Oscillator and audio harmonic filtering...

Better to use a sinewave generating IC. Intersil 8038 as I recall; maybe some others available today.

Luhan

You dont need a notch filter just a low pass RC filter. If you can ensure you have a square wave output the even harmonics are supressed anyway.

Google on "Wien Bridge Oscillator". Built right one of those would have very low harmonics so you could really pack the tones in, and a diode-stabilized version, while not having the very lowest harmonics, would be smaller than the harmonic filter you'd need to put after a 555.

You either need to generate a sine wave to start with or use the output of your square wave oscillator and filter it right at the source. Either with LC or with an active filter around opamps. I don't think a microphone/speaker combo will cut it, plus it would likely consume more space than a real filter and pick up surrounding audio noises.

Using square waves for that is one of the best ways to make your life harder unnecessarily. I would use sine waves generated and added digitally (with an MCU or a DSP). You said "audible" frequencies, so you should be able to also include some noise shaping, in case you need higher resolution. You may not even need a DAC. Just a PWM output and an external filter could work.

However, I think the whole strategy is bad. If the only information that your tones carry is their frequency, why don't you just store, for each tone, a value that is proportional to its frequency? Your file will be much smaller.

What do your sensors provide? A variable voltage? A wave with variable frequency? In the first case, you should digitize the voltage and store it. In the second case, you should measure the frequency (easy with an MCU) and store that measurement. Forget about storing the whole wave.

Best,

Hello Tim,

Such as this one here:

Take a look on XR2206 , generate a sinewave with low distortion.

Eduardo.

J.Shrum wrote:

Why not periodically sample the signals and perform an FFT on the samples? You could then just store the single number (for each "track") that represents the frequency of the fundamental.

With a square wave, the first harmonic is the third. Depending on the magnitude of the expected frequency deviation, you might be able to put the frequencies of several sensors in the space before the third harmonic of the lowest frequency one.

-- john

Excellent advice everyone. I'm going to carefully review all of the suggestions and decide what is most feasible.

To describe my exact useage: I am using thermistors to vary the tone freq of the 555. I then record that over a period of hours. Then I can go back and view the spectral readout of the wav/mp3, and look for any temperature dips, and spikes. So far its worked very well, ... just gotta get rid of the harmonics so I could monitor more locations w/ the use of the same recorder.

Greg, If I understand you correctly, are you suggesting use a circuit to alternate between each sensor and record a 1-2 second bit of each one? I assume if that was done, I would be able to identify the current sample, because it would be the current lowest frequency in that particular time space. Did I understand this right?

Well, that should work. But the main point I was making was that if you perform on-the-fly FFT's on the samples that you gather from the sensors, you don't have to record all of the sound samples but only the single number that represents the fundamental frequency for each sampling period. You then have many choices of how to arrange the numbers in your data file, including timestamped records or even as a serial stream if, as you say, the individual sensor's range of frequencies do not overlap each other.