Micro Speaker SPL Requirements: How Loud Does Your Device Really Need to Be?
1. Why the Loudest Datasheet Number Is the Wrong Starting Point
Short answer: A published sensitivity figure is measured at a fixed distance, voltage and frequency, and often in a defined cavity. It describes the component, not the requirement, and it does not survive transfer to a different distance, cavity or measurement basis.
Micro speaker sensitivity is usually published in the form of a level at a stated frequency, at a stated distance, at a stated test voltage and often inside a stated cavity volume. All four qualifiers matter. Change the distance and the level follows the inverse distance relationship. Change the cavity and the level moves again. Change the test voltage and you move along the component's own level-versus-input curve rather than along a fixed value. A figure carried from one datasheet to a different product without these qualifiers is not a specification.
There is a second reason the loudest number misleads. Because sensitivity and rated power are largely independent properties, the highest published sensitivity in a catalogue frequently belongs to a low-power configuration, and vice versa. Selecting on maximum sensitivity alone tends to pull the programme toward a part that cannot deliver sustained output, which shows up later as distortion at normal listening levels rather than as insufficient volume.
2. The Three Inputs a Real SPL Requirement Needs
Short answer: A defensible target needs listener distance, the noise floor at that distance, and the content type. Without all three, any SPL figure is an assumption rather than a requirement.
1. Listening distance: the intended position of the listener, measured from the acoustic opening rather than from the product centre. This is the first number to fix, because everything else scales from it.
2. Noise floor at that position: the ambient or mechanical noise already present where the sound has to be understood. This sets the floor the signal has to clear, and it is usually measured rather than assumed.
3. Content type: what has to be understood at that position. Speech announcement, music playback and warning tones do not require the same level, and a voice-only product can often specify a lower target than a music product of the same size.
Once distance and noise floor are known, the required level follows from them rather than from a catalogue. The inverse distance relationship is the useful part: level falls by 6 dB for every doubling of distance, so a target set at 1 m and used at 30 cm carries roughly 10 dB more level than the original figure, while the same target used at 3 m carries about 10 dB less. This is why a requirement written for a bench condition can be simultaneously optimistic and pessimistic in a real product, and why it has to be recomputed rather than adjusted by feel.
3. Level, Intelligibility and Where the Level Should Sit
Short answer: Level is not the same as intelligibility. What carries intelligibility is the level in the critical speech band relative to the noise floor; low-frequency output contributes voice body and naturalness rather than comprehension.
Two separate questions are often merged into one specification. The first is whether the output is loud enough to be perceived and to cover the noise floor; the second is whether what is said can be understood. Merging them leads to over-specified levels, because raising overall level does not reliably improve comprehension once the speech band is already above the noise floor — it raises the whole spectrum, including parts that contribute nothing to intelligibility and increase distortion risk.
The more useful framing is a margin in the band that matters. Low-frequency output contributes voice body and naturalness, and matters in music content. Speech intelligibility depends much more on preserving the level in the critical speech band at a moderate overall level. Where speech intelligibility has to be measured, ITU-T P.807 is the appropriate recommendation; the P-series recommendations on speech quality address quality rather than intelligibility and should not be cited as intelligibility methods.
The practical consequence for a requirement is that the band and the level should be stated separately: a target level for the speech band at the listener position, and a target for overall level where music is expected. Programs that write only one number usually discover during acoustic testing that the level was adequate and the band was not.
4. Sensitivity Does Not Track Rated Power
Short answer: Across a published catalogue the same rated power class spans a wide sensitivity range, and the highest rated power does not produce the highest level. Rated power describes thermal and mechanical endurance, not loudness.
Table 1 groups published configurations by rated power and shows the sensitivity range inside each group. The 2.0 W group alone spans 93 to 108 dB, a 15 dB range across comparable power ratings. The 3.0 W group spans 95 to 105 dB. At the other end, the single 10.0 W configuration publishes 95 dB — the same figure as several 1.0 W configurations. If rated power determined level, these numbers would not behave this way.
Table 1: Published sensitivity range within each rated power class, with the underlying configurations.
Rated power | Sensitivity range as published | Configurations in the group | Diaphragm diameter / size range | F0 range |
0.8 W | 90–97 dB (7 dB span) | HS001534H39, HS150727H, HS150827H, HS002038H, HS160930H, HS241540H42, HS203045H45 | φ15 to 20 × 30 mm | 600–1050 Hz |
1.0 W | 90–96 dB (6 dB span) | HS204130H30, HS220942H42, HS280935H35, HS402060H, HS253540H, HS121722H, HS201623H, HS151130H, HS241534H34 | 12 × 17 mm to 40 × 20 mm | 700–1000 Hz |
2.0 W | 93–108 dB (15 dB span) | HS005717H, HS003021H, HS004528H, HS005017H, HS0066165H, HS007625H, HS-BX-0045-KT5, HS341135H, HS361331H, HS251233H, HS203595H30, HS250926H and others | φ18 to φ76 mm | 180–800 Hz |
3.0 W | 95–105 dB (10 dB span) | HS004023H, HS0066165H, HS007625H, HS284011H, HS-BX-3520, HS-BX-2514-YJT01, HS-BX-203008H | φ40 to 35 × 20 mm BOX | 500–1250 Hz |
5.0 W | 105 dB (single value) | HS004528H, HS005017H | φ45 × 28 mm to φ50 × 17 mm | 180–220 Hz |
10.0 W | 95 dB (single value) | HS00663H | φ66 × 32 mm | 180 Hz |
Table 2 lists the measurement basis behind each group, because the ranges above are only meaningful once the basis is visible.
Table 2: Published measurement basis by rated power class.
Rated power | Published measurement basis | Test voltage | Impedance class in the group |
0.8 W | SPL at 2 kHz / 10 cm, various cavity volumes (1 to 3 cc BOX or free volume as published) | 2.53 Vrms | 8 Ω ±15% |
1.0 W | SPL at 2 kHz / 10 cm, 1 cc BOX or free volume as published | 2.83 Vrms | 7 Ω ±15% and 8 Ω ±15% |
2.0 W | SPL at 2 kHz / 10 cm, cavity volume as published per configuration | 2.83 Vrms (4 Ω class) and 4.0 Vrms (8 Ω class) | 4 Ω ±15% and 8 Ω ±15% |
3.0 W | SPL at 2 kHz / 10 cm, cavity volume as published per configuration | 3.46 Vrms | 4 Ω ±15% |
5.0 W | SPL at 2 kHz / 10 cm | 4.47 Vrms | 4 Ω ±15% |
10.0 W | SPL at 2 kHz / 10 cm | 6.32 Vrms | 4 Ω ±15% |
5. When the Requirement Is a Limit Rather Than a Target
Short answer: Some products carry a maximum level instead of a minimum, because a regulation caps the sound a product may produce. In that case the compliance requirement runs through the component's frequency response, not through its overall level.
Not every programme needs the device to be louder. Products aimed at young users frequently carry a regulatory limit on maximum sound level at a defined measurement distance, and the engineering task becomes keeping the product below a ceiling rather than pushing it toward a floor. That inverts the selection logic, and it changes which parameter matters.
In a limit-driven programme, reducing overall power is the obvious approach and frequently the wrong one. Maximum level measured at a distance is set by the highest point in the frequency response, so lowering power reduces level everywhere at once and does not change the shape of the response. If the response has a narrow peak in the midrange, the peak is what the limit measures, and the perceived loudness of a broadband signal falls faster than that peak does. The result of the obvious approach is a product that passes the limit and sounds quieter than the alternative did before it was derated.
The effective approach is to change the response shape rather than the level. A configuration with genuinely wider frequency response delivers more perceived loudness for the same maximum level, because the energy is distributed instead of concentrated. In one such programme the original 29 mm driver could not meet the limit at a subjective level that was audible, and the customer's attempts to fix this through software settings, driver substitution and cavity modifications were all ineffective, because each of them changed level without changing the shape of the response.
Project Case Study (Hongsheng) — A Narrow-Band Mid-Range Design for Outdoor Announcement
The requirement came from an outdoor driving-test ground, where units are installed in roadside posts and announcements have to be understood at 3 m in daylight. The customer's specification included long-term outdoor exposure, moisture resistance and a warm-up under direct sun. The starting point was a 40 mm 4 Ω 10 W full-range driver, and more than ten full-range 40 mm candidates were evaluated without meeting the requirement.
The reason repeated substitution failed became the central finding of the project. The requirement was a voice band target, roughly 1 to 4 kHz. General-purpose 40 mm drivers are full-range designs whose energy concentrates in the low midrange, so the share of output inside the actual speech band is low. At matched power the midrange level was therefore insufficient, and no amount of choosing a different full-range model changed that, because the constraint was the frequency band rather than the model. Hongsheng therefore specified the driver for the band the product actually uses.
Four engineering changes were applied. The diaphragm material was changed to a polyester film, raising level in the 1 to 4 kHz band. The free-air resonance was designed to 900 Hz, placing the driver's resonance peak in the midrange instead of below it — for voice announcement at distance, narrow-band midrange output serves intelligibility better than extended low-frequency output, and this is the reverse of the usual expectation that a lower resonance is always better. Test points were then placed inside the speech band at 0.5, 0.8, 1.0 and 1.2 kHz, rather than at a conventional single 2 kHz point, so the supplier's published figures describe the band the product uses. Finally, three changes were made without altering the 40 mm exterior or the structure: a secondary magnet added 2 dB, a shim adjustment reducing the magnetic gap from 0.25 mm to 0.18 mm added 1 dB, and a copper-clad aluminium voice coil added 3 dB, for 6 dB in total from magnetic circuit and moving-system optimisation alone.
Testing produced a sensitivity of 110 ± 3 dB average, measured at 10 cm at 0.1 W input. The 0.1 W basis is the correct one for a sensitivity figure and is stated explicitly here to avoid comparison against power-driven measurements. The unit is rated 5 W with a 6 W maximum. Environmental qualification followed the product definition: temperature shock between −20 and +70 °C, static damp heat for 48 h, 0.75 m drop × 10, a 48 h load test, plus lead-pull checks at 3.0 N horizontal and 2.0 N vertical, with the acceptance baseline that all testing returns the response within a stated deviation of the baseline. The magnet selected is N45H neodymium, whose material rating extends to 120 °C, while the product's specified operating range remains −20 to +70 °C.
The result is in production with 15,000 units shipped. Alongside the acoustic work the programme reduced the rating from 10 W to 5 W and the unit price from 10 RMB to 6 RMB, so the customer saved 60,000 RMB across the shipped quantity while the measured level rose. The transferable conclusion is narrow: for announcement at distance, specify the frequency band the product uses, and measure the supplier's output at test points inside that band.
Ten full-range drivers failed one requirement, because the constraint was the frequency band and not the model. Specifying the band delivered 6 dB without changing the 40 mm exterior.
This is the standard approach Hongsheng applies when a level requirement is driven by content rather than by a target figure: specify the band the product uses, and verify the result at test points inside that band.
6. How to Derive the Target for Your Product
Short answer: Fix the listening position and the noise floor first, derive the level from the distance relationship, then state the band and the level separately. Confirming the measurement basis with the supplier closes the specification.
4. Write down the listener position relative to the acoustic opening, not the product centre, and state the distance the requirement is written for.
5. Record the noise floor at that position, measured rather than estimated. Where the product is quiet, this may be the dominant term in the requirement.
6. Derive the target level from the distance relationship rather than from a catalogue figure, and note which distance the supplier's published measurement used.
7. State the band and the level as two separate requirements. For voice, a level in the critical speech band at moderate overall level; for music, the overall level and low-frequency expectation.
8. Ask the supplier for the measurement basis behind the published figure — frequency, distance, test voltage and cavity condition — and confirm it matches your own test conditions.
9. If the product carries a maximum level rather than a minimum, confirm whether the compliance measurement responds to the highest point of the response, and treat response shape as a design parameter rather than only level.
7. FAQ: Setting Micro Speaker SPL Requirements
Q1: What SPL should a voice announcement product specify?
Specify the level in the critical speech band at the listener position rather than an overall figure. A common working range for close voice announcement is the low-to-mid 90 dB at 10 cm in published datasheet terms, but the requirement has to be written for the actual listening distance and noise floor, because those two numbers usually dominate.
Q2: Is a higher rated power always louder?
No. In one published catalogue a 10.0 W configuration publishes 95 dB while several 1.0 W configurations publish 95 to 96 dB, and the 2.0 W group spans 93 to 108 dB. Rated power describes how much the part can take, not how much it produces at normal listening levels.
Q3: How does distance change the required level?
Level falls by 6 dB for each doubling of distance. A target written at 1 m therefore carries about 10 dB more level when used at 30 cm, and about 10 dB less at 3 m. Recomputing the requirement from the intended position is usually more accurate than adjusting a bench figure.
Q4: Does more output improve intelligibility?
Once the speech band is above the noise floor, raising overall level does not reliably improve comprehension, because the added level sits in bands that do not carry speech information. Where intelligibility has to be measured, ITU-T P.807 is the appropriate recommendation.
Q5: We need to meet a maximum sound level limit. Should we lower the power?
Lowering power reduces level across the whole spectrum and does not change the shape of the response. If the limit is measured on the highest point of the response, a configuration with wider bandwidth delivers more perceived loudness at the same measured maximum than a narrow-response part simply derated.
Q6: Why do two datasheets give different SPL for the same model?
Almost always because the measurement basis differs: frequency, distance, test voltage or cavity volume. Published figures in the range typically state 2 kHz at 10 cm at a rated test voltage, often inside a defined cavity. Values measured at a different distance, free-air or at a different voltage are not comparable.
Q7: Should the test point be the conventional 2 kHz?
Only if 2 kHz represents the band the product uses. For voice products, test points inside the speech band give a more meaningful figure, and selecting them is a way of stating the design intent, since the supplier then publishes values measured where the product operates.
8. Summary: Derive the Requirement, Do Not Adopt the Figure
A defensible sound pressure requirement has three inputs and one discipline. Listener distance, noise floor and content type determine the target; the measurement basis determines whether a catalogue figure can be used to check it. Selecting on the loudest published number instead tends to produce parts that cannot sustain their output at normal levels, and produces requirements that are wrong in opposite directions at different distances.
The published range in this article makes the point concretely: within a single 2.0 W rated class, published sensitivities span 93 to 108 dB, while the highest rated power in the catalogue does not produce the highest level at all. Sensitivity and rated power are independent properties, and the product requirement belongs to the first of them. Hongsheng's catalogue publishes configurations across voice announcement, portable listening and industrial housings with the measurement basis stated for each, so a required level can be checked against a published configuration under comparable conditions rather than estimated.
Next step If you are setting a sound level requirement for a device, the shortest route to a configuration worth testing is to state five things: the space available for the driver and its cavity, the impedance the amplifier will drive, the power the rail can supply, the level required at the intended listening position, and the frequency band the content actually occupies. With those specified, our engineering team can recommend a suitable configuration for evaluation, or state plainly which part of the acoustic design has to change first.
More in This Series
· How to Compare Micro Speaker SPL, Power and Impedance Correctly →https://www.hsdz-spk.com/news/581.html
· Micro Speaker Enclosure Volume: How Much Cavity You Actually Need → https://www.hsdz-spk.com/news/583.html