Linear array beam forming and phase response troubleshooting

[nauman]

Hi all

I am working on beam forming of high frequency (>400 KHz) linear array for imaging sonar. I have gathered sensor data of 80 channel linear array sub merged halfway in 8 m deep water tank with a transmitting probe 6 m away from it at same depth. I recorded multiple pulses and perform offline analysis on it. The problem i am facing is very high side lobe levels in a beam pattern developed using recorded data. Can anyone kindly guide me what could be possible causes of higher side lobe levels?

Thanks & Regards

 

[BvU]

Something like this (but with source and receiver interchanged) ?
You could be a bit more specific (drawings?) about your setup: any reflections against the tank walls ? How do you analyze the data, define side lobes, etc ?

 

[nauman]

Something like this (but with source and receiver interchanged) ?
You could be a bit more specific (drawings?) about your setup: any reflections against the tank walls ? How do you analyze the data, define side lobes, etc ?

Thanks for reply. A rough sketch of the test setup is attached. I performed base banding for each channel, apply focusing coefficients corresponding to 6 m distance on complex output of each channel (to compensate near field) and then performed delay and sum beam forming in frequency domain on the data chunk containing pulse. Then, I performed IFFT on beam formed output and plot the beam levels (in Log scale) corresponding to time instant of receiving pulse.

 

[tech99]

Low side lobes require a tapered illumination of the array; binomial distribution gives zero sidelobes. So for your active array I presume you need to apply weightings to the individual sensors.

 

[berkeman]

A rough sketch of the test setup is attached.

It's best to convert Word documents to PDF form for attaching to your posts. You can use free PDF writers like PrimoPDF to do the conversion. Many users are reluctant to download a Word document, but usually PDF documents are much safer. FYI

 

[tech99]

Thanks for reply. A rough sketch of the test setup is attached. I performed base banding for each channel, apply focusing coefficients corresponding to 6 m distance on complex output of each channel (to compensate near field) and then performed delay and sum beam forming in frequency domain on the data chunk containing pulse. Then, I performed IFFT on beam formed output and plot the beam levels (in Log scale) corresponding to time instant of receiving pulse.

I notice that the tank will provide a prolific source of spurious reflections, in particular the bottom left corner, which will reflect all incident energy back to the receiver.

 

[Baluncore]

The rectangular tank will be the problem. Are you sure they are side-lobes and not three-corner reflections?
What will you be imaging with the array?
Will it be done in that test tank, or can it benefit from doppler in moving water?
Can you benefit from a swept frequency or chirp?
Going to a 2D array with irregular low-correlation sensors would remove sidelobes and improve resolution.

 

[sophiecentaur]

A rough sketch of the test setup is attached.

Unfortunately my version of MSW doesn't show the image. Perhaps PFD would solve the problem?

If the tank in which you are doing the tests is small, the reflections could perhaps be reduced by a suitable coating on the walls. I have only experience of measuring RF antenna arrays and the solution is often to use a big area of uncluttered ground and a well elevated mast. Compact antenna measurement facilities tend to be very expensive and I guess the same thing could apply to Ultrasound systems.

 

[Baluncore]

Attached is a pdf export of the word document, done from my free LibreOffice.

 

[sophiecentaur]

Thanks to @Baluncore for the document translation.
One simple test of your test setup would be to move the array to different vertical positions and see how the pattern is affected. If that effect is gross then you could perhaps put a rough surface on the bottom of the tank to break up any specular reflections. A similar thing on the surface could also help. Rough surfaces would (I think) reduce gross effects on the pattern (main beam and side lobe levels) but would probably fill in the nulls and minor lobes.

what could be possible causes of higher side lobe levels?

Can you post some images of the actual performance and also the expected performance, given an idealised layout? How much worse is the result than you expected?

 

[Baluncore]

Maybe it will need shaggy curtains of carpet hanging near the walls, to prevent the reflection of ultrasonic energy that causes constructive and destructive interference.
What range of wavelength in water is expected?

 

[nauman]

Can you post some images of the actual performance and also the expected performance, given an idealised layout? How much worse is the result than you expected?

Sorry for delayed response. The measured beam pattern and estimated desired beam pattern are attached.

 

[nauman]

Maybe it will need shaggy curtains of carpet hanging near the walls, to prevent the reflection of ultrasonic energy that causes constructive and destructive interference.
What range of wavelength in water is expected?

Sorry for delayed response. Wavelength will be around 4mm for 420 KHz frequency and 1500 m/s sound speed

 

[sophiecentaur]

Sorry for delayed response. The measured beam pattern and estimated desired beam pattern are attached.

All the main features are present in your practical version but the side lobes are about 3dB higher than calculated. That suggests a significant amount of spurious additions. The main beam seems to have been subjected to an additional set of contributions (a reflection from the floor?) that have put a dip where you might have expected the main peak. I suggest that moving the array up or down by, say λ/4 could significantly alter the main beam pattern. Would that be possible for you to try?

It may be easier to repeat the theoretical calculation, including an 'image' array, reflected in the floor. Your experiment is, unfortunately, very much near field so the calculations are not as simple as for the far field radiation pattern of an antenna array. (Which can be a piece of cake, when you are used to it.)

 

[Tom.G]

From another viewpoint, I suspect you are in realm of experimental uncertainity.

Taking from the beam pattern plots, I get the response peaks at 3.6° and 3.4°. The path length difference between them based on your sketch showing 8m distance is about 1.7mm, or 0.425λ. This could be a vertical position discrepancy of 28mm between the Transmit and Receive transducers. Is your positioning that accurate?

Oh, and a 10°C water temperature difference causes roughly a 2% to 3% change in transmission speed (wavelength).

 

[sophiecentaur]

@Tom.G Some good points there. There are a lot of things the OP can experiment with to home in on the source of the problem. Hopefully it can be sorted but these things always need a few passes through, before getting the best result.

 

[nauman]

Taking from the beam pattern plots, I get the response peaks at 3.6° and 3.4°. The path length difference between them based on your sketch showing 8m distance is about 1.7mm, or 0.425λ. This could be a vertical position discrepancy of 28mm between the Transmit and Receive transducers. Is your positioning that accurate?

My position accuracy is not that accurate. Actually, before performing any beam forming, i monitor individual channel outputs of my array (in time domain) and manually adjusting the position of transmitting probe vertically (as shown in my setup) until i got the maximum levels at individual channels.

 

[nauman]

One more problematic finding i want to share is that this measured beam pattern is not stable over multiple pings i.e. for each ping, there is significant variation in measured beam pattern w.r.t side lobe levels. To analyse this issue, i measured phase/gain response of each array channel w.r.t reference channel (which in my case is center channel) for a typical frequency and there was lot of phase and amplitude variation among multiple pings.

i.e. if for example phase of channel 10 is 80 degree w.r.t. reference channel #40 for one ping, for next ping it will change dramatically to say 300 deg and in next ping it will again change to some random value although experimental Tank environment is very stable, no water movements, receive array and transmitting probe are fixed etc.

Kindly put some light on this issue also.

 

[Baluncore]

If echos from the walls are present then you requires twice the positioning accuracy.
1. How flat is the surface of the pool?

In post #1 you refer to pulses. In post #18 you refer to pings. We need to know the exact characteristics of the pulses as they echo within the known pool dimensions.
2. How many cycles or how long is the ping?
3. What is the rise and fall envelope of the pulse?
4. Is there any frequency chirp or sweep during the pulse?

The huge phase differences you report will be due to timing or positioning errors. They could also be due to phasor addition of reflections if the ping is long.
5. How do you perform and time the parallel A to D conversion of all the sensors?
6. What decides and how is the time zero reference determined?

 

[nauman]

If echos from the walls are present then you requires twice the positioning accuracy.
1. How flat is the surface of the pool?

In post #1 you refer to pulses. In post #18 you refer to pings. We need to know the exact characteristics of the pulses as they echo within the known pool dimensions.
2. How many cycles or how long is the ping?
3. What is the rise and fall envelope of the pulse?
4. Is there any frequency chirp or sweep during the pulse?

The huge phase differences you report will be due to timing or positioning errors. They could also be due to phasor addition of reflections if the ping is long.
5. How do you perform and time the parallel A to D conversion of all the sensors?
6. What decides and how is the time zero reference determined?

Sorry for confusion. As it is direct reception of a pulse here by receive array(no echo reception except unwanted), i think pulse term is more relevant here than ping.

1. During experimentation, pool is kept in calmed state

2. Pulse is 1ms (420 Cycles) long

3. As receive pulse is being analyzed using individual channels measurement, it takes 4 to 5 cycles to reach maximum level and in approximately same no of cycles it decays. However, no shading is being applied on pulse itself before transmission.

4. For beam pattern measurements, only CW Pulse is utilized

5. I am using National Instrument's Simultaneous Data Acquisition System (PXIe Chassis based DAQ) in external trigger mode. This chassis based solution claims that all ADCs starts conversion simultaneously at receiving external trigger. Sample rate of system is being generated using system internal clock.

6. I am generating external trigger pulse (DC) from one channel of a function generator (FG) and CW Pulse from second channel of FG which goes to transmitting Probe. This second channel is being used in external trigger mode. The trigger pulse generated from first channel goes to FG's second channel trigger input as well as DAQ system trigger input so that both transmission and reception starts at same time. Time zero reference is the instant at which transmission and reception starts simultaneously.

Thanks & Regards

 

[Baluncore]

i.e. if for example phase of channel 10 is 80 degree w.r.t. reference channel #40 for one ping, for next ping it will change dramatically to say 300 deg and in next ping it will again change to some random value although experimental Tank environment is very stable, no water movements, receive array and transmitting probe are fixed etc.

There is something very wrong with the timing if you are getting 220° phase jitter.

Maybe you could cut the ultrasonic components from the path and digitise an attenuated signal output of the pulse generator with all converters in parallel. That will test the phase shift of the A-D converters, and the trigger synchronisation.

 

[nauman]

There is something very wrong with the timing if you are getting 220° phase jitter.

Maybe you could cut the ultrasonic components from the path and digitise an attenuated signal output of the pulse generator with all converters in parallel. That will test the phase shift of the A-D converters, and the trigger synchronisation.

Sorry for delayed response. I have tested the phase response of A-D converters by using a Function Generator in external trigger mode and with Burst mode. By applying external trigger to ADCs and F.G at same time, i apply CW burst from F.G to all channels simultaneously with some delay (simulating the distance b/w transmitter and receive array) and measure the phase difference b/w reference channel and all other channels which is approx 0 degree. I even measured the beam pattern with this data which coincides with desired specifications.

 

[sophiecentaur]

Sorry for delayed response. I have tested the phase response of A-D converters by using a Function Generator in external trigger mode and with Burst mode. By applying external trigger to ADCs and F.G at same time, i apply CW burst from F.G to all channels simultaneously with some delay (simulating the distance b/w transmitter and receive array) and measure the phase difference b/w reference channel and all other channels which is approx 0 degree. I even measured the beam pattern with this data which coincides with desired specifications.

When you get really odd results like this, with strange phases, it can be a signal to noise ratio problem. We could be down to precise details about signal levels and / or interference. Do you have images of actual waveforms, for instance? How sure are you about the measuring equipment? Does it give the results you would expect from straightforward arrangements?
It's the old Sherlock Holmes thing about eliminating all the impossibles and the result, however unlikely it may seem, is what you are looking for,