when I took some statistics coursework and ended up working fairly closely with a statistics professor during my thesis. Please keep in mind that the following recommendation comes from this amateur level of knowledge.

I read a HN comment that mentioned "Hypothesis Testing: An Intuitive Guide for Making Data Driven Decisions" by Jim Frost and picked up a copy. This book is part two of a three book series by Frost, first of which is an introductory statistics book and the last of which concerns linear regression.

I'm about 3/4 through the Hypothesis Testing book and am so impressed I just ordered the other two in the series. This book is the companion I wished I had during my undergraduate work. Reading it, I realized how much I partially digested or glossed over during my stats course. It also is a good way to reassess those statistical practices that I was taught during non-stats courses as "just how we do things." Frost is good at concisely explaining the appropriate uses of different kinds of statistical tests and their pitfalls.

What it lacks is implementation details- the book contains very few equations and no code (although I believe there is downloadable example data available). For my purposes, this works great as I most needed an overview of the different hypothesis tests and their suitability for different situations. I now have more confidence about choosing and understanding a test at a high level and can dive into specific implementation details in other sources as needed.

To be perfectly clear, I have no designs on becoming a professional statistician and I'm re-learning stats just to support a couple side projects and generally broaden my capabilities. Someone whose primary interest is the math and implementation of statistical calculations will be better served by another book.

I'm not who you are replying to, but I would recommend the textbook "Machining Fundamentals" by Walker and Dixon. It's the best 101-level introduction to the topic that I have found.

The tool manufacturer Sandvik also has a decent free online set of courses that I found helpful.[0] I have also heard people mention the Titans of CNC online training courses,[1] although I have not checked them out personally.

Finally, "Machinery's Handbook" is also a venerable resource that deserves mention, but more a desk reference than something to read through as a learner. They come out with new editions regularly, but older ones are still very useful (and have better quality bindings).

Overall I would recommend seeking out resources that are addressing a professional audience, rather than a hobbyist one. Hobby-level discourse and instruction about machining varies widely in quality.

[0] https://www.sandvik.coromant.com/en-us/services/education/pa...

[1] https://academy.titansofcnc.com/

Fadal VMC40 CNC mill. Mine is a 1984, making it one of the first ones made.

This was one of the first wave of CNC machines made and priced for a smaller job shop environment. It's not large (axis travels are 20" X, 16" Y) and it's not fast by modern standards (slow travel, max 10k rpm spindle speed). However, it's easy to use, not over-complicated, rigid with box ways, and DIY repairable. There are several places to get affordable parts and support. I have yet to encounter a component that is "no longer available," a common occurrence on other machinery. I've had a specialist out to inspect it when I got the machine, but other than that I've been able to fix and maintain it easily. Besides, very little has gone wrong with it, and I started running it hard straight out of a 20 year hibernation.

The control electronics are built with a backplane design that makes it extremely simple to swap and upgrade components. The control boards themselves are made with all through-hole components, making board-level repairs easier. I upgraded the main control board from a '84 version to a '95 version in about 30 minutes (gaining speed and a lot of UI improvements). If I wanted, I could retrofit a fully modern control system in a couple hours. If you're not in the industry this may not sound significant, but if you're familiar with CNC machines this is huge for prolonging the life of a machine. For comparison, older Haas control boards are known to fail and the only current recourse is to pay Haas about $20k for a full retrofit by a factory tech.

The interface is text based and runs on a CRT but it's incredibly simple, ergonomic and easy to learn. Discoverability is excellent, with built-in documentation and useful error messages. If Siemens and Fanuc CNC controllers are emacs and vim, Fadal is nano. If the battery backup goes bad, I can restore the control parameters using a human-readable menu with meaningful variable names in about 5 minutes from memory. On my Mazak lathe, this takes hours of entering hex codes into numbered fields. In comparison to other machines I've used, the Fadal feels like it's made to be used by machinists, not gatekept by factory service techs and expensive training courses.

The worst part of the machine is that it has a total loss mist lubrication system for the spindle, which requires added ventilation otherwise it will fill the shop up with oil mist. I would also like a full enclosure, which mine doesn't have but later ones do. But other than that it's incredibly resilient, high quality and simple machine, the likes of which simply aren't being made anymore.

I'm also a fan of hand planes. It's worth noting that Lie-Nielsen Toolworks currently makes very, very nice updates of these old designs with blades made of modern tool steel. Arguably they are the best bench planes ever made, even though old Stanley planes are superb. Here's Lie Nielsen's version of the No. 8.[0]

[0] https://www.lie-nielsen.com/products/no-8-jointer-plane?path...

I spent a bit working at a custom stairbuilder. Handrails that smoothly curve in three dimensions, either to connect straight sections or along a curved stair, are called "tangent" handrails. Constructing these tangent handrails is challenging. It requires both geometrical rigor in devising the plans as well as an individual craftsperson with a very high level of manual/visual skill. While we would work out the curvature of the rail in CAD or paper plans, a huge amount of implementation detail was left to a traditional woodworker, such as carving the profile of the tangent rail to create transitions pleasing to the eye and touch. Additionally, the variable structure of wood grain meant that the choice and orientation of the wood stock often had to be done piece-by-piece, often with custom fixtures.

I was a resident computer nerd there, so one of my projects was adapting a 5-axis CNC router to rough out these tangent rails. We were successful to some degree, but the fixturing and programming for individual tangent sections was very complex. It worked out that the CNC approach was only economical when the design of the stair required many multiples (rare in our case). Otherwise the skill and adaptability of individual craftsman was much more efficient, and the CNC was relegated to constructing jigs or 2D pieces.

My training was in high-end artisan furniture making, and the design challenges I saw in custom stairs were way more complex than almost any other field of woodworking I am familiar with, including the more technically avant-garde furniture makers. The only clear exception is wooden boat building, which appropriately enough was the background of many of my coworkers at that company.

Wrote a python script to extract data and produce spreadsheets/reports from the database of an ancient, custom, closed source FoxPro-based ERP software. By being able to create custom reports, I easily reduced the workload during tax time by a factor of 10. This replaced a process that involved manual review of hundreds of sheets of dot-matrix printouts for certain edge cases.

The design would be first more-or-less worked out on paper, using various drafting tricks such as french curves or tensioned wooden strips.

This can then be translated to a series of full-scale wooden templates, either interlocking 2D sections, or 3D carved from solid. Alternatively, small diameter metal rods can be bent and welded together for a same purpose. Either technique allows a craftsperson to combine precisely formed 2D curves into a 3D shape. Small design changes would also be made at this stage, usually by eye rather than to a particular measurement. This full-scale reference model is called a body buck or in Italian "manichino."

Once you have the complete buck, craftspeople would shape flat aluminum panels into to fit on various sections of the buck. The craftsperson works iteratively until the shape is right, using tools such as an English wheels, bending brakes, mallets and sandbags as well as cutting tools like shears and hole punches.

Once the shape of each individual panel is correct, they are fastened to the car chassis by welding or a mechanical fastener like a screw. The appearance and surface of the body is still quite rough at this point, so there is an extensive manual finishing process before paint is applied. Panel gaps are adjusted, bulges and dents are tapped out with hammers, lots of sanding. Soft filler materials are applied to fill and smooth the body surface- we use "Bondo" for this now but back then they used a lot of lead. The "accurate" appearance of a final painted car body comes from this hand finishing process. However, accurate is a bit of a misonmer as even though the body may look smooth and shiny there are often a lot of geometrical imperfections and asymmetries. For example, headlights in slightly different positions, replacement windshields won't fit etc. For certain models, you can even use predictable "imperfections" in the bodywork to determine whether the body was replaced or majorly repaired at some later date.

This technique was in wide use in Italy (probably other places too, but idk) through the mid-1960s, at which time modern stamping processes became able to cheaply mass produce panels with much less manual labor. As this process is so labor and skill intensive, it's now a specialty. Even among auto body repair professionals, it's uncommon to find someone with the skill to shape large panels from scratch to a high standard.

If you're interested in this type of stuff, there's a great youtube channel called "cbrwatahiki"[0] where a guy uses manual techniques to reproduce aluminum car bodies in his garage. Language is Japanese but it's still possible to watch and learn a lot if you can't understand. While his methods are not quite the same as what a 1930s-50s era production shop would use, it still shows a lot of what is involved to fabricate a body based on a pattern.

[0] https://www.youtube.com/channel/UCZjqr9ScOlWSkcnTK_mLiSA

Pricey, but priceless for someone who is spending even a moderate amount of time enjoying the backcountry. In context, a typical REI-equipped backpacker is spending $100+ on boots, $200+ on a tent, $200+ on a pack, $100+ per piece of high quality outerwear, not to mention hundreds or thousands more on skis/bikes/climbing gear if they do more than just hike/backpack. When you're laying out that amount of cash just to go out and do your favorite activity, it makes sense and is fairly common to spend an additional $400 for a convenient portable distress/backcountry connectivity plan.

I own a late 40s Italian bicycle equipped with Cambio Corsa. It's now a display piece because I don't want to maintain the glued on tubular tires or risk breaking something irreplaceable. However in the past I have taken it out on a couple short, flat rides to test the system, enough to make a couple observations about using it.

When changing gears, it doesn't feel like there is much risk of the wheel slipping forward out of the dropouts. The dropouts themselves are angled up and back, which helps prevent that. Also, when changing gears you are sitting back, putting weight on the seat and changing the weight distribution of the bike to favor the rear. This also counteracts the tendency of the wheel to come forward. The action of the chain only pulls the axle forward in the dropouts when shifting to a lower gear (higher tooth count cog on the freewheel, closer to the wheel spokes). There isn't much risk of "overshifting" the chain to a point where it will pull the axle out of the dropouts, simply because the chain would just contact the spokes or a spoke protector and slide back down into gear (unless something is really broken).

Also, you can see from the photos in the article how differently the dimensions of racing bicycle frames were fitted to riders back then. Modern racing bikes tend to have a lot of seatpost extension and longer handlebar stems, which shift the rider's weight forward on the bike. A simplified explanation is that this improves aerodynamics and certain handling characteristics, at the cost of requiring the rider have better core strength and flexibility. In contrast, the photos in the article show typical mid-20th century bicycle fit, with much less seatpost extension and shorter stems. Compared to modern racers, Coppi and Bartali sit father back, with their butt closer to the top tube. This puts the Cambio Corsa system in closer reach and the weight distribution has a more stable, rearward bias. The action of reaching back to change gears would feel treacherous and destabilizing on a modern racing bike, but would be a bit more comfortable on the bikes of the time.

This is all said with the big caveat that I have never ridden my Cambio Corsa bike in a long distance or competitive scenario. Undoubtedly it's still a challenging system to master. I can imagine the extreme difficulty of operating it on the unpaved, bumpy climbs typical of that era.

Edit: Here is a link to scans of the original patent drawings for the Cambio Corsa. It shows the mechanical details- click on the "Pages" tab to see all the scanned pages of the patent.

https://www.disraeligears.co.uk/site/french_patent_899402_-_...

There is also an official documentary about the making of episode 1 that is viewable on YouTube.[0] Even though this was produced as promotional material for the film's DVD release, Lucas' total control and hubristic self regard are barely concealed subtexts. There is a great scene where Spielberg is visiting Lucas on set and they are just repeating to each other "It'll be great!" or some such platitude with forced enthusiasm. Even Lucas seemed to know he was out of his depth.

[0] https://youtu.be/da8s9m4zEpo

Industrial equipment like CNC machines.

The most interesting part of the story is the $3900 electric crate motor that can now be ordered direct from Ford. I'm interested in the impact this will have on the electric motor retrofit scene. I'm not familiar with all the options out there, but it seems like current retrofit solutions are either very DIY-engineering (i.e. salvaged Tesla drivetrains) or very expensive integrated systems (i.e. electric GT's offerings). OEM availability of retrofit motors (and eventually batteries) could shake this up majorly. A comparison might be how the ubiquitous availability of new Chevy/Ford high performance V8 crate engines has fed the American hot-rodding scene for decades.

The PC Engines site says ALIX is EOL, and recommends using apu2 instead. Is there a particular reason you would recommend ALIX over apu2?

Is there some service provider/tool for this that is widely used in the used car industry? Or just convergence on a common approach?

I think the cabin is primarily for weatherproofing purposes. In my experience with shipping/receiving big boxes and pallets of stuff (in the US), shrink wrapping is not a given. Even most shrink wrapped shipments are not weatherproof to the degree where they would remain totally dry when flown under a drone during a rainstorm. Usually the shrink wrapping makes it more "drip proof," for brief periods of sitting outside while being transferred between warehouses/vehicles.

IMO, holding customers to a higher standard of weatherproofing is a non-starter. Adoption of drone transport will be more popular if customers can basically assume the weather exposure is the same as with conventional transport.

If you have a vehicle that needs high octane racing fuel, it's already necessary in some places to go to a specialist shop and buy the fuel in individual cans like you would buy motor oil. This would just become the norm.

Marelli makes some electronics for Ferrari, but Bosch makes a lot as well. For example, Ferrari has used Bosch fuel injection systems and ECUs since the 80s.

I used to live with a dog who really, really loved to be petted. He would hop up next to you on the couch while you were reading a book, on a laptop/phone etc and very deliberately hook your forearm with his "wrist" (not sure the right term, first major leg joint above the paw) and pull your hand/arm towards him for pets. Cute the first few times but he was so persistent and deliberate about it, it became a little annoying!

Are those co-mingling issues actually resolved now? I still avoid purchasing anything on Amazon where receiving a counterfeit or slightly not-as-described item would be consequential. I haven't personally had a bad experience, but have read enough anecdotes and reviews about it over the past year or two that it has strongly shaped my Amazon purchasing habits.

You're describing a system that is operating beyond capacity. There's no special fix required other than increasing capacity, by adding more buses to existing routes or new routes. There may be political issues with allocating funding or getting authorities to acknowledge/address lack of capacity. But it's not something that requires redesigning the bus.

In my experience with tech-heavy luxury vehicles like BMWs, dealerships have a relatively small time window from when the vehicle is introduced to when their knowledge about that vehicle drops off sharply. For example, I would trust a BMW dealer right now to work on a 2018 car, but I would take a 2000 car to an independent specialist.

One reason is turnover in dealer mechanic staff. Independent specialists tend to be more stable and have a genuine interest in and affection for the older cars of a particular brand.

Another reason is that most new luxury brands are highly focused on leasing brand new cars or getting customers to trade in on the latest model. They really only want to sell brand new cars at the top of the depreciation curve. People who own older cars will usually not buy them from the dealer and tend to not use dealer mechanics (often because the dealer is the most expensive option around). So dealer mechanics see less older cars, and service managers largely don't consider owners of older cars to be good/profitable customers that they need to keep happy.

A related issue that causes poor experiences like you describe is the "book time" job scheduling/pay system used by most dealer mechanics. This system favors a "parts swapping" approach where mechanics try to fix problems by quickly installing new components with a minimum of diagnosis time. This gets expensive for the customer due to dealer parts costs and strongly disincentivizes thoughtful, time-consuming diagnostic procedures and research. There is also usually an informal seniority system in dealers where the senior mechanics will take the easy "gravy" jobs that can be completed quickly but cost a lot. Junior mechanics will get stuck with the difficult, hard to diagnose jobs that suck up a lot of time and reduce a mechanic's earning potential. So you end up getting the least qualified mechanics working on tricky, unprofitable issues like yours.

To be fair, these issues aren't exclusive to high end vehicles, but they are especially noticeable at this market segment.

I appreciate the reply. How did you get a background in toolpath generation? I am curious because I have a background in both software dev and machining, but haven't combined them much. Was this something you became familiar with in school or only when working at those companies? Any resources you would recommend?

Not to mention the github repo for the CAM software is currently empty except for a readme and license.[1]

I would love to see a functional open source CAM software that would compete with basic tiers of Mastercam/Fusion360/etc. I think it would be of huge benefit to the manufacturing industry, education and hobbyists. There are a lot of machinists/CNC programmers very worried about Autodesk's dominance over the field and the consequences of further consolidation and crapification. Many machinists are currently learning hard lessons about SaaS.

Unfortunately in my observations (as a user and CAM software purchaser, not a CAM dev) writing toolpathing engines seems to be a Hard Problem with most of the expertise concentrated within entrenched companies. There are a few smaller companies developing CAM software but most seem to build either on one of a couple licensed CAM engines or very old legacy software from back before the big players dominated the field. Lots of the functional requirements for a CAM system have been stable for years (decades even), so a casual observer would think it is ripe for an open source alternative. However I haven't yet seen an open source alternative that could plausibly compete or even achieve feature parity with any of the commercial CAM software I use.

Best of luck to the PolarCam dev. I don't have high expectations given the magnitude of the task and the hobbyist nature of the associated hardware, but I'd love to be proven wrong.

[1] https://github.com/kadirilkimen/PolarCam

There are some industrial machines that compensate for chatter by slightly varying the spindle speed when chatter is detected. However machines so equipped are fundamentally very rigid and usually experience chatter occurs when pushing end mill performance to its limits. I suspect the adaptive spindle speed strategy would be much less useful on a machine like the Polar Bear.

While not directly answering your question, here is a suggestion that may improve your understanding of CB and help inform your decision about what radio service to use.

There are a number of WebSDRs available that provide a public web interface to radio receivers located in various places. They are live streams of radio traffic that can be tuned to your frequency of choice. If you go here[1] and filter by the 10m band and the North America region, you can find several that cover the 26-27MHz frequencies used for CB channels[2]. While none of these are located in the Bay area, it should give you a sense of how active CB is and what typical usage looks like. You can also use web SDRs to listen to ham (look up 2m/70cm calling frequencies as well as local repeater frequencies) and FRS/GMRS frequencies[3] in your area.

To be perfectly honest, I feel like GMRS may be best for your use case, but the beauty of WebSDRs is that you don't have to invest in any equipment to listen. This way you can learn more about these different types of radio services and decide which is best for your use case, without lots of expensive and time consuming buy-in.

[1] http://websdr.org/

[2] https://en.wikipedia.org/wiki/Citizens_band_radio#Standard_c...

[3] https://en.wikipedia.org/wiki/General_Mobile_Radio_Service#F...

There a few patterns I have observed in the vintage car parts industry, depending on the demand for the parts (tied to collector demand for the cars). I'm sure these apply to other industries like the Atari guy.

Consistent demand, med-high value: the business' inventory is bought up by a larger business that supplies a related niche as part of an expansion plan. This can result in poorer availability/service as the tribal knowledge of inventory organization and item details is lost in transition.

High value but demand is inconsistent/slow: parts bought up by a collector or restorer, they go off the market into storage in order to keep a specific collection going. They may in theory sell parts to others but usually difficult to deal with and almost always requires an "in".

Low demand low value: Scrapped, or stored in indifferent conditions (outside or piled up in storage units) until everything degrades.

Of course, if another "that guy" appears there is a possibility for a stable transition. But as other commenters alluded to, these folks tend to have idiosyncratic personalities and lifestyles so that happens very rarely in my experience.

According to SEMA,[0] in 2019 there was $890 Million in retail sales of parts for pre-1974 vehicles. Size of businesses encompasses everything from large operations like RockAuto or O'Reilly to small ultra-specialized one-man operations.

[0] https://mcusercontent.com/8af94f2880dbe283f4188cd3d/files/b8...

This rings true for me, and reminds me of my early experience learning about machining and CNC machines. There are a lot of very low quality googleable sources out there for that field (mostly dealing with the hobbyist side of it) that crowded out the information I needed for professional development and troubleshooting.

Eventually, as my knowledge of the field increased I learned the "pro" terminology for certain objects and processes, which improved my ability to search useful info. I also discovered through chance or personal recommendation some very useful pro-level online sources, which don't have a large presence on search engines. There was a long grind of several years before I got to the point where I could search online for machining information and solutions online with confidence that I could find useful results and assess their reliability. Even then, there is a lot that I need to find in books/manuals or in conversation with experts.

While this is a process that takes place when learning about any technical topic, in my experience its easier in programming (and to a lesser degree general IT topics) than in fields that exist primarily in the world of physical objects. IMO, this has something to do with the tendency of programming problems to "self-define" themselves in specific textual language (I.E. compiler errors you can copy and paste into google) and also with origins and focuses of the largest internet knowledge bases.