Sonar Observations

A) Echo Amplitude

The amplitude of the echo significantly drops quadratically as the distance to an object increases.  If the gain of the receiver amplifiers was fixed at a gain high enough to pick up the weak echoes from small objects at 60” away, echoes from a large object at 20” would saturate both the input amplifiers and the analog to digital converter ( ADC ).  Even if the gain was reduced so the input amplifiers didn’t saturate, a 16 bit ADC doesn’t have enough dynamic range to pick up echoes from small objects 60” away without saturating with echoes from large objects at close distances.  In my biosonar system I use information such as peak amplitude and pulse width to determine if two echoes are from a single object or multiple objects.  When the amplifiers or the ADC saturate much of this information is lost.

To solve this problem in Sonar 3, I changed the gain of the receivers on the fly while collecting echo data.  This helped but also had problems.  The two input stages are high gain, high Q, notch filters and it doesn’t take much to make these filters ring at 40KHz.  Making a large change in the gain causes the amplifiers to ring for about 1mS, which looks exactly like an echo pulse.  Small gain changes cause small ring signals.  I started with four gain changes between 10” and 60”, eventually increased that to seven gain changes and more recently increased that to sixteen gain changes.  I can still see small ringing each time the gain changes, and some echoes are so strong that they still saturate the second stage amplifier, but this was a big improvement.  Now when the transmitter is pulsed, the receiver gain is set to 516 and increases in sixteen steps until the gain is 2,460 at 9mS or about 60”.

I am redesigning the receiver circuits on my Sonar 4 system, reducing the Q of the receivers and adding many more gain steps.  I added a high speed eight-bit digital to analog converter (DAC) between the two receiver stages and another eight-bit DAC between the detectors and low frequency amplifier.  This will give me a total of 512 gain steps and with the 16-bit ADC the dynamic range will now be 4,294,967,296.  This might seem like a lot, but based on all the problems I have had with saturation, I suspect it will adequate.

B) Two Echoes from One or Two Objects?

How can we be sure two echoes are from a single object instead of two separate objects?

If there is an object in front of the sonar at a location of X = 8” and Y = 35” the left and right echo distances will be: left = 36.4” and right = 34.1”.

Figure #1

But couldn’t these two echoes be from objects anywhere in front of the sonar?

Figure #2

The ultrasonic transducer I am using for the receivers has a beam width of 30° or +/-15°.  Echoes from objects outside this beam are very weak and difficult to pick up.  With my sonar system with the two receivers separated by 12” and the transmitter in the center, I am able to pick up echoes from most and object within +/-20° to the left or right of either receiver.  Therefore an object must be within +/-20° of the receivers’ centers to be detected

But the only area where two echoes could come from a single object is the area where the two receiver’s beams overlaps.  This area is centered on the center line and in my sonar system is about +/- 20°.  Echoes from objects outside this area will be detected by only one receiver.  See Figure #3.

Figure #3

However, within this +/-20° area how do we know the two echoes are from a single object and not two separate objects?  To start with echoes from any object within this area will be detected by both receivers.  Therefore if there is a single object within this area each receiver will detect an echo from that object.  If there are two objects within this area each receiver will pick up at least two echoes.  Then it becomes a matter of determining which echo is from which object.

One easy way of weeding out some wrong combinations is by checking the Object’s X value.  Some combinations of echoes will result in an object with an X value that is outside the +/-20°area which is not possible.  Therefore one of the first checks I run on an Object is to make sure its X value is within +/-20° of the center line.

Echo amplitude and pulse rise time can also help determine which two echoes came from a single object.  In general the amplitude of an echo is greater from an object directly in front of the receiver than one off to either side of the receiver.  But not always.  A better indicator is rising slope of the echo pulse.  The rising edge of an echo pulse from an object directly in front of the receiver is all most always much sharper than the rising edge of an echo from an object on either side of the receiver.  As X distance from the center line increases the slope of the rising edge rapidly decreases.  In almost all of the tests I have run the rising slope of the shortest distance echo from an object is much sharper than the rising slope of the longer distance echo.  Therefore if two echoes appear to be coming from a single object and rising slope of the shortest distance echo is not as steep as the slope from the longest distance echo it’s a strong indicator that these two echoes are not from the same object.

In the algorithm I am using to identify objects from echoes I first combine all echoes to create every possible object from all of the echoes received from one transmit pulse.  Before combining the echoes I perform no checks.  Once all possible objects are identified the algorithm then runs a number of checks on each object and discards objects that fail these checks.  I got much better results running the checks after the objects were created than I did when I ran the checks before the objects were completed.  The checks performed on each object are:

  1. Check if X distance is within the area where the two beams overlap.
  2. A rough amplitude check. The shorter distance echo’s amplitude should be greater than the longer distance echo’s amplitude.  Some exceptions.
  3. Verify the shorter distance echo’s rising slope is steeper than the longer distance echo’s rising slope. Some exceptions, but not many.
  4. A combination of amplitude and rising slope.
  5. Check for any echo that is associated two objects. One echo cannot come from two objects.  Keep the object with the echoes that are a closer match to the rising slope and amplitude checks.

C) Resolving Close Objects

The biggest problem with trying to identify two close objects is ringing: transducer ringing and circuit ringing.

The receiver circuit ringing is caused by the high Q of the two band pass filters in the receiver circuit.  When a single cycle is applied to the input of the first stage, the output of the second stage rings for 31 cycles, which results in a pulse width of about 700uS on the output of the detector.  I redesigned these filters in Sonar 4 and reduced the Q of both stages.  Now the output of the second stage rings for only three cycles.

However, the receiver transducer ringing is even worse.  When this transducer receives a single cycle pulse it rings for about 56 cycles.  I have tried a lot of things to reduce this ringing, such as changing the load on the output of the transducer and clamping the output (shorting the output with an analog switch) of the transducer, but nothing seemed to have any effect.  When I clamped the output for five cycles the ringing was gone for the five cycles, but when the clamp was released the ringing came back as if the clamp had never occurred.

I am currently experimenting with two single cycle pulses within the transmit pulse.  The second pulse is delayed from the first pulse by 5.5 cycles.  The echoes from the second pulse hits the receiver five and a half cycles after the first pulse, which is 180° out of phase from the first pulse.  The purpose of the second pulse is to stop the receiver transducer ringing.  So far this looks very promising.  It reduces the amplitude of the ringing by about 50% and makes it possible to see an echo from a second object less than two inches behind the first object.  With a single transmit pulse two objects six inches apart, one behind the other, appeared to be a single object.

The pictures below are echo signals from a small object 20” away.  In all of these pictures the echo signal is the output of the detector.

In Picture #1 the large pulse on the top trace to the left is the receiver picking up the transmitted pulse.  The pulse to the far right is the echo from the small object 20” away.  The lower trace is the transmitter signal.  In picture #1 the transmitter pulse is a single one-cycle pulse.  Even though the transmitted signal is a single cycle pulse, 25uS wide, the output of the receiver transducer is about 60 cycles and the output of the detector shown in the picture is about 1.5mS wide.

Picture #2 is exactly the same conditions as Picture #1 except the transmit signal is now two single-cycle pulses.  The second pulse is delayed 5.5 cycles, 137.5uS, from the start of the first pulse.  As you can see the second pulse canceled out much of the ringing of both the picked-up transmitted pulse and the echo pulse from the object 20” away.  The echo pulse, about 3.4mS from the first transmit pulse, is now only about 300uS wide instead of 1.5mS in Picture #1.  There is however a bump about 500uS after the main echo pulse which is what is left of the ringing.  This pulse location and amplitude is very consistent which makes It easy to eliminate in firmware.

Picture #3 is the same conditions as Pictures #2 with another small object placed about two inches behind the first object.  There are now two echo pulses: the first echo pulse at 3.4mS is the echo from the first object and the second echo pulse at 3.7 mS is the echo from the second object which is two inches behind the first object.  If you compare this picture to the first picture you can see how the pulse from the second object would be completely buried in the ringing when a single transmitted pulse is used.

I have run a lot of tests with different objects at different distances and using two transmitted pulses seems to work very well at eliminating most of the ringing and makes it much easier to resolve close objects.  The only downside of this method is amplitude of the echo pulses is about 40% lower than echoes from a single transmitted pulse.

Picture #1

Picture #2

Picture #3