Ultrasound beamforming

Hi,
A typical beamforming method widely used in ultrasound technology is to apply delays between channels and then summing the signals of all the channels. However, when properly calculated delays are applied, it could happen that the phase of the arriving signal is 180 deegres (half the ultrasound wavelength) shifted between channels and summed amplitude is decreased. This results in decreasing SNR and dynamic range. Does anybody here have an idea how to solve this problem? 

On 2/16/2019 18:53, stchebel@gmail.com wrote:
> Hi,
> A typical beamforming method widely used in ultrasound technology is to apply delays between channels and then summing the signals of all the channels. However, when properly calculated delays are applied, it could happen that the phase of the arriving signal is 180 deegres (half the ultrasound wavelength) shifted between channels and summed amplitude is decreased. This results in decreasing SNR and dynamic range. Does anybody here have an idea how to solve this problem?
> 
Are you receiving or transmitting?  The behavior is the same but for the 
  sake of discussion, assume you are receiving, and that the source is 
in the far field.  The angle between the source and all of the sensors 
is (nearly) the same.

Consider an impulse from the source.  In general it will arrive at some 
of the sensors before others.  Pick one arbitrary sensor as your 0-delay 
channel, then the others will have some relative physical delay, 
positive or negative, based on the speed of sound and the distance in 
the direction of the source between that sensor and the "0-delay" 
sensor.  That delay gets added to the delay that is programmed into your 
beamformer.  I'm sorry for the wordy response if the is obvious to you.

If the delay is 180 degrees, that means that there is a null due to the 
sensor in question in the direction of the source.  If the total from 
all sensors when the phase shifts due to the physical and programmed 
delay are taken into account equals 0, your beamformer has a null in 
that direction.

There are two reasons you might want a null.  First, you may have a 
known interferer (noise source) in that direction.  Secondly, you may 
not care about that direction, but really want gain in another direction 
where the delays are all 0.  You can't get something for nothing.  More 
gain in one direction means less gain in other directions.

-- 
Best wishes,
--Phil
pomartel At Comcast(ignore_this) dot net

On Saturday, February 16, 2019 at 3:53:36 PM UTC-8, stch...@gmail.com wrote:
> Hi,
> A typical beamforming method widely used in ultrasound technology is to apply delays between channels and then summing the signals of all the channels. However, when properly calculated delays are applied, it could happen that the phase of the arriving signal is 180 deegres (half the ultrasound wavelength) shifted between channels and summed amplitude is decreased. This results in decreasing SNR and dynamic range. Does anybody here have an idea how to solve this problem?

When we sample a single time domain channel, assuming the baseband (DC to maximum desired frequency) case, we must satisfy the Nyquist criteria which requires that we sample for a rate of more than 2 samples per cycle in time of the highest desired frequency. A beamformer takes in samples from a set of locations in space and calculates samples for a signal from a selected bearing. There is a Nyquist criteria for the location of the sample positions in space. Consider a linear array of equally spaced sampling sites. The Nyquist criteria is that there must be more than 2 sample locations Array sensors_ for every wavelength (for the signal propagating in the media you work in) of the highest desired frequency to avoid spacial aliasing. If you satisfy that, you cannot have an unexpected 180 degree error.

Dale B. Dalrymple

<stchebel@gmail.com> wrote:

>A typical beamforming method widely used in ultrasound technology is to
>apply delays between channels and then summing the signals of all the
>channels. However, when properly calculated delays are applied, it could
>happen that the phase of the arriving signal is 180 deegres (half the
>ultrasound wavelength) shifted between channels and summed amplitude is
>decreased. This results in decreasing SNR and dynamic range. Does
>anybody here have an idea how to solve this problem? 

If you are trying to obtain maximum ratio combining, applying the
appropriate phase shift (perhaps with a Hilbert transformer) to
each channel should work.

How to obtain the needed phase shifts would depend on application,
but a training signal should work.

Steve

On Saturday, February 16, 2019 at 6:53:36 PM UTC-5, stch...@gmail.com wrote:
> Hi,
> A typical beamforming method widely used in ultrasound technology is to apply delays between channels and then summing the signals of all the channels. However, when properly calculated delays are applied, it could happen that the phase of the arriving signal is 180 deegres (half the ultrasound wavelength) shifted between channels and summed amplitude is decreased. This results in decreasing SNR and dynamic range. Does anybody here have an idea how to solve this problem?

but that means that the delays were NOT "properly calculated"

if the delays were properly calculated ___for the given wavelength and angle of arrival___  , the signal will be enhanced, not degraded.

maybe i don't understand the question?

maybe the point you miss is that beam forming is NOT a wideband process but is wavelength (frequency) dependent.


mark

On Tuesday, February 19, 2019 at 10:13:37 AM UTC-5, mako...@yahoo.com wrote:

> 
> maybe the point you miss is that beam forming is NOT a wideband process but is wavelength (frequency) dependent.

Beamforming can be either narrowband or wideband. Narrowband beamforming applies phase shifts to the signals - this equates to a time delay at a particular frequency. For narrowband systems you're assuming that the phase delay value doesn't change much over the frequency range of interest.

Wideband beamforming is achieved using time delays. Time delays can be achieved by upsampling the signal to achieve higher time resolution needed for beamforming - usually some form of interpolation is also used. It can also be achieved by applying a tapped delay line filter to each element of the array. The filter is trying to delay the signal by the appropriate amount. 

Most systems are assuming a constant speed of propagation, but in some environments the speed of propagation may vary.

David

> 
> Wideband beamforming is achieved using time delays. 
> 
>

I was hoping someone would chime in here.

A time delay that is constant over frequency will STILL yield a phase shift varies over frequency and therefore not be suitable for a wideband beamformer?


Perhaps you meant to say DSP can be used to create wideband beamforming by providing a time delay that varies over frequency such that the PHASE shift is constant over frequency, example Hilbert xform?

Mark


 

On 3/4/2019 12:13, makolber@yahoo.com wrote:
> 
>>
>> Wideband beamforming is achieved using time delays.
>>
>>
> 
> I was hoping someone would chime in here.
> 
> A time delay that is constant over frequency will STILL yield a phase shift varies over frequency and therefore not be suitable for a wideband beamformer?
> 
> 
> Perhaps you meant to say DSP can be used to create wideband beamforming by providing a time delay that varies over frequency such that the PHASE shift is constant over frequency, example Hilbert xform?
> 
> Mark
> 
> 
>   
> 
It looks like you are assuming that beamforming must be done in the 
frequency domain.  If your receivers feed delay lines with a relatively 
fine delay resolution, you can beamform entirely in the time domain.

-- 
Best wishes,
--Phil
pomartel At Comcast(ignore_this) dot net

> > 
> >>
> >> Wideband beamforming is achieved using time delays.
> >>
> >>
> > 
> > I was hoping someone would chime in here.
> > 
> > A time delay that is constant over frequency will STILL yield a phase shift varies over frequency and therefore not be suitable for a wideband beamformer?
> > 
> > 
> > Perhaps you meant to say DSP can be used to create wideband beamforming by providing a time delay that varies over frequency such that the PHASE shift is constant over frequency, example Hilbert xform?
> > 
> > Mark
> > 
> > 
> >   
> > 
> It looks like you are assuming that beamforming must be done in the 
> frequency domain.  If your receivers feed delay lines with a relatively 
> fine delay resolution, you can beamform entirely in the time domain.
> 
>

I agree with your statement, but

 I think the beam created by this "time domain method" will still vary as a function of frequency?

Because delay lines  create a phase shift that varies with frequency.

Because it is phase shift (not time delay)  that controls the beam

Am I wrong?

m




 

On Tuesday, March 5, 2019 at 7:51:47 AM UTC-8, mako...@yahoo.com wrote:

> I agree with your statement, but
> 
>  I think the beam created by this "time domain method" will still vary as a function of frequency?
> 
> Because delay lines  create a phase shift that varies with frequency.
> 
> Because it is phase shift (not time delay)  that controls the beam
> 
> Am I wrong?

Yes.

> 
> m

In the broadband time delay beamformer, the delays set the beam pointing angles, which are constant in frequency. The beam widths will vary with frequency as the effective size of the array, in wavelengths, changes with frequency.

Dale B. Dalrymple