Ideas

I was thinking about another instrument for audio - an accurate differential dB meter. This might be helpful in making gain or loss measurements. Also, measuring amplifiers with balanced outputs or other non-ground referenced signals.

Old-school VU meters use a very simple circuit with full-wave rectifiers to drive the meter mechanism. They depend on the non-linear behavior of the diodes to create a logarithmic response, but this is not very accurate. Of course, a small VU moving needle meter is not expected to be very accurate but they were fine for things like analog tape decks.

Professional-grade instruments are very pricey since they are so specialized. 

Several functions are required: an input attenuator (perhaps with auto-ranging), a differential amplifier (with adjustable gain), an RMS converter for AC signals, a log converter to get dB readouts, and a display of some sort. The display could be as simple as 7 segment digits or something fancier like an LCD readout (and maybe a touchscreen for controls).

Another useful feature might be to include peak detection.

So I was wondering about a cost-effective way to build one. The heart of the device is the RMS conversion device. Years ago there was a chip from Linear which used thermal techniques to do true RMS conversion. Unfortunately, that chip became obsolete. Some are available on the used market for big bucks.

More recently, I came across the Analog Devices (originally Linear) LT1968. Datasheet is here. This chip has been around for many years. Jim Williams wrote an app note describing various circuits using the devices in the family. It can be used in either single-ended or differential mode, so may provide the differential function out of the box. This may not be the ideal chip for the job, but since it has good bandwidth and low noise may be a good starting point. What is good about it is that it uses a delta-sigma conversion technique so it is accurate for different waveforms. Another possibility is the AD736/AD737, but it uses the "average responding" technique, meaning full-wave rectification, which is only very accurate for sine wave inputs. That may be fine, but it is not great for things like noise measurement. 

There are several log converter ICs on the market, but many appear to be aimed at the RF market. One possible candidate is the TI (previously Burr-Brown) LOG104 chip. Datasheet is here. It operates on current inputs, so would require V/I conversion, but it can compute log ratios as well, making it more versatile. Another possible chip is the LOG114 which appears to be fully differential and can be used for voltage inputs. Most of the applications I have seen for these devices involve photometrics. These devices can be problematic when implementing practical circuits since they need to operate over a wide range, and become inaccurate at the extremes, so that needs to dealt with.

So there are some possibilities for prototyping the core of such a device.

It might be feasible to make the instrument "intelligent" by adding e.g., a Raspberry Pi or Arduino and display. It could then have a display configured in software and provide logging capability for measurements - that would require some extensive coding and may be overkill, but cool.

The attenuator I used in my preamp project had one flaw: the relays are driven by a linear binary code sequence and therefore,  at any given volume step, anywhere from one to 8 relays are changing state. This results from a great deal of clattering on some of the steps. The worst case is that ALL the relays will change on particular steps. There a=occaisionl small glitches in the audio stream when this happens. It is not hugely objectionable, but does not seem like the best engineering approach.

There is an old encoding technique called Gray Code encoding where only one-bit changes with each step. I came across this post in DIY Audio by contributor jpk73 who had the same idea and simulated it here. The binary to Gray Code conversion is fairly simple - it is done with a set of exclusive or gates controlling a set of DPDT relays. These should be latching, as is the case with my existing setup.

There is an older patent, now expired, covering this idea as well.

It would be much nicer to have only one relay changing state as the volume is ramping up or down.

This should be trivial to implement since the conversion is done by just a few gates.

While watching a golf tournament on TV - specifically the Charles Schwab Challenge in Fort Worth TX May 28-31, 2026 - an idea popped into my head.

The Problem:

The Colonial Country Club, where the tournament is held, is located directly across the river from a very large and active railroad yard. There is a consistent screeching noise caused by steel wheels on rail cars coming across the river and getting picked up by the many sensitive microphones scattered all over the golf course. The noise is probably worse on the broadcast than in person.

At professional televised golf events there are typically 150 to 200 microphones distributed on tees and greens, plus roving mic attendants, as well as broadcast towers.

All these are fed back to the control building and mixed in a complex control board; these days computer controlled.

What if the mics could be equipped with adaptive DSP filtering devices to tune out unwanted noises? For example, typically the mics on tees are used to pick up the "thwack" of a drive. This audio has a very distinct spectrum and could easily be filtered in real time using DSP techniques to eliminate all but the sound of the ball strike.

An array of distributed mics could be centrally monitored and controlled with back channels to each DSP-equipped mic location. Such a controller could utilize AI techniques to optimize the settings on each mic as well as the overall audio collection scheme.

There are of course, other applications for intelligent filtering of distributed audio pickup devices.

Here are some examples:

1. Military and Tactical Communications

• The Problem: In active combat zones, soldiers wearing tactical communication headsets must communicate clearly while surrounded by the deafening, high-pitched sounds of armored vehicle tracks, fighter jet engines, or sirens.

• The Application: Shrinking an edge-processing DSP down to an ASIC (Application-Specific Integrated Circuit) housed directly inside a soldier's radio preamp or headset line. The look-ahead transient detector would protect critical voice transients (or the sound of incoming gunfire), while the LMS adaptive filter dynamically erases the shifting mechanical shrieks of nearby heavy machinery.

2. High-Speed Rail and Industrial Maintenance

• The Problem: Rail operators like Amtrak or freight companies spend millions manually inspecting tracks and train wheels for defects. The deafening squeal of the wheels makes acoustic data collection nearly impossible.

• The Application: Placing these adaptive preamps on microphones mounted underneath test trains. By selectively filtering out the expected, dominant harmonic frequencies of the normal track friction, engineers can use the back channel to expose hidden, low-amplitude acoustic anomalies—such as hairline fractures in the steel rails or failing wheel bearings—before a catastrophic derailment happens.

3. Smart City Infrastructure and Shot Detection

• The Problem: Systems like ShotSpotter (SoundThinking) use acoustic sensors scattered across cities to detect gunfire. However, urban construction, squealing car brakes, and subway noise frequently cause false positives or mask the actual gunshots.

• The Application: Upgrading municipal sensor nodes with a look-ahead transient bypass. The edge DSP would continuously neutralize the noises of city transit and traffic. When a gunshot occurs, the transient detector immediately flags the acoustic shape, locks the filter, and ships a clean, isolated audio packet back to police dispatch via the cellular/wireless back channel for instant AI verification.

4. Aeromedical and Emergency Services (HEMS)

• The Problem: Flight paramedics and pilots operating inside medical helicopters struggle to communicate over the high-frequency whine of the turbine engines and the heavy thump of the rotor blades. Critically, checking a patient's heartbeat or breath sounds with a stethoscope is impossible in mid-air.

• The Application: Integrating this preamp directly into digital medical equipment and flight intercoms. The central AI can track the known harmonic footprint of the specific helicopter model, push those coefficients to the edge preamps, and erase or attenuate the aviation noise—allowing a doctor to clearly hear a patient's chest sounds while flying at 120 knots.

5. Open-Air Music Festivals and Live Concerts

• The Problem: Outdoor music festivals often struggle ambient crowd noise and nearby stage bleed can pollute a performer's microphone.

• The Application: Deploying these preamps across the festival grounds. The bidirectional network would allow a front-of-house engineer to use central AI to monitor bleed from the EDM stage leaking into the acoustic stage's microphones, dynamically targeting and eliminating the intrusive audio frequencies at the preamp level before it hits the main speakers.

And possibly many more.

Implementation Ideas

An obvious place to start the design is with microphone selection. Recently a class of digital microphones has become popular. These eliminate the need for an A/D converter, since it is built-in. An example is the INMP441 mic. The are really cheap. Here is an example on Amazon. These output an I2S bus signal that can interface to any number of digital processors. The design should also accommodate conventional analog mics. Furthermore, the design needs to allow for multiple local arrays. Directional mics can be positioned in opposed pairs or arrays for noise canceling and sound direction discrimination.

A robust approach to edge DSP implementation would likely be one of the newer TI chips like the AM6X Sitara series. These chips are extremely fast (up to 40 GFLOPS) and feature additional Cortex CPUs for management and communication. They also have a "kitchen sink" of interfaces for various protocols, including digital audio such as SPDIF, IEC60958-1, and AES-3 Formats. The automotive series is designed for extreme temperature limits, so they are suitable for outdoor use.

Prototyping is a challenge, since this chip is packaged on a 484 ball array carrier. Beyond an evaluation board, sophisticated multilayer PCB techniques are needed to build a custom board. However, some suppliers like PCBWay will mount SMD devices for you.

Next, a bi-directional network interface is required for remote digital audio feed and reverse control of filter parameters. The physical link could be wireless, fiber optic, or Ethernet. Each of these has plusses and minuses. One advantage of ethernet is PoE availability, which could power the edge units, but Gigabit Ethernet is limit by cable length to about 600 meters. The other two options require local power supplies. Fiber offers complete immunity to interference, can stretch for miles, and would be secure. Wireless is convenient but subject to interference and distance constraints. An ideal system would accommodate all 3 options.

The combination of remote, local arrays with localized DSP make up what we could call smart acoustic devices.

A central AI system is envisioned for gathering feeds and sending filter parameters back down to distributed smart devices. This AI system should analyze the multiple feeds and make corrections to the overall array field in real time. This is what I would call "Super DSP".