# Use cases

Use compressionKIT when you need to retain or transfer ECG or PPG waveforms and want to compare a smaller payload against reconstruction quality.

## Where compression can help

Store compressed frames on a wearable or gateway and reconstruct them for later analysis. Include timestamps and framing overhead in the storage budget.Explore PPG configurations →
Reduce the signal payload sent to a phone or server. Measure total transmitted bytes and encoder compute cost in your own transport workflow.Explore ECG configurations →
Compare reconstructed waveforms with the original recordings. Examine the measurements and signal conditions that matter to your application.Evaluate your recordings →
Compare neural, DSP, and hybrid methods using shared evaluation and packaging tools.Compare methods →

## Estimate the payload

For a single channel, the uncompressed payload rate is:

```text
raw bytes/second = samples/second × bits/sample ÷ 8
compressed payload ≈ raw payload ÷ compression ratio
```

The examples below assume **16-bit samples**, continuous recording, and an **8× payload compression ratio**. They are arithmetic sizing examples, not measured device results.

| Signal | Sample rate | Raw payload | At 8× | Raw per hour | At 8× per hour |
| --- | ---: | ---: | ---: | ---: | ---: |
| PPG | 64 Hz | 128 B/s | 16 B/s | 460.8 kB | 57.6 kB |
| ECG | 256 Hz | 512 B/s | 64 B/s | 1,843.2 kB | 230.4 kB |

Here, 1 kB is 1,000 bytes. Add packet headers, timestamps, metadata, and any padding to budget a complete recording. Model weights and codebooks consume additional storage independently of the recorded payload.

A smaller payload does not directly establish battery life or total system savings. Those depend on encoder execution, buffering, radio behavior, and where decoding runs.

## Choose quality requirements before a ratio

1. Identify what the reconstructed waveform must preserve, such as peak timing or pulse shape.
2. Compare several ratios on representative recordings, including noisy segments.
3. Inspect waveform error and downstream measurements together. An average can hide difficult cases.
4. Measure complete payload size, memory use, and runtime on the intended integration.

The [customer evidence](https://ambiqai.github.io/compressionkit/customer-evidence/) page provides recorded results and their limits. The [validation scorecard](https://ambiqai.github.io/compressionkit/validation-scorecard/) explains the metrics. Neither a codec ratio nor a single score establishes suitability for a clinical application.

## Try the workflow

Start with the [installation and round-trip guide](https://ambiqai.github.io/compressionkit/getting-started/), then use the [evaluation notebook](https://ambiqai.github.io/compressionkit/guides/02_evaluate_on_your_data/) with your signal. See [deployment](https://ambiqai.github.io/compressionkit/deployment/) when you are ready to integrate a validated package.