CamillaFIR User Manual
CamillaFIR User Manual
This manual is the practical end-user guide for CamillaFIR. It explains what the program does, which files it expects, how to choose the main settings, what gets exported, and how to deploy the result to a convolution-capable playback system.
CamillaFIR generates room-correction FIR filters from REW measurements. The typical workflow is:
- Measure the left and right channels in REW.
- Load the measurement files into CamillaFIR.
- Choose a target curve and operating mode.
- Generate filters.
- Export and deploy the output to CamillaDSP, Roon, Equalizer APO, or another FIR-capable DSP.
1. What is CamillaFIR?
CamillaFIR is a room-correction FIR filter generator for loudspeaker and listening-room optimization.
It is designed to:
- read REW measurement exports
- compare the measured response against a target curve
- create left and right FIR filters
- export ready-to-use files for convolution playback systems
- keep correction behavior safer and more repeatable through built-in guard rails
CamillaFIR can work in several styles:
- fully automatic with preset search
- guided manual operation with safety clamps
- advanced manual tuning with more direct control
2. Requirements
Before you start, you need:
- REW (Room EQ Wizard)
- a calibrated measurement microphone
- separate left and right channel measurements
- a playback system that supports FIR convolution
Common deployment targets:
- CamillaDSP
- Roon convolution
- Equalizer APO
- other DSP or player software that accepts FIR WAV impulse files
Recommended preparation:
- use the same microphone position and measurement method for both channels
- keep the sample rate consistent through your measurement and playback chain
- verify that REW exports include phase when using text files
3. Workflow Overview
The normal workflow inside CamillaFIR is:
- Files tab Load left and right measurement files.
- Target tab Select a built-in target curve or load a custom target.
- Basic tab Choose the operating mode, filter type, sample rate, taps, and the main safety settings.
- Advanced / Window / XO tabs Refine advanced correction, timing, windowing, and crossover behavior when needed.
- Run tab Start the generation process and review the produced output.
For most systems, start simple:
- load measurements
- choose
CamillaFIR automatic modeorBasic (recommended) - choose
AsymmetricorMixed Phase - generate filters
- listen and re-measure before making more aggressive changes
4. Input Files
CamillaFIR supports the two common REW export workflows.
4.1 REW text export (.txt)
This is the most direct input format for frequency-response processing.
Expected content:
- frequency in Hz
- magnitude in dB
- phase in degrees
Important notes:
- the left and right channels must be loaded separately
- phase data is required if you want full phase-aware correction behavior
- headers are acceptable as long as the export is a normal REW text export
4.2 REW impulse export (.wav)
CamillaFIR can also import impulse-response WAV files and convert them internally for processing.
Use this when:
- you prefer an IR-based workflow from REW
- you want CamillaFIR to derive the response using its own windowing path
Practical guidance:
- export mono impulse files for each channel
- keep the export settings consistent between left and right
- match the sample rate to the rate you expect to use in playback when practical
5. Operating Modes
CamillaFIR has three main operating modes. The DSP engine is the same, but the workflow policy changes.
5.1 CamillaFIR automatic mode (recommended)
Use this when you want the program to search for a technically strong preset automatically.
What it does:
- evaluates multiple preset combinations
- scores candidates
- picks the best candidate before final export
- keeps the workflow guarded compared with unrestricted expert tuning
Best for:
- first-pass results
- users who want a strong starting point quickly
- rooms where you do not want to hand-tune every parameter
5.2 Basic
Use this when you want manual control but still want guard rails.
What it does:
- applies conservative defaults
- keeps important safety clamps active
- reduces the chance of extreme or unstable correction
Best for:
- most home systems
- users learning the software
- repeatable day-to-day filter generation
5.3 Advanced (expert)
Use this when you already understand the tradeoffs and want broader control.
What it does:
- exposes more direct tuning freedom
- relaxes mode-based safety constraints
- makes you responsible for the result
Best for:
- experienced users
- deliberate A/B testing
- expert tuning of difficult systems
6. Target Curves
CamillaFIR includes a large set of built-in target curves, including:
- Harman variants
- Dr. Toole
- B&K variants
- Flat
- cinema and speech-oriented targets
- nearfield and studio-style tilts
You can also load a custom target file.
General guidance:
- a mild house curve is often more natural than perfectly flat in-room playback
- bass-heavy targets can sound impressive, but they increase headroom demand
- if you are unsure, start with a moderate built-in target and listen before pushing bass harder
In automatic mode, CamillaFIR can also select a target automatically when that workflow is enabled.
7. Filter Types
CamillaFIR supports four main filter types.
7.1 Linear Phase
- strongest timing accuracy
- can produce the highest latency
- can introduce audible pre-ringing in some situations
Latency estimate:
- latency is approximately
taps / (2 * sample_rate) - at 44.1 kHz and 65,536 taps, linear-phase latency is about 743 ms
Best for:
- offline listening
- systems where latency is not important
- users prioritizing timing accuracy
7.2 Minimum Phase
- very low latency behavior
- magnitude-focused correction
- avoids linear-phase pre-ringing behavior
Best for:
- low-latency systems
- conservative playback chains
- cases where phase correction is not the main goal
7.3 Mixed Phase
- linear-style behavior in bass
- minimum-phase style behavior at higher frequencies
- good balance between correction quality and practical use
Best for:
- general home listening
- users who want strong bass correction without full linear-phase behavior everywhere
7.4 Asymmetric
- near-linear low-frequency control with earlier impulse placement
- lower perceived delay than full linear phase
- often the most practical all-round choice
Best for:
- mixed music, movie, and desktop use
- users who want strong correction with more practical timing behavior
For many systems, Asymmetric is the most sensible first choice.
8. Key Settings
These settings have the biggest effect on the result.
8.1 Taps
Taps define FIR length.
- more taps improve low-frequency resolution
- more taps also increase latency and CPU load
- short filters are faster, but they can lose low-frequency precision
8.2 Sample Rate
Choose a sample rate that fits your playback chain.
- match the playback system when possible
- avoid unnecessary sample-rate conversion if you do not need it
- use
Generate All Rateswhen you want one export package for multiple playback rates
8.3 Correction Range
The correction band defines the range where CamillaFIR is allowed to work.
Typical use:
- lower limit protects against unnecessary infra-bass correction
- upper limit helps avoid over-correcting poorly measured high frequencies
If the result sounds unnatural, narrowing the correction range is often safer than increasing algorithmic aggressiveness.
8.4 Maximum Boost and Cut
These settings limit how strongly the filter can move the response.
- high boost increases headroom risk
- strong cuts can sound overdamped if overused
- safer values are usually better for first runs
8.5 Level Matching
Level matching aligns the measured response to the target before correction is generated.
CamillaFIR supports:
- smart automatic scan
- manual window selection
If the target sounds consistently too lean or too heavy, check the leveling window before changing multiple other settings.
8.6 HPF and Low-Bass Protection
Low-frequency protection matters for real loudspeakers.
Relevant controls include:
- HPF
Low-bass boost lockExcursion Protection
These are especially important for:
- small woofers
- bass-reflex systems
- ambitious house curves
9. Advanced Features
9.1 Temporal Decay Control (TDC)
TDC is designed to reduce excessive ringing and slow modal decay, especially in bass.
In practice it can:
- tighten bass decay
- reduce boom
- improve bass articulation
Use it carefully:
- stronger settings can sound drier
- if bass becomes too lean or over-damped, reduce TDC strength first
9.2 Adaptive Frequency-Domain Windowing (A-FDW)
A-FDW changes how finely different frequencies are corrected based on measurement confidence.
In practice:
- high-confidence regions can follow the target more closely
- low-confidence regions are smoothed more conservatively
This helps avoid over-correction in uncertain data.
9.3 Bass-first AI
Bass-first AI tells the system to treat low-frequency room modes as meaningful correction targets instead of smoothing them away too aggressively.
Use it when:
- room modes dominate the bass
- you want stronger intentional low-frequency cleanup
Be more conservative if:
- measurements are noisy
- the speakers have limited bass capability
9.4 Excursion Protection
Excursion Protection is a dynamic safety feature.
It helps prevent dangerous low-frequency boost by reacting to the measurement and target relationship.
Use it when:
- you are using bass-heavy targets
- your speakers have limited low-frequency headroom
- you want safer automatic bass behavior
9.5 Stereo Link
Stereo Link helps preserve left/right balance and phantom center stability.
Keep it enabled unless you have a specific reason to let each channel behave more independently.
9.6 IR Windowing
Additional controls affect timing and final impulse behavior:
- IR window style and shape
- asymmetric window placement controls
These matter most when you are refining latency behavior, channel alignment, or export impulse shape.
10. Running Filter Generation
When the main settings look correct:
- verify that both channels are loaded
- confirm target curve and mode
- review sample rate, taps, correction range, and safety limits
- start the run from the
Starttab
During generation, CamillaFIR evaluates the measurements, builds the target relationship, applies safety logic, and exports the final result.
If you are using automatic mode, the program may take longer because it is comparing multiple candidates instead of generating only one direct configuration.
11. Output Files
After a successful run, CamillaFIR produces a result bundle. Common output files include:
L_...wavandR_...wavLeft and right FIR impulse-response filesSummary_...txtHuman-readable report with scoring, settings, and diagnosticscamilladsp_...ymlCamillaDSP-ready configuration file- ZIP package Collected export bundle for convenient deployment
Depending on the workflow, the export package can also contain additional configuration files alongside the WAV filters.
12. Reading the Summary
Summary_...txt is the main report for understanding what CamillaFIR produced.
Useful sections include:
- acoustic score and target match
- confidence metrics
- core settings used for the run
- gain, headroom, and clamp behavior
- TDC, A-FDW, bass-first, and protection status
- alignment and stereo-related diagnostics
Use the summary to compare runs, especially when you change:
- target curve
- filter type
- correction range
- taps
- protection settings
When comparing runs, change one or two variables at a time. This makes the summary much more useful.
13. CamillaDSP Integration
CamillaDSP deployment is the most direct workflow.
Typical steps:
- copy the exported FIR WAV files to your CamillaDSP coefficients folder
- copy or adapt the generated
camilladsp_...yml - point your CamillaDSP setup to the new filters
- verify gain staging before normal listening levels
Practical notes:
- confirm that the sample rate matches your playback path
- if you hear clipping or overload, reduce master gain or convolution gain
- always test at moderate volume first
14. Roon and Equalizer APO Integration
CamillaFIR also works with other convolution-capable systems.
14.1 Roon
- load the exported FIR WAV files into Roon’s convolution engine
- make sure the sample-rate handling matches your playback chain
- leave enough headroom if the result includes bass lift
14.2 Equalizer APO
- use the generated FIR WAV files in the convolution block
- apply enough preamp reduction to avoid clipping
- re-check channel assignment and sample-rate behavior on Windows
15. Health Checks
CamillaFIR performs validation before and during a run.
Typical result classes:
WARNNon-critical issues worth reviewingCRITCritical issues that may block the run, especially in guarded workflows such as Basic or automatic mode
Warnings can indicate:
- risky boost/headroom conditions
- incomplete inputs
- setting combinations that are technically possible but not recommended
If a health check fails, fix the root cause instead of trying to bypass the warning blindly.
16. Troubleshooting
16.1 The run will not start
Check:
- both channels are loaded
- the files are valid REW exports
- the sample-rate and mode settings are sensible
- any
CRIThealth checks have been resolved
16.2 The correction sounds too thin
Check:
- target curve choice
- leveling window
- correction upper range
- TDC strength
- low-bass protection settings
16.3 The correction sounds too bright or aggressive
Try:
- lowering the correction upper limit
- using a gentler target
- increasing smoothing or using safer defaults
- moving from Advanced to Basic settings for comparison
16.4 Bass is still boomy
Check:
- whether TDC is enabled
- whether Bass-first AI is appropriate for the room
- whether taps are long enough for bass resolution
- whether the measurements themselves are reliable
16.5 Bass disappears or becomes weak
Check:
Low-bass boost lockExcursion Protection- HPF frequency and slope
- overly aggressive TDC or conservative target selection
16.6 Stereo image feels unstable
Check:
- left/right measurements were taken consistently
Stereo Linkis enabledAuto-Align L/Rbehavior is sensible- one channel is not using a different file or rate accidentally
17. Best Practices
For the most reliable results:
- start with clean measurements
- use one target change at a time
- avoid extreme boost on first runs
- keep Basic mode as your reference workflow
- listen, then re-measure
- compare summaries instead of relying only on memory
The best correction is usually not the most aggressive one. A stable, repeatable, and well-protected filter is usually the better long-term result.
18. Recommended First Run
If you want a safe starting point, use this approach:
- export left and right REW measurements
- load them into CamillaFIR
- choose
BasicorCamillaFIR automatic mode (recommended) - choose a moderate built-in target
- choose
AsymmetricorMixed Phase - keep boost conservative
- generate the filters
- deploy them at reduced volume first
- re-measure with the filters active
That is the fastest path to a useful and defensible first result.