only do-able with digital filtering?

[sliders_alpha]

hi

I want to embed 16 signal into a song, those signal are 100Hz apart from each other and start at 18KHz.

Each one is used to control led projector, and you cannot hear them
But how to retrieve them?

I was like, yeah, don't worry, I'll use some narrow filter.
But the thing is, it's not narrow enough.

I then found this extremely narrow filter (3Hz passband): http://www.alg.myzen.co.uk/radio/qrp/af_filter.htm
But I can't tune it for frequency higher than 5KHz.

So here I am, I never used a digital filter, I only know their name.
I did some Laplace transform and Z transform in school, and i still do, but I'm not really good at Z-one.

I could use those PIC:

PIC32MX795F512L-80I/PF
http://ww1.microchip.com/downloads/en/DeviceDoc/DS-61156B.pdf

DSPIC33FJ32GS610-I/PT
http://ww1.microchip.com/downloads/en/DeviceDoc/70591C.pdf

So, must I use digital filtering? Any good link? (I've found a book, but it's more about math than actual filter).

Thanks
_________________
yup, i know, i don't speak english very well

CCS V4.057

[FvM]

You should use digital filters, if the signal is already in the digital domain, otherwise analog ones, I think. Tone decoders like LM567 are another way to detect control signal with simple on-off keying.

A dsPIC or PIC32 should be able to handle e.g. 44 or 48 kHz sample rate required to reproduce the tones and a basic filter algorithm.

B.T.W., did you check if your sound recorder respectively player reproduces above 18 kHz signals with sufficient level?

[Ttelmah]

First, 18K, is possibly a bit low.
Though you can't here it, you may well find some teenagers, can. I used to always 'annoy' people, by knowing when a particular bit of kit started working, by hearing when the monitor timebase switched up from just over 18KHz. If the tones are at all frequent, you will find some youngsters getting really annoyed. Also you need to be very careful in choosing tones round here, since a surprising amount of kit accidentally produces tones in the area. A bit of testing may be needed here....
Now, that being said, the easiest 'hardware' way to 'identify' tones (rather than 'filter' them), is a PLL, assuming that you send a 'key' tone first to lock the PLL (like the colour burst on a TV system). Using a DSP, is also a good way to go, but will require a lot more code, but be cheaper in a production designs. In a DSP, the Goertzel algorithm, would be the 'choice' for speed of response in such detection, involving less maths than a full Fourier analysis. Do a search for DSP DTMF decoders, for which a lot of code has been published, and which you should be able to scale to the higher frequencies, provided you can get the sample rates high enough. This is going to be a significant amount of code though, and remember that the tighter the tones are together, the slower the response time. You could actually send more data, with less problems, and with easier decoding, by sending a continuous tone, and just modulating this, using NRZ encoding, then at the receiver, just have a PLL locked onto the tone, and feed the voltage controlling the PLL, into a comparator. It'll go up when the tone increased, and down when it drops, reflecting the phase changes in the incoming stream...

Best Wishes

[sliders_alpha]

[SherpaDoug]

I would tend toward the second method, maybe 3 or 4 bits per light would give plenty of luminosity levels. The time slots will be MUCH easier to troubleshoot with a simple scope.

Or consider the signal used by radio control models. Each channel is encoded as a 1 to 2 ms long pulse, the first pulse for the first servo, the second pulse for the second servo, etc. If the pulses were 18kHz bursts instead, it might work for you.
_________________
The search for better is endless. Instead simply find very good and get the job done.