This discussion paper compares the use of CTDs and sound velocimeters to collect sound velocity data.  The paper identifies discusses major differences between the two approaches, including the following:

• Accuracy
• Spiking
• Equation of state errors
• Fragility
• Acclimatization time

Improving the accuracy or reliability of any underwater acoustic system depends on understanding the acoustic environment and minimizing all sources of error in the system.  Knowledge of the ambient sound speed conditions is one of the most critical and universal sources of error in underwater acoustics.  Some example relationships between sound speed and various acoustic systems are:

• Echo sounders require an average sound speed estimate from the surface to the bottom.
• Acoustic systems utilizing hydrophone arrays require an accurate knowledge of sound speed at the array face to calculate transmission and reception angles of the acoustic pulses.
• Acoustic positioning systems and multi-beam sonar systems require accurate profiles of the sound speed from the surface to the bottom to correct for refraction.
• Acoustic doppler current profilers require sound speed to calculate the depth of each current speed measurement.
• Acoustic modems require sound speed profiles to calculate effective ranges.
• Naval sonars require knowledge of the sound speed profile to determine convergence zones, shadow zones, and ducts (both surface duct and SOFAR channel) when hunting for, or trying to hide, submarines.

Errors in measuring sound speed result in proportional errors in both range and direction.  This can be seen in the three equations below.

• The acoustic distance traveled is d=t*c where t is the sound speed and t is the time taken to travel the distance.  Note that for an active system the acoustic distance is twice the distance between the sonar transducer and the target and the equation becomes d=t*c/2.
• The angle between an arriving acoustic signal and a line normal to the transducer array plane is Ɵ=sin-1((t*c)/d)) where c is the sound speed, t is the phase delay between transducer elemnts and d is the transducer element spacing.
• The refraction (bending) of an acoustic ray as it travels through the ocean is described by Snell's law. sin(Ɵ1)/c1 = sin(Ɵ2)/c2.  where Ɵ1 is the propagation angle off vertical at depth 1, Ɵ2 is the propagation angle off vertical at depth 2, c1 is the sound speed at depth 1, and c2 is the sound speed at depth 2.

Since sound speed accuracy is critical to the accuracy of underwater acoustic systems it is important to understand where the sources of error in sound speed can arise.

CTD & SV Comparison

	CTD	SV
Total Accuracy	+/- 0.25 to 0.50 m/s	+/- 0.025 m/s
Number of Sensors	Three	One
Equation Errors	± 0.19 m/s, although equations may differ by as much as ±0.30 m/s	No equation of state is used and hence there is zero equation error
Sensor Errors	Depends on the manufacturer; Total sensor error of ± 0.06 m/s likely	+/- 0.025 m/s
Response Times	Market leading temperature sensors have response times of 100 milliseconds	47 microseconds
Sound Velocity Spiking	Significant, given multiple sensors with varying response times	Zero
Acclimatization	20 minutes	5 seconds (SV Xchange sensor)

Sound speed - also called sound velocity - can be measured directly using a sound velocimeter or it can be calculated from salinity temperature and depth measurements, assuming standard ocean salt ratios.

The direct measuring devices are known as sound velocimeters. These instruments measure the actual sound speed of the water at the location of the instrument.  Sound velocimeters measure sound velocity using a time-of-flight methodology.  A single acoustic pulse is transmitted into the water.  The pulse travels a fixed, calibrated distance to a reflector plate and then returns through the water to the transducer. The fixed distance is achieved using rods which are thermally stable and have zero thermal response time. A high resolution, high stability timing circuit is used to measure the acoustic travel time.  With travel distance and elapsed time known, sound velocity is determined.  Applied Microsystems sound velocity sensors offer accuracies of up to ±0.025 m/s.

Calculation of sound velocity based on CTD results is a second, less accurate approach to sound velocity.  The CTD uses three separate sensors to measure the conductivity, temperature and pressure at the location of the instrument. The conductivity, temperature and pressure are used to compute salinity using an equation of state. Then the salinity, temperature and pressure are used to calculate the sound speed of the water, using another equation of state.  Thus CTD based sound speed data is a combination of three sensor measurements and two equation of state calculations.  This pyramid like approach - calculations based on calculations based on measurements - has a dramatic impact on overall accuracy, as seen in the table to the right.

When considering whether to measure sound velocity with time-of-flight sound velocity sensors or to calculate using conductivity, temperature, depth and salinity, two other factors should be taken into account: 1) the possibility of sound velocity spiking; and 2) acclimizatization times.

With a CTD, total sound speed accuracy is a combination of the sum of the potential individual errors as follows:

• Sensor errors from individual conductivity, temperature and pressure readings
  
• Response time / spiking errors
• Equation of state errors

Individual sensor errors will depend upon stated accuracies, calibration dates, and drift rates for all three CTD sensors: conductivity, temperature and pressure.  Response time, spiking, and equation of state errors are addressed in the following sections.

In theory, CTDs are generally estimated to generate sound velocity readings to a best case accuracy of 0.25 m/s.  In reality, CTD accuracies for sound velocity are often around 0.50 m/s.    In contrast, time-of-flight sound velocity sensor accuracies range from +/- 0.025 m/s to +/- 0.05 m/s.

A further source of error is the mis-match in response times of the three parameters which must be measured when calculating sound velocity: conductivity, temperature, and pressure.  These differences in response times will cause sound velocity spiking as the instrument traverses picnoclines.

	Sound Velocity	CTD
Sensor response times (time constant)	SV: 47 microseconds	C: 25 millisconds T: 100 milliseconds P: 10 milliseconds

Sensor response times vary dramatically when comparing an SVP to a CTD.  The near zero SV response time of the SVP means that every sound speed data sample will be accurate.  The slower response time of the CTD means that there will be a lag before the CTD reports the correct sensor values and therefore the correct calculated sound speed.

In addition , the mismatch in sensor response times of the CTD causes anolmalous spikes in the calculated parameters of salinity and sound speed.  This phenomenon is typically referred to as 'salinity spiking'.  The magnitude of these spikes is dependent on the sensor response time mismathes and the gradient of the change in water conditions.

The sound speed profile data from a CTD is recognizable by two characteristics.

1. A separation of the picnocline depth between the downcast and the upcast.  This is shown in the first data plot below. It is a result of a) the temperature sensor response time and b) the thermal lag in the conductivity cell.
2. Large sound speed spikes which occur at the thermoclines as a result of the sensor response time mismatches of the CTD.  This is shown in the SV Error plot, the second figure below.

The UNESCO Technical Paper Marine Science #44  recommends either Chen & Millero's or Del Grosso's NRL II equations of state for calculating sound speed from CTD data.  These equations are assumed to be accurate in the field to within 0.25m/s.  Funnily enough, the following chart illustrates that the two equations can generate differences of greater than 0.25 m/s between them, yet both are believed to be accurate.  To view the graph in greater detail, click here. 

When a CTD is first placed in the water the conductivity cell requires a 'wetting time' before it achieves full accuracy. The wetting time for a dry conductivity cell is approximately 20 minutes. The wetting time can be reduced by pre-wetting the conductivity cell for two minutes in water containing a small amount of soap and alcohol. The soap and alcohol reduce the water's surface tension and speed up the wetting process.

Analog sound speed sensors also have a wetting time before achieving full accuracy, The wetting time depends on both the sensor and the measurement technique.
- Invar sound speed sensors require about 20 minutes
- Analog composite sound speed sensors require about 2 minutes.
- SV.Xchange sensors require about 5 seconds.
As with the conductivity cell, the wetting time can be reduced dramatically by pre-wetting in water containing soap and alcohol.

Conductivity cells are made of glass and are fragile in comparison to direct measuring sound speed sensors.  Invar direct measuring sound speed sensors are more fragile than composite sound speed sensors.  The best sensor on the market for shock is the SV.Xchange which has been tested to 500 Gs.

Given this fragility, CTDs require the most care in handling.  The glass conductivity sensors and tiny high speed temperature probes are both fragile.  A hand in the wrong spot on a CTD can damage the instrument. Shock from the instrument being dropped or hitting the side of the ship can also damage a CTD.

Invar sound speed sensors are robust in terms of humans handling the instrument.  However these sensors can still be damaged by shock.  The composite sound speed sensors are by far the most robust.  The SV.Xchange sensors have been tested to 500 Gs.  An SV.Xchange sensor dropped on the deck, even landing on the sensor, will not damage the sensor or affect its calibration.