All posts Technology

104, 208 or 416? Choosing how often a pen should look at your hand

MKMalli KosuriMay 20268 min read
In short

Handwriting's useful motion lives below roughly 20 to 30 cycles per second, so sampling theory says anything above ~60 captures the signal. The pen runs 208 Hz, comfortable margin for sharp events like tip impacts, without the power, bandwidth and memory bill of 416.

The post walks the pros and cons of each rate honestly, including what would genuinely improve at 416 and why it is not free.

The short version. The full post has the detail and the why.

In this post

  1. How fast does a hand actually move
  2. The case for 104, and its ceiling
  3. Why 208 is the sweet spot
  4. What 416 would buy, and charge

How fast does a hand actually move

Handwriting feels fast, and is slow by electronics standards. The strokes of ordinary writing contain nearly all their energy below twenty to thirty cycles per second; the sharpest common event is the tip striking paper. Sampling theory's rule is to sample at least twice your highest frequency of interest, so around sixty times a second would technically capture handwriting. Technically is doing heavy lifting in that sentence, margin matters, which is why the menu starts higher.

The case for 104, and its ceiling

Our sensors offer standard rates that double: 104, 208, 416 samples per second. At 104 the battery smiles, the data is light, and steady writing reconstructs well. The ceiling shows at the edges: tip impacts land between samples and lose their sharpness, and the crisp force transitions that anchor stroke boundaries blur by a few milliseconds. For a product whose analysis leans on exactly those boundaries, the edges are not a good place to save.

low rate: peaks slip between samples 208 Hz: peaks and valleys both captured
The same stroke sampled two ways: at a low rate the sharp peaks fall between samples and vanish; at 208 Hz the shape survives, with margin to spare.

Why 208 is the sweet spot

At 208, a sample lands every 4.8 milliseconds. Stroke boundaries locate crisply, impacts are caught, rapid scribbles stay well inside margin, and every downstream computation gets clean input. Meanwhile the costs stay civil: the radio can carry the stream comfortably, the offline log stores a school day's writing without strain, and the power budget still delivers the all-day promise. Every number in the pipeline, packet sizes, buffers, storage, is tuned around this rate.

What 416 would buy, and charge

Honesty about the next doubling: 416 would time impact events more finely and please future research models that mine micro-structure, and we can enable it for lab captures. As a shipping default it charges twice the radio load, twice the storage burn, meaningfully more power, and, by sampling theory, adds nothing to the frequencies where handwriting actually lives, while sampling the sensor's own noise floor more often. Rates are a dial, not a destiny; today every doubling past 208 buys mostly heat.

Key takeaways

  • Handwriting's energy lives below ~30 Hz; margin, not theory, sets the practical rate.
  • 104 Hz blurs exactly the stroke boundaries the analysis depends on.
  • 208 Hz captures every edge crisply while power, radio and storage stay in budget.
  • 416 Hz doubles the bills to sample noise more often; useful in the lab, wasteful in the pocket.

The best sampling rate is not the biggest number the datasheet offers. It is the smallest number that loses nothing you care about.

Read next