Speed against control, and picking the machine for the job

Choosing between a machine you can adjust and one you cannot.

Picking between a fast, closed machine and a slower, adjustable one comes down to whether the part in front of you is an ordinary case the closed machine's fixed settings were already tuned for, or an awkward one that needs a parameter nobody thought to expose, since speed and adjustability turn out to trade directly against each other rather than one simply being the newer, better version of the other.

A newer, considerably faster printer arrived in the lab locked down to a small handful of settings, sitting next to the older, slower machine that exposed nearly every parameter a person could want to change, and it took only a couple of awkward parts before the two machines settled into being for genuinely different jobs rather than one simply replacing the other.

The short version

A closed machine's manufacturer has already done a great deal of the tuning work on a person's behalf, choosing a temperature, a speed and a set of defaults that work reliably across the overwhelming majority of ordinary prints, and locking those choices down is exactly what lets the machine run fast and consistently without asking an operator to understand what any of the settings actually do. An open machine hands all of that tuning back to the operator, trading away some of that speed and some of that hand-holding for the ability to change literally anything the moment a part turns out to need something the defaults were never built to handle.

The point-and-shoot-camera comparison

A fully automatic camera reads the light, sets the exposure and takes a sharp, well-balanced photograph in well under a second, faster and more reliably than most people could manage adjusting the settings themselves, for the overwhelming majority of ordinary daylight scenes it was actually built around. Hand that same camera a genuinely unusual scene, a subject lit from directly behind, a fast-moving object in near darkness, and its fixed logic has no way to be told what it is actually looking at, producing a confidently wrong exposure a manual camera's operator could have corrected in seconds simply by overriding the one setting the scene actually needed changed. A closed 3D printer behaves like the automatic camera, fast and reliable across the ordinary case and silently unable to adapt the moment a part falls outside it, while an open printer behaves like the manual one, slower to operate well but able to be told exactly what an unusual part actually needs.

Why exposing every parameter is not automatically an advantage

An open machine's adjustability only pays off in the hands of someone who actually understands what each parameter does, since a person who does not know why a print temperature or a retraction setting matters gains nothing from being able to change it beyond a longer list of ways to make a print worse than the closed machine's defaults ever would have allowed. This is the part of the trade-off that gets missed most often, treating "adjustable" as a strictly better property regardless of who is doing the adjusting, when in practice an open machine handed to an inexperienced operator can produce worse, less consistent results than a closed machine's defaults simply because there are now more knobs available to turn incorrectly.

A shared lab bench makes this worse rather than better, since a machine's settings left changed by whoever solved yesterday's awkward part do not reset themselves before the next ordinary part arrives, and a print that fails mysteriously the next morning is very often failing because of a setting nobody remembered to put back. A closed machine has no equivalent failure mode, since there is nothing left to leave in the wrong state, which is its own quiet argument for keeping it around specifically for the ordinary work a shared lab does most of the time.

One figure worth keeping in mind

The closed machine finished an ordinary bracket in a fraction of the time the open machine needed for the identical part, a genuine and repeatable speed advantage on exactly the kind of print its defaults were tuned around. On the handful of unusual parts that needed a specific setting changed to succeed at all, the closed machine simply could not produce a working part regardless of how long it was given, while the open machine, slower throughout, still eventually delivered one.

What this changes in practice

Sending a part to the right machine starts with asking whether it is ordinary or awkward for that specific lab's usual work, since an ordinary part sent to the open machine wastes time relearning settings the closed machine already had correct, and an awkward part sent to the closed machine simply fails regardless of how many times it is reprinted. Keeping both machines available, rather than replacing the older one the moment the faster one arrives, turned out to matter more than either machine's individual specifications, since the two together covered a range of parts that neither one covered alone.

Where this stops being true

A closed machine's manufacturer occasionally does expose enough adjustment, through material profiles or a limited set of exposed settings, to handle a fair number of the awkward cases without needing the open machine's full range of control, in which case the choice stops being as sharp as fast-and-fixed against slow-and-flexible and becomes a genuine question of exactly how much of that flexibility the closed machine was actually given. A lab that only ever prints ordinary parts, meanwhile, may never need the open machine's full range at all, and keeping it around purely out of habit, rather than because it is actually solving a real recurring problem, spends bench space and time on flexibility nobody in that particular lab is using.

My key error with this

I had a part that needed a slot cut in it and a deadline that made the faster machine look like the obvious choice, so I ran it there without thinking particularly hard about the decision, on the general principle that faster is better when time is short. The slot came out within its dimensional tolerance and with a finish that was nowhere near good enough for a surface that had to seal against it, which I did not discover until the part was assembled and would not hold pressure. Recutting it properly on the slower machine took a fraction of the time that finding the leak, stripping the assembly and starting again had cost. What replaced the belief is that speed and control are separate axes rather than two ends of one, that every machine is fast at some things and untrustworthy at others, and that the honest question is not which machine finishes soonest but which requirement on the part is the one that will actually decide whether it works.