Using the TMS320F2812 DSP Board to locate the Ping
Overview:
Timers to set our observation window to only be open for 1.3 Msec every 2 Seconds, synchronized to the ping burst of interest.
By looking at the journals of the six schools that received any points at all for locating the ping transmitter during the 2006 event, you can see that they all used some sort of passive sonar hydrophone array to accomplish this task (four of the six schools used arrays of (4) hydrophones, and the other two used arrays of (3) hydrophones). It can further be seen, by examining various references called out in the journals such as U.S. patent 4,622,657 that by measuring phase differences (time that the ping tone arrives at each hydrophone) that the bearing, and perhaps even elevation angle, to the ping transmitter can be determined. The following sections provide a preliminary description of the hardware and software that I think we will need to accomplish this task.
1.0 Hydrophone selection and Array Geometry:
The hydrophones we choose should have frequency response that is flat to at least 40 KHz. They should be omni directional in the horizontal field and possibly also in the vertical field . They should be capable of withstanding depths of at least 32 feet, and they will need to be small enough in diameter to accommodate a 30 KHz underwater wavelength/2 spacing (approximately 1 inch center to center). For the purpose of this discussion I will reference the TC4013 made by Reson (PDF spec is on group website), which meets or exceeds all of the above requirements.
The hydrophone array (the number of hydrophones we need and their placement) must also be determined. Team member Colin Bradbury in his January 9th, 2007 Google Group email has already started to address this issue.
The hydrophone selection and understanding of the array geometry required are going to be extremely important in determining our success in this task.
2.0 Signal conditioning preamps for hydrophone output signals:
The ping transmitter puts out a 187db signal and the TC4013 hydrophone has a receiving sensitivity of –211db, therefore the output of the hydrophone is –211 + 187 = -24db which equates to approximately .14 volts peak to peak. Since the A/D converter channels on the DSP have a range of 0 to 3 Volts (full scale) we need to amplify the hydrophone signal by 3.0/. 14 = 21.5.
It is important to have the input to the A/D be as close to its full scale range as possible since this gives our circuit it’s best signal to noise ratio. If we use programmable amplifier stages for the hydrophone inputs we can have the DSP digitally “calibrate” our hydrophones to the pinger during our practice run and we can get the best signal to noise ratios possible.
It will also be very important to mount these hydrophone amplifiers as close as possible (less than a foot away) to the hydrophones themselves since they will be high impedance circuits and very susceptible to electrical noise.
3.0 Analog to Digital (A/D) Converters:
The TMS320F2812 DSP contains a single 12 Bit A/D converter and (16) sample and Hold (S/H) devices that can be multiplexed to sixteen separate analog input channels.
For our “ping finder” we will most likely use three or four of these channels, depending on the hydrophone array geometry we choose.
There are sixteen individually addressable result registers, which store the conversion value. The digital value of the input analog voltage is derived by:
- Digital Value = 0 when input <= 0
- Digital Value = (4096 x Input Voltage)/3 when 0V < input > 3V
- Digital Value = 4096 when input >= 3V
The conversion (or sample) rate of the A/D converter can be programmed, and can be as fast as 80 Ns (nanoseconds). Obviously, the faster our sample rates the more accurate our phase delay measurements (between hydrophones) will be. The downside of a fast sample rate is that there will be more data to analyze. For this discussion I will assume a worst case (as far as DSP loading is concerned) of 80 Ns. The number of samples to be analyzed per two-second-ping burst period, assuming the worst case of a (4) hydrophone array, and a 1.3 Ms ping burst interval can be calculated as follows.
(1.3 Ms/80 Ns) x 4 Channels = 65,000 samples / 2 seconds
Since our EzDSP board only contains 64K x 16 of off-chip Static RAM ( SRAM) memory it can be seen that an 80Ns sample rate for (4) channels can not be accomplished. We will therefore do our analysis assuming a sample rate of 160 Ns.
(1.3 Ms/160 Ns) x 4 Channels = 32,500 samples / 2 seconds.
In a 2 second time period our 150 MHz DSP is capable of executing approximately 300 Million single cycle instructions, which would equate to:
300,000,000/35,000 = 8597 instructions available to analyze each sample.
The phase detection accuracy we can expect assume a worst case ping frequency of 30 KHz (33 Usec period) is:
160 Ns / 33 U sec = .00000485 %
