rigol2spice

Convert, transform, analyze, and plot real oscilloscope captures.

Converts Rigol oscilloscope CSV and WFM exports to:

• PWL (piece-wise linear) files for LTspice, Cadence, ngspice, and other SPICE simulators.
• MATLAB vectors for numerical analysis and further processing.
• Mono WAV files in 32-bit floating-point or 16-bit PCM format.
• NumPy arrays and archives for Python-based processing.

Supports multi-channel captures, channel selection, and channel math.

For convenient waveform processing, it provides an ordered transformation pipeline, measurements, and FFT analysis, and optional SVG plots.

Written in Swift as a compiled native binary for high performance; runs on Windows, macOS, and Linux.

Watch on YouTube

Supported Input Formats

Input format is detected from the file contents, not its filename extension.

Format	Oscilloscope families
Legacy Rigol CSV	Legacy Rigol oscilloscopes
Centaurus CSV	Newer Centaurus-platform oscilloscopes, including DHO800/DHO900
Rigol WFM	DS1000B, DS1000C, DS1000D/E, DS1000Z, DS2000/MSO2000, DS4000/MSO4000

DHO .wfm, logic-only WFM payloads, and Rigol .trc files are not supported.

Quick Start

Download the latest release, then:

rigol2spice.exe input.csv output.txt

The output .txt file can be loaded directly as a PWL source in LTspice and other SPICE simulators.

WFM captures use the same command:

rigol2spice.exe capture.wfm output.txt

Output Formats (-f)

Format	Output	Requirements
pwl (default)	Tab-separated timestamp/value pairs for SPICE	None
matlab	MATLAB column vector points containing vertical values	None
wav32	Mono 32-bit IEEE floating-point WAV	Finite values in -1...1; at least two uniformly spaced samples at a whole-number sampling rate
wav16	Mono signed 16-bit PCM WAV	Finite values in -1...1; at least two uniformly spaced samples at a whole-number sampling rate
npy	One-dimensional little-endian Float64 array of vertical values	None
npz	Uncompressed archive with little-endian Float64 arrays timestamps and values	None

SPICE:

* LtSpice
V1 n+ n- PWL file="output.txt"

* PSpice
V1 n+ n- PWL(FILE="output.txt")

* HSpice
V1 n+ n- PWL PWLFILE="output.txt"

* Ngspice
V1 n+ n- PWL file output.txt

* QUCS-S
V1 n+ n- PWL file=output.txt

MATLAB:

run("output.m");
plot(points);

NumPy:

import numpy as np

values = np.load("output.npy")

capture = np.load("output.npz")
timestamps = capture["timestamps"]
values = capture["values"]

Channel Selection (-l And -c)

List available channels in a capture with -l or --list-channels:

rigol2spice.exe -l input.csv

Select a specific channel (default is CH1) with -c or -channel:

rigol2spice.exe -c CH2 input.csv output.txt

Combine channels with +, -, *, and / (standard precedence; parentheses allowed). Transformations then apply to the result:

rigol2spice.exe -c "CH1+CH2" input.csv output.txt
rigol2spice.exe -c "(CH1-CH2)/CH3" input.csv output.txt
rigol2spice.exe -c "CH1-CH2" -t "RemoveDC" input.csv output.txt

Serial Decoding (--decode)

Decode UART, I²C, and SPI signals directly from analogue capture channels. UART accepts an explicit baud rate or automatic detection:

rigol2spice.exe capture.wfm --decode "UART rx=CH2, baud=115200, threshold=1.65"

Set baud=auto to estimate the baud rate from the captured signal:

rigol2spice.exe capture.wfm --decode "UART rx=CH2, baud=auto, threshold=1.65"

An explicit baud rate is preferable when it is known. Automatic detection requires enough clean transitions in the capture to determine the bit period reliably. threshold is the analogue voltage used to distinguish low and high logic levels.

I²C requires SDA and SCL. It reports START/repeated-START transactions, 7-bit addresses, read/write direction, bytes, and ACK/NACK:

rigol2spice.exe capture.wfm --decode "I2C sda=CH1, scl=CH2, threshold=1.65" --decode-format csv --decode-output i2c.csv

Use sda-threshold and scl-threshold instead of the shared threshold when the channels use different logic levels.

SPI supports modes 0–3, MSB/LSB-first words, optional chip select, and either or both data directions:

rigol2spice.exe capture.wfm --decode "SPI clk=CH1, mosi=CH2, miso=CH3, cs=CH4, mode=0, bits=8, threshold=1.65" --decode-format csv --decode-output spi.csv

order defaults to msb; cs-active defaults to low. Omit cs for captures that contain only the intended transfer. Every relevant protocol channel receives the same ordered transformations before decoding. The output-only --downsample step runs later and therefore cannot alter decoded frames.

When --decode and --plot are combined, the SVG includes a time-aligned protocol lane below the waveform. UART bytes and parity/framing errors, I²C addresses/data/ACK status, and SPI MOSI/MISO words are labelled over their decoded time intervals; each interval also includes an SVG tooltip. Decoded values below 0x80 also show their ASCII representation. Printable characters appear directly (0x48 · ASCII "H"); control characters use escapes such as \n, \r, \t, or \x1B.

Use --decode-format to select the decoder output and --decode-output to write it to a file:

rigol2spice.exe capture.wfm --decode "UART rx=CH2, baud=115200, threshold=1.65" --decode-format csv --decode-output uart.csv
rigol2spice.exe capture.wfm --decode "UART rx=CH2, baud=auto, threshold=1.65" --decode-format bin --decode-output uart.bin

Decode format	Output
text (default)	Human-readable decoded events, including timing and UART status information
csv	The same event information as text, represented as CSV with a header row
bin	Raw decoded octets, with no text, timestamps, headers, separators, or error metadata

Text and CSV output include the same ASCII representation for decoded values below 0x80. I²C address bytes are not presented as ASCII because they encode the address and R/W bit rather than payload data.

For UART, bin contains valid frames only. For I²C it contains data bytes and omits address bytes. For SPI it contains MOSI words, or MISO when MOSI is absent. SPI words wider than eight bits cannot be represented by bin; use text or CSV instead. Use text or CSV when timing, addressing, acknowledgement, or error details need to be retained.

If --decode-output is omitted, text and csv are written to standard output. Binary output requires --decode-output, preventing raw bytes from being mixed with the normal console report.

Transformations (-t)

Pass an ordered list of transformation commands with -t or --transformations. Separate commands with semicolons or line breaks (CR, LF, or CRLF); empty commands and blank lines are ignored. Enclose the complete string in double quotes for the Windows command line:

rigol2spice.exe input.csv output.txt -t "RemoveDC; ClampMax 0.7; Offset -1.2"

Use -tf or --transformations-file to read the same command text from a UTF-8 file. If -t is also present, file commands run first and inline commands follow:

rigol2spice.exe input.csv output.txt -tf transforms.txt -t "Normalize"

Commands use the syntax OPERATION argument. Operation names are case-insensitive. Scalars accept decimal (0.7), engineering (3n, 10u), and scientific (5e-3) notation. Use - for negative values. Transformations run strictly from left to right.

Amplitude & Level

Operation	Example	Effect
RemoveDC	RemoveDC · RemoveDC Avg · RemoveDC Median · RemoveDC Mid	Estimate and subtract the DC component¹
Offset	Offset -1.5	Add the scalar to every value
Min0	Min0	Shift so the lowest sample is 0 (subtract min)
AddNoise	AddNoise 10m	Add zero-mean Gaussian noise with standard deviation equal to the scalar
TVDenoise	TVDenoise 50m	1D total variation denoising with weight λ (piecewise-constant bias; preserves edges)⁶
Multiply	Multiply 10	Multiply every value by the scalar
Invert	Invert	Multiply every value by −1 (alias of Multiply -1)
PeakTo	PeakTo 3.3	Scale so the peak absolute value equals the scalar
PeakToPeak	PeakToPeak 2	Scale so max − min equals the scalar
Normalize	Normalize	Alias of PeakToPeak 1
ScaleRMS	ScaleRMS 707m	Scale so the sample RMS equals the scalar
dB	dB 6	Scale amplitude by the given voltage dB (×10^(dB/20))
dBmW	dBmW 10 · dBmW 0, 75	× volts for that power into R Ω (e.g. 10 dBm @ 50 Ω ≈ 0.707 V)³
dBm	dBm 10	Alias of dBmW
dBW	dBW 0 · dBW -30	Same as dBmW but relative to 1 W (0 dBW = 30 dBm ≈ 7.07 V @ 50 Ω)³

Clipping, Gates & Shaping

Operation	Example	Effect
ClampMin	ClampMin 0	Clamp values below the scalar
ClampMax	ClampMax 3.3	Clamp values above the scalar
Limit	Limit -0.7, 0.7	Clamp values between low and high
Gate	Gate 0.6	Zero values below the threshold; keep the rest
DeadZone	DeadZone 0.1	Zero values inside ±threshold; keep the rest
Digitize	Digitize 1.5 · Digitize 1.5, 0, 3.3 · Digitize 1.2, 1.8, 0, 3.3	Map each sample to one of two levels (hard threshold or Schmitt)⁵
Threshold	Threshold 1.5	Alias of Digitize
SlewLimit	SlewLimit 100M	Limit
SoftClip	SoftClip -0.7, 0.7	Soft-clip into [low, high] via tanh
Fade	Fade 100n	Linear fade-in and fade-out over the given duration
FadeIn	FadeIn 50n	Linear fade-in only
FadeOut	FadeOut 50n	Linear fade-out only
Quantize	Quantize 8, 3.3 · Quantize 8, -1, 1	Round to N-bit levels over [0, full-scale] or [low, high]
Abs	Abs	Replace every value with its absolute value
Rectify	Rectify	Half-wave rectify (keep values ≥ 0, zero the rest)

Time Domain

Operation	Example	Effect
TimeShift	TimeShift -5m	Shift timestamps; negative values shift left
TimeScale	TimeScale 2 · TimeScale 0.5	Scale the time axis (preserve start time)
Seamless	Seamless · Seamless 100n	Force last value = first (or append a ramp of the given duration)
MatchEnds	MatchEnds	Alias of Seamless
Pad	Pad 5m · Pad 5m, 0	Extend by a duration, holding the last value (or a given level)
HoldLast	HoldLast 5m	Alias of Pad
PadPoints	PadPoints 100 · PadPoints 100, 0	Append exactly N samples, holding the last value (or a given level)
HoldLastPoints	HoldLastPoints 100	Alias of PadPoints
ExtendTo	ExtendTo 10m · ExtendTo 10m, 0	Extend to an absolute end time, holding the last value (or a given level)
ExtractPeriod	ExtractPeriod · ExtractPeriod 0.5	Keep one cycle from the first rising crossing (auto or given threshold); shift to t=0
CutBefore	CutBefore 5m	Discard samples before the timestamp and shift the first retained sample to t=0
CutAfter	CutAfter 10u	Discard samples at or after the timestamp
DropLast	DropLast 10u	Remove samples inside the final duration; retain a sample exactly on the start boundary
DropLastPoints	DropLastPoints 10	Remove exactly the final N sample points
Trim	Trim 1m, 10m	Keep samples with start ≤ t < end and shift the first retained sample to t=0
Repeat	Repeat 2.5	Append copies of the capture⁴
Oversample	Oversample 5	Divide the capture into 5 equal segments and average their aligned samples
Forecast	Forecast 5m · Forecast 5m, 500	Automatically choose a forecast model and append a duration, optionally using an exact number of samples

Oversample N performs coherent ensemble averaging for a repeating capture. The sample count must be exactly divisible by N. It returns one segment starting at t=0, preserves the original sample interval, and averages each output value from the N corresponding segment values. If the division is not exact, the error reports how many samples—and how much capture time—to add or remove. Segment alignment is essential: drift or a non-repeating signal will blur the result.

Forecast duration[, samples] is a self-contained forecast: it trains on the current waveform and appends uniformly spaced future samples, so no model file or extra dependency is needed. It automatically backtests causal temporal convolution, repeating-pattern, linear-trend, stable-mean, and hold-last strategies on a known part of the capture. The best strategy is retrained on the complete capture; the console reports its name and a 0–100% backtest confidence. Confidence below 20% also emits a warning and uses a conservative competitive forecast. It works best when the capture contains several examples of the pattern to predict. The input must contain at least 8 finite samples. Forecast aligns data by sample position and uses the average interval between the first and last timestamps for its output, so small or deliberate timestamp irregularities are accepted. Long, chaotic, or changing signals remain inherently uncertain, so treat the result as an estimate.

Resampling

ResampleF

Field	Description
Syntax	ResampleF frequency[, interpolation]
Frequency	Target sampling frequency; accepts engineering notation
Result	Creates a uniform grid whose samples are exactly 1 / frequency seconds apart
Timestamps	Preserves the first timestamp and stops at the last complete interval inside the original capture
Interpolation	linear (default), pchip, sinc

Downsample

Field	Description
Syntax	Downsample factor[, interpolation]
Factor	Must be greater than 1 and may be fractional
Result	Reduces the point count to original count / factor, rounded to the nearest sample
Timestamps	linear and sinc preserve both endpoints; fast retains original timestamps and only preserves the final timestamp when it belongs to the selection sequence
Interpolation	linear (default), sinc, fast

Upsample

Field	Description
Syntax	Upsample factor[, interpolation]
Factor	Must be greater than 1 and may be fractional
Result	Increases the point count to original count × factor, rounded to the nearest sample
Timestamps	Preserves both endpoints and the capture duration
Interpolation	linear (default), pchip, sinc

Interpolation methods

Interpolation	Behavior
linear (default)	Interpolate linearly between adjacent samples; fastest and suitable for most PWL workflows
pchip	Monotonic piecewise-cubic interpolation; produces smoother curves without overshooting monotonic data
sinc	Finite windowed-sinc interpolation; best suited to sufficiently sampled, band-limited signals
fast	Select original samples at source positions 0, factor, 2 × factor, and so on; performs no interpolation

Scalar arguments accept an optional physical-unit suffix after the engineering prefix. The unit is ignored, so Pad 5ms is equivalent to Pad 5m, and ResampleF 10MHz is equivalent to ResampleF 10M. Engineering prefixes remain case-sensitive (m is milli and M is mega).

ResampleF and the Downsample transformation do not apply an anti-alias low-pass filter automatically when reducing the sampling frequency. Add LowPass earlier in the transformation chain when anti-alias filtering is required. The post-processing option --downsample N is separate: it reduces only a converted signal output, after transformations, analyses, and protocol decoding have used the full processed waveform.

Triggers

Trigger transformations find an event using interpolated crossings between samples and align the resulting waveform. Directions are rising, falling, or either where supported. An optional final after argument ignores events before that capture time.

Operation	Example	Effect
Trigger	Trigger 0.5 · Trigger rising, auto · Trigger falling, 50%, 2m	Shift the first level crossing to t=0
TriggerSchmitt	TriggerSchmitt rising, 1.4, 1.8 · TriggerSchmitt falling, 1.4, 1.8, 2m	Arm at the opposite hysteresis level, then trigger at the requested edge
TriggerNth	TriggerNth rising, 0.5, 3 · TriggerNth either, auto, 5, 2m	Shift the Nth matching crossing to t=0
TriggerCapture	TriggerCapture rising, 0.5, 100u, 500u	Keep the requested pre/post window around a level crossing; optional fifth argument is search-after
TriggerWindow	TriggerWindow rising, 0.5, 100u, 500u	Alias of TriggerCapture
TriggerPulse	TriggerPulse high, 0.5, 10u · TriggerPulse low, 0.5, 10u, 20u	Trigger at the start of a complete high/low pulse meeting a minimum or inclusive min/max width
TriggerBand	TriggerBand enter, 1, 2 · TriggerBand exit, 1, 2, 5m	Trigger on enter, exit, rising above, or falling below a band
TriggerSlew	TriggerSlew rising, 10, 90, 500k · TriggerSlew falling, 10, 90, 100k, 500k	Trigger at the start of a percentage-level transition meeting a minimum or inclusive min/max slew rate
TriggerDropout	TriggerDropout rising, 0.5, 2m · TriggerDropout either, auto, 2m, 5m	Trigger when no matching edge occurs for the duration after an initial edge
TriggerRunt	TriggerRunt rising, 1, 2, 20u	Rising: cross low but return/fail to reach high; falling: cross high but return/fail to reach low

Levels may be numeric, auto, or a percentage such as 50% for Trigger, TriggerNth, TriggerCapture, TriggerPulse, and TriggerDropout. auto is the midpoint of the oscilloscope-style Top and Base; P% is that percentage of the same span.

Most triggers discard samples before the aligned event. TriggerCapture deliberately preserves context before it:

rigol2spice.exe input.csv output.txt -t "TriggerCapture rising, auto, 100u, 500u"

This keeps up to 100 µs before and 500 µs after the event. The output starts at t=0, so the trigger normally occurs at t=100 µs. The pre/post window is clipped to the samples available in the original capture.

Filtering & Calculus

Operation	Example	Effect
LowPass	LowPass 1k	Apply a low-pass filter²
HighPass	HighPass 100	Apply a high-pass filter²
BandPass	BandPass 900, 1.1k	Apply a band-pass filter²
BandStop	BandStop 48, 52	Apply a band-stop filter²
Notch	Notch 50, 4	Reject a narrow band with a zero-phase IIR biquad²
MovingAverage	MovingAverage 5	Centered moving average over N samples
Median	Median 5	Centered median filter over N samples (spike-resistant)
Detrend	Detrend	Remove the least-squares linear trend (slope and offset)
Diff	Diff	Numerical derivative dv/dt
Integrate	Integrate	Cumulative trapezoidal integral (starts at 0)

For a capture with baseline drift and 50 Hz interference, both operations can be applied in one pass:

rigol2spice.exe input.csv output.txt -t "Detrend; Notch 50, 4"

¹ RemoveDC subtracts a DC estimate. Optional method (same names as analysis measurements; case-insensitive):

Method	Meaning
DC (default; aliases Cluster, KMeans)	1D k-means with k = 3 on finite sample values; several deterministic initializations; pick smallest squared error; DC = midpoint of lower and upper centroids
Avg (aliases Average, Mean)	Arithmetic mean of sample values
Median	Median of sample values
Mid (alias Midpoint)	Midpoint of min/max: (max + min) / 2

Bare RemoveDC is equivalent to RemoveDC DC.

² Filters are automatic zero-phase IIRs built from stable biquad sections and applied forwards and backwards. LowPass and HighPass use a 16th-order Butterworth response per pass (effective 32nd-order magnitude response); BandPass combines both in series, and BandStop sums independent high-order low-pass and high-pass paths outside the rejected band. Notch uses eight compensated sections with zeros exactly at its center frequency, which remains effective for extremely narrow low-frequency rejection at high sample rates. Their cost is O(N) and does not grow when the requested frequency is small relative to the sample rate. The supplied cutoff or band edges are approximately −3 dB after the two passes. Frequencies must be greater than 0 and below Nyquist (fs/2); band and notch edges must satisfy 0 < f1 < f2 < fs/2.

³ dBmW and dBW convert an absolute power level to a voltage scale factor: V = √(P · R), with P = 1 mW · 10^(dBmW/10) or P = 1 W · 10^(dBW/10). Optional second argument is load impedance in ohms (default 50). Unlike dB, these are not relative gains — they multiply the waveform by that absolute voltage (useful for unit-amplitude templates).

⁴ Repeat requires a value greater than zero and accepts fractional repetitions; its final point is interpolated when needed.

⁵ Digitize turns an analog capture into a two-level digital PWL. Argument forms:

Form	Meaning
Digitize T	Hard threshold: output 1 when the sample is ≥ T, else 0
Digitize T, low, high	Same compare, but output low / high instead of 0 / 1
Digitize fall, rise, low, high	Schmitt trigger: go high when the sample is ≥ rise, go low when ≤ fall; stay put while between the two (hysteresis). fall and rise may be given in either order

The first sample seeds the state (≥ rise → high, otherwise low). With a single threshold the compare is hard (no band between levels). Use hysteresis when the analog edge is noisy so the digital output does not chatter.

⁶ TVDenoise λ solves min_x ½‖x − y‖² + λ · TV(x) with TV(x) = Σ |xᵢ₊₁ − xᵢ| (Condat’s direct 1D algorithm). Larger λ flattens plateaus more aggressively while keeping large jumps; λ = 0 is a no-op. λ has the same units as the sample amplitude. Not an inverse of AddNoise.

Modulation & Demodulation

The analog modulation transformations turn the current waveform into a carrier-modulated signal while preserving timestamps and sample count. AM automatically centers and peak-normalizes its message to −1…1 because its argument is a dimensionless modulation depth. FM and PM deliberately use the sample values unchanged, allowing earlier transformations such as RemoveDC, Offset, Multiply, or PeakTo to define the desired message scale.

Operation	Example	Effect
AM	AM 100k, 0.8 · AM 100k, 0.8, 2	Amplitude modulation: carrier frequency, modulation depth, and optional carrier amplitude (default 1)
FM	FM 100k, 10k · FM 100k, 10k, 2	Frequency modulation: carrier frequency, sensitivity in Hz per input unit (normally Hz/V), and optional carrier amplitude (default 1)
PM	PM 100k, 1.2 · PM 100k, 1.2, 2	Phase modulation: carrier frequency, sensitivity in radians per input unit (normally rad/V), and optional carrier amplitude (default 1)
DemodAM	DemodAM 100k, 0.8, 20k	Recover AM baseband: carrier frequency, modulation depth, and baseband low-pass cutoff
DemodFM	DemodFM 100k, 10k, 20k	Recover FM input units: carrier frequency, sensitivity in Hz/unit, and quadrature low-pass cutoff
DemodPM	DemodPM 100k, 1.2, 20k	Recover PM input units: carrier frequency, sensitivity in rad/unit, and quadrature low-pass cutoff

For FM, the instantaneous frequency is carrier + sensitivity × value; a sensitivity of 10k applied to a ±2 V message therefore produces a ±20 kHz frequency deviation. For PM, the phase offset is sensitivity × value. The FM instantaneous frequency must remain above 0 and below Nyquist. PM changes can also alias if the signal produces phase changes that are too fast for the sample rate.

ModulateAM, ModulateFM, and ModulatePM are aliases of the short modulation names. AMDemod, FMDemod, and PMDemod are aliases of the corresponding demodulators. Demodulation uses coherent quadrature mixing followed by the automatic zero-phase low-pass filter, so its carrier and sensitivity parameters should match the modulator. Its cutoff must cover the occupied modulated baseband while remaining below both the carrier and its distance from Nyquist. A carrier-frequency error appears as a DC error in FM; an unknown carrier phase appears as a DC ambiguity in PM. For AM, depths up to 1 avoid overmodulation; values above 1 are accepted deliberately.

For example, modulate a captured message in FM and then recover its baseband:

rigol2spice.exe input.csv fm.txt -t "RemoveDC; PeakTo 1; FM 100k, 10k, 2"
rigol2spice.exe fm-capture.csv recovered.txt -t "DemodFM 100k, 10k, 20k"

Analysis (-a)

Pass measurement commands with -a or --analysis. Syntax matches -t (commands separated by semicolons or CR/LF/CRLF line breaks; empty commands and blank lines ignored; case-insensitive; engineering scalars). Results print to the console with one fractional digit and combine the engineering prefix with the physical unit, for example 2.4ms, 8.4mV, 1.2MV/s, or 2mW. Signal amplitudes are reported in volts; ratios such as duty cycle and THD are displayed as percentages, while crest factor uses ×. Analyses always run after transformations on the full processed waveform, before the output-only --downsample step. When -a is used, the PWL output file is optional (same as -l / -p).

Use -af or --analysis-file to read analysis from a UTF-8 file. When combined with -a, file analysis come first, so dependents such as an inline THD can reuse an FFT declared in the file.

rigol2spice.exe input.csv -a "Max; Min; RMS; PkPk; Frequency"
rigol2spice.exe input.csv output.txt -t "RemoveDC" -a "DC; Avg; FrequencyAtZero"
rigol2spice.exe input.csv -p -a "FFT 1024; Frequency"
rigol2spice.exe input.csv -a "Basic; Timing"

Frequently used measurements are also available as presets. A preset expands to the listed analyses and can be mixed with individual operations or other presets:

Preset	Expansion
Basic	All operations in Capture and Amplitude & level below (aliases omitted)
Timing	All operations in Timing below that accept no arguments (FrequencyAt omitted)
Spectrum	FFT; THD
WaveType	SineWaveType; SquareWaveType; SawtoothWaveType; TriangleWaveType (requires a preceding FFT)

For example, -a "Basic; SlewRise" prints the basic summary followed by the rising-edge slew rate. Preset names, like all analysis operations, are case-insensitive.

Capture

Operation	Example	Result
Duration	Duration	Capture time span t_last − t_first
Points	Points	Number of samples
SampleRate	SampleRate	Estimated sample rate (N−1) / Duration
Interval	Interval	Mean sample interval 1 / SampleRate
Start	Start	First sample timestamp
End	End	Last sample timestamp

Amplitude & level

Operation	Example	Result
Max	Max	Maximum sample value
Min	Min	Minimum sample value
HiPeak	HiPeak	Alias of Max
LowPeak	LowPeak	Alias of Min
PkPk	PkPk	Peak-to-peak (max − min, same as PeakToPeak)
Peak	Peak	Peak absolute value max|v| (same as PeakTo)
Amplitude	Amplitude	Half of peak-to-peak (PkPk / 2)
Mid	Mid	Midpoint of min/max: (max + min) / 2
Avg	Avg	Arithmetic mean of sample values
DC	DC	DC estimate (same default k-means method as bare RemoveDC)¹
RMS	RMS	Sample RMS (same definition as ScaleRMS)
ACRms	ACRms	AC RMS: √max(0, RMS² − Avg²)
StdDev	StdDev	Population standard deviation (same value as ACRms)
Stdev	Stdev	Alias of StdDev
Crest	Crest	Crest factor: peak absolute / RMS
Median	Median	Median sample value
PeakTime	PeakTime	Timestamp of the first maximum sample
MinTime	MinTime	Timestamp of the first minimum sample
MeanAbs	MeanAbs	Arithmetic mean of the absolute sample values
Top	Top	Upper histogram mode (scope-style high level)
Base	Base	Lower histogram mode (scope-style low level)
Overshoot	Overshoot	(max − Top) / (Top − Base) (fraction)
Undershoot	Undershoot	(Base − min) / (Top − Base) (fraction)

Timing

Operation	Example	Result
FrequencyAt	FrequencyAt 0 · FrequencyAt 1.5	Average period/frequency of complete waves at that level (rise+fall crossings; first wave needs 3 crossings, each next wave +2, sharing the boundary; partial start/end ignored)
FrequencyAtZero	FrequencyAtZero	Same as FrequencyAt 0
Frequency	Frequency	Same as FrequencyAt at the min/max midpoint (Mid)
RiseTime	RiseTime · RiseTime 20, 80	First rising edge time from low% to high% of min/max span (default 10 → 90)
FallTime	FallTime · FallTime 20, 80	First falling edge time from high% to low% of min/max span (default 90 → 10 via the same percent pair)
SlewRise	SlewRise · SlewRise 20, 80	First rising-edge slew rate Δv/Δt over the same percentage levels as RiseTime (default 10 → 90)
SlewFall	SlewFall · SlewFall 20, 80	First falling-edge slew rate Δv/Δt over the same percentage levels as FallTime (default 90 → 10; reported as a positive magnitude)
PulseWidth	PulseWidth · PulseWidth 0.5	Average high pulse width (rise→fall) at threshold (default Avg)
LowPulseWidth	LowPulseWidth · LowPulseWidth 0.5	Average low pulse width (fall→rise) at threshold (default Avg)
Duty	Duty · Duty 0.5	Average duty cycle as a fraction 0…1 (high time / period) at threshold (default Avg)
EdgeCount	EdgeCount · EdgeCount 0	Number of level crossings at threshold (default Avg)
RiseCount	RiseCount · RiseCount 0	Number of rising crossings at threshold (default Avg)
FallCount	FallCount · FallCount 0	Number of falling crossings at threshold (default Avg)
Jitter	Jitter · Jitter 0	Std. dev. of complete-wave periods at threshold (default Avg; needs ≥ 2 periods)
PeriodStd	PeriodStd	Alias of Jitter
PeriodMin	PeriodMin · PeriodMin 0	Minimum complete-wave period at threshold (default Avg)
PeriodMax	PeriodMax · PeriodMax 0	Maximum complete-wave period at threshold (default Avg)
PeriodPkPk	PeriodPkPk · PeriodPkPk 0	Range of complete-wave periods (max − min) at threshold (default Avg)

Integrals & Power

Operation	Example	Result
Integral	Integral	Trapezoidal ∫v dt over the capture
Power	Power · Power 75	Mean RMS power Vrms²/R in watts (default 50 Ω)
Energy	Energy · Energy 50	Trapezoidal ∫v²/R dt in joules (default 1 Ω, preserving the original bare Energy)
dBm	dBm · dBm 75	RMS power in dBm into R Ω (default 50): 10·log₁₀(Vrms²/R / 1mW)

Spectrum

Operation	Example	Result
FFT	FFT · FFT 1024 · FFT 1024, middle · FFT 1024, end	Calculate and retain a real FFT for subsequent spectral analysis
THD	THD	Total harmonic distortion from the retained FFT: √(Σ Aₕ²) / A₁ for harmonics 2…10; displayed as a percentage
Fundamental	Fundamental	Frequency and Hann-corrected peak amplitude of the retained FFT's strongest AC component
Harmonic	Harmonic 3	Hann-corrected peak amplitude of harmonic N of the retained FFT fundamental
SineWaveType	SineWaveType	Similarity to a sinusoid: fundamental with no ideal higher harmonics
SquareWaveType	SquareWaveType	Similarity to a 50% square wave: odd harmonics proportional to 1/n
SawtoothWaveType	SawtoothWaveType	Similarity to a sawtooth: every harmonic proportional to 1/n; SawWaveType is an alias
TriangleWaveType	TriangleWaveType	Similarity to a triangle wave: odd harmonics proportional to 1/n²
WaveType	WaveType	Preset that runs all four wave-type similarity analysis

FFT-dependent analysis are sequential. THD, Fundamental, Harmonic, and all wave-type analysis must appear after an FFT in the same analysis list; otherwise parsing fails. They reuse that exact spectrum instead of calculating another FFT and therefore do not accept a point-count argument:

rigol2spice.exe input.csv -a "FFT 1024, middle; THD; Fundamental; Harmonic 3"
rigol2spice.exe input.csv -a "FFT 4096, start; WaveType"

If another FFT appears later, subsequent dependent analysis use the new result. Other analysis between them do not discard the retained spectrum.

FFT syntax is FFT [points][, position]. The position may be start, middle, or end; start is the default. The position can also be supplied without a point count (FFT middle), although it only changes the selected samples when the requested window is shorter than the capture. If the capture has fewer than the requested number of points, all available samples are used.

Before the FFT, the selected values have their mean removed and receive a Hann window. They are then zero-padded to the next power of two. The reported peak is the strongest local AC maximum with sub-bin refinement. Console output includes the actual point count and window position; -p adds the retained magnitude spectrum in dB to the SVG plot. Its axes use rounded 1–2–5 frequency divisions, labeled dB grid lines, and a maximum 120 dB visual range; the title also shows the FFT bin resolution Δf.

Wave-type analysis compares measured harmonic-amplitude ratios, up to harmonic 10 or Nyquist, with the four ideal profiles above. SineWaveType is max(0, 100% − THD%) over those harmonics. Each other score is 100% minus the harmonic-profile vector error relative to the energy of its ideal higher harmonics, limited to 0–100%. The four scores are independent and are not normalized to sum to 100%. This allows an unknown waveform to score poorly in every analysis and an ambiguous waveform to match more than one profile. At least the third harmonic must fit below Nyquist. Noise, PWM duty cycles other than 50%, clipping, asymmetry, frequency drift, too few periods, analogue bandwidth, and spectral leakage can reduce the match. The classifier deliberately uses magnitudes rather than phase, making it insensitive to time reversal and waveform polarity.

Post-Processing Options (-d, -k, -p)

Option	Effect
-d, --downsample N	Reduce only the converted signal output by factor N, after transformations, analyses, and decoding
-k, --keep-all	Disable removal of redundant (collinear) points
-p, --plot [file]	Write an SVG plot of the processed signal (default: plot.svg)

The plot uses 1 pixel per sample, auto-scaled Y (min / avg / max markers), and decade-spaced X time markers. When -a is combined with -p, analysis results are listed in a text block under the time plot (not drawn over the waveform). When --decode is combined with -p, a time-aligned decoded-protocol lane is added below the time plot. An FFT analysis also appends a spectrum panel (dB vs frequency, peak marker). A console warning is emitted above 10 000 points (use a Downsample transformation or a shorter time window when the plot itself should contain fewer samples).

Full Usage Reference (-h)

USAGE: rigol2spice.exe [<options>] <input-file> [<output-file>]

OPTIONS:
  -l,  --list-channels            List channels and exit
  -c,  --channel <channel>        Channel or math expression (default: CH1)
  -t,  --transformations <value>  Ordered transformations separated by semicolons or line breaks
  -tf, --transformations-file <value>
                                  Read transformations from a file before inline commands
  -a,  --analysis <value>         Ordered analyses separated by semicolons or line breaks; FFT dependants follow FFT
  -af, --analysis-file <value>    Read analyses from a file before inline commands
       --decode <value>           Decode UART, I2C, or SPI signals
       --decode-format <format>   Decoder output: text, csv, or bin (default: text)
       --decode-output <file>     Write decoded output to a file
  -d,  --downsample <ratio>       Downsample ratio for the converted signal output
  -f,  --format <format>          Output format: pwl, matlab, wav32, wav16, npy, or npz (default: pwl)
  -k,  --keep-all                 Keep all sample points
  -p,  --plot [<file>]            Write SVG plot (default: plot.svg)
  -h,  --help                     Show help

output-file is optional with -l, -p, -a, -af, or --decode.

Building from Source

Requires Swift 6.2+.

git submodule update --init --recursive
swift build

Builds on macOS, Windows, and Linux.

Acknowledgments

Special thanks to Scott Prahl, author of RigolWFM, for publishing the reverse-engineering research and Kaitai Struct schema that made DS1000Z WFM support possible.

Legal

This is an independent project, not affiliated with Rigol Technologies, Inc.