[SOLVED] How accurately can a sound frequency be generated or measured?

Ulysses_ · 07-14-2016, 06:03 PM

Need a very accurate sound waveform at 2345.6 Hz. Any noise or distortion leading to harmonics is ok, it is just the baseline frequency that has to be accurate, to less than 1 Hz.

Can a regular sound device be this accurate? How accurate can they be? +/- 0.1 Hz?

Can a regular sound device capture a sound like this coming from another computer, and measure its frequency with an accuracy as high as 0.1 Hz? How accurate can this measurement be?

Ulysses_ · 07-14-2016, 06:13 PM

Alternatively, can a cpu make a LED flash 2345.6 times a second with a frequency accuracy as high as 0.1 Hz or better? Maybe the electrical signal that feeds the LED can then go to an amplifier.

jefro · 07-14-2016, 06:52 PM

Note, try to use edit instead of adding to post. Takes the question out of zero replies.

.1 hz isn't too precise or difficult if the golden or tested value can be used. Your problem is you don't have a calibrated source.

Running an LED may be possible they they don't have sharp cut on and cut off specs, they'd run more like a rlc circuit.

Many musical devices can produce tones and at some frequency, many can also read the frequency. What exactly do you want with this odd frequency?

Emerson · 07-14-2016, 07:13 PM

I just used Audacity to generate 2345.6 Hz tone. Have you tried Audacity?

jpollard · 07-14-2016, 07:30 PM

Accurate sounds can be generated as long as the sampling rate is above the nyquist limit for the given frequency. For 2345.6Hz, that would be around 5000 Hz. Standard audio generators use samples rates around 33KHz, and some go higher.

https://en.wikipedia.org/wiki/Nyquist_frequency

michaelk · 07-14-2016, 07:38 PM

Depends on what you mean by a regular sound device? As stated 0.1 Hz is not that precise by today standards so it should be possible with average equipment. What are you trying to accomplish?

Ulysses_ · 07-14-2016, 07:53 PM

Could sample 128k times a second which is a much better precision but how do we know the sampling is really exactly 128k times a second?

Don't the crystal clocks of the sound cards have a variation? How much is that variation compared to an atomic clock say, at an ntp server?

Ulysses_ · 07-14-2016, 08:01 PM

By the way that frequency is just an example. I want to build a Rife machine out of a laptop and a stereo. And a couple of electrodes made of copper, covered in napkins wet with salt water.

A Rife machine kills specific microbes by passing a current through the body at a specific frequency for each microbe. An accuracy better than 0.25 Hz is given in the specs of some such machines. The most basic machines cost $800.

jefro · 07-14-2016, 08:04 PM

Any computer or musical device could be off. Without any calibration you can't tell. If you need it to be calibrated then you have to take it to a service or compare it to a nationally known calibration source. If this needs to be precise then you'll need to get some reference or a device that has calibration stamp on it.

I kind of doubt you'd need a clock reference that is off one beat every three years. I used to use a clock like that for very precise times.

Ulysses_ · 07-14-2016, 08:13 PM

Surely these things are quantifiable. Like error = 1 second per hour. That would be 1/3600, but I want 1/23456 or better, so not good enough.

Ulysses_ · 07-14-2016, 08:23 PM

Maybe the error can be discovered by capturing one hour's worth of this sound from the microphone of another computer and then writing a program to count the number of peaks in that file.

jefro · 07-14-2016, 08:24 PM

Still no reference.

Still using clock in computer.

What is this for anyway?

I suspect that 90% of the computers that you may have access to are good enough. Computers have fixed many of the older issues.

Ulysses_ · 07-14-2016, 08:27 PM

Rife machine as I wrote above. If the capture is started and stopped based on the cmos clock, then don't we know for sure it lasts one hour? To a reasonably high precision (or the cmos clock would be useless).

Ulysses_ · 07-14-2016, 08:51 PM

Got another idea. Have two computers play the same frequency and capture from a microphone. If the capture file has a constant amplitude on a graph, the frequencies are very close and the error negligible. If they go out of phase every 10 seconds resulting in zero amplitude, the error is 1/10 Hz.

michaelk · 07-14-2016, 09:13 PM

Just been reading up on Rife machines. What are you going to use for an output transducer? Impedance might be a problem and you could destroy your stereo.