Ultrasonic transducer help

[[email protected]]

I'm looking for two piezo transducers between 500KHz and 1MHz for use
in doppler velocity measurement. I'd like to find really cheap piezos
and put them in my own plastics assembly with polyurethane epoxy to
match the water's acoustic impedence. Precision is definitely not
important, does anyone know where to find cheap piezos in this range
with wire leads? All I am finding are high end transducers with highly
focused beams, which actually the opposite of what I need. Thanks in
advance.

 

[Don Pearce]

I'm looking for two piezo transducers between 500KHz and 1MHz for use
in doppler velocity measurement. I'd like to find really cheap piezos
and put them in my own plastics assembly with polyurethane epoxy to
match the water's acoustic impedence. Precision is definitely not
important, does anyone know where to find cheap piezos in this range
with wire leads? All I am finding are high end transducers with highly
focused beams, which actually the opposite of what I need. Thanks in
advance.

It is not being high end that causes them to have highly focussed
beams. At these frequencies, with corresponding short wavelengths, you
would find it hard to make a transducer any other way.

d

Pearce Consulting
http://www.pearce.uk.com

 

[Winfield Hill]

Don Pearce wrote...
You can get the raw custom disks from Channel Industries in
Santa Barbara, although it's rather technical and you have to
know what you're doing. http://www.channelindustries.com/ Give
them a call and ask what they have left around from other jobs.

It is not being high end that causes them to have highly focussed
beams. At these frequencies, with corresponding short wavelengths,
you would find it hard to make a transducer any other way.

Yes, and no. Most commercial transducers in this frequency range
that one easily come across are intended for use in the metallurgy
industry, and they have special short-range focusing structures
mounted on the piezo element. If instead you simply use a small
disk, say 10mm dia, the acoustic beam will diverge. The beam-width
in water is a function of the frequency and the width of the disk.

The disk's acoustic vibration mode may be complex, depending on
its thickness, e.g., see http://www.ift.uib.no/~jankoc/thesis/ but
a piston radiator is an appropriate approximation for many cases.
Even so, the acoustic directivity may be complex to analyze, e.g.,
http://www.ndt.net/article/v07n06/karpelson/karpelson.htm however
you can simplify and assume a 5-degree beam width for a 10mm disc
at 1MHz, and scale from there.

To understand acoustic transmission calculations, beam-width and
the sonar equation, attenuation and other issues, I recommend
Robert Urich's excellent book, Principles of Underwater Sound.
On pages 102-111 of the third edition, you'll learn that the
attenuation in water can be severe, increasing with the square
of frequency. At 1MHz it's about 0.4 dB/meter (that's 0.8dB/m
for a round trip). This is in addition to spreading losses.

Have fun, Kelly, and report back to us as you proceed.

 

[James Meyer]

It is not being high end that causes them to have highly focussed
beams. At these frequencies, with corresponding short wavelengths, you
would find it hard to make a transducer any other way.

You can use most of the optical methods, lenses and mirrors, to expand
or contract ultrasonic beams.

There is a good source for piezo transducers on eBay. Check out
http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=42899&item=3865272506&tc=photo.

You will have to add your own wires, but that's extremely easy to do.

Go to their eBay store for more items. Their web site has data sheets
for their transducers. I've ordered from them before and can recommend them
highly.

Jim

 

[[email protected]]

Thanks all for the help, the resources, this is exactly what I was
looking for. One more question: are there any good
websites/books/tutorials that describe how to make a transducer from a
piezo disk? I've it is rather difficult and thought I should try to
find something with wire leads, but I'd like to know more. Thanks
again.

 

[[email protected]]

Sorry, I answered my own question pretty quickly. It seems they all
come with electrodes that just require a good soldering connection.

 

[Winfield Hill]

[email protected] wrote...

Sorry, I answered my own question pretty quickly. It seems they all
come with electrodes that just require a good soldering connection.

And you'll notice that Channel Industries' downloadable
info booklet has detailed soldering instructions.

What are you working on?

 

[[email protected]]

What are you working on?

I am taking a stab at making a flow meter using continuous wave
doppler. Mainly to avoid ones on the market that are problematic and
pretty expensive for what you get. If I can make a transducer
successfully (and test it out with an oscilloscope), I think I can make
the A/D, DSP, and interface work pretty easily. Suggestions are always
appreciated, thanks again for your help.

 

[James Meyer]

I am taking a stab at making a flow meter using continuous wave
doppler. Mainly to avoid ones on the market that are problematic and
pretty expensive for what you get. If I can make a transducer
successfully (and test it out with an oscilloscope), I think I can make
the A/D, DSP, and interface work pretty easily. Suggestions are always
appreciated, thanks again for your help.

I don't think A/D and DSP are entirely necessary. Everything you need
should be "do-able" with analog devices. Take a look at the spec sheets for the
LM2907 F/V converter. www.national.com/pf/LM/LM2907.html

Jim

 

[Winfield Hill]

James Meyer wrote...

I don't think A/D and DSP are entirely necessary. Everything you
need should be "do-able" with analog devices. Take a look at the
spec sheets for the LM2907 F/V converter.
www.national.com/pf/LM/LM2907.html

Tell us more, please, oh Jim, the ultra-sound man!

 

[James Meyer]

Tell us more, please, oh Jim, the ultra-sound man!

OK, but remember, you asked for it.

Start with a clean source of voltage to drive the transmitt element.
Harmonic distortion doesn't matter much, but amplitude and frequency "noise"
should be kept as small as you can. Make it easy to vary the frequency a little
to hit the "sweet spot" of the transducers, ie. be able to tune for maximum
smoke.

The receiver section needs some gain at the ultrasonic frequency with a
little selectivity. A tuned step-up transformer feeding a JFET with an LC tank
in its output should be enough.

Follow that with a simple signal diode envelope detector. The received
signal will contain a lot of the un-dopplered transmitt signal with the
dopplered part being 40 to 60 db lower, so most of the system gain needs to be
after the envelope detector. The un-dopplered carrier will create a large DC
value at the detector. You don't need that for anything, so throw it away with
a coupling capacitor. Make the RC roll-off low enough to catch the slowest flow
rate you are interested in.

From here on, it's a normal audio frequency signal.

Next, pass the signal through an active low-pass filter with the upper
cut-off set to pass the doppler for the highest flow rate of interest. Gain can
be incorporated in the filter or place a gain stage before the filter with,
perhaps, another after the filter.

The F/V converter, if you use a LM2907, only needs a 25 to 50 mV P-P
signal, so distribute the overall gain to get that value. The output of the F/V
converter is, of course, an analog version of the flow rate.

I have done the job by putting the raw amplified ultrasound through a
fast A/D into a DSP, but I can't recommend that for a DIY project.

Jim

 

[[email protected]]

Jim/Winfield,
Thanks both for the help. The problem with Jim's approach is I am
much more of a programmer than an electrical enginner and really only
understood the parts that pertained to signal processing. Here's my
planned approach, let me know what you think.

If you sample at the transmit frequency (F), exactly, a velocity of 0
corresponds to no frequency exactly. A doppler shift of 100 Hz will
result in a strong frequency around 100Hz in the FFT. You can
theoretically detect a shift of F/2, but that would be an
unrealistically high velocity. The other problem is that with a
relatively small FFT, say 2048, one bin corresponds to a large change
in velocity.

Therefore, if you sample at F/2, a doppler shift of 100HZ will result
in a strong frequency around 50Hz in the FFT, and the maximum
detectable doppler shift is F/4 (Still very fast in terms of water
velocity). So my plan is to sample at somewhere around 50KHz, which
will still give me span of F/20, which in practical terms is about 30
m/s, depending on the angle. A benefit of this approach is that the
resolution of each bin is quite, which is desireable.

Naturally, the signal will have to be amplified before DSP, I will
determine that gain using an oscilloscope. Any thoughts? I am going
to try using a rabbit core module (http://rabbitsemiconductor.com), as
they are pretty easy to program. Please let me know if you see any
problems, or have suggestions. Thanks again.

 

[Winfield Hill]

[email protected] wrote...

Jim/Winfield,
Thanks both for the help. The problem with Jim's approach is I am
much more of a programmer than an electrical enginner and really
only understood the parts that pertained to signal processing. ...

Ahem!

 

[James Meyer]

If you sample at the transmit frequency (F), exactly, a velocity of 0
corresponds to no frequency exactly. A doppler shift of 100 Hz will
result in a strong frequency around 100Hz in the FFT. You can
theoretically detect a shift of F/2, but that would be an
unrealistically high velocity. The other problem is that with a
relatively small FFT, say 2048, one bin corresponds to a large change
in velocity.

That's fine as far as it goes. Did you notice that the doppler shifted
energy will be 40 to 60 db lower than the total signal? That places very harsh
requirements on the A/D process and the digital processing subsequent to that.

Jim

 

[Winfield Hill]

James Meyer wrote...

That's fine as far as it goes. Did you notice that the doppler
shifted energy will be 40 to 60 db lower than the total signal?

Isn't it often worse than that? Jim, for your nice return signals
have you been using well-chosen send-receive transducer placements,
with a shield, and with dirty water?

 

[James Meyer]

James Meyer wrote...

Isn't it often worse than that? Jim, for your nice return signals
have you been using well-chosen send-receive transducer placements,
with a shield, and with dirty water?

My major effort so far has been with blood flow using reflection from
red blood cells. Dirty water? I guess that's a pretty good description. The
transducers are side-by-side and oriented so that the beams converge on an
artery about 10mm under the skin's surface.

Jim

 

[Winfield Hill]

James Meyer wrote...

Winfield Hill wroth:


My major effort so far has been with blood flow using reflection
from red blood cells. Dirty water? I guess that's a pretty good
description. The transducers are side-by-side and oriented so that
the beams converge on an artery about 10mm under the skin's surface.

No, I suspect even dirty water is a poor backscatterer compared to red
blood cells. Clean water, even worse unless there are turbulence cells.
Ahem, the O.P. did say, "with polyurethane epoxy to match the water's
acoustic impedance," so I'd imagine poor backscatter signal strength
compared to the transmit signal could be a significant issue.

 

[Glen Walpert]

My major effort so far has been with blood flow using reflection from
red blood cells. Dirty water? I guess that's a pretty good description. The
transducers are side-by-side and oriented so that the beams converge on an
artery about 10mm under the skin's surface.

Jim

Nice project, Jim. I hope it works out to be usable for continuous
monitoring of blood flow to the brain, usable by first responders to
detect the reduction of brain blood flow which often occurs after head
trauma. The current method used by first responders, the Glasgow
scale, is applied correctly less than half of the time (detects less
than half of the instances of high Intra-Cranial Pressure (ICP)
causing loss of blood flow to the brain, in a timely manner). Since
loss of brain blood flow always results in death or severe brain
damage if not treated promptly, such a device has the potential to
save many thausands of lives. High ICP, typically caused by
hemorrhage of small blood vessels when the brain moves in the skull
due to impact or brain "bruising" and swelling, can be properly
treated only at major trauma centers having MRI and a surgical team
prepared to enter the skull, and sucessful outcome is dependent on
prompt recognition of the problem and rapid transit to an appropriate
facility. Since there is often no immediate outward sign of this
problem, and there may not be any other serious injury, appropriate
treatment is often not provided even if readily available.

This is just the sort of thing that NIH likes to fund.

Best regards,
Glen

 

[[email protected]]

The water is dirty, wastewater (i.e sewage) if you really want to
know, which should help a bit with the recieve. I am aware the return
signal will be pretty faint compared to the transmit, I'm not sure how
faint but am prepared to invest the time to make a transducer, connect
to an oscilloscope and play around with it until I find out if it is
possible. My guess is the first few bins of FFT will be all bleed from
the transmit, so near zero velocity detection will be impossible. I'm
also trying to read up on transducer design to get a good recieve
signal. Thankfully it is not as critical as blood flow, but should be
a good way to learn and possibly produce something useful.