dB re 1 V/µPa: Reading Hydrophone Sensitivity Without a Lab

Why a -174 dB spec tells you nothing until you convert it, pair it with self-noise, and confirm it in the field

Hydrophone datasheets are the least readable documents in underwater acoustics. A page that opens with "sensitivity: -174 dB re 1 V/µPa" and "self-noise: 31 dB re µPa²/Hz" has told a buyer almost nothing, because the two numbers live in different reference systems and neither one is a voltage. Buyers end up comparing a -170 dB device with a -180 dB device and concluding that the second is better, when the two are separated by nothing more than where the manufacturer chose to put the reference point.

This note is the arithmetic we use when a datasheet arrives and someone asks which unit is larger: the three conversions, why self-noise decides whether you hear the target, and how to verify a hydrophone without a calibration tank. Our range is listed on the hydrophone category page if you want the numbers alongside this text.

Why a hydrophone sensitivity number looks strange

A microphone datasheet gives you millivolts per pascal. A hydrophone datasheet gives you decibels referenced to one volt per micro-pascal, which is a unit almost nobody uses outside this field. The convention exists for one reason: underwater sound pressures are tiny. A quiet sea state produces pressure fluctuations of a few hundred micro-pascals, and the loudest biological click might reach tens of pascals. Referencing to one micro-pascal keeps the numbers in a range where a human can compare them at a glance.

The cost of that convention is that the decibel value is not proportional to output. It is logarithmic, and it is referenced to a voltage per pressure, not a voltage. Two devices that read -170 dB and -180 dB differ by 10 dB, which is a factor of 3.16 in output voltage, not a factor of ten. That factor-of-three confusion causes more bad purchasing decisions than any other misreading in the category.

There is one more complication. Some datasheets reference to one volt per pascal instead, which shifts the number by exactly 120 dB. A device specified as -54 dB re 1 V/Pa and one specified as -174 dB re 1 V/µPa are the same instrument. Compare a pascal-referenced sheet against a micro-pascal-referenced one without converting and the gap looks like a factor of a trillion.

Decoding dB re 1 V/µPa

The definition is literal. Take the transducer's open-circuit output voltage in volts, divide it by the pressure in micro-pascals, and express that ratio in decibels:

Sensitivity (dB re 1 V/µPa) = 20 log10( V_out / (p / 1 µPa) )

Read backwards, that means: for a sensitivity of -174 dB, one micro-pascal of pressure produces 10^(-174/20) volts, which is 2.0 nanovolts. That is the whole meaning of the spec. Everything else is bookkeeping.

The identity that saves the most time:

dB re 1 V/Pa = dB re 1 V/µPa - 120

Because 1 Pa equals 10⁶ µPa, and 20 log10(10⁶) = 120 dB. So -174 dB re 1 V/µPa is -54 dB re 1 V/Pa, and a hydrophone calibrated by a European lab that reports in pascal-referenced decibels is not on a different scale at all, just a different offset.

Two anchors are worth memorising: -180 dB is 1.00 mV/Pa and -174 dB is 2.00 mV/Pa. Every 6 dB doubles or halves the millivolt figure, which makes any spec sheet quick to sanity-check.

The arithmetic: from decibels to millivolts per pascal

The conversion most engineers actually want is decibels to millivolts per pascal, because that is the unit their preamplifier and ADC care about.

Sensitivity (dB re 1 V/µPa)Equivalent (dB re 1 V/Pa)Output at 1 Pa
-170-503.16 mV
-174-542.00 mV
-180-601.00 mV
-186-660.50 mV
-200-800.10 mV
-215-9517.8 µV

The rule is mechanical: subtract 120 to switch reference, then take 10^(dB/20) for volts per pascal. Write it into a spreadsheet once.

The table also exposes something the raw spec hides. Between -170 dB and -215 dB there is a 45 dB span, a factor of 178 in output voltage. A -215 dB hydrophone is not a "less sensitive" device in any useful sense; it is a device whose specification is written at a different gain setting, which is exactly what our broadband unit does when it lists selectable gains of -20, 0, +20, +30 and +40 dB. When you see a number that far below the family, ask for the gain and reference conditions before comparing. A supplier who cannot answer that question has not specified the part.

Self-noise: spectrum level is not band level

Sensitivity says how much voltage a given pressure produces; it says nothing about whether that voltage clears the electronics' own noise. That second number is self-noise, written in a unit that looks similar yet behaves differently: dB re µPa²/Hz.

The "/Hz" is the whole point. This is a spectrum level, a density, not a level within a band. To turn it into a pressure you take 10^(L/20) and get micro-pascals per root-hertz:

  • 27 dB re µPa²/Hz → 22.4 µPa/√Hz
  • 31 dB re µPa²/Hz → 35.5 µPa/√Hz
  • 37 dB re µPa²/Hz → 70.8 µPa/√Hz

Then, to get the noise actually present in a measurement band, multiply by the square root of the bandwidth, or add 10 log10(BW) in decibels. A 31 dB/Hz device observed across 10 kHz carries 31 + 40 = 71 dB re 1 µPa of total noise; the same device across 100 kHz carries 81 dB. Both are correct for the same hydrophone. This is why a datasheet that quotes self-noise without stating the band and the reference frequency is unusable for system design: the number changes by 10 dB every time the bandwidth grows tenfold.

DH-LN50 low-frequency digital hydrophone

Our low-frequency unit, the DH-LN50, is specified at 27 dB re µPa²/Hz at 1 kHz across 0.1 Hz to 50 kHz. That is the lowest spectrum level in our range, and it is why the model exists: for ocean-noise baselines, 4 dB of extra self-noise is the difference between recording a shipping lane and recording your own electronics.

Four hydrophones read as a set

Here is the useful exercise. Convert all four units into the same two figures, output per pascal and input-referred noise, and the range stops looking like a ladder of "better" and "worse" and starts looking like four different reference points.

ModelBand (-3 dB)Sensitivity @1 kHzOutput at 1 PaSelf-noise @1 kHzInput-referred noiseDepth rating
DH-LN500.1 Hz – 50 kHz-170 dB3.16 mV/Pa27 dB re µPa²/Hz70.8 nV/√Hz6,000 m
DH-LN10010 Hz – 100 kHz-174 dB2.00 mV/Pa31 dB re µPa²/Hz71.0 nV/√Hz6,000 m
DH-LN15020 Hz – 150 kHz-180 dB1.00 mV/Pa37 dB re µPa²/Hz70.8 nV/√Hz500 m
DH-405920 Hz – 200 kHz-215 dB ±317.8 µV/Panot publishednot derivable300 m

The last column pair is the finding. Multiply the spectrum level by the output per pascal and the input-referred noise collapses to roughly 71 nV/√Hz for all three LN units. The three models share an acoustic front end of the same quality; what differs is bandwidth and depth rating, and the sensitivity figures differ only because the reference is the same while the source level of the element is not.

This matters when you choose. If someone calls the -180 dB unit the most sensitive, note that it produces the least voltage per pascal of the three and carries 10 dB more spectrum-level self-noise. It suits dolphin clicks and 150 kHz survey work; it is wrong for a 100 Hz ambient baseline. Sensitivity, alone, ranks nothing.

Matching the hydrophone to the ADC (worked example)

Take the DH-4059, whose -215 dB figure looks alarming. Its maximum sound pressure before clipping is 220 dB re 1 µPa, which is 10^(220/20) micro-pascals, or 10⁵ pascals — 100 kPa.

Multiply by the sensitivity in volts per pascal:

1.78 × 10⁻⁵ V/Pa × 1.0 × 10⁵ Pa = 1.78 V

So at the stated clipping limit, at zero gain, the element drives about 1.78 V into the digitiser. Against a typical 24-bit converter running a couple of volts full scale, that is roughly 70–90 percent of full scale, which is exactly where a manufacturer wants the loudest permitted signal to land. The -215 dB figure is not a weak hydrophone; it is a hydrophone specified at the point where its own clipping limit and its converter's full scale line up. Both numbers are needed to see that.

Two more checks follow from the same logic.

Converter resolution is not the limit. A 24-bit converter has a theoretical quantisation floor near 6.02 × 24 + 1.76 ≈ 146 dB below full scale, comfortably under the 140 dB analog range, so the front end sets the floor. Buying 32-bit conversion, as the DH-100 offers, buys averaging headroom, not a quieter hydrophone.

Derive the noise floor, then verify it. If the clipping limit is 220 dB re 1 µPa and the dynamic range is 140 dB, the implied noise floor sits near 80 dB re 1 µPa. That is arithmetic on published figures, not a measurement, and it should be treated as a target to confirm against a real quiet-water recording rather than a guarantee.

DH-4059 broadband digital hydrophone

A last practical note on gain. Setting +40 dB on a hydrophone whose clipping point maps to 1.78 V pushes the loudest allowed signal past full scale, and the recording clips on the first close-range source. Gain lifts quiet signals above the converter floor; it is not a way to compensate a sensitivity spec you find inconvenient.

What changes once the hydrophone is in the water

A calibration certificate describes a bare element in a controlled field. In service, four things move the effective sensitivity, and none of them appear on the datasheet.

Mounting and baffle. An element pressed against a hull, a mooring frame or a floating body sees a reflected field, so the measured pressure can rise or fall by several decibels with geometry and frequency. A hydrophone calibrated free-field is not automatically calibrated in your frame.

Cable and connector capacitance. A long cable with a high-impedance element forms a low-pass filter with the amplifier input. At 150 kHz, a few metres of poor cable can cost more than the entire tolerance band of the specification. Wet-mateable connectors help with mating, not with capacitance.

Directivity. A small cylinder is close to omnidirectional at low frequency and becomes directional as the wavelength approaches the element size. At 200 kHz the acoustic wavelength in water is about 7.5 mm, so a 105 mm element is many wavelengths long and its response off-axis is nothing like its on-axis response. Ratings quoted on-axis are on-axis.

Flow noise from the mount. A hydrophone in a current generates turbulence noise in the same band as the signal, often the dominant term on a mooring in a tidal channel.

If your application needs the pressure gradient rather than the pressure, a vector hydrophone changes the reference entirely: it reports particle acceleration or pressure difference, its sensitivity is specified against a different unit, and the millivolt-per-pascal table above does not apply. Do not mix the two families in one comparison sheet.

DH-LN100 low-noise integrated digital hydrophone

FAQ

Can I check sensitivity in the field without a calibration lab?

Yes, to a useful but limited accuracy. Three methods are practical:

Charge injection. Feed a known AC voltage through a capacitor of known value into the element's terminals. The injected charge is C × V, and comparing the resulting output against the expected charge sensitivity tells you whether the element and its preamplifier are still in the state they were calibrated in. It verifies the electronics chain and detects cable faults; it does not verify the acoustics.

Comparison against a reference. Put the unit under test and a second hydrophone with a current certificate in the same tank, at the same distance from the same source, and divide the two responses. This is a real acoustic check, and its accuracy is bounded by your reference and by tank reflections, not by your procedure.

Reciprocity. With two or three reversible transducers you can derive sensitivity without a reference, which is the method behind the standards. It needs a tank, a stable source and more time than most field teams have.

The full laboratory method is defined in IEC 60565-1 for free-field calibration, with ANSI/ASA S1.20 as the corresponding American procedure. Neither can be reproduced on a boat. What you can reproduce is a trend: measure the same unit the same way every six months and you will catch a failing cable or a water-filled connector long before it ruins a deployment. We keep a short discussion of the field options in the hydrophone forum thread.

Why does my recording show 40 dB more noise than the datasheet?

Usually because the datasheet quotes a spectrum level and you are reading a band level, and because your gain and the ambient noise both sit above the floor. Convert first: 31 dB/Hz across 10 kHz is 71 dB re 1 µPa. Compare that with typical open-ocean spectrum level near 1 kHz, in the 50–70 dB re µPa²/Hz range. If the sea is louder than the instrument, you are recording the sea, as intended.

Does a lower dB number mean a better hydrophone?

No. Sensitivity, self-noise and bandwidth trade against each other, and a figure written 10 dB lower can simply mean a different gain or a different reference point. Convert everything to output per pascal and input-referred noise first, then compare within the same band and the same depth rating.

What depth can I actually deploy to?

Read the rating as a rating, not a margin: LN50 and LN100 to 6,000 m, LN150 to 500 m, DH-4059 to 300 m with custom options. Exceeding it produces a flooded element, not a graceful failure.

How often should a hydrophone be recalibrated?

A 3 dB shift changes every absolute level in your dataset, which a long monitoring programme cannot absorb. Many operators work on a 12 to 24 month cycle for reference instruments and verify in between with charge injection. If the data goes into a compliance report rather than research, the reporting standard sets the interval, not the manufacturer.

More hydrophone notes, including how we treat bandwidth versus depth in selection, are collected on the hydrophone blog.

dB re 1 V/µPa: Reading Hydrophone Sensitivity Without a Lab
China Sonar November 16, 2026
Archive