300baud FSK over GSM

Started by Denis Gleeson September 29, 2004
Hello All

We are attempting to get 300baud FSK data over a GSM link.

Ive seen different postings on this in the past and the concensus
is that it may be possible given that its FSK and at a slow rate.
Although the GSM codecs distort the phase of audio tones they should 
not effect the frequencies.

Essentially the requirement comes from the fact that the equipment
at the far end can only communicate at 300baud FSK over the PSTN network.
Our end will work with GSM and needs to communicate at the 300 baud FSK.

Has anybody had any experience with this?

Many thanks for any help in advance.

Regards

Denis

_____________________________
http://www.CentronSolutions.com
"Denis Gleeson" <dgleeson-2@utvinternet.com> wrote in message
news:184c35f9.0409290304.1fed30b4@posting.google.com...

> Although the GSM codecs distort the phase of audio tones they should > not effect the frequencies.
I think you are doomed ;-) The GSM Codecs work by building a model of the human "speech machinery", determining the model parameters and the initial stimulus parameter and then send the parameter set across; not the actual data: The voice at the recieving end is merely a simulation of the speaker! The Codecs are speech-only!!
> Essentially the requirement comes from the fact that the equipment > at the far end can only communicate at 300baud FSK over the PSTN network.
Why not use the Modem in the Mobile Terminal? The modem can negotiate with the other end and will do the right thing - om most terminals it will even support the normal AT command set.
"Denis Gleeson" <dgleeson-2@utvinternet.com> wrote in message 
news:184c35f9.0409290304.1fed30b4@posting.google.com...
> Hello All > > We are attempting to get 300baud FSK data over a GSM link. > > Ive seen different postings on this in the past and the concensus > is that it may be possible given that its FSK and at a slow rate. > Although the GSM codecs distort the phase of audio tones they should > not effect the frequencies. > > Essentially the requirement comes from the fact that the equipment > at the far end can only communicate at 300baud FSK over the PSTN network. > Our end will work with GSM and needs to communicate at the 300 baud FSK. > > Has anybody had any experience with this?
If you need to have a 300 baud FSK modem at both ends, using audio, I'd be inclined to simply use a standard 9600 baud GSM modem like those made by Siemens and convert the 300 baud to 9600 baud with an MCU. A small PIC or AVR will do the job. Leon
Frithiof Andreas Jensen wrote:
> The GSM Codecs work by building a model of the human "speech > machinery", determining the model parameters and the initial stimulus > parameter and then send the parameter set across; not the actual > data: The voice at the recieving end is merely a simulation of the > speaker! The Codecs are speech-only!!
I'm no expert but http://www.commsdesign.com/story/showArticle.jhtml?articleID=16501605 suggests that there's a PCM component which attempts to correct the difference between the generated signal and the actual signal is 156 bits per frame (in the standard Full Rate codec), and one frame is 20ms, that means a PCM component of 7.8kb/s. There are of course timing and phase issues that come about through chunking but Denis mentions that he's already aware of those. Leon's suggestion of using a GSM modem for transmission and a converter is probably a better one though. Geoff
"Leon Heller" <leon_heller@hotmail.com> wrote in message 
news:415a9c6e$0$20247$cc9e4d1f@news-text.dial.pipex.com...
> "Denis Gleeson" <dgleeson-2@utvinternet.com> wrote in message > news:184c35f9.0409290304.1fed30b4@posting.google.com... >> Hello All >> >> We are attempting to get 300baud FSK data over a GSM link. >> >> Ive seen different postings on this in the past and the concensus >> is that it may be possible given that its FSK and at a slow rate. >> Although the GSM codecs distort the phase of audio tones they should >> not effect the frequencies. >> >> Essentially the requirement comes from the fact that the equipment >> at the far end can only communicate at 300baud FSK over the PSTN network. >> Our end will work with GSM and needs to communicate at the 300 baud FSK. >> >> Has anybody had any experience with this? > > > If you need to have a 300 baud FSK modem at both ends, using audio, I'd be > inclined to simply use a standard 9600 baud GSM modem like those made by > Siemens and convert the 300 baud to 9600 baud with an MCU. A small PIC or > AVR will do the job
Or assuming the modems are external, just set the baud rate on the serial interface to 300baud and let flow control do the rest.
> > Leon >
"Frithiof Andreas Jensen" <frithiof.jensen@die_spammer_die.ericsson.com> 
wrote in message news:cje5sc$g9b$1@newstree.wise.edt.ericsson.se...
> > "Denis Gleeson" <dgleeson-2@utvinternet.com> wrote in message > news:184c35f9.0409290304.1fed30b4@posting.google.com... > >> Although the GSM codecs distort the phase of audio tones they should >> not effect the frequencies. > > I think you are doomed ;-) > > The GSM Codecs work by building a model of the human "speech machinery", > determining the model parameters and the initial stimulus parameter and > then > send the parameter set across; not the actual data: The voice at the > recieving end is merely a simulation of the speaker! The Codecs are > speech-only!!
It is a pretty good simulation, since one can normally recognise the voice of the caller.
> >> Essentially the requirement comes from the fact that the equipment >> at the far end can only communicate at 300baud FSK over the PSTN network. > > Why not use the Modem in the Mobile Terminal? > > The modem can negotiate with the other end and will do the right thing - > om > most terminals it will even support the normal AT command set. > >
"Denis Gleeson" <dgleeson-2@utvinternet.com> wrote in message 
news:184c35f9.0409290304.1fed30b4@posting.google.com...
> Hello All > > We are attempting to get 300baud FSK data over a GSM link. > > Ive seen different postings on this in the past and the concensus > is that it may be possible given that its FSK and at a slow rate. > Although the GSM codecs distort the phase of audio tones they should > not effect the frequencies. > > Essentially the requirement comes from the fact that the equipment > at the far end can only communicate at 300baud FSK over the PSTN network. > Our end will work with GSM and needs to communicate at the 300 baud FSK. > > Has anybody had any experience with this? > > Many thanks for any help in advance. > > Regards > > Denis > > _____________________________ > http://www.CentronSolutions.com
Is this going to work? You have one advantage in that one end at least is PSTN. FSK works by counting the zero transitions in the detected waveform. There are two frequencies, the clock is abstracted from the higher. Bit stuffing ensures that there is always a full cycle within a set period. At the lower frequency you get half a cycle, but at the higher frequency you get a full cycle of carrier (the numbers may be greater for 300 as apposed to 1200baud) corresponding to the state of a bit. So the question is will the waveform be sufficiently well transported over GSM for the reconstructed waveform at the other end (of the air segment) to be able to count the transitions accurately. In my opinion on a good quality GSM connection to a static GSM set you probably will, but you will probably want to use a suitable protocol to minimise errors: - 1. Data should be packetised with reliable checksums (min 16bit) and sequence numbering. 2. The conversation should be essentially half duplex, even though GSM is full duplex, so that echo in the network does not confuse one station's receiver while it is transmitting. HDLC will probably be quite handy. If your GSM station is mobile, poor quality, distant, or in a an area with reflections and interference then you may be in trouble. As a guide 1200/2400FSK works fine on airband in 5kHz AM bandwidth.
"R. Mark Clayton" <nospamclayton@btinternet.com> wrote in message
news:cjeg5d$15q$2@titan.btinternet.com...

> It is a pretty good simulation, since one can normally recognise the voice > of the caller.
That's because humans are pretty good at recognising voices in difficult circumstances - a computer will have a much harder time. ;-) Still, there *should* be a Modem in the GSM terminal; That Modem will use the "proper methods" which is either to disable the GSM codecs and use PCM instead, possibly using several timeslots, or to "tunnel" the digital data to the PSTN where there will be a device dealing with data comms - i.e. A Modem. Depends on the Mobile terminal and the Network how it is done - "userside" sees a vanilla 9600 bps AT compatible modem. PS: Trying to send the modem sounds trough a GSM 'phone sounds really wierd and hack'ish - like those "phone suction pads" one used in the 70'ies movies whenever there was a hacking run in the script - on GSM voice, Morse might be be the best!
dgleeson-2@utvinternet.com (Denis Gleeson) wrote:

>Hello All > >We are attempting to get 300baud FSK data over a GSM link. > >Ive seen different postings on this in the past and the concensus >is that it may be possible given that its FSK and at a slow rate. >Although the GSM codecs distort the phase of audio tones they should >not effect the frequencies. > >Essentially the requirement comes from the fact that the equipment >at the far end can only communicate at 300baud FSK over the PSTN network. >Our end will work with GSM and needs to communicate at the 300 baud FSK.
Given the earlier thread, this is not a credit card terminal by any chance!? Linus -- Linus Surguy - Magrathea Telecommunications Ltd. Wholesale and retail telephone services. www.magrathea-telecom.co.uk www.uknumber.co.uk www.callthrough.co.uk www.telesave.co.uk: UK 2.5p/1.5/1p South Africa 6p US,France,Germany,Eire 2.5p Looking for VoIP? We will gateway SIP & IAX to and from the PSTN. -----= Posted via Newsfeeds.Com, Uncensored Usenet News =----- http://www.newsfeeds.com - The #1 Newsgroup Service in the World! -----== Over 100,000 Newsgroups - 19 Different Servers! =-----
> whenever there was a hacking run in the script - on GSM voice, Morse might > be be the best!
Actually, maybe speech synthesis and recognition would be the best!