50 years later, how would you improve this.

A DSP cookbook approach won't work. Radar, or ultrasound, deals with a source-poor narrowband environment. If you want a voice from a crowd, your digitizer will have to gain-ride to avoid saturation, and the 'cost' won't be about software, it'll be about the miserable dynamic range imposed by the front end.

Upfront cost of 36 microphones plus 36 20-bit converters. The time-delay part, though, just requires a lot of pointers into a circular buffer/accumulator.

That certainly depends on the exact problem, now doesn't it? Useful array microphones *are* cookbook.

Again, where is your specification.

My specification is replicated in DNA strands, can't really post it on this froup.

As for my (possibly unclear) remarks, I'm familiar with ultrasound pickups with scores of elements, but those are resonant, so give a good signal/noise ratio because the response is allowed to be poor outside the relatively narrow band of interest. There is also repetition of pulses, and some time-gating of the receiver, so any background noises have little chance of getting into the digitized data.

The DSP has a lot of advantages in that case, that don't apply to a constellation-of-microphones in a noisy environment, with wideband input. I'd expect a voice-from-a-crowd scenario would waste dynamic range (require more input resolution than one gets out). In particular, 20-bit converters are rather expensive, and matching a 36-tube array might take 36 ADCs. That might give (after processing) a 12-bit audio stream.

The aluminum tubes are bulky, but so are battery packs for a DSP microphone processor.

I don't really care about your specification but you made sweeping generalizations that just aren't true except, perhaps, in your little world.

Don't need scores to be useful.

Again, it comes down to your specification. There are many applications that aren't at all tough.

I musta missed the specification that included "battery pack" weight.

Recall, though, this is on the subject of 'how would you improve' the 36-tube original.

Same one, 'how would you improve'...

The WebSDR

radio receivers work surprisingly well. They digitize a part of the radio spectrum with 14 or 16 bit ADCs. Multiple digital downconverter units then extracts a user specified section of the spectrum and send the slice over internet to the client.

The best unit in the Netherlands digitize the whole 0-30 MHz spectrum. A few minutes ago, it had more than 100 clients, each listening to separate frequency slice. They had initially problems with a pager transmitter on site, but notching it out before the ADC, the system works quite well. I also tried to listen to obvious intermodulation frequencies, but it was surprisingly quiet.

If you are interested in speech only, say 0-5 kHz and use 16 bit 48 kHz ADC, you are effectively oversampling and get two extra bits in the speech range.

Already a few years ago, you could get dual (stereo) "24 bit" ADCs, with 120 dB SNR measured with 20 kHz bandwidth. If you are only interested in 5 kHz bandwidth, you should expect some improvement in SNR.

How about five Cirrus CS5368 (8 ch/chip). OK, they claim only 114 dB dynamic range, so this is just 19 bits.

Don't use 46 tubes.

I didn't see anything about battery power at all but this thread has been wandering all over so who knows where it's gone.

OK, move it this direction. Highly directional to pick up voice at long distances, or from a window vibrating from a voice. Seems like low noise is key, the electronics can be made low noise. But what mic(s) and/or arrangement would have the least noise. Also what mechanical arrangement, 36 tubes, or 16 tubes going with the Bark Scale Bands

Or a single tube.

Mikek

If random noise is the issue, why not use a parabolic reflector (as in bird recording) to collect as much sound power as possible.

This is not new. I have seen drawing from the 1800's of a woman standing on the balcony in front of a 2 m mildly curved mirror. She is standing at the focal point in front of the mirror, holding a cone against her ear. Apparently the purpose was to spy on neighbors.

I think so. I always questioned the directivity of the tube version. It only amplifies sound to cover up residual noise. Has anyone actually used one ?

Greg

noise from the side is delayed by differing amounts (in the several different tubes) and the received signals cancel, basically it's a wideband phase array.

--
This email has not been checked by half-arsed antivirus software

So, tubes, shotgun mike, parabolic reflector, and phased array are all candidate solutions. One more, I recall the SF Exploratorium used a sonic lens to focus sound...

DSP on big phased array is awkward. It's possible to make a many-element array of (roughly) cardioid capsules, inexpensively, and choose a direction with software... old principle, that, the Germans used Wollenweber antennas seventy-some years ago.

Mathematically, the solution for directional gain will be a linear combination of the microphone inputs, and instead of 2pi steradians (cardioid), will select a smaller solid angle. To select a 0.5m circle at 30m (about right for a single person in an auditorium), is to make a 7200:1 reduction in solid angle. Since the phased array can use DSP to direct the focus in the whole of a hemisphere, that means 7200 different mainly-non-overlapping linear combinations (they're orthogonal) which means a phased array of 7200 different inputs is required.

7200 microphones. That's a LOT for a DSP to handle. Interestingly, one could produce the same time-delay/weight/sum with many small speakers along a transmission line, with a single pickup. That'd be a 1-tube sonic processor, and this tube doesn't have to point alarmingly at the spokesperson...

Scaling the 7200 mics down to a smaller count is fairly easy, if you give up the electronic-aiming-at-multiple-directions. I'm thinking of N pickups, in cylinder or cone array, because symmetry around the axis is simpler than remedial calculation. Now instead of one direction in 7200, the

0.5 meter-at-30 meters spot is the center of 60pi rings; the microphone count need only be 188!

The HAARP program has been resurrected. It has been turned over to the University of Alaska, Fairbanks. Prior news stories said the transmitters wer e solid state. They are tube.

The U.S. Air Force had removed all of the finals, but only have half to turn over to the University. It is scheduled to be back in operation, next year.