From: Jadran on
Hello,

I m implementing fax/modem detection. So far it is based on CNG and CED
tones. Goertzel's algorithm is used for tones detections and seems to work
fine. However I would like to increase reliability, specialy for cases when
such tones are not present or missed. Idea is to do it by detecting
presence of V.21 FSK sequence. This could be done with solution based on
FSK demodulator. But as my goal is only reliable fax/modem presence
detection, not FSK decoding, I would like to keep my FSK detection code
simple as possible.

Do you have any suggestions how can I improve my fax/modem detection code
based on Goertzel without adding unnecessary complexity.


From: Rune Allnor on
On 20 Des, 20:36, "Jadran" <jsiro...(a)yahoo.com> wrote:
> Hello,
>
> I m implementing fax/modem detection. So far it is based on CNG and CED
> tones. Goertzel's algorithm is used for tones detections and seems to work
> fine. However I would like to increase reliability, specialy for cases when
> such tones are not present or missed. Idea is to do it by detecting
> presence of V.21 FSK sequence. This could be done with solution based on
> FSK demodulator. But as my goal is only reliable fax/modem presence
> detection, not FSK decoding, I would like to keep my FSK detection code
> simple as possible.

To detect a fax, you will have to decode the fax signal.

Only if the signal can be decoded to a valid fax transmission
can you be reasonably certain that it is a fax signal. Of course,
you don't have to decode *all* the signal; a couple of seconds
worth of data ought to suffice, and you don't have to print the
contents of the fax. So in that sense your decoder might be
a bit simpler than the regular decoder stuff.

Rune
From: Vladimir Vassilevsky on


Jadran wrote:

> Hello,
>
> I m implementing fax/modem detection.

You are implementing fax/modem detection to do what?

> So far it is based on CNG and CED
> tones. Goertzel's algorithm is used for tones detections and seems to work
> fine. However I would like to increase reliability, specialy for cases when
> such tones are not present or missed. Idea is to do it by detecting
> presence of V.21 FSK sequence. This could be done with solution based on
> FSK demodulator. But as my goal is only reliable fax/modem presence
> detection, not FSK decoding, I would like to keep my FSK detection code
> simple as possible.
>
> Do you have any suggestions how can I improve my fax/modem detection code
> based on Goertzel without adding unnecessary complexity.

Fax/modem signal = wideband noise-like signal with high energy. It is
very distinctive from anything else. You just have to detect such noise
without attempting decoding.

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
From: Vladimir Vassilevsky on


Rune Allnor wrote:

> On 20 Des, 20:36, "Jadran" <jsiro...(a)yahoo.com> wrote:
>
>>I m implementing fax/modem detection.

> To detect a fax, you will have to decode the fax signal.
>

That depends on why would they need fax/modem detection.

BTW, in Russia, they tried to impose special tax on fax/modem owners.
They even created line circuitry which blocked data transmission and
allowed voice communication only. Fortunately, that didn't happen.

VLV
From: steveu on
>
>
>Jadran wrote:
>
>> Hello,
>>
>> I m implementing fax/modem detection.
>
>You are implementing fax/modem detection to do what?
>
>> So far it is based on CNG and CED
>> tones. Goertzel's algorithm is used for tones detections and seems to
work
>> fine. However I would like to increase reliability, specialy for cases
when
>> such tones are not present or missed. Idea is to do it by detecting
>> presence of V.21 FSK sequence. This could be done with solution based
on
>> FSK demodulator. But as my goal is only reliable fax/modem presence
>> detection, not FSK decoding, I would like to keep my FSK detection
code
>> simple as possible.
>>
>> Do you have any suggestions how can I improve my fax/modem detection
code
>> based on Goertzel without adding unnecessary complexity.
>
>Fax/modem signal = wideband noise-like signal with high energy. It is
>very distinctive from anything else. You just have to detect such noise
>without attempting decoding.

Its only wideband once the image modem starts. I assume he is trying to
detect if its a FAX near the beginning of a call for some selection reason.
All he will have at that stage is the initial 1100Hz or 2100Hz tones, or
the V.21 FSK, which isn't awfully wideband.

Demodulating the FSK, and looking for the initial flags of the HDLC data
is a fairly simple thing to do.

Steve