Hi,
We are two weeks since the last software-update of the gmskmodem; so ...
I just pushed a new version onto github:
Changes:
- some additional error-checking for ALSA
- There is now an option "-id" which enables the modem to play out a
wav-file after every "n" transmissions. The reason is that all
transmissions of hams have to be "in the clear" and hams have to give
their callsign every "x" overs.
As we are dealing with developing a new protocol -which cannot be
understood by other hams- it might be usefull to add a small
audio-fragment with your callsign in plain audio or CW after a
transmission so that people are able to identify the source of "this
strange broadcast".
- I've also changed the Makefile so that the code can be compiled with
floating-points instead of integer operations.
This may sound a bit strange; but this is the result of helping out some
people to port the "WSPR" application to the raspbery pie.
It turns out that -on some devices, like the RasPi- calculating a filter
in DSP is faster when using floating-point operations. The ARM1176 in
the Pie seams needs more CPU cycles to do a 32 bit * 32 bit integer
operations; then doing the calculation using the onboard VFP unit.
By modifying a flag in the Makefile, one can now select if the code is
compiled using integer or floating-point math.
As usual, the code can be found at github:
https://github.com/on1arf/gmsk/tree/master/gmskmodem_codec2
Planned for the next releases:
- implement the header. (who knows where I can find code for
implementing golay FEC encoding)?
- do some further experimenting with optimizing the integer math.
After doing some research on the assembler-code generated by gcc when
doing code-optimalisation of the integer math; it turns out that on a
ARM9 and a ARM11, every 32 * 16 multiplication in C is actually
translated into three multiplications on assembler level: two "MUL"
(multiply) and one "MLA" (Multiply-and-add) instructions.
As some devices (like almost all D-STAR gmskmodems out there) actually
operate with audio using only 10 bits (instead of 16), I want to
experiment if it not possible to reduce the resolution of the audio and
of the filter FIR coefficients; so that this 16*32 bit multiplication
will be limited to an instruction that fits in 32 bit; hopefully
reducing the CPU-power needed on certain devices without a FPU.
73
Kristoff - ON1ARF
------------------------------------------------------------------------------
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
