Recording Underwater Ambiences
October 15, 2007

	Like a blanket of glass, the surface of calm water acts as a reflector, inverting the phase of sounds that hit it from above and beneath.

For a transition between two music pieces I needed to record the sound of ocean waves crashing continually over the listener so that they eventually become fully submerged in the surf. Recording the waves from above was simple enough, but from below—this required a special microphone, or at least a special way to keep my condensers from getting ruined. Information on the creative aspects of underwater recording is sparse and there are a lot of contrary opinions about which method works best. I did some tests to find out for myself which techniques offer optimal sound quality and then utilized my findings to create the required sound effect. For anyone that is interested in designing underwater ambiences, this article will provide some interesting insight on underwater recording and related sound design techniques.

Sound Propagation

The speed of sound is completely determined by the physical properties of the material it is traveling through. Studying the behavior of water ripples is a good way to understand this. Imagine dipping the end of a stick into a pool of water. This causes a circle of ripples to expand and dissipate—think of this as sound radiating from a stationary point in space. If the stick is moved slowly in one direction the shape of the expanding ripples is no longer a circle but an oval. Now imagine that this is sound emanating from a moving source. The ripples move at their own speed which is determined by the viscosity of the water, not the speed of the stick. As the stick moves faster it will start to catch up with the ripples in front and the ripples behind spread further apart. This is why a police siren rises in pitch when it approaches and drops when it has passed, an effect called Doppler shift.

The only way a sound wave can move faster is by moving through more dense materials. Sound travels at higher speeds in materials such as liquid and solids. Compare the speed of sound in air: 1128 ft/sec (344 m/sec), seawater: almost 5000 ft/sec (1500 m/sec) and steel: over 20,000 ft/sec (6000 m/sec). A sound wave can quickly change speeds by jumping from one material to another and in the process it can even bend, just like light will refract as it passes through a magnifying glass. In fact, it is possible to a create a sound magnifying lens. In order to achieve this, fill a balloon with carbon dioxide (CO2) and listen to sounds traveling through it. Sound travels more slowly through CO2 than it does through air. The shape of the sphere in combination with the density of the CO2 bends sound into a concentrated point on the opposite side. The principles of this are exactly the same as the way a glass sphere in the sun produces a magnified point of light. This is because light travels more slowly through the glass, which is more dense than air. When light or sound waves move from one material to another, unless the collision is at a right angle, the waves bend. The curvature of the glass ball and CO2 balloon cause the waves to bend toward a singular point.

The reason that thunder booms and rolls is because of the speed that sound travels at through air. When lightning strikes, a channel of superheated air forms that can stretch out for several miles in many directions. This causes audible shock waves that emanate from segments so many miles apart that portions of the sound are greatly delayed. Thunder will sometimes start with a slow ripping or peeling sound. This is caused by differences in distance that the listener is to each part of the bolt, the sound of the closest portion reaching the listener first, followed by the furthest part, typically toward the top. Sounds traveling long distances become more filtered on the high end, the lowest frequencies traveling the furthest and sustaining the longest. This filtering is caused by such things as wind, turbulence, temperature and humidity. Because temperatures close to the ground are generally warmer, sound waves traveling horizontal to the ground can refract upward. Sound also travels faster downwind, so turbulence scatters sound and heavy crosswinds can laterally sweep it away.

The speed of sound in water is mostly determined by molecular viscosity and is further effected by pressure (depth), temperature, salinity and even water movement. Sound travels almost five times faster in water than air and can reach much greater distances. Acoustic energy does not travel far in shallow water because much of it is absorbed by the bottom but in deep ocean water sound can travel for thousands of miles.

The speed of sound underwater is faster at higher pressures and slower at colder temperatures. In the ocean, at a depth somewhere between .4 to .6 miles, the pressure is not too great and the temperature is not too low which creates an efficient layer for acoustic propagation. Sound originating from this depth refracts towards the horizontal axis never reaching the surface or the sea floor and is trapped in what is called the deep sound channel. This channel is useful for monitoring the voice patterns of whales from distances of over 1000 miles. Deep water whales actually use it for echolocation. By making sounds at these depths and listening for echos they are able to sense the shape and position of geographical structures such as mountains hundreds of miles away.

The difference in the way a human perceives a sound in water vs. air has largely to do with the increased sound speed. If sound moves 4.32 times faster, then perceived distances seem to be that many times shorter. To a swimmer, sounds in a 20 ft x 20 ft x 8 ft (6m x 6m x 2.5m) deep pool sound like they are occurring in a 4.6 ft x 4.6 ft x 1.85 ft (1.4m x 1.4m x .6m) box. The water surface traps sound beneath so acoustically it is like a lid. We are accustomed to hearing diffused sounds that scatter around in the air and bounce off of objects before reaching our ears. In water, because of the increased speed of sound, there is less time delay between direct sound and reflected sound. Therefore sounds heard at medium distance tend to sound extremely close. Also, since much of our directional hearing sense is achieved through time delays between our two ears, stereo imaging is greatly reduced.

Many other factors influence the way things sound underwater but most of these acoustical phenomena are of more concern to scientists than sound recordists. For example, one could come up with an exhaustive diagnosis of the way an unsettled water surface scatters sound and modulates its amplitude and frequencies. But these affects are generally inaudible and the recordist is after something much more apparent, such the way a passing boat roars or a creaking hull resonates. If it is desired to characterize these sounds as they occur in unsettled water, then better results can be achieved through sound design techniques.

Some of the tools I used for this project include the Aquarian Audiio H2a-XLR, Zoom H4, Rode NTG2 and Sound Devices MP-1.

There are a number of ways to record sounds in water. Some sound designers have chosen to sidestep the process and make above water sounds seem like they are underwater. The argument is that true underwater recordings sound too strange. Films have taught us that underwater sounds muffled, echoing and bubbly. In actuality water is alive with high frequencies but a bright sounding recording tends to come off as less realistic. Therefore, the proven approach is making the material sound less like real life and more like what we have grown accustomed to in movies. Most people recommend covering microphones in condoms and dipping the capsule just below the water surface. I was eager to compare the various ways underwater recording is approached and choose the best one for my project.

I purchased an Aquarian Audio H2a-XLR, which is a newly designed low-cost hydrophone advertised as being highly sensitive with low noise and a wide frequency response. It has an optional internal phantom powered preamp, balanced 3 pin XLR cable up to 15 meters long, and a durable steel housing that can withstand pressures of up to 80 meters. Before making the purchase, I compared the H2a with a similarly priced hydrophone made by DolphinEAR. The standard DolphinEAR hydrophone is packaged with a special headphone amplifier and walkman style headphones. The "pro" model features a balanced 3 pin XLR cable and standard 10 meter cable up to 100 meters long. The Aquarian Audio website did not offer any specifications for me to compare the H2a with the DolpinEAR/Pro but the lower price and phantom power made the decision easy.

To power the H2a, I purchased a Sound Devices MP-1 portable preamp. Disappointed by my Zoom H4's noisy preamps and buzzing phantom power feature, I was eager to eliminate this problem once and for all by letting a professional preamp handle the power and gain. Besides having a very low noise floor, in comparison to the the H4's internal preamps, the MP-1 sounds warm and detailed, lending much more personality to my microphones. I am also impressed with the physical construction of it. Sound Devices gear is built to withstand harsh field recording conditions and it really feels like it is in a different league than my fragile home studio gear. The MP-1 seems right at home in bad weather and it can survive heavy impacts without a scratch.

I added an Rode NTG-2 shotgun microphone to my kit for recording above water splashes and I also thought to try out a Shure SM-58 dynamic mic and Rode NT1 condenser for dunking in the water with condoms. The H4 handled recording, headphone amplification and even some above water stereo miking with its built in electret condensers.

Recording Methods

"I was responsible for a lot of the underwater propeller sounds, the caterpillar motor, some of the torpedo sounds, the underwater ambiences. It was one of the biggest jobs I've ever done. Water is very difficult to record. We invented an underwater microphone which was a film can, filled with oil with a Crown (PZM) pressure zone microphone. We also had one in an oil can which was a little more dense and one in an air can. Sound Designer John Fasal and I would take underwater air tools and go to supervising sound editor Cecelia Hall's house to record sounds in her pool. I'd swim around for the torpedo, and then I'd jump into the pool and record the water from the outside and then underwater on two separate tracks. So you've got all the airy and splashing sounds outside the pool and all the underwater sounds recorded with the oil can mikes." —Frank Serafine on sound design for The Hunt for Red October

The condom fit over the SM-58 easily and I used a zip-tie to secure it to the cable and keep water from seeping in near the XLR connector. I recorded some flowing water by dipping it into waterfalls. The SM-58 was surprisingly resilient to rumbling noise as water hammered against it. The sound is rather muffled but it seemed to do a good job of recording a direct hit of cascading water. This vocal microphone is not very sensitive and underwater with a condom on the sensitivity is much further reduced. Dipping it in a few feet away from a small waterfall produced a very muffled and quiet signal. Some of the best underwater gushes were captured by submerging it half way into the water at an angle which produced a nice blend of the open air sound and showering bubbles.

I attempted to capture some big underwater splashes with the SM-58 by hurling huge rocks into a large natural pool. When I tossed the first boulder in, something very unexpected happened: while the above water splash sounded very powerful to my ears, with a deep "boosh" and shower of droplets, the underwater version sounded like a sharp slap followed by many smaller slaps and an ineffectual clunk as the rock hit the bottom. I tried tossing variously sized rocks into the water at different distances from the mic but they all sounded very harsh and completely unlike the deep underwater explosion I wanted. I was curious what was causing this strange effect and if it was simply the insensitivity of the condom-wrapped SM-58.

Next I connected the H2a hydrophone to my recorder and dipped it into the same pool (the MP-1 preamp had not arrived in the mail yet but as these were just tests I was not too concerned with the quality). This was my first time using a hydrophone and I really did not know what to expect. Outside of the water, the H2a is very quiet but it does pick up some ambient sound through the air. Tapping lightly on the housing with my finger produced a very powerful low frequency thud, an indication of wide dynamic range. When I dipped the hydrophone into the pool I could clearly hear trickling water from ten feet away. Comparatively to the SM-58, the hydrophone revealed a lot of high frequency detail and I could even hear a couple of tiny air bubbles slide off the hydrophone and pop at the surface.

Dipping the hydrophone into rushing water created a horribly distorted sound unless the water had only minor amounts of air bubbles. It is too sensitive to be submerged in a white torrent but produced excellent recordings when placed close-by the activity. The H2a is silent when calm water is rushing over it—this is something I wondered about since many people "drag" hydrophones from the backs of sailboats. The only potential problem is that the point where the cable meets the water can transfer vibration to the hydrophone. This is more of a problem if there is a wake behind the boat and the solution is to keep the cable from rubbing up against anything and also submerging it more deeply. The H2a is so sensitive that in a quiet setting one can actually hear the water surface sliding up and down the cable. Submerging the hydrophone more deeply always seems to cure it.

The hydrophone captured details of rushing water that the condom-wrapped condenser could not. However, it would distort if placed directly within a plume of bubbles.

I was impressed with what I was hearing through my headphones and decided to give the rocks another go. The H2a revealed much more detailed underwater splashes and I could even hear a rush of bubbles as the rock descended to the bottom. Unfortunately, it captured the same harsh slapping noise as the rock hit the water and a whole series of ugly pops as the rain of droplets followed. The H2a recording also seemed to resonate in the upper-midrange, something that I could not fix regardless of where I placed the hydrophone in relation to the splashes.

Next off I put a condom on a Rode NT1 large diaphragm condenser and ran it through the same tests as the SM-58 and H2a. The NT1 has a heart-shaped cardioid polar pattern and is very sensitive, picking up a wide range of frequencies, especially in the low end. This mic proved to be much less muffled than the SM-58 and perhaps even more realistic sounding than the hydrophone just because it captures the deep warm sound that we would expect to hear underwater. Overall, the NT1 sounds smoother than the hydrophone. However, it does not pick up even a fraction of the detail that the hydrophone does. The SM-58 rolled off at about 200 Hz, the NT1 at 500 Hz and the H2a sloped off above 3000Hz. Comparing the three, it was evident to me why most sound recordists use microphones in condoms—because it produces muffled and therefore "realistic" underwater recordings. The hydrophone seemed more promising, but there was something strange about the frequency response that I could not put a finger on. In addition, all three methods produced very unpleasing splash sounds with extremely loud slaps.

The sun was starting to set and I was running out of time so I quickly recorded all the same sounds above water with a Rode NTG2 shotgun microphone. Later, I would see how well I could make these recordings sound underwater. After doing so much underwater recording, it was pretty dramatic to experience above water sounds with the NTG2. The crickets were getting pretty loud, birds were singing, I could hear waterfalls from 50 feet away and the wind was blowing leaves around. This lively setting posed its own recording difficulties—all these background noises were not an issue when recording underwater.

Note to Zoom H4 users:
I recently performed a modification to my Zoom H4, installing a 1500uf capacitor, in an attempt to eliminate the whining noise created by running the phantom power feature on batteries. This brought the noise floor down 12 dB but the whining sound is still very present, especially when recording quiet sounds. If this oscillation stayed at one frequency one could just notch it out with a filter but it goes up and down in pitch over time. As a matter of fact, upon closer examination I found that its not just one oscillation, but a whole cluster of them bunched together. If you are recording mono, there is a way to cancel this sound out. The buzzing noise is identical in both channels, as long as the L-M-H gain switches are set the same (this switch sets the preamp gain level in steps of 6dB.) Even if you are only using one XLR input, the H4 records the unused input, which will sound like a whistling noise if the H4 is running on batteries and you have the phantom power feature on. To eliminate this sound from the used channel, just invert the waveform on the unused channel and then mix the two sides together (most audio editing software can handle this task). This procedure cancels out the anomaly completely. Bose Quiet Comfort headphones work in this way, by sampling ambient room noise, inverting the waveform and combining it with the music.

Hydrophone Anomalies

Close examination of the hydrophone recordings in my studio revealed some interesting things. In quiet moments I could actually hear crickets, which is pretty remarkable considering that nearly all the above-water sounds can not penetrate the water's surface. Underwater reverberation was very quiet, a bit more noticeable when recording something from a distance. The shallow water combined with a bed of leaves at the bottom were keeping reflected sounds to a minimum. I also heard some faint irregular static, evidence that the H2a is noisy. I was able to reproduce this in the studio with different preamps so I sent a copy of it to Robb Nichols, the owner of Aquarian Audio, asking if this was normal. As it turns out, my H2a was defective and it is the only one reported to have this problem. He said I could immediately send it back for a free repair but as the noise was faint, I decided to hang on to the defective H2a until I had completed the recordings for this project. (After the H2a was repaired the static was greatly reduced but still slightly present.)

Curious about the resonant peak I heard while recording, I did some tests to determine the H2a's frequency response. I went back to the pool and brought the H2a and a shotgun microphone. With the hydrophone on one channel and the shotgun on another, I dragged a stick through the water and recorded this on both channels. Then I pulled out my laptop and, using Waves Paragraphic Eq, adjusted the equalization on the hydrophone recording to more closely match the shotgun recording. I ended up with a big dip at 3000Hz which was compensating for an unflat frequency response on the hydrophone. I reproduced this in my studio by hanging the H2a about 3 feet from an Event 20/20 bas reference monitor and blasting it with a loud white noise signal. This allowed me to make a frequency response plot (pictured). The plot is not clinically accurate because the reference monitor is only as flat as ±3dB between 38Hz and 20kHz. Nonetheless, I believe that this was a valid method for evaluating the H2a's flatness.

My tests revealed a peak at about 3000Hz...Nothing that a little EQ can't flatten out.

I sent an email to Robb Nichols about this and he said that the peak is an adverse characteristic of most hydrophones. He went into detail this: "The reason why many hydrophones have a pronounced peak frequency is that the mass that a sensor assembly must have to work in a high-pressure, corrosive, conductive environment has resonance. This resonance can be easily designed to be well above the human auditory range. But in order to maximize sensitivity, a hydrophone designer often wants to put the piezo crystal (that almost all hydrophones use) in a bending mode. If you compress a cylinder, or in our case, laminate the sensor to a plate, you can design the mechanical forces to be amplified--stretching the crystal. If you can make the hydrophone more sensitive, you obviously don't need as much electrical gain. And without getting into it, you can design a lower-capacitance sensor with the same given sensitivity. This usually means smaller, wider polar response, and a lower-noise preamp. All great! But as soon as we put some mass in motion by bending it, it lowers the resonance—usually into our auditory range. This plays hell with the frequency response above the first resonance. To make short of it, designing a hydrophone, like many things, is an exercise in compromise. In the design of our hydrophones, I also factored in cost--which immediately also brings up manufacturing issues as well. I have chosen with this design to use a stock piezo bender in an alignment that has a first resonance at about 3KHz, and apply a few tricks to dampen the resonance."

He concluded that a linear frequency response was sacrificed in favor of increased durability, better noise performance and sensitivity to low frequency sounds.

After flattening out the response of all my H2a recordings with an equalizer, they sounded much more natural, without any unfavorable side-effects from the signal processing. In my opinion, this procedure is necessary in order to maximize the usefulness of hydrophones as a sound design tool. Without applying the correction curve to your recordings, you may decide that hydrophones sound too weird and instead go the route of condom dunking which does not offer the same quality and flexibility.

Getting the Right Sound

	We have been brainwashed by the entertainment industry into thinking that underwater sounds beautiful and reverberant. In actuality, the beauty is most often achieved through signal processing.

A very common technique in film sound for creating underwater ambiences is to drop the pitch of water recordings by up to 2 octaves and add some reverb. This creates a dreamy and muffled effect that fits the way we expect things to sound underwater. Since I recorded everything at a sample rate of 96KHz, I had only to change the sample rate rate lower. I tried this on recordings from all microphones and it seemed to work well on most—the corrected hydrophone waveforms sounded particularly nice as they contain so much high frequency information. The SM-58 recordings could not handle too much downward pitch shifting because they are very muffled to begin with.

An issue yet to be solved was the undesirable slapping noise that occurred when objects hit the water's surface, which included water droplets. I have determined that the direct sounds travel so fast to the hydrophone that usual recording distances sound too close, and one would never want to record explosive sounds so closely. This can be demonstrated by tapping two pieces of metal together a foot away from a shotgun microphone and then four or five feet away. When the two sounds are normalized, the close recording appears to have a short loud spike in the front and the distant recording is smooth, without a spike. If you remove the spike from the close recording and normalize it again, it has a better balance of direct and indirect sound like the more distant recording. Therefore, I figured that by simply clipping these loud transient sounds out of the splashes the problem would be fixed and I was correct. The splashes sounded great after this and since splashing water is so random sounding to begin with, cutting out little pieces here and there does not make the waveform sound tinkered with. After I did this, the amplitude could be brought up much higher without clipping, which made some of the recordings more noisy. I used noise reduction software to fix this but my recommendation is to set the levels pretty high when recording closely with a hydrophone, ignoring the surface slaps, then take the clipped transients out with audio editing software. Alternatively, one can record water sounds at a greater distance and turn up the gain.

I experimented with the shotgun recordings by pushing back the high end with a low-pass filter and dropping the sample rate. I was able to get some decent sounding effects even though I could not easily replicate the underwater bubbles that the hydrophone had picked up. The shotgun waveforms have an entirely different sound, even when processed to sound like underwater. They are very pleasing on the ears to begin with, with lots of high end detail and when dropped in pitch they retain a lot of texture. The processed above water recordings produced familiar sounding underwater ambiences with minimum effort. There are many things I prefer about the hydrophone recordings though and my experiments showed that a mixture of both produces the most realistic and detailed results.

My final decision was to use the hydrophone to record underneath the ocean waves and combine this with above water recordings. My only concern was that the hydrophone would be too sensitive to handle the torrent but I hoped to find a solution by positioning the H2a properly.

Recording the Waves

I chose to do the recording at Parker River National Wildlife Refuge, an 8 mile barrier island with 4,662 acres of protected wetlands in Newburyport MA. My experience recording nature sounds has taught me to get as far away from civilization as possible, to minimize human noise and magnify the sound of natural habitats. I would have no choice but to cope with the constant sound of recreational boats, an unregulated free-for-all of marine propulsion that makes any coastal recording a challenge. I went to do the recording in early October and the weather was unusually balmy. Unfortunately, the waves were pretty gentle that day and this particular beach turned out to be completely sandy. I was hoping for some big plunging waves and the sound of crushed shells or pebbles being pulled back by the backwash. I would instead have to design these characteristics with Foley effects.

The tide was out, revealing a reef and this turned out to be a good area to record because it broke the waves a little and provided me with areas of calm and lively water. I dipped the hydrophone into the most active surf and was greeted by a load of distortion. I dialed the gain back on the preamp but it still sounded distorted. I moved closer to the shelter of the reef and was able to get some good recordings of waves washing more lightly over the hydrophone. I tried getting the sound of the sand moving around underneath the waves but it was so quiet in relation to the waves that I could not hear it. The above-water waves did not sound like I wanted them to. It was apparent that I would have to come up with ways of making this ambience sound larger than life.

About 10 minutes into recording, the signal coming from the hydrophone turned to fuzz. Switching over to 12V phantom power and cranking up the gain seemed to cure the problem. I later found out that the Sound Devices MP-1 48V phantom power feature failed. It was as dry as a bone in its carrying case so I am not sure what triggered the mishap. Anyway, Full Compass immediately sent me another one, without even requiring me to send back the defective one before the new one arrived. Kudos to Full Compass for making the replacement process so easy. (Later I tested the replacement out and it worked perfectly.)

I made some above-water stereo recordings with the Zoom H4, attempting to capture the full ambience of the waves, including the foam sound. There was a pretty heavy breeze hitting the condenser capsules so I slipped an Audio Technica AT8120 windscreen over it and the problem was cured. The H4 worked well for this recording, and its high noise floor was inaudible recording the surf. I could hear the foaming backwash bright and clear through my headphones. I was wishing that the waves would crash more loudly and the boats were not so noisy but I was already dreaming up ways to solve this with signal processing.

Mixing and Enhancing

Manufacturers of equipment often ask me questions like, 'Don't you think it would be great if we could get this multichannel microphone and take it to Niagara Falls or the Amazon or where icebergs are breaking off in Glacier Bay and make a recording of that space so that you could actually capture it in multi-channel sound—wouldn't that be the ultimate in film sound?' Er, no, it wouldn't be. It would mostly be very boring. Because movies are not about depicting reality, no matter how dramatic that reality is. Movies are about playing with reality and telling lots of little lies that in the end point to a more interesting truth than you could ever get at by depicting actual reality. If feature films were about depicting reality, the average actor wouldn't be better looking than the average person on the street...In the United States we wee these National Geographic documentaries where you hear little prairie dogs' feet skip across the prairie as seen through a telephoto lens, you know, 500 yards away. I want that microphone. But it's a lie. Those are not prairie dog feet; those are someone's fingers in a studio somewhere in Los Angeles. (Thom 124)

There were two main enhancements that I needed to perform: make the waves crash more and also create a more convincing underwater backwash. I went back to the pool and again threw rocks in, this time recording them above water. My plan was to mix enough of these splashes together to make a continuous breaking wave sound and then mix this into the ocean wave recording. I estimated that I would need about 4 or 5 big splashes for each wave and ended up recording about 40 big rocks plunging into the pool. Then I dragged a long tree branch through the water and recorded this with the hydrophone. The plan was that I could use this to create a more pronounced backwash sound. I also wanted the sound of pebbles or sand getting dragged back into the surf. I did this in my studio by recording sunflower seeds spilling slowly out of my hand onto a towel. There were a few other fun sounds I recorded as well, in the off chance that they would be useful enhancements. This included the fizzing interaction of mentos candy and soda—I captured this with the hydrophone.

Finished

Here is a clip of the completed sound by itself and mixed with the soundtrack:

Wave Transition

Wave Transition (mixed)

I'm glad that this project prompted me to investigate underwater acoustics because I now have access to a whole new world of sounds—splashes and bubbles are just the beginning. After thorough experimentation, I firmly believe that the hydrophone is the tool of choice for recording sounds in water. In order to make the most of it as a sound design tool the recorded waveform may need some tweaking but the results seem much better than what can be achieved with a condom-protected condenser.

References:

Nichols, Robb. E-mail to the Author. 8 Oct. 2007.

Serafine, Frank. Interview. LoBrutto, Vincent. Sound-on-Film: Interviews with Creators of Film Sound. New York: Praeger, 1994.

Thom, Randy. Designing a Movie for Sound. Soundscape: The School of Sound Lectures 1998-2001. ed. Larry Sider, Diane Freeman, and Jerry Sider. London: Wallflower Press. 2003.

Sound Movement. Discovery of Sound in the Sea. n. pag. Online. Internet. Available: dosits.org

[back]