High frequency microphone amplifier

Hi all -
I am working on a project that involves capturing sonar data, and it involves some pretty hefty amplification.
I am using preamplified hydrophones (underwater microphones) with an OCV of -167 dB//1V/μPa. (http://www.itc-transducers.com/store/catalog.taf?function=detail&product_id=30) The data acquisition card (just in case anyone is interested, http://sine.ni.com/nips/cds/view/p/lang/en/nid/13677) takes signals in the range of -5V to +5V, which is obviously a much larger range than the hydrophones can produce. Also, to make things more difficult, our frequency range is 50 kHz to 112.5 kHz with peak sensitivity at 79 kHz.
I have acquired a few INA217 IC's, and I've been trying to hack an amplifier, but I haven't made much progress. Does anyone have any suggestions on how to build an amplifer for these high frequencies and with this much gain? Any response is appreciated.

-Mike
 
Analog Devices and Texas Instruments are your best bets. Guess where Uncle Sam gets his milspec low noise opamps. ADI has a number of white-papers on amplifiers for ultrasound, hydrophones and sonobuouys.

Take a look at the AD797 and the SSM2019. The INA166 from TI's BurrBrown is probably better suited for Sonar than the INA217.

btw, the INA217 is a good replacement for those who need the old (now discontinued) SSM2017.
 
I have a couple of newbie type of questions for this project.

First off, do I just wire the Vout from the transducer (microphone) to the Vin on the amplifier? The transducer does not have differential outputs, only Vout, and ground, plus something called B+ and a calibration wire.

Also, it seems that I will be needing a decent power supply to do this. Any recommendations on what to use for this? I'm currently using some random 5V AC to DC adapter with the end of the wire cut off, split, and wired into a breadboard. I've seen some tutorials saying that you can use 9V batteries wired together in series. What sort of gains should I really expect from using these amplifiers? Can I expect to see output voltages in the volts and get out of the range of millivolts?

Lastly, and the most embarassing question to ask... how do I properly ground my connections? The IC's require a ground, as well as the transducer.

Thanks for your help jackinnj.
 

jcx

Member
2003-02-17 7:38 pm
..
"The noise from a typical fishing vessel will have a
spectral level of 150-160 dB re. 1 uPa/Hz at 1 m
distance."

You really need some more # before any intelligent choices about additional amplification can be made

Expected peak spl * sensitivity determines the gain to match your adc input dynamic range

Finite adc resolution can also make extreme low noise pointlessly below lsb resolution -16 bits is very good resolution today for 1 Mhz sample rate which you may need, so 96 dB is all the dynamic range you may be able to encode

The transducer spec sheet claims it has internal preamp

Output noise (vs freq?) from transducer preamp determines how low noise your amplifier has to be – there is little advantage in having your input noise << ~1/3 of the preamp’s output noise since noise adds vectorially

Transducer preamp high pass corner frequency determines importance of 1/f noise in your amp – or the position of any desired band limiting filtering in your electronics - to avoid the low frequency trawler noise for instance

Sample rate determines required anti-alias filter frequency, built into or following the amp
 
I am not especially interested in high quality audio. My major concerns are: 1 - speed of the signal and 2 - a discernable difference between background noise and interesting noise. Let me elaborate.

The goal of this project is to put together a 3D array of microphones with the intention of being able to determine the direction of a sound. I will be producing the sound, so it can be fairly loud. The clarity of the sound is not important. I am only concerned with the output voltage of a loud sound being significantly higher (5 to 10 X) than the background noise, and this amplification needs to be fast, since the sounds will be short bursts, not sustained sounds.

I really want to be able to amplify the sound by 20 to 30 dB in the range of 50 to 112.5 kHz. It would also be nice to have a filter (band pass or high pass) to focus in on that range.

I hope that is enough information to understand our application. Let me know if you need more information.
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
Just for my perspective, I need to translate to "air mike" customary specifications. And I'll do it out-loud so others can spot my mistakes.

ITC-8084A hydro mike: Midband OCV -167 dB/1V/µPa

SM-58 stage mike: Open Circuit Voltage: –54.5 dBV/Pa* (1.9 mV)

It takes 1,000,000 µPa to make 1Pa, and this is pressure, so add 120dB. The wet-mike's -167dB spec is equivalent to -47 dBV/Pa in an air-mike, or 7.5dB hotter than an SM58.

On a rock-n-roll stage, an SM58 needs a preamp. Maybe as much as 30dB, maybe less than 10dB. This is to get to ~1V RMS nominal, which would leave 10dB headroom into your +/-5V digitalizer.

I don't know if your underwater sound source can match the levels of a rock drummer. But I do think your 20-30dB gain goal is not-unreasonable.

But there are bigger problems in your path.

For one: sound pressure declines 6dB for each doubling of distance. Unless your source is constant radius, or its level is adjusted to give a good level at the mike, the preamp needs a wide dynamic range, and very possibly gain adjustment. If your source is an electric speaker, and 1 Watt of power makes a nice signal at 10 feet, at 100 feet you need 100 Watts and at 1,000 feet you need 10,000 Watts. Whether you use clap-sticks or sticks of dynamite, the economics are the same. And if the source is fixed-level (presumably at the maximum level), then at closer range the signal at the mike will be VERY high, enough to overload the preamp and digitalizer.

(Yes, in typical ocean the 6dB/double-distance rule fails at very long range, greater than ocean depth. It also goes screwy as you cross thermoclines. Finding the enemy sub outside your harbor is tough. But by mentioning "fishing vessel" I suspect your scale is mostly inside the 6dB/doubling rule.)

> this amplification needs to be fast, since the sounds will be short bursts, not sustained sounds.

If it will pass 100KHz, it will pass the burst-rate.

The digi-card is rated 250K samples per second. IF it will really do that, it is just barely fast enough to capture a -112KHz sine wave.

If you need a gain of 1:30 (about 30dB) at 112KHz you need an amplifier with an open-loop gain of 30*112K= 336KHz, plus more for feedback. That's almost trivial. A 741 will be down a few dB at the top of the band.

A 741 will not swing +/-5V in 1/112KHz. Using a generous rule of audio-thumb, you want an amp with a 25V/uSec slew rate. This is not hard: TL071 may come close enough for your purposes.

I'm not sure you want to be capturing raw sound. If you want direction, and the ping is significantly higher than background noise, you just need to note the time of the pulse rise.

To put your work back on dry land where most of us live: JFK is about to be shot again. We put one mike in Dealey Plaza, wire it so a sharp loud sound will punch a hole in a slowly moving paper-tape. That tells when the shot happened. For HiFi Stereo we put up two mikes 10 feet apart. If the paper-tape holes are at the same point in time, the gun was on a line perpendicular to the line between the mikes. If the holes are 10mS apart (in air, 1 foot ≈ 1mS) then the gun was along the line of the mikes, and the first hole points toward the gun. 7mS apart defines two half-lines 45 degrees off the mike line. The dualities can be resolved by a third mike; triangulation. In a 3-D world such as deep water, you need four mikes in a tetrahedron.

But all you need for direction pointing is arrival time, not all that other stuff.

I ferget the speed of sound in water; say it is 1 foot ≈ 0.1mS (10X speed in air). If the mikes are 10 feet apart, a 7.07mS gap points to 45 degrees. Your proposed sampler quantizes time in 0.004mS (4uS) units, so "7.071mS" could really be 7.066 or 7.074mS. This gives 0.063 degree error, which I suppose is fantastically good compared to thermoclines and other watery distortions. And note that the only data you record is four time-stamps, not millions of samples that need further crunching to extract data.

What bothers me most is the utter lack of useful interface info on ITC's site. It has a preamp inside; I do wonder if you should be paying for that when you have to build a preamp anyway. Perhaps you do: these are ceramic mikes with odd output impedance. In audio, ceramics mean short cables or onboard preamps. However, what is the supply voltage, what is the preamp output noise, what is its output impedance, is it unbalanced or balanced? Unless ALL hydrophones use the same interchangeable interface, these are things that should be right on the web site, but are not (that I could find in 10 minutes). I could understand if they sold interface boxes, but I don't see those either.

Some of your electronic questions are at a level that makes me suggest: you are spending quite some cash on mikes, and boats etc, on a project that can't be bench-tested even in a bathtub. I think you need to recruit someone who knows more than a little electronics, and preferably someone who has dabbled in hydrophones and the whole sonar thing. This basic problem was wrestled 60+ years ago. While it is outside my personal experience, it must be daily-chore to some electronics geek. (If those mikes cost as much as they look, ITC's support staff should help, and may have designs on file.)
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
> how do I properly ground my connections? The IC's require a ground, as well as the transducer.

Run a 2,000 mile green wire back to shore. :( :rolleyes: :D

You are confused because "ground" (or in the UK: "earth") is a very poor word for electrical stuff. It means many things, and almost none of them involve a rod stuck in the dirt.

Power supplies and the stuff they power have (at least) two wires. For a number of dull practical reasons, one of them is often common to most of the system, and improperly (but almost unavoidably) called "ground".

Signals can be caried "balanced", as the difference between two wires, without reference to anything else in the world. However it is often cheaper and sometimes cleaner to connect one signal wire to a "signal common" that all signals in the system are referenced to.

Low-level signals need shields, and these need a place to dump any interfering crud they get, without crudding-up the signal.

In most home HiFis, power signal and shield commons are the same thing. It works. It avoids complications which might dirty the sound. It is cheap, which leaves money for other improvements.

In Pro Audio, the mike lines and often the high-level signal lines are balanced, two signal wires and a separate shield. Inside a box, the shield usually connects to power-common and that may connect to internal signal common. These days, even external balanced signals may really reference power common (the good old stuff didn't).

So keep your various commons in mind. They may all come together at some point, but should not share wires unless you estimate the consequences.

The outside case of the microphone (a shield) will generally come to the next big box's shield, perhaps your PC (which is always a bad idea around low-level signals). The PC's box shielding goes to shipboard power. That is surely bonded to the hull, and consequently to "the sea and the whole world", but that "grounding" is not needed. As long as all the commons connect to each other, you could beach the ship on a pile of glass bottles, or lift it into the air or into deep space, no contact at all with the "earth"; it would still work fine. Same way as your rubber-tired car radio and engine-computer work without a rod stuck in dirt.
 
PRR said:


Some of your electronic questions are at a level that makes me suggest: you are spending quite some cash on mikes, and boats etc, on a project that can't be bench-tested even in a bathtub. I think you need to recruit someone who knows more than a little electronics, and preferably someone who has dabbled in hydrophones and the whole sonar thing. This basic problem was wrestled 60+ years ago. While it is outside my personal experience, it must be daily-chore to some electronics geek. (If those mikes cost as much as they look, ITC's support staff should help, and may have designs on file.)

Or, since the Defense Department is closing the Groton Sub Base, you could just post in the help wanteds in the Groton Times!

Suggest you pick up one of the DSP whitepapers which specifically deal with sonobuoys on the Texas Instruments website.
 
Ok, it would seem that I need to elaborate our designs here.

First off, this is not a system that is going to be tested in the ocean, or anything close to it. Our test area is a pond, that we can assume is about 4 meters deep and is generally an elipse with diameter1 of 30 meters and diameter2 of 20 meters.

We won't be needing any powerboats to hold up the contraption. The pond has a floating... thing in the middle of it. It's not really a raft, but it isn't much different than a large raft, so I guess we can just call it a raft to make things simple.

We are indeed using a tetrahedron of hydrophones. It is a cube of PVC (edge length = 2 meters) with the hydrophones mounted at 4 corners (forming a tetrahedron). The cables will be run back to shore (which is not 2000 miles away, more like 20 meters) to a computer which will acquire the data. I have already written the software to just record the time of the first signal arrival.

This isn't a massive budget project. Other than the hydrophones, our expenses are practically nill, aside from some cabling and this amplification sub-project. The project isn't really to test the hardware system. This is more of a proof of concept to test our software on the recorded data.

Thus, all I really need is a method to amplify these audio signals (which can be in the range 50kHz to 112.5kHz, but will most likely be a loud sound covering a large portion [60%?] of that range) with a gain of 20 to 30 dB. I have been trying to test some instrumentation amplifiers (INA217) with a normal microphone (if anyone is interested http://www.radioshack.com/product.asp?catalog_name=CTLG&product_id=33-3022).
I am just having problems getting this amplifier to actually work. I was hoping someone would be able to post (or link) a tutorial on assembling an amp similar to this.

PRR, thank you very much for your help. I currently just have all of the grounds wired to a single point but that point isn't going anywhere from there. Could I just connect the ground(s) to the DC source V- and that would be good enough?

Also, the speed of sound in water is (generally) 1490 to 1510 meters/second.
Sorry for the really long post.
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
> Ok, it would seem that I need to elaborate our designs here.

No. At least I don't care if you use a battleship or a rubber ducky in an ocean or a bathtub. To repeat: What bothers me most is the utter lack of useful interface info on ITC's site. .... what is the supply voltage, what is the preamp output noise, what is its output impedance, is it unbalanced or balanced?

I found www.dolphinear.com and www.AquarianAudio.com offering hydrophones with generic consumer audio plugs. That isn't what you have, and anyway implies maybe-only-20KHz response.

www.reson.com has hydrophones with connectors very much like your ITCs. I do NOT know if the wiring is universal!!!!

[IMGDEAD]http://headfonz.rutgers.edu/hydro1.gif[/IMGDEAD]

Looking at that interfacing (assuming internally preamped hydrophones), I tossed off this ten-cent idea:

[IMGDEAD]http://headfonz.rutgers.edu/hydro2.gif[/IMGDEAD]

A pair of 9V batteries will power it for many-many hours; I've lived with HiFi preamps powered this way.

The chip should be TL072, available anywhere in the world for $0.32 (except $1.69 at Radio Shack). The sketch has LF412 because this stupid stimulator did not come with any "common" opamp model, and I have not yet herded my collection into it. LF412 is, for a first impression, "equal" to TL072.

Top-end roll-off varies with gain, but at maximum gain is only -1dB at 112KHz.

Bottom roll-off varies with Cin. Omit that and you are good to DC. I suspect for radar-ranging work, you want as little bass response as possible, which suggests Cin could be 1 or 2 nFd (1,000pFd-2,000pFd) for -3dB cutoff at 15KHz to 180KHz (3.5 octaves).
 
Allow me to but in a bit...

PRR,
If the TL072 is used at, say, 100kHz, a small variation in gain between the four amps is likely to produce a variation in the output signal phase angle. This may jeopardize accuracy of the measurement. INA 217 seems much better suited for the job.

Mike,
It might be a good idea to place one transducer into the tetrahedron for calibration purposes.

Regards,
Milan
 

jcx

Member
2003-02-17 7:38 pm
..
a good implementation should have anti-alias low pass filtering, the common approach is to use active filters

since anti-alias filters are often rather high order phase/group-delay stability is an issue - both passive component matching between channels and finite/variable op amp gain-bandwidth can be problematic

I have used tl074 in complex composite phase compensating arrangement for a 20 KHz low phase shift measurement device, only because I didn't have time to get anything faster on the corporate approved component list - use fast op amps, its easy today to get op amps fast enough to put the active pole >> 20x the filter corner frequency
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
> INA 217 seems much better suited for the job.

Agree that an experienced tinkerer will turn to something snazzier than the '072. However Mike has had trouble getting anything out of the 217, and hasn't said why. I suspect he should start with a KISS design, if only to narrow his misunderstanding. When he gets a signal, he can determine if the marginal speed of a '072 is an issue. If it needs something faster, much faster opamps are easy to find. With a working '072 prototype, spiffier chips are semi-drop-in. (Hmmmm.... I should note that even the mild '072 would like bypass caps on its pins, and faster chips demand good bypassing...)

And the INA217 just has too many pins. The differential input appears to be more complication than feature for this unbalanced pre-amplified application, inputs are liable to oscillate, etc.

> should have anti-alias low pass filtering

Yes. But here the sample rate comes out about twice the nominal low-pass of the mikes and the amps. The roll-off may not be very sharp, aliasing is possible, but probably reduced enough to allow proof-of-concept. Also, for direction location we really want the first large sample (that isn't a noise-fluke), so the sampling theorems we rely on in steady-state and in audio are perhaps not applicable.

> a small variation in gain between the four amps is likely to produce a variation in the output signal phase angle. This may jeopardize accuracy of the measurement.

Quite true. I will offer several points in refutation; any, all, or none may be applicable.

The gross phase error seems to be "small", 18 degree, 1uSec, or 0.0015 meters, sub-millimeter (if my sliderule didn't drop a decimal place). Now the onus is on the PVC pipe and mike-clips to maintain millimeter accuracy. (I didn't get much sleep last night, so if I've spilled some zeros and screwed-up bad, sorry.)

The differential error between two '071s from the same crate is smaller; among the units in a dual '072 or quad '074 even less.

The time response of the hydrophones is not specified except as a typical frequency curve. I don't know how they make these things; I know in air-mikes with a top-ring, control of the resonance frequency is tough, and low tolerance means high dollars. There may well be more phase variability among the four "identical" mikes than across all the '072s ever sold. ('072 GBW can be 10%-30% below "typical", no upper limit, but you won't find any vastly-fast '072s.)

This project is about the computer software, not the hardware. And in fact cheap smart CPUs can reduce the need for expensive precision hardware. Calibrate! In context of a raft-test, probably the mikes will be "loose" at first. Hold all four in your hand, rap their noses with a book to align them, have a source a few yards away. The resulting recording instantly and simply gives you the correction for each mike and amp. As moamps says, hanging a small source in the exact center of the tetrahedron (perhaps on four matched bungee-cords) allows re-calibration any time.

Being a proof-of-concept, modest repeatable errors are not fatal. They give you something to do with any funding you can scare-up, or be the core of a thesis for your next degree. If the prototype came out perfect, you'd be down to the deadly-dull business of Marketing.
 
ITC sent me data sheet with the pinouts, and they are (seemingly) identical to the pinouts from RESON. The only difference I see is that the DC in is called B+. I have contacted ITC asking "what is the supply voltage, what is the preamp output noise, what is its output impedance, is it unbalanced or balanced". I'll put that information up as soon as I receive it.

Here is the schematic provided by ITC (http://www.simoniac.com/~welsh/img/ITC8084A.jpg)

I think I may have finally gotten some output from the INA217. I have my sound card wired into the Vin (playing a constant 440 Hz tone at max volume) and gain set to 40dB (110 ohms across Rg). I am getting about 0.45 Vrms on the input terminals (pins 2 and 3), and 2.0Vrms on the output (I am measuring the voltage across Vin- and Vout, pins 2 and 6).

I am also going to try to build TRR's circuit. One other thing I would like to know is how will things change when I make the adjustment from differential inputs to single ended inputs? Can the INA217 even work on a single ended input?