Version 2.30.1
Design the filter.
Then print the page
that goes in the rack.
Alpha Sound Lab takes a measurement, aligns a multi-way system, fits the correction, and works out the limiter settings each driver actually needs. Then it prints all of it on one page you can type into any processor — including every processor that has never heard of FIR.
Windows 10 1809 or later · measuring free permanently · no account, no telephone home
The workflow
Nine stages, in the order you actually work
Take a measurement or import one. Align the ways. Set the crossovers. Fit the correction by hand or automatically. Listen to it. Export it. Then measure again and see whether it did what it promised. Each stage shows you its result before you move on — there is no button that does everything, on purpose.
- 1Import
Smaart, REW, or any file with frequency, magnitude, phase and coherence columns. Mic calibration is applied once, and the tool records who applied it.
- 2Measure
Take your own, with an interface, a microphone and a loopback cable. Swept sine, several sweeps averaged, a level meter before the measure button rather than a warning after it — and an impulse view that shows when the sound arrived, in milliseconds and metres. Two live modes sit in front of it, for watching the system while you move things rather than after.
- 3System
Sub, low, mid, high. Delay, polarity and gain per way, with the alignment worked out from where the ways actually overlap.
- 4Target
Flat, tilted, shelved — the curve you are correcting towards, owned in one place.
- 5FIR Design
Windowed-sinc and linear-phase Linkwitz-Riley legs, with the delay and polarity carried in the taps.
- 6IIR / PEQ
Parametric bands by hand, with Q and bandwidth in octaves side by side.
- 7Auto Fit
Levenberg-Marquardt fitting that prefers few bands to many, and refuses to correct what it cannot see.
- 8Audition
Hear the correction on your own material before it goes anywhere near a rig.
- 9Export
Coefficients, processor presets, and the settings sheet.
New in 2.19
Two ways of watching it happen
A capture tells you what the system did. Neither of these waits for that. One window, two modes, and they answer different questions — so the program ships both rather than picking.
Live IR — where, and how far
A short sweep, fired over and over. The impulse response and the arrival distance move while you move the microphone or aim the loudspeaker, in milliseconds and in metres, measured against the loopback rather than against your driver’s buffers.
This is the one for finding a cluster’s arrival time at the mix position, and for the delay between an array and its subs. A sweep puts all its energy at one frequency at a time, so it buys 20–30 dB of noise rejection over a broadband signal and its impulse peak is sharp enough to read.
It is honest about the cost of being quick: a short sweep spends more of its octaves inside its own fades, so a live frame is trustworthy over a narrower band than the capture that follows it. The window prints both frequencies and draws nothing outside them.
Live TF — what it is doing, right now
A dual-FFT transfer function. It compares the reference and the microphone continuously and reports the difference, so it works on pink noise, on the walk-in music, or on the band with the room full. You do not fire a sweep at an audience; this is what you use instead.
Magnitude, phase and coherence, on five transform lengths at once so the bottom of the trace resolves and the top still keeps up. Same loopback the sweep already needs — no new wiring.
And it says what it does not know. The phase trace is marked not aligned until the flight time has been found, because until then it is describing the distance rather than the loudspeaker. Coherence is not drawn at all until two frames have been averaged: after one, the arithmetic gives exactly 1 for any two signals whatsoever — including a microphone that is not plugged in.
Live IR is free, with the rest of measuring. Live TF is what a licence buys — it is not another measurement so much as a different kind of tool: a continuous analyser you work with while a show is running, rather than something you take and put down.
Measuring a system
The part you used to finish on paper
Microphone at the mix position. Measure the left array, the right array, the subs, the front fills and each delay tower, one at a time. The differences between their arrival times are the delays to set — and now something holds them and does the subtraction.
It knows which way round the answer goes
You can only ever add delay, so the source that arrives last is the reference and takes none; everything earlier is held back to meet it. Delaying the far source instead of the near one produces numbers that look entirely reasonable and put a tower further out of time than it started.
Every delay is shown in milliseconds and in metres, because a delay is a distance and “the tower is 34 metres behind the main” is something you can check by walking it.
Temperature cannot move it
The delay is a difference between two measured times, and neither was worked out from a distance — both were measured against the loopback. A cold hall and a hot afternoon give an identical answer. Only the metres beside it move.
And it refuses to subtract across two microphone positions. Two arrivals from different seats differ by the seat as well as by the source, and no arithmetic separates them afterwards, so a row measured somewhere else is listed and left out rather than quietly folded in.
On the bench
What did the new woofer change?
The program works on one measurement at a time, which is right for designing a filter and wrong for the question people spend afternoons on. Keep a measurement on the shelf and the next one no longer destroys it.
Measure the box. Change the driver. Measure again. Then overlay the two, or hold them against each other in numbers — with their mean levels aligned first, so what is reported is a difference in shape. A replacement driver of a different sensitivity would otherwise show a flat few decibels across the whole band: true, and not the question. The sensitivity difference is reported on its own line so it is not lost.
And if the program ever closes without saving, it keeps a copy aside and offers your work back when it next starts — still unsaved, still pointing at the file it came from.
The part nobody else does
A page you can type into any processor
Every FIR designer assumes the job ends with a coefficient file loaded into a FIR-capable box. Most boxes in the world are not that, and the engineer standing in front of a DriveRack is not going to buy a rack to use a measurement tool.
So the tool prints the numbers instead: crossover points and slopes, delays in milliseconds and samples and metres, polarity, gain, the band list with Q and bandwidth both, and the limiter settings for every way. A system set up from that sheet and a DriveRack gets the alignment, the crossover and the EQ. It gives up only the linear phase.
- Prints, emails, or survives being photographed on a phone
- Readable six months later without this program installed
- States what it does not model, so nobody treats it as a guarantee
================================================================
SYSTEM SETTINGS SHEET
Main hangs - 4-way line array
48000 Hz 343 m/s assumed for distances
================================================================
WAY 4 High 1.4" horn
----------------------------------------------------------------
Crossover high-pass 1.60 kHz Linkwitz-Riley 24 dB/oct
(no low-pass - this way runs to the top)
Delay 4.104 ms 197.0 samples 1.408 m / 4.62 ft
Polarity normal
Gain -4.5 dB
RMS limiter threshold -1.72 dBu -3.94 dBV 0.635 V
attack 0.62 ms release 100 ms
Working driver 80 W AES / 8 ohm -> 25.30 V
amp 700 W / 8 ohm, 32.0 dB gain
timing from one period of 1600 Hz
at the threshold the driver sees 80 W
Processor limits
It designs what your processor can actually take
Fitters do not know what your box accepts. On a narrow room mode one will happily return Q 8, and on a deep dip it will ask for +15 dB. Plenty of processors stop at Q 4 and +13 dB, and every brand draws that line somewhere different.
Tell it what yours takes and Auto Fit works inside those limits. Not fitted freely and clamped afterwards — that would leave a curve nobody optimised, with no record of what moved. Bounded, it places bands differently: two gentle ones where a narrow one was not allowed.
And when a limit genuinely bites, it says so by name rather than quietly handing you a gentler correction than the measurement justified.
- Pick your processor from a list, or type four numbers off its manual
- Works for the legacy box no shipped device list would ever include
- Separate limits for FIR — taps, rate and latency budget
- Boost and cut are separate, because real processors are lopsided
- Nothing is clamped or shortened for you — it tells you, you decide
Processor RHAON STLA/9R (fw 1.10)
Boost / cut +13.0 dB / -18.0 dB
Q 0.10 to 4.00
Source reported by the owner of the unit
---- after Auto Fit -------------------------------------------
Fitted 6 bands in 25 iterations: 2.19 dB RMS error
reduced to 0.16 dB. 1 band was held back by RHAON
STLA/9R - the fit wanted more than it accepts
---- after designing a FIR ------------------------------------
RHAON STLA/9R holds 256 taps but this design is 1025.
Shortening it costs resolution at the low end first,
which is usually where the correction was needed.
---- on the settings sheet ------------------------------------
* Band 2 (420.2 Hz) asks for 15.00 dB of boost;
RHAON STLA/9R stops at +13.00 dB.
Limiters
The same rig. Thirteen decibels apart.
An 18″ sub and a compression driver on one system want completely different limiter thresholds, and a single setting is wrong for both. The threshold is arithmetic — the driver’s rating referred back through the amplifier’s voltage gain — so the tool computes it per way rather than leaving you to guess.
18″ sub · 1600 W AES · 4 Ω
+8.28dBu
+6.06 dBV · 2.010 V
attack 33.3 ms, release 333 ms
1.4″ horn · 80 W AES · 8 Ω
−1.72dBu
−3.94 dBV · 0.635 V
attack 0.62 ms, release 100 ms
Give the horn the sub’s number and it sees
13.01dB too much
which is 10 × log₁₀(1600/80), exactly.
Three units, always, on the same line
Processors disagree about whether their threshold field means dBu, dBV or volts, and dBV sits exactly 2.2185 dB below dBu. Reading one into the other’s field is 2.2 dB hot — which a sub shrugs off and a compression driver does not. So every threshold is printed in all three and nobody converts anything at one in the morning.
And it refuses to guess
No driver rating, no suggestion. Thermal time constants, excursion below box tuning, and what a real amplifier does into a real load are named on every sheet as not modelled, because a threshold presented without them is a number people treat as a promise.
Several positions
What EQ can fix, and what it cannot
A measurement taken at one place describes one place. Part of what it shows is the loudspeaker, and EQ fixes that everywhere. The rest is the room, and EQ “fixes” that at the microphone while making every other seat worse. One measurement cannot tell them apart — not because it is a poor measurement, but because the information is not in it.
Measure the same source from several seats and the tool separates them. What is the same everywhere is the loudspeaker; what moves as you move is not.
| Frequency | What it is | Spread across seats | Coherence | Verdict |
|---|---|---|---|---|
| 94 Hz | port tuning | 2.1 dB | 0.97 | EQ it |
| 331 Hz | reflection null | 20.8 dB | 0.97 | leave it |
| 1804 Hz | horn dip | 6.4 dB | 0.97 | EQ it |
Coherence is 0.97 at all three. Coherence tells you how good the measurement is, not whether the thing it measured exists anywhere else — a 20 dB null in one seat can be a flawless measurement. Without the position check, a fitter would happily correct all three.
Pricing
Measuring is free. Correcting is what you buy.
Finding out what a system is doing costs you nothing and has no time limit. Designing the correction is the part that was paid for.
Free, permanently
Measure & Align
No cost, no account, no expiry
Everything up to the filter
- Measurement, with your own interface and microphone
- Live IR — watching the arrival while you move things
- The arrivals table — several sources, one microphone, the delays
- A measurement library, to hold one against another
- Measurement import and microphone calibration
- Multi-way system tree with per-way delay and polarity
- Alignment, and how exact it has to be
- Several-position analysis
- The settings sheet
- Saving and reopening your projects
One edition
Suite
USD 360 or PHP 20,500 within the Philippines
Everything above, plus the correction
- Live TF — the continuous analyser, for working on a system while it plays
- Target curves and FIR design
- IIR / parametric fitting
- Auto Fit
- Per-way limiter and compressor thresholds
- Coefficient and processor preset export
- Every future 2.x update, at no further cost
Bought once. Yours permanently. It does not lapse, it does not phone anywhere, and it keeps working whatever happens to us.
Buying one
PayPal internationally, GCash within the Philippines. Email masterpro0115@gmail.com with the name the licence should be issued to and you will be sent payment details and, on payment, your licence file. Licences are issued by hand, by a person, usually within a day.
Your licence is a small file. Keep it — it is the licence, it is not tied to any one computer, and the same file works wherever you install the program. If you lose it, ask and it will be sent again.
The name is inside the file and cannot be changed afterwards without a new one being issued. If the licence is for a company, send the company name.
For help afterwards, open a ticket on the issue tracker or email the same address — whichever you prefer. Either reaches the person who wrote the program, and there is no queue in front of you.
There is no money-back guarantee. That is exactly why the free half exists and why it has no time limit: measure your own loudspeakers, with your own microphone, in your own room, before you pay anything. Nobody should buy a measurement tool on a promise when they can buy it on a measurement.
Version 2.30.1
What changed
The full notes ship with the release on GitLab. This is the short version.
New
- Live IR A short sweep fired over and over, so the impulse and the arrival distance move while you move the microphone. In milliseconds and in metres.
- Live TF A dual-FFT transfer function that works on pink noise, the walk-in music or the band. Magnitude, phase, coherence and an impulse, at 1/3 to 1/24 octave resolution. Licensed.
- The arrivals table Several sources measured from one microphone position, and the delay each of them needs. The part of the job that used to be finished on paper.
- A measurement library Keep one so the next does not replace it, then overlay or compare. Measure the box, change the driver, measure again.
- It gives your work back If it ever closes without saving, the work comes back — still unsaved, still pointing at the file it came from.
- Air temperature, entered once Every distance in the program uses the figure you type. Between a cold load-in and a full house a delay tower reads two metres further off. It cannot change a delay: those are differences between measured times.
- It refuses to measure when Windows is set up wrongly The recording and playback halves of one interface can be set to different formats, and measuring through that is not reliable. The program now stops and says which part differs.
- The Phase tab draws the measurement Not only the filter being designed.
Fixed
- The delay finder could not align a subwoofer It looked only between 100 Hz and 8 kHz, so a sub-only stimulus returned a confident 0.00 ms against a real 5.00. It now follows the signal's own band. If you aligned subs with a 2.19–2.22 build, re-check them.
- It could invent a delay out of two unrelated signals The peak of correlated noise looked like an arrival; it is now judged against what chance would produce.
- The input meters froze after a measurement The panel went on saying it was listening. They now run during a sweep too, which is when clipping happens.
- An interface unplugged mid-session said nothing Meters froze, traces stopped, every panel claimed to be running.
- The live analyser could not run the machine out of memory Its audio queue is bounded, and an overrun restarts the average and says so.
- Live TF never played its own pink noise On drivers that kill the first playback after a device is reopened, it stopped a third of a second in — every time. It looked like a cable fault, because the frame counter runs on the recording side and climbs happily on silence. If the stimulus fails to start, it now says so.
- A crash took your measurements with it Recovery returned the arrivals table and discarded the captures underneath it. Worse, a session spent only measuring never counted as unsaved work, so no autosave ever ran. Captures now come back with their impulse and their signal-to-noise intact.
- The logo blew up on a slow connection Every inline graphic now carries its own size, so nothing balloons before the stylesheet arrives.
Download
Try all of it for sixty days
The demo is the whole program, unlocked, for sixty days. When they are up it does not shut down and does not take your work away: measuring and aligning stay free permanently, saving included. Only the correction — the filter design half — asks for a licence.
Download for Windows Read the manual
Windows 10 1809 or later, 64-bit · 30 MB installer
This program is not code-signed, so Windows SmartScreen shows a “Windows protected your PC” notice on first run. Choose More info → Run anyway. A certificate costs more each year than it would be worth to the people this is for, and it is not planned.