Previous Page RF Capture Guide

Sub-Page Speed & Decoding Testing

Next Page RF Decoding Guide

RF captures of tape formats are normally oversampled, meaning in the case of formats like VHS the total actual signal range tops out at 8mhz or 16msps so when using RF capture hardware at rates of 28-40msps we are technically oversampling, but this is not a bad thing due to filtering and aliasing of side bands on initial sampling.

Capturing more then the minimum allows for some redundancy from minor drops during capture to bit-flips in active storage devices but for archives going to solid mediums like optical discs, ware currently space is at a premium, down sampling allows users to cut RF captures down on a per format basis to the minimum needed for that signal.

The benefits of this also include, better SNR as there is less noise outside of the signal range, faster decoding times and smaller final files as there is less data to look though.

You can read the study here in the technical breakdowns doc.

FM RF captures commonly called samples or sampled data, is just like standard common PCM audio data just electrical signal values digitised into binary bits of information, but this is brightness and colour information in a different waveform pattern than normal sound waves, but due to being the same data format as PCM audio, we can use lossless audio codecs such as FLAC to compress captured FM RF data down A LOT or lossy ones to just break the files in fun ways too!

In terms of software support 8-bit and 16-bit singed/unsigned are what's common and universally used by most tools, including the decode suit, these data files are equivocal to audio files but with MHz of information instead of KHz.

But lets say our ADC captures 10-bit / 14-bit etc then normally you get scaled 16-bit data, then you have the MISRC which uses the 12-bit format.

FLAC compression essentially drops those extra 4 values to 0 in the case of 16-bit files alongside its standard compression methods.

Now in the real world 8-bit is more then enough for tape formats initial sampling, it's the MSPS or a million samples per second rate that actually matters and as low as 6-bit is enough to store VHS and Video8 at 20MSPS.

RF Data at 40msps in RAW or Bit-Packed --> FLAC-Chop (unpacks, cuts, down-samples & bit-crushes) --> 16msps 8-bit FLAC output

Down-Sampling is a method of cutting file size down with controlled loss of data more accurately removing background noise or dead space for example a 10~16bit ADC only has 10~16-bits of information but can only be saved in a 16-bit format FLAC will remove those 6 extra bits of essentially 0 value information or lower it to 8-bits or crush it to 6-bits, alongside filtering cut off removing anything else recorded outside the real-world usable signal range for said particular audio or video format, simmer to how a XLR microphone will have a cut-off for 75hz to remove buzzing sounds from lights or power transformers.

This process can be done for low bandwidth formats like HiFi FM signals with little worry but with video formats its more a stricter practice and the bandwidth also changes between PAL/NTSC and tape formats.

A modern full 20msps 8-bit Video / 5-10msps HiFi - VHS RF + Audio set will be today under 40-50GB/h pre-compressed with Level 8 FLAC.

The Qt6 GUI (Windows / macOS / Linux) runs the exact same probe -> cut pipeline with simple tabs:

1. Chop tab — Browse or drag-and-drop a capture, the probe runs in the background and shows the real HH:MM:SS duration (never the 1000x-wrong header value), set the IN (green) / OUT (red) markers with the time box + Set IN / Set OUT buttons or the dual-handle slider, then Process writes <input>-cut.flac sample-exact next to the source.

2. Batch Task tab — add many files (or drop them anywhere on the window) to re-compress / down-sample them whole-file with the same output settings, in parallel (one worker per CPU thread) into an optional output folder, with per-file status, Stop and skip-existing / overwrite control, a file that fails individually (rate below the selected output mode for example) is reported and the rest of the batch continues.

3. Sync Edit - Allows for chopping of full-set captures such as those made via MISRC GUI from Clockgen and or MISRC hardware workflows.

4. Metadata Editor tab — view & edit the Vorbis tags on FLAC files (Ogg .ldf inputs are read-only here, cut outputs are native FLAC and fully editable).

Since this is uncompressed data it is easy to handle, the captures save to .raw but should be renamed to .s16.

ffmpeg INPUT.s16 -i -f s16le -ar 40k -ac 1 -acodec flac -compression_level 11 -f ogg OUTPUT.flac

Drag and Drop Script 16-bit-singed-to-flac.bat

@echo off
title Compressing : %~n1
ffmpeg -i "%~1" -f s16le -ar 40k -ac 1 -acodec flac -compression_level 11 -f ogg "%~dp1%~n1.ldf"

FLAC-Chop replaces the old ld-lds-converter | SoX pipe scripts, one command unpacks the 10-bit .lds, applies the wiki-aligned sinc -n 2500 0-<cutoff> low-pass profile for the target rate and writes the FLAC, with no intermediate raw files.

# 40msps .lds -> 24msps 8-bit FLAC - NTSC/PAL (SVHS/Video8 class)
flac-chop DdD-capture.lds . 0 999999 --rate 24000 --bits 8

# 40msps .lds -> 20msps 8-bit FLAC - VHS (PAL/NTSC safe minimum)
flac-chop DdD-capture.lds . 0 999999 --rate 20000 --bits 8

# 40msps .lds -> 10msps 8-bit FLAC - VHS & SVHS HiFi FM
flac-chop DdD-capture.lds . 0 999999 --rate 10000 --bits 8

Raw 8-bit .u8 captures down-sample the same way, the file name must carry the rate (i.e capture_8-bit_40msps.u8) as headerless raw has no header to read it from.

The built-in rate presets & their sinc profiles:

Omitting --rate keeps the source rate, --no-filter skips the sinc profile, --bits 12 writes true 12-bit FLAC (for 12-bit MISRC captures) and --bits 6 writes the 6-bit crush in an 8-bit container.

The Batch Task tab does all of this whole-file through the GUI, pick the output mode / bit-depth / sinc filter, add the files, go.

Initial application call command

Linux: sox

Windows "C:\Program Files (x86)\sox-14-4-2\sox.exe"

sox Input.flac -n spectrogram

SoX can export a handy spectrogram chart of RF & Audio waveform data.

sox -r 28636 -b 8 -c 1 -e unsigned -t raw infile.u8 -b 8 -r 18000 -c 1 -t flac outfile_18msps_8-bit.flac sinc -n 2500 0-8670

sox -r 40000 -b 8 -c 1 -e unsigned -t raw infile.u8 -b 8 -r 18000 -c 1 -t flac outfile_18msps_8-bit.flac sinc -n 2500 0-8670

sox -r 40000 -b 16 -c 1 -e signed -t raw infile.s16 -b 8 -r 10000 -c 1 -t flac outfile_10msps_8-bit.flac sinc -n 2500 0-8670

The minimum you can down sample data to is 7-6 bit before visual degradation in a video signal this is a extreme compression method

Applies to VHS/Beta/Video8/EIAJ level formats and has not been tested on Hi8/SVHS etc.

1. Run ffmpeg -i input.flac -af astats -f null - and take note of RMS level dB

2. if you're using a CX card and the number is less than -9.12, or you're using a DdD and the number is less than -21.2 then stop, you have less data than this'd keep, so it would only increase size.

3. Run ffmpeg -i input.flac -af "volume=precision=fixed:volume=<your level without the minus here>dB*0.7071*0.5,aformat=sample_fmts=u8" -compression_level 12 small.flac (you can skip -compression_level 12 for much faster compression and very slightly bigger filesize)

Ensure the following is stated

• Sample Rate
• Bit Depth
• Media Format

Optionally but very useful

• Time & Date
• Capture Device Used
• Gain Setting Used
• Playback Hardware Used

You can also rename the .flac to your media format name using standard designators & standard naming guide that can be very useful for mass .xxx extension lookup for archives, if you want to keep things simple ignore this and ensure format is in file name.

Sub-Page Speed & Decoding Testing

Previous Page RF Capture Guide

Next Page RF Decoding Guide