I started a discussion about the sensitivity of
the fdmdv modem to A/D resolution. I have finally gotten
down to testing this and the result is amazing (to me).
I wanted to understand if a mcu with a 12 or 10 bit a/d
converter would be sufficient for a modem.
To test this I have taken a dv encoded file, and running
it through Davids fl_fdmdv GUI program I get the spots
on the scatter map about 1.5 mm as displayed on my laptop
without maximizing the screen. Thus the SNR as judged from
the scatter map is about 50:1, almost 37dB. (1.5 mm in a 80 mm square)
So, I wrote a small snippet of code that did the following:
real[i] = (real[i] + 8) & 0xfff0;
initially with 16, but then with 32 etc all way up to 1024.
Very little expansion of the scatter "spots" occur until I have actually
reduced the resolution from 16 bits down to 6 (!) bits.
My conclusion from this experiment is that reducing the final wordlength
down to this small size does NOT deteriorate the SNR significantly,
so most probably an 8-bit D/A and 8-bit A/D would be sufficient.
Thus it could even be done with an R-2R ladder.
Spot size IS 1.5 mm, indicating that the "math" processing of the signal
although *good* is not perfect, i.e. since I never get a single pixel
the scattering is created by math "imperfections".
So, halving the word size has minimal effect on the result.
I even had to do a od -x | more and inspect the wav file to convince
myself this is actually true.
Regards,
Gullik /SM6FBD
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
