wireless voice using FSK

[silk]

I thinking of building a wireless link for audio (just like a wireless
phone), using FSK on the ISM band (using one of the many transceivers
that are on the market)
Can anyone give me some guidance as to what data rate would I need for
voice communication (full duplex or not).
Also, I would like some guidance on how to do it: what vocoder to use,
if I should use PWM, etc.

Any help will be appreciated.

 

[Joseph Goldburg]

WHY!

Digital FSK example
-----------------------
Telephone voice band 300Hz to 3.4KHz ... so let say 4KHz for speech
You must sample a twice the highest freq - (Niquist limit) Thus 8KHz
sampling rate.
lets use an 8bit ADC thus each sample is 8 bits thus the tranmission rate
is 8bits x 8KHz = 64Kbits per second
Thus you need ADC, moduator + Tx then Rx demodulator then DAC all at
64Kbits per second

Thus the bandwidth of Tx and Rx must accomodate 64Kbits per second.
(I think the wide bandwidth filters are expensive and hard to get)

** This system takes up a lot of bandwidth just for speech!

** And long range the signal will drop off sharply becuase of the digital
nature (1 or 0)

Standard Narrow band FM exmaple
---------------------------------------------
modulator+Tx then Rx demodulator
Bandwidth only needs to accomodate speech of 4KHz not 64Kbits per sec

No ADC or DAC needed


JG

 

[John Fields]

I thinking of building a wireless link for audio (just like a wireless
phone), using FSK on the ISM band (using one of the many transceivers
that are on the market)
Can anyone give me some guidance as to what data rate would I need for
voice communication (full duplex or not).
Also, I would like some guidance on how to do it: what vocoder to use,
if I should use PWM, etc.

---
Check this out:

http://wireless.fcc.gov/rules.html


and, in particular:

http://a257.g.akamaitech.net/7/257/...ss.gpo.gov/cfr_2002/octqtr/pdf/47cfr15.13.pdf

 

[Joel Kolstad]

WHY!

As a building block to something Really Cool? It's true of course that
straight digitization and binary encoding is arguably worse from a
reliability and bandwidth usage point of view than FM, but on the other
hand, you need to start at some low level like this to start building
fancier digital communication schemes. For instance, CDMA telephones use
something like 5-10x less bandwidth per user than the old AMPs system (which
was narrowband FM) did while simultaneously providing much better signal
quality, on average.

I imagine there are also a lot more jobs out there these days designing
digital communication systems than FM systems, BTW.

As for the guy directing him to the FCC specs... I don't advocate breaking
government regulations, but assuming this guy is starting with, e.g., 100mW
output through a random length wire antenna, it's a safe bet that the storm
troopers won't be knocking down his door any time soon. People get overly
concerned with this at times -- I'd bet a nickle that you couldn't show me a
good RF designer out there who hasn't built a transmitter at some point that
violated some FCC regulation or other.

---Joel Kolstad

 

[Active8]

WHY!

Digital FSK example
-----------------------
Telephone voice band 300Hz to 3.4KHz ... so let say 4KHz for speech
You must sample a twice the highest freq - (Niquist limit) Thus 8KHz
sampling rate.
lets use an 8bit ADC thus each sample is 8 bits thus the tranmission rate
is 8bits x 8KHz = 64Kbits per second
Thus you need ADC, moduator + Tx then Rx demodulator then DAC all at
64Kbits per second

But if you use a Delta-Sigma Codec you can get by with 8kbps. try

http://www.aerocomm.com/

Not a lot of range, but they're FCC approved.

you'll find others if you search. I called these guys. Don't let
the specs fool you. They may say 800mbps data rate but the commo
protocol eats up a lot of it leaving you with 100kbps IIRC from my
phone discussion.

 

[silk]

I thinking of building a wireless link for audio (just like a wireless
phone), using FSK on the ISM band (using one of the many transceivers
that are on the market)
Can anyone give me some guidance as to what data rate would I need for
voice communication (full duplex or not).
Also, I would like some guidance on how to do it: what vocoder to use,
if I should use PWM, etc.

Any help will be appreciated.

Thanks for all you help guys.
I am just trying to play with it for my own fun.
I have no problem with finding the RF parts (FSK transceiver in this
case). The problem is with the actual audio interface. I am trying to
see how feasible it will be to transmit audio or even higher quality
audio. With most cheap FSK transceivers are about up to 200kbps it
gives me a lot of head room for voice but limits my sampling spectrum
for audio.

I guess I am not clear whether or not I will have to use a DSP or how
to sample the audio and move it in a serial data to a transmitter and
vice versa on the receive side.


Any thoughts?

Thanks.

 

[no_email_addy@no_email_addy.com]

I am just trying to play with it for my own fun.

Yeah, go for it I say.

I have no problem with finding the RF parts (FSK transceiver in this
case). The problem is with the actual audio interface. I am trying to
see how feasible it will be to transmit audio or even higher quality
audio. With most cheap FSK transceivers are about up to 200kbps it
gives me a lot of head room for voice but limits my sampling spectrum
for audio.

Well, a simple audio compression method is 4-bit ADPCM, the quality is very good
- you can try it on your pc simply by recording some audio to a file using the
built in ADPCM codecs (assuming your using a windows OS) and then playing it.

This would give you 32Kbps with a sampling frequency of 8KHz. Theirs little
difference between uncompressed 16-bit samples and 4-bit ADPCM compression at
the same sample rate.

If you have 200Kbps to play with, then you could use a sampling rate of 44Khz -
assuming you only want to tx a single audio channel (mono).

I guess I am not clear whether or not I will have to use a DSP or how
to sample the audio and move it in a serial data to a transmitter and
vice versa on the receive side.

Depends on the sample rate and the complexity of digital compression etc.

A simple ADPCM routine with a packetizer and crc error checking on each packet
would be easily written for say the Atmel mega 8-bit risc series of cpu's (16MHz
clock, single cycle instruction set) - assuming your working with voice
bandwidth. These chips have built in 10-bit ADC's too - which makes sampling
the audio a doddle. You could use one of the PWM outputs as an audio output
DAC, but a proper external DAC is best.

However, a compression scheme like G.723.1 (which gives you a bit rate of around
6Kb/s for normal voice audio) takes a lot of processing - a DSP would certainly
be required here.

Clive