Separate DSP latency from acoustic travel time by writing a complete timing budget for each signal path, then subtracting the two total arrival times. Distance predicts only the air-propagation term; a properly referenced acoustic measurement can already contain both propagation and equipment delay.
Open the Audio Delay Calculator.
How to Separate DSP Latency from Acoustic Travel Time
Choose a common starting event, such as the same electrical input signal, and a common ending location, such as one fixed microphone position. Every time value must refer to those same boundaries before the paths can be compared.
For a simplified path, arrival time is the sum of electronic latency, acoustic propagation, and any intentional adjustable delay. Electronic latency can include conversion, buffering, processing, transport, and the enabled operating mode of powered equipment.
Acoustic propagation is distance ÷ sound speed. At the calculator’s 20°C model condition, sound speed is 343.42 m/s, so each meter contributes approximately 2.912 ms. That conversion does not describe how fast a DSP executes a filter or how long a digital network buffers audio.
Room EQ Wizard’s group-delay documentation notes that a measurement can include computer, interface, processing, and source-to-microphone travel delays. The measured value therefore needs its reference method documented before any term is subtracted.
Draw both signal paths before doing arithmetic
List the devices and routes from the shared input to each loudspeaker. Mark any components common to both paths and identify the point where the routes diverge. Shared latency usually cancels when comparing the relative arrival of the two paths.
For example, if both outputs pass through the same mixer stage before one receives an additional processor, the common stage is not an extra difference. Counting it on only one side creates a timing error even if the copied manufacturer latency figure is accurate.
Also identify intentional output delay that is already active. A processor may display a user-set delay separately from its fixed internal latency. A budget that omits the existing setting can incorrectly recommend adding the same correction again.
Keep a record of device modes and presets. Different processing options, routing, sample-rate conversion, or transport arrangements may have different latency. Use specifications for the actual enabled configuration rather than treating one product name as one timeless delay value.
Build an illustrative timing budget
Suppose the main speaker is 18 m from the chosen microphone and a fill speaker is 6 m away. Assume the main electronic path contributes 1.2 ms and the fill path contributes 5.0 ms, with no intentional alignment delay initially active.
| Timing term | Main path | Fill path |
|---|---|---|
| Electronic latency, assumed | 1.20 ms | 5.00 ms |
| Distance to microphone | 18 m | 6 m |
| Acoustic travel at 20°C model | 52.41 ms | 17.47 ms |
| Predicted total arrival | 53.61 ms | 22.47 ms |
The fill arrives about 53.61 − 22.47 = 31.14 ms earlier. Adding approximately 31.14 ms to that path makes its predicted total match the main at the selected point under these assumptions.
The geometric path difference alone is 12 m, equivalent to 34.94 ms. The fill already has 3.8 ms more electronic latency, so the additional adjustable delay is 34.94 − 3.8 = 31.14 ms.
These equipment latencies are invented for the worked example. They are not specifications for a real mixer or powered loudspeaker, and a real alignment should use documented values or appropriately referenced measurements.
Enter relative electronic latency in the calculator
The embedded tool converts distance difference to milliseconds using its temperature model, then adds or subtracts the entered processing value. It cannot identify which physical path has the extra latency from a positive number alone.
For the example, enter 12 m, 20°C, and 3.8 ms, selecting subtraction because the earlier acoustic path is already electronically later than its counterpart. The result is approximately 31.14 ms of extra delay for the fill.
If the extra 3.8 ms were instead on the main path, the fill would need the geometric 34.94 ms plus 3.8 ms, giving about 38.74 ms. Write the path names and sign before selecting the mode.
A negative result means the selected path is already later overall than the target under the entered assumptions. A normal delay block cannot advance it by applying negative time. Recheck the path selection and consider which path actually needs added delay.
Do not feed the complete acoustic arrival time into a field intended for electronic difference. A measured 22.47 ms fill arrival includes its 6 m travel term in this example; entering all of it as processing would subtract acoustic time twice.
Understand sample-based latency specifications
Some equipment or software describes a delay in samples. Convert with milliseconds = samples ÷ sample rate × 1,000, using the actual rate in samples per second.
At an illustrative 48,000 samples per second, 64 samples equal 1.333 ms and 256 samples equal 5.333 ms. At 96,000 samples per second, the same sample counts correspond to half those durations.
| Illustrative sample count | At 48 kHz | At 96 kHz |
|---|---|---|
| 32 samples | 0.667 ms | 0.333 ms |
| 64 samples | 1.333 ms | 0.667 ms |
| 128 samples | 2.667 ms | 1.333 ms |
| 256 samples | 5.333 ms | 2.667 ms |
This table converts counts only. It does not establish that a real device’s complete latency is exactly one selected buffer or that changing sample rate halves every stage. Conversion, algorithm, transport, and additional buffering may contribute separately.
Do not equate a computer’s selected buffer size with complete one-way system latency without the software and interface documentation. The setting is one input to a larger path, not necessarily the entire measured delay.
Distinguish one-way, round-trip, and relative measurements
A round-trip interface measurement can include output conversion, input conversion, buffering, and software routing. Dividing it by two assumes symmetry that may not exist. It does not automatically establish the latency from a mixer input to one powered speaker’s acoustic output.
For speaker alignment, relative arrival at the same microphone can be sufficient even when absolute component latencies are unknown. The two paths still need a stable common timing reference so their measured difference is meaningful.
If measurements report main arrival at 54 ms and fill arrival at 23 ms using a compatible reference, their 31 ms difference already includes the measured acoustic and electronic relationship. Do not add another geometric 35 ms correction on top simply because a tape measure predicts that component separately.
Use the geometric budget as a reasonableness check on the measured difference. A large unexplained discrepancy can indicate incorrect reference settings, another active delay, a wrong path distance, or a different processing mode.
Preserve the measurement’s time reference
Measurement software may reposition time zero to an impulse peak, response start, estimate, or explicit reference. If each speaker is independently shifted to its own zero, the displayed traces can look aligned while their original arrivals were different.
Room EQ Wizard’s measurement guidance describes timing-reference and time-zero options. Use a method appropriate to the system and keep its reference unchanged across the measurements being compared.
The microphone should remain at the same position when the objective is the relative timing at one listener. Moving it changes the propagation terms and makes the two measured arrivals belong to different geometries.
Save unmodified measurements or preserve the applied offsets in the record. This allows a later reviewer to distinguish actual arrival differences from display shifts introduced during analysis.
Keep pure time offset separate from filter behavior
An ideal constant delay shifts all frequency components by the same time, producing a linear phase slope. Real crossovers and loudspeaker responses can also introduce frequency-dependent phase and group delay.
Subtracting one geometric number does not remove those effects. A subwoofer or band-limited system may have a broad impulse response, and the position of its largest peak is not a universal definition of arrival that can be compared casually with a full-range speaker’s sharp peak.
The REW impulse documentation explains how the displayed response and its timing offsets are represented. Use the relevant frequency response, phase, and summation around the overlap region rather than trying to turn every visible peak into one fixed DSP-latency number.
If a preset or crossover changes, recheck the alignment. The cabinet position may be unchanged while the electrical and acoustic response relevant to integration has changed.
Use a repeatable verification sequence
First document routing, active delays, modes, sample rate, microphone position, and temperature. Calculate the propagation difference, then add the known unequal electronic terms with an explicit sign convention.
Next measure each path using the same valid reference, compare the arrivals, and assess combined response through the relevant overlap. Keep the original setting available before applying an adjustment so the comparison can be repeated.
Finally check nearby listening positions and the intended coverage area. A setting optimized at one point may be a compromise elsewhere because relative distances change. Do not describe a one-point arithmetic match as complete room-wide alignment.
Should every device latency be added to both paths?
For an absolute arrival budget, include each stage that lies between the chosen boundaries. For relative alignment, identical common stages cancel; count only genuine differences once.
Can an acoustic measurement reveal DSP latency by itself?
It can reveal complete system timing relative to its reference. Isolating DSP latency requires knowing or eliminating other terms, including propagation and measurement-chain delay, under a controlled comparison.
References
- Room EQ Wizard: Group delay and contributing timing terms
- Room EQ Wizard: Making measurements and using timing references
- Room EQ Wizard: Impulse graphs and time offsets