Printer Harmonics and a Hypothetical ZT231

What We Worked On

Philip brought two YouTube Shorts with a deceptively specific acoustic question: one printer starts with a backfeed, the other does not, and their motor tones live at audibly different frequencies.

We downloaded both, extracted their soundtracks as 48 kHz stereo PCM WAV, measured their event envelopes and harmonic structure, then built seven procedural recordings of a hypothetical ZT231 performing different jobs. The generated files contain no copied reference samples.

Reading Motion from Sound

The short ZT230 recording has two almost identical transport bursts at about 2.995–3.853 s and 5.010–5.858 s. Both sit on a harmonic grid spaced around 405 Hz, with strong lines near 405, 812, 1,620, 2,432, 3,243, and 5,675 Hz. Their envelopes are orderly: roughly 150 ms of acceleration, a stable run, and about 100 ms of release. There is no separately detectable reverse tail.

The longer ZT410 clip is more complicated. Its early stable family has roughly 243.16 Hz spacing, with energy near 973, 1,216, 1,945, 3,891, and 5,839 Hz. One main run decelerates sharply for about 80 ms and then trails into roughly 0.4 seconds of intermittent mechanical motion. That event order is consistent with the backfeed Philip identified. Later, the same recording moves into a louder family dominated by 1,216 and 2,432 Hz.

A Useful 203-dpi Coincidence

Two measured ridges line up almost perfectly with common transport speeds if we use a 203-dpi mechanism as the working hypothesis:

That makes 6 ips and 8 ips line or step cadence a useful basis for sound design. It does not prove that either ridge is the electrical full-step or microstep rate. Gear, platen, case, and table resonances can suppress a fundamental and make one of its harmonics sound dominant.

Building the Little Mechanical Orchestra

The synthesizer integrates a smoothly changing cadence through acceleration, steady transport, and deceleration. Around it sit subharmonics and harmonics, low-rate speed jitter, roller eccentricity, load-dependent paper and ribbon noise, weak chassis modes, dry engagement impacts, paired backlash clicks, fan and holding-current beds, and short asymmetric reflections.

The reverse oscillator changes phase direction, but phase sign alone is not a perceptual cue. The pause, backlash, new ramp, changed speed, and shifted harmonic balance are what make the hypothetical backfeed read as reverse.

Listen to the Hypothetical ZT231

Every render is 48 kHz stereo, signed 16-bit PCM. Peaks are held at -5 dBFS, there are no clipped samples, and deterministic regeneration produced identical SHA-256 hashes. Begin playback at moderate volume; narrow mechanical tones can feel louder than their meter level suggests.

What Went Well

The analysis separated comb spacing from the loudest harmonic, and it separated direction claims from observable timing. That let the synthesis borrow useful acoustic structure without pretending to know undocumented motor poles, gearing, firmware ramps, or driver chopping.

The render pipeline checked WAV headers, duration, channel count, sample format, DC offset, spectral centers, peak level, clipping count, and deterministic hashes after generation.

What Didn’t Go Well

The first sandboxed command never reached yt-dlp because a local sandbox mount configuration failed. Scoped execution proceeded only after explicit approval. That is an environment defect, not an audio failure, and it deserves repair rather than normalization.

A first batch-conversion loop also dropped the leading slash from one source path. ffmpeg correctly refused the nonexistent relative path; rerunning the second conversion with an exact absolute path completed cleanly.

Most importantly, these clips are ZT230 and ZT410 phone/video recordings—not isolated ZT231 measurements. Room reflections, automatic gain control, media, ribbon, wear, mounting, and enclosure coupling all contaminate model identity. The results are illustrative sound design, not a service diagnostic.

Takeaways

1. 1
   Direction lives in the event sequence

   Pitch has magnitude, not a forward or reverse sign. Pauses, lash, ramps, and visible transport motion carry the directional evidence.

2. 2
   The loudest tone may be a harmonic

   A 1.620 kHz whine can sit on a 405 Hz grid. Both numbers matter.

3. 3
   Mechanical plausibility comes from layers

   Cadence alone sounds like a synthesizer. Media drag, case modes, impacts, load modulation, and restrained reflections make it read as machinery.

4. 4
   Hypothetical should stay in the filename

   The jam/retry render is an illustration and must never become a reference for diagnosing real hardware.