Solved: Speech 2

[neochrome32]

this is meant to sing daisy daisy.

it doesn't sound QUITE right// the PWM appears to be correct. as the samples come out perfect.. but for some off reason...

it sounds lame

eg.

int8 a, b;
uint16_t c = a + (b*2)/a;

this seems to be screwing up..

CPU, PIC18F46k22, running 64Mhz

[Ttelmah]

Don't set variables, except to things that are constant, in the declaration line.

For example:

[neochrome32]

TTelmah,
Thank you again!

code update;

im still not quite clear on the "Shortened" part tho?

i deftyped all to signed, and specified uint16_t
with things

the bit that worries me is the initialiser

a = b = c;
sounds like i have to
a = b = 5; // forexample

if thats the case.. it seems to work on SOME parts and not others

the NOISE part still doesnt get heard, and the T sounds dont work still...

i found "ROM" and CONST pretty much do the same, for speed interest i am using const, but will change to ROM later (Speed meaning, upload time) lol

it sounds much clearer tho...

i'll upload a mp3

http://www.electronscape.co.uk/daisy_2.mp3

[Ttelmah]

You are still ignoring what I'm saying. First still using & where you shouldn't - look at the original code.
Then still using the read byte functions. These are not needed if you use rom. The avr, like the PIC needs different code to access things in the program memory, but CCS can do this for you. Look at the example I posted.

a=b=c; is perfectly legitimate C, and works fine in CCS.

Where there are problems, is the assumption that variables will contain something when they are declared. On things like the PC, RAM is zeroed at boot. You have variables being set to variables that are not initialised. If you are going to do this (bad practice anyway...), then add #zero_ram to the code so the values will be zeroed.

Then the other killer is the interrupt frequency. Every 256 instructions. It takes over 60 instructions on the PIC just to get into and out of an interrupt. Each program memory read takes a dozen instructions. Add the other stuff being done in the interrupt, and it is doubtful that it can be done in time.

[neochrome32]

Ttehmah thanks again..

i thought read everything didn't think i was ignoring anything! ok im gonna go through your example (i've noticed you turned the noiseMod back to an uint8...

#zero_ram is in my main.h, will this protocol happen on any NEW variable declared AFTER this?

[neochrome32]

WHAT THE

am i missing something else???

[neochrome32]

seems that i got it working, but again its not a great sounding system... im gonna have a crack at making my own speech engine instead... converting code appears to have taken more time than it would have if i made one from scratch....

ALSO the speech chip 1: has some interesting ways to do things so im gonna use that phoneme system.


Wish me luck, when done i'll be posting the code here for everyone to either improve or just have :D

[neochrome32]

for those interested: i've done the speech chip. its by no means perfect.


http://www.electronscape.co.uk/myprojects/robospeak_v1/index.asp

the source code can be had and a demo sound from here...