Serial Studio audio pipeline validator

Overview

This project is a utility for converting audio data recorded via Serial Studio's Audio I/O driver into playable WAV files. It's for developers testing Serial Studio's data pipeline integrity using audio signals like voice, tones, or music. If the reconstructed .wav file sounds like the original, the I/O pipeline works.

No extra hardware needed. Just Serial Studio, a microphone or audio source, and this Python script.

Serial Studio's audio features may need a paid license. See serial-studio.com for details.

Audio format

Serial Studio logs audio samples to CSV like this:

Elapsed (s),Audio Input/Channel 1,Audio Input/Channel 2,...
0.000000,0.000000,0.002541,...

• First column. Elapsed (s) relative timestamp, used to auto-detect the sample rate.
• Remaining columns. Audio samples for each channel; any header ending in Channel <n> counts.
• Channel count. Mono, stereo, surround, and so on, depending on how many channels are logged.

Features

• Converts any Serial Studio CSV into a valid .wav file.
• Supports mono, stereo, and multichannel audio.
• Auto-detects the sample rate from the Elapsed (s) column (fallback 44,100 Hz).
• CLI options for:
  • Sample rate override (--rate).
  • Input sample format (--in_format: float32, int16, int24, int32, uint8).
  • TPDF dither before quantization (--dither).
• Removes DC offset and normalizes each channel automatically.
• Helps debug Serial Studio pipelines by letting you listen to the reconstructed audio.

How to use

1. Record audio in Serial Studio

• Set the input source to an Audio I/O device.
• Configure the number of channels (1 = mono, 2 = stereo, and so on).
• Enable data logging to a .csv file.
• Speak, play music, or record any sound.

2. Convert CSV to WAV

python3 csv2wav.py path/to/audio.csv

You can optionally specify an output path, sample rate, and format:

python3 csv2wav.py path/to/audio.csv output.wav --rate 48000 --in_format int16

3. Listen to the result

• Open the generated WAV file in any media player.
• Compare it to your original source (mic recording, test tone, and so on).
• Clean, accurate playback means the data path works end to end.

Supported input sample formats

Default is float32. The format describes the values stored in the CSV; the output WAV is always 16-bit PCM.

Example use case

As a developer or tester:

• Inject an audio signal into Serial Studio using a virtual mic or file playback.
• Serial Studio logs the raw data into a CSV file.
• Run csv2wav.py to convert it back into a WAV.
• Listen to check whether the WAV matches the original audio.

If it sounds correct, the pipeline works. If it's clipped, noisy, or silent, check device config, sample rate, or formatting.

Files

• csv2wav.py: the core CSV to WAV conversion script.
• README.md: project documentation and usage guide.

Dependencies

• Python 3.x.
• numpy (install via pip).

pip install numpy

Command-line usage

python3 csv2wav.py input.csv [output.wav] [--rate <hz>] [--in_format <type>] [--dither]

• input.csv: CSV file exported from Serial Studio.
• output.wav: optional name for the output WAV file.
• --rate: optional sample rate in Hz; when omitted it is derived from the Elapsed (s) column, falling back to 44100.
• --in_format: input format (float32, int16, int24, int32, uint8).
• --dither: apply TPDF dither before int16 quantization.

[!TIP] For advanced resampling, route audio through a virtual loopback device and set the target frequency with Serial Studio. Then run csv2wav.py to export at the new frequency. This produces clean downsampling results suitable for slowed and reverb style mixes.

Troubleshooting

• WAV plays silence. Check whether your CSV has audio values, or just zeros.
• Distortion. Input values may exceed [-1, 1]. Normalize or rescale properly.
• Wrong channel order. Check that your audio source is mapped correctly in Serial Studio.