Question: underwater locator beacon project

[smashedProton]

Hello all,
I just started my freshman year of college and I joined a robotic submarine team.

I'm working on a hydrophone ULB system.  The current system isn't working but it consists of three directional hydrophones with preamplifiers integrated in them that are read into a sound card and then sent for processing on a computer.  The pings are ~20khz so the sound card isn't good enough to reconstruct the waveform for the program on the computer.  I believe that I have three options:

-pick a different ADC and use that to send data to the computer
-use interrupts on a PIC microcontroller and do sound ranging
-use a DSP and just send the distance and direction data to the computer

I would like to use the second option because sound ranging is easy to do computationally, and it would be far easier to get the distance to the target.  I was thinking that I could use a peak detector and a comparator in series and send the output straight into the interrupt of the PIC.  I dont know if this will actually work, will the rising edge off the peak detector be consistent enough to rely on it for nanosecond level measurement?  The PIC has 5 16 bit counters.  16 bit resolution is enough but I think it could only be used in microsecond increments which isnt enough for this application.

What do you think?

[Precipice]

A 4th option might be to get a sound card with a higher sample rate? (although you'd need to confirm that your sound card has at least three actual ADCs, rather than mixing inputs into two).
(My elderly UA-101 is good for 24 bit 192Ksa/s, with 20Hz-40KHz analogue paths. I hope things have only got better in the last decade).

Is your team better at hardware or software?

Edit: Nanosecond measurements? That seems ambitious and possibly unnecessary. 1500m/s suggests that 1mm resolution is 2/3 us, and you're not seriously hoping for 1mm, are you? Where on the wave are you planning to take your 'distance' point? If your transducers are at all resonant, there'll be a slope from 'no sound' to 'sound'.

Edit2: I think that, whatever solution you pick, ease of prototyping, debugging and changing things trumps pretty much any other requirement. Do you have scope plots of what the current system sees?

[Marco]

How about AM modulating the waveform? In principle you could demodulate that in analogue so your sound card could capture the demodulated signal (or do the peak detector thing on the envelope ... it's going to be far more accurate than trying to do it on a single cycle of the ultrasonic signal).

[smashedProton]

How about AM modulating the waveform? In principle you could demodulate that in analogue so your sound card could capture the demodulated signal (or do the peak detector thing on the envelope ... it's going to be far more accurate than trying to do it on a single cycle of the ultrasonic signal).

How would AM modulation work in this context?  I'm not used to doing it with signals of the same frequency.  Thanks

[HackedFridgeMagnet]

A 4th option might be to get a sound card with a higher sample rate? (although you'd need to confirm that your sound card has at least three actual ADCs, rather than mixing inputs into two).
(My elderly UA-101 is good for 24 bit 192Ksa/s, with 20Hz-40KHz analogue paths. I hope things have only got better in the last decade).

Surely that is the obvious solution.

Or change the beacon transmission frequency, but maybe that is a requirement.

[Marco]

How would AM modulation work in this context?  I'm not used to doing it with signals of the same frequency.  Thanks

I assume you're just exciting the piezo speaker with a pulse consisting of a square wave right? Multiply that square wave with a half sine of much lower frequency ... multiplication == AM modulation. That would add complexity to the piezo driver though ... you could gate the square wave with a lower frequency PWM encoded sine, which would allow you to use a simple square wave driver but which could be detected in much the same way as AM (but I expect I've lost you now).

BTW even if you do get a sound card which supports 96/192 kHz sampling (I think even cheap USB devices based on say VT1620A do) you will probably need to mod the input low pass filter ... and hope the codec in and of itself doesn't already limit bandwidth too much.

[zapta]

Why did you pick 20khz? You mentioned nanosecond accuracy which surgery very fine distance resolution. You may need to use higher frequency.

Is this a demo student project or a real system that need to perform? If the later you will need to invest more in the signal processing. It also depends on the range and acoustic environment.

[David Hess]

A 4th option might be to get a sound card with a higher sample rate? (although you'd need to confirm that your sound card has at least three actual ADCs, rather than mixing inputs into two).
(My elderly UA-101 is good for 24 bit 192Ksa/s, with 20Hz-40KHz analogue paths. I hope things have only got better in the last decade).

Many (most?) higher sample rate sound cards still implement 20kHz anti-aliasing filters.  There are ones that do not though:

http://www.clarisonus.com/Research%20Reports/RR001-SoundCardEval/RR001-PCsoundCards.html

[EvilGeniusSkis]

I would suggest using 37.5kHz watch this @MikesElectricStuff video for an explanation

EDIT: oops forgot link here it is http://youtu.be/mQehX0rVYuY?t=2m56s

[nfmax]

Another approach would be to use a local oscillator and an analogue multiplier to frequency shift the incoming pulses into the sub 20kHz range, then feed them into the standard sound card. Depending on the bandwidth of the transducer, you should be able to get all 3 of them into say 5 - 15 kHz. Then use DSP to get the first break times for each channel. I assume the transmitter is on the object being tracked, so you have no control over it, right?

[zapta]

1500m/s suggests that 1mm resolution is 2/3 us,

And double that time for round trip so this helps. On the other hand,  20Khz is relatively low frequency for short distances and between it and the signal distortion in the water and time constants in the detector the practical resolution will be much worse than that.

I would start with recording real signals with a good DAC and playing with them with some math package running different scenarios and solutions before building anything. If you don't have experience in this field, your initial assumptions are likely to be wrong.

[Charles Creations]

Another approach would be to use a local oscillator and an analogue multiplier to frequency shift the incoming pulses into the sub 20kHz range, then feed them into the standard sound card. Depending on the bandwidth of the transducer, you should be able to get all 3 of them into say 5 - 15 kHz. Then use DSP to get the first break times for each channel. I assume the transmitter is on the object being tracked, so you have no control over it, right?

You can read more about this at the following:
http://en.wikipedia.org/wiki/Heterodyne