I should've opened my sx353 earlier... found a "SAM". And that's what I found about SAM:
Siemens uses the "PSB 2168" IC for the answering machine part of the devices since quite some time. It is also called "SAM" (Sophisticated Answering Machine).
The manual for the SAM (documents are online at Infineon site) describes it in such detail that one can connect, configure and use it, but audio compression is left unexplained as a black box named "DigiTape(tm)":
"10.3 kBit/s, 5.6 kBit/s and 3.3 kBit/s are supported. [...] The coder can optionally use silence gap coding."
A document mentions Windows tools for creation (and encoding?) of prompt & phrase files for the answering machine, but there's nothing available for free download.
Given the high quality at such low bit rates, it probably uses a linear prediction codec (CELP or more advanced). But 10.3 kbit/s is no standard rate... And that now is where I have to resign with my decoding experiments - because for that kind of algorithms it's not so easy to reconstruct the exact math and its parameters from just the data. It would be easier just to buy a SAM (or take one from a broken Siemens phone), add some RAM and simple microcontroller with PC interface, and attach this as an external codec unit to the PC - that's actually feasible! 8)
But maybe we're lucky and someone finds a paper online describing some CELP derivative, producing bit streams with 10.3 kbit/s? Or knows someone who knows someone who knows someone who (co-)developed "DigiTape(tm)"?
Good luck;
Kolja
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I should've opened my sx353 earlier... found a "SAM". And that's what I found about SAM:
Siemens uses the "PSB 2168" IC for the answering machine part of the devices since quite some time. It is also called "SAM" (Sophisticated Answering Machine).
The manual for the SAM (documents are online at Infineon site) describes it in such detail that one can connect, configure and use it, but audio compression is left unexplained as a black box named "DigiTape(tm)":
"10.3 kBit/s, 5.6 kBit/s and 3.3 kBit/s are supported. [...] The coder can optionally use silence gap coding."
A document mentions Windows tools for creation (and encoding?) of prompt & phrase files for the answering machine, but there's nothing available for free download.
Given the high quality at such low bit rates, it probably uses a linear prediction codec (CELP or more advanced). But 10.3 kbit/s is no standard rate... And that now is where I have to resign with my decoding experiments - because for that kind of algorithms it's not so easy to reconstruct the exact math and its parameters from just the data. It would be easier just to buy a SAM (or take one from a broken Siemens phone), add some RAM and simple microcontroller with PC interface, and attach this as an external codec unit to the PC - that's actually feasible! 8)
But maybe we're lucky and someone finds a paper online describing some CELP derivative, producing bit streams with 10.3 kbit/s? Or knows someone who knows someone who knows someone who (co-)developed "DigiTape(tm)"?
Good luck;
Kolja
Thanks a lot Kolja for the effort you put into that. Maybe it will provide a starting point for someone else to pursue this further.
Best regards
Tilman