I have this connected to analog input of a spark core (similar to arduino) and I am trying to figure out a formula to measure decibels. The formula I'm using must be incorrect because I am not getting readings that correspond with decibel ranges. Can you tell me how you are measuring decibels with your sound detector? FYI, here's the formula I'm using:

(20 * ( log ( voltage1 / voltage 2) ) ) -- voltage 2 is a constant, voltage 1 is the analog read. Thanks!

It is 1.7" by 0.9".

The mounting holes are at (0.1, 0.8) and (1.6, 0.8).

The 5-pin connector is from (1.65, 0.2) to (1.65, 0.6) on 0.1" intervals

Yes, by all means. You can put in a pot, rheostat, decade box, rotary switch with discrete resistors, etc.

If you look at the gain range table with R3 removed, you'll notice that R17 doubles for every ~6 dB added gain - you'll want to use a configuration that matches that resistance increase.

Ideally, a 1 Meg reverse-audio taper pot would be the best fit, and would give you smooth control over the gain range. If you set up an audio taper pot (AKA log taper), and put it in backwards, you'd have similarly smooth control, but you've have to turn it counter-clockwise to increase the gain.

With a linear taper pot, you'd find that most of the gain range was all bunched up at one end of the throw, and only changes a tiny bit over the rest of the rotation.

To do exactly what they're doing in that video (performing audio cross-correlation), you could use the audio output of the sound detector - at that point, it's essentially a fancy microphone, and a plain electret mic would do the same job.

If you're trying to localize a transient sound, you could time the difference between the gate outputs of two sound detectors (perhaps using the input capture feature of a microcontroller) - but that won't work for steady-state sounds, because the gate would always be active.

If you want a voltage that corresponds to the sound amplitude, the envelope pin should do that.

There's an example sketch using an Arduino to read the envelope pin in the hookup guide.

You're probably right about adding resistors. It will help keep loud sounds from overloading the output. If it's possible, you can test out resistors by tack-soldering them in, or even just holding them carefully in place.

The envelope is an approximation of volts peak-to-peak. It's not calibrated to any particular unit or absolute reference. The inexpensive microphone on the front end has fairly wide tolerance on its sensitivity, meaning every unit responds a little differently.

To make a simple relative dB meter, you could read the envelope using the ADC on a microcontroller. Binary encoding actually forms a shorthand for decibel levels - each successive bit position in the conversion value equates to another 6 dB SPL. It's coarse, but not very hard to do.

Yes, that will work. I've seen one of our QC techs do exactly that.

You'll also need to take off R1, the 2.2K resistor, or else you'll be feeding Vcc to your headphone socket.

Since your question is pretty tightly related to this product, this is an OK place to ask. You might also check out our support forums...I've learned a lot through the years just by hanging out on forums, and watching the conversation unfold.

The sound detector should work to detect the buzzer sound - just put it near the buzzer. You may need to adjust the sensitivity, as covered in the hookup guide.

One thing to consider is that it will also pick up other sounds - if the buzzer is the loudest thing in the room, you won't have to worry about false positives from music or the TV or conversation. Just set the gain of the detector to pick up the buzzer.

What you do from there depends on your interests, abilities and comfort level.

You could actually ring a mechanical bell - you can still find doorbells with real bells on them at hardware stores...or if you can't find one, you could make your own with a small bell (google for "desk bell", about $5 at office shops), and rig a solenoid to strike it. Either way, you'll probably need to use a transistor or MOSFET to drive the solenoid, as the sound detector probably can't provide the current needed to drive it directly. Reference circuit 12 in the SparkFun Inventors Kit booklet for a sample transistor circuit - replace the motor with the solenoid.

An alternative would be to do everything purely electronically, but you probably need to generate a tone to drive a speaker. You could use a small microcontroller to generate a PWM wave, or a simple oscillator circuit like a 555, either activated by the gate from the sound detector.

Also, with a microcontroller in the system, you could generate ring or beep patterns when the buzzer sounded.

You'll also need a power source of some sort - batteries or maybe a wall-wart. Personally, I'd also include a switch so it can be turned off. I wouldn't it to beep or ring if I wasn't outside.

Hey - are you trying to build an automatic scoring system for standard 72 inch target frames? I've been pondering building the same thing. Have you gotten very far?

Subbing 80 nF for R17 would give you 19.9 Hz lowpass function. If you really want 1Hz lowpass, 1.6uF will do it.

But you'll also need to scale the hipasses formed by C2 & R2 and C4 & R4. At the moment, they're both at 15 Hz. There isn't really much of a passband with a 15 Hz hipass followed by a 1 Hz lopass.

The filter equations are:

• First stage Hipass Fc = 1/(2piC2*R2) = ~15 Hz
• First stage Lopass Fc = 1/(2piR3*(Cap subbed for R17)) = ~20 Hz with 80 uF cap
• Second stage Hipass Fc = 1/(2piC4*R4) = ~15 Hz

The peak detector is probably better expressed in terms of rise/fall time, than pass frequency.

• Rise time (in seconds) is R8*C1 = 1 millisecond
• Fall time (in seconds) is C1*R9 = 100 milliseconds

R8 could even be 0 Ohms, if you want really fast peak detection. R8 and R9 were selected based mainly on aesthetic behavior. If R9 is too small, you may not see the LED blink, even when the output trips, and if R8 is small, the output becomes fairly glitchy. It acts as a degree of noise reduction.

Ahh, feedback. My old nemesis. We meet again.

Your analysis is accurate - it's the sound coming into the speaker, then back into the mic. From the perspective of feedback, gain is gain. It doesn't really matter which stage applies it...the preamp on the sound detector, the Pandora, or the mono amp. Another decibel louder is another decibel louder.

There are several approaches used to circumvent feedback. They aren't always universally applicable, but each is worth experimenting with, or perhaps trying in tandem.

First, adjust the physical proximity of the speaker and mic, if you can. Break the path that the sound waves are traveling. If you could put the mic inside the helmet, seal it sufficiently, then place the speaker outside (like in Vader's chest pack or belt), connected in such a way that the sound and vibration don't come back to the mic. It's also worth a mention - it might not be sound waves traveling in air causing the feedback, but instead vibration coming physically through the structure. The tighter you connect things together, the more prevalent this phenomenon will be - likewise, looser, more resilient connections will tend to help prevent mechanical transmission. Of course, this approach runs somewhat counter to your thought of using the surface transducers.

Second, we get into filtering. If we can remove the problem frequency, it won't feed back. On large sound systems, 31-band graphic equalizers are used - you find the frequency of the feedback, and turn down the corresponding slider. The EQ is being used as a bank of notch filters. If the feedback is always the same frequency, you might be able to make a passive notch filter to remove that frequency.

Can the Pandora do both your FX and a notch filter at the same time? That would avoid adding any more components.

If not, there's a simple passive notch in Horowitz and Hill's Art Of Electronics, or see what Google comes up with - the twin-T or bridged-T topologies might be worth hunting for. If the frequency is extremely high or extremely low, a simple hipass or lopass filter might remove it.

Filtering can be a little bit tricky. If there's enough loop gain for one frequency to ring, there's probably enough that other frequencies also want to ring...you solve one only to reveal the others. Also, depending on conditions, the frequency might not always be the same. Other factors like humidity, the exact orientation of the speaker and mic, and the acoustics of the room you're in might alter the frequency.

Third, a somewhat more esoteric approach would be to add a bit of delay. Feedback occurs when the sound travels around the loop almost instantaneously. Even a short delay, like 50 ms delay might keep it from happening. You might also find a short delay like that really disorienting and hard to talk through. Worth a try, though.

You might work through various parameters inside the Pandora. If you've got any distortion, guitar amp modeling or compression running, try turning them off. Guitarists sometimes like it when their amp is right at the edge of feedback - it allows them to get long sustain. Compression can make quiet sounds louder, thus more likely to feedback. Also, try adding a noise gate - when you stop talking, it should mute the input, to help avoid feedback.

In prototyping this board I had some real hassles getting it working at 3.3V. Here's what I learned:

1. Make sure that your 3.3V supply is stable. The first supply I tried actually output a 500 mV triangle wave, riding on top of a 3V DC offset. The noisy supply made the whole thing really touchy.
2. Adjusting the gain for a specific input should help you suit your application more precisely. This portion of the hookup guide will walk you through that.

There are a couple operating points that shift as the supply voltage changes. The DC bias on the mic capsule changes, which alters the sensitivity a bit. Also, the trip voltage for the Schmitt trigger is a function of the supply voltage. On a lower supply voltage, the threshold is lower, making it appear more sensitive.

How are you adjusting the supply voltage?

Ideally, the LED brightness control would be independent of the supply to rest of the circuit.

I'm not sure this circuit it especially suitable for an application like that - it might require significant tuning.

First, you'd have to find a microphone element more suited to that frequency range. Many microphones are tailored to human hearing - the response is generally specified as starting at 20 Hz. The specs on the capsule here don't even specify anything below 50 Hz. The matching tolerance from capsule to capsule isn't particularly tight, either. I don't offhand know of a similar infrasonic microphone that would just drop in...googling leads me to very high-spec Bruel and Kjaer capsules.

A first-order lowpass filter isn't especially hard - just put a cap in for R17. Using C = 1/(2pi * Rf * Fc), I get 80 nF, a reasonable value. Of course, this will change if you adjust the gain using the R3 || R17 combination.

A higher order filter probably requires additional amplifier stages.

The circuit also contains a couple of first-order highpass functions. The signal from the microphone is AC-coupled into the preamp stage. C2/R2 form a highpass filter at 16 Hz. The low frequency response would likely need to be extended for your application. The second filter is formed by R8/C1, to slow the rise-time of the peak detector a tiny bit - to make things more responsive, reduce R8.

I take it that the projectiles make a measurable sound as they pass, and that the trajectory is predictable enough that they'll pass through the areas covered by the dishes?

We usually leave reference designators on the PCBs in layers that don't get printed in the silk.

The label on the SJ is an example of one reason we don't: the boards are fabricated and populated as a panel with many boards. Each instance on the panel takes the next set of unique sequential numbers. "SJ11" tells us it was the 11th board on the panel.

The label on the SJ has been fixed. The next batch of boards will all read as "SJ1."

The filtering is definitely an interesting idea, and I've been kicking a state-variable filter around on the workbench a bit. I'll keep it in mind!

Yes. They both use the same capsule, and the preamplifier circuit of the Sound Detector is similar to the Electret Breakout. If you only used the audio output of the Sound Detector, it's similar to the electret breakout, the major difference being the gain.

I'd recommend a buffer in the application, perhaps with some additional gain.

Powering the circuit from 5.2V, and driving the mic until the audio output starts to clip, the envelope output swings from 0V to short peaks of maybe 3.25V. With a lower supply voltage, it won't swing quite so far.

There are several factors that make the violet LED a bit problematic.

• The envelope output is protected from short circuits with a 100 Ohm resistor. If it could hit the 5.2V ral, it could potentially source 52 mA, but...
• ...the LMV324 has a stated output sourcing current limit of about 34 mA into a dead short, constraining the current even further.
• The Violet LED's forward voltage is pretty high. I just measured one at about 3.2VDC.

So the output barely rises high enough to overcome the forward voltage, and when it does, there's only a trickle of current available.

Knowing that things were current limited like that, I just tried that violet LED to see what would happen. It flickers a tiny bit in response to the loudest sounds. In comparison, a plain red LED (Vfwd of around 1.8V) lights right up.

Are you hoping to sample at 8 kHz, or do you want the FFT to resolve frequencies including 8 kHz?

The Arduino documentation states (I believe they mean this for the AVR-based boards)

It takes about 100 microseconds (0.0001 s) to read an analog input, so the maximum reading rate is about 10,000 times a second.

So you could sample at 8 kHz. however, to resolve 8 kHz, Nyquist's Theorem tells us you need to be sampling at more than double the highest frequency that you're trying to capture. The ADC isn't fast enough.

That also assumes that there is enough horsepower to run the FFT - they're computationally intensive. You'd probably have to do some work to allow the ADC to trigger a conversion-complete interrupt, so the FFT can run in parallel with the conversion. The Arduino library busy waits while the conversion is running.

An Arduino Due is probably a better candidate - it can sample more quickly, and the processor is much faster.

The thinking on the gain staging was as follows: if the output from the preamp stage swung from rail to rail (0V to 5V), the rectifier then chops it in half. So the rectifier has a gain of 2.2, to restore the missing half. So rail-to-rail stays rail-to-rail there.

To use it with an external audio source, the recipe might be as follows

• Remove the mic capsule.
• Remove R1.
• Inject the new signal at the positive pad where the mic was.
• Populate R17 for the desired overall gain. If you have a source that's already in the 0V-5V range already, 1K will set the preamp at unity (there's a spreadsheet in the Git repo that you can use to calculate for other values of R17).

The opamp doesn't lock up when it clips, but it's not good practice to overdrive the input too significantly. You can protect the input by configuring a pair of diodes as a clamp.

The attack time is determined by R8. We want the circuit to be responsive, so it's pretty small at 100 Ohms. If it gets much bigger, it completely misses short bursts. The decay time is set by R9.

A number of guitar envelope followers are similar to this circuit - track down the Mutron 3, for instance.

Hmmm...not an application that I anticipated, but I can see how it might be useful.

If you really do mean 125 Hz, I'd say it's worth a try. I have zero experience in that domain, but if you can induce a voltage into the first stage, this will work from roughly 16Hz to 12 kHz. If you want to extend the LF response, bump up the value of C2...you could even solder a "piggyback" cap right onto C2

However, if that's a typo for 125 kHz, this isn't going to work so well. The Amplifier has a rated gain-bandwidth-product of 1.2 MHz. When we ask it to do 100X gain, the bandwidth drops to 12 kHz, well below 125 kHz.

we don't make these without the mic, but it adds a negligible cost, so you could just desolder it and attach your own wires.