Getting analog input from a microphone and FFT transform.

oxo · Joined: 13 Nov 2012 Posts: 219 Location: France

Yep. I'm guessing some kind of ultrasonic receiver, although 18khz is a little low.

Ttelmah · Joined: 11 Mar 2010 Posts: 19541

Agreed.

Start at the beginning. The microphone is not likely to produce enough output to drive the ADC of the PIC directly, and at the very least needs level shifting. So you are going to need some form of signal processing in front of the PIC.
As OXO says, if this is simple tone detection, then make this a sine to square converter, and use te PIC CCP, to measure the time between edges.

However PLL's are not expensive. Every medium to fast PIC, produced for the last few years, has one inbuilt for it's master clock. The point about the PLL, is it directly decodes the frequency changes to a voltage, which directly reflects the frequency shift. 100* easier to work with. The cost of the PLL, is tiny compared to the cost of the amount of RAM needed for DSP decoding of the direct data stream, and is do-able, rather than probably being close to impossible. A standard TV these days probably contains a dozen PLL's. FM radio's all use them, etc. etc.. A basic PLL frequency decoder, is a few cents. An LM567 for example is about $0.1, in small quantities, and if you monitor the voltage on pin 1, instead of just looking at the square wave output, you have a direct indication of the frequency changes in your signal. Buy these in bulk, and they are about $0.01!....

Now assuming you need to actually do something with the decoded stream, this is now at a much lower frequency, so can be handled by a much more basic PIC. You have saved more than the cost of the PLL...

The DSP idea, is definately in 'sledgehammer/nut' territory, and has potential costs to go with this.

Best Wishes

ranii · Joined: 11 Sep 2013 Posts: 16

zero crossing is for children (no offend but thats true .)
When i said one signal at a time i mean 1 main signal , if there will be some noise-the zero crossing is worth nothing. an fft is a must at the mathematical aspect. or some kind of auto-corrolation function if you know them .

Its not ultrasonic but a part of a more sophisticated technology .

For the one who said dtmf is at low f's , thats true, but not everything has to be like what you already know. dtmf and fsk are more likely the same, and frequencies can be changed as you wish .

So ,Ttelmah you talked about pll, do you have one in mind for these f's that is not expensive ? because what i have seen is not cheap for mass .
(of course there will be pre amp-i wrote that in my first post )

ranii · Joined: 11 Sep 2013 Posts: 16

lm567 by the way, does not have 200hz resolution and can't be tuned a head for a few frequencies . (i must be tuned to f1 f2 f3 f4.. than looking for each to come) .

ranii · Joined: 11 Sep 2013 Posts: 16

the internal pll - i did not heard about that , so if its exist- why people still performing an fft, or at least trying to do that ?

and how the internal pll has the enough space for all the samples ? something is missing to me with that ..

oxo · Joined: 13 Nov 2012 Posts: 219 Location: France

So make your own pll. 18 kHz is not very high, and you might learn something about electronics.

Ttelmah · Joined: 11 Mar 2010 Posts: 19541

ckielstra · Joined: 18 Mar 2004 Posts: 3680 Location: The Netherlands

matheuslps · Joined: 29 Sep 2010 Posts: 73 Location: Brazil

hum.....

In my routine of studies, I generally read a lot of forums to learn something new...

Normally I do not enter someone's topic and start my own question but since this is a recent topic and I do not know why and how I found this topic but the subject discussed here is interesting.

If I understand right, according to some members on the latest replies, I can use a PLL chip like the CD4046 (74HC4046) to measure frequency?

Example: Is it possible to insert a sound signal, after I shift it up to make it "monopole" (ranging around 2.5V for example), in the PLL chip and measure the frequency(ies)?

I was not searching about this subject but this can be so interesting that yesterday I take a couple of hours to study how a PLL chip works.

I always tried to learn a simple way to measure the frequencies from a sound file in real life but always ended on the FFT field. I said real life because I can use MatLab to do this, but this is not an embedded way to this. Always the same: DsPic.... float point arithmetic..... ect.... And even tried to follow the FFT example of CCS with a DsPIC33.....

For now, the best and simpler way that I found to accomplish this task is using a CI called MSGEQ7:

https://www.sparkfun.com/products/10468

Datasheet: https://www.sparkfun.com/datasheets/Components/General/MSGEQ7.pdf

So, more ideas about this?