Voice without intellivoice

[artrag]

This is one of the samples that I thought converted pretty well. See if you can improve on it.

Thanks, these are my rom and player with 5 channels (so ECS only, do not work without it).

I fixed a bug, so evaluate these new files.

Press 4 to hear your sample

intyLoFi_5channels.rar

ecsLoFi5.bas

Edited July 7, 2016 by artrag

[freewheel]

Cool. They sound the same to me, maybe I'm crazy. I think I basically did the same thing with the player as you did.

[artrag]

This version is using 6 channels

BAS and ROM included

intyLoFi_6channels.rar

[artrag]

Cool. They sound the same to me, maybe I'm crazy. I think I basically did the same thing with the player as you did.

I think I did an error, let me fix it

I've changed the files in attach to the 5 channel version

 

Basically 5 channels and 6 channels sound very close

I would say that after having added 4/5 channels, the distortion is more due to the square wave and to the phase errors than to the missing tones

Edited July 7, 2016 by artrag

[artrag]

This version of the voice encoder can produce any number of channels

[artrag]

I know it is not intellivision but I'm too proud of it to keep it for myself

This is the first application if my new voice encoder:

 

 

The voice patch to the original rom is only 8KB and works on a dedicated scc audio chip (5 channels).

Needless to say, the original msx version of Salamander has no voice, the speech we added is borrowed from the arcade version.

Edited July 10, 2016 by artrag

[artrag]

 

I was thinking about saving 1/2 of the room taken by the data by packing the volumes in the top 4 bits of the periods.

Is there a good way to extract the 4 top bits and shift them in the lowest 4 bits in intybasic?

[+nanochess]

I was thinking about saving 1/2 of the room taken by the data by packing the volumes in the top 4 bits of the periods.

Is there a good way to extract the 4 top bits and shift them in the lowest 4 bits in intybasic?

I would suggest:

output = input / 16

IntyBASIC optimizes most power of 2 division operators.

[artrag]

In this case I should do

 

output = input /4096

 

Is it optimized as well?

[+nanochess]

In this case I should do

 

output = input /4096

 

Is it optimized as well?

Not that constant but try "output = input / 256 / 16"

[artrag]

Is there a way to allow a rom array of varptr x(0) where x is a rom array as well?

If I do:

Data varptr x(0)

I get an error even if x() is an array of constant values in rom

 

This could simplify the huge

on (nsfx) goto ...

I use now in the player

I would do

#frames = fsxpntr(nsfx)

with

fsxpntr:

Data varptr voice0(0), varptr voice1(0)... etc

Edited July 11, 2016 by artrag

[+nanochess]

Is there a way to allow a rom array of varptr x(0) where x is a rom array as well?

If I do:

Data varptr x(0)

I get an error even if x() is an array of constant values in rom

 

This could simplify the huge

on (nsfx) goto ...

I use now in the player

I would do

#frames = fsxpntr(nsfx)

with

fsxpntr:

Data varptr voice0(0), varptr voice1(0)... etc

This is a very much requested feature, I'll try to put it into next version of IntyBASIC

[artrag]

If the things are in the right order, the change would probably consist of adding an extra conditional test where now you generate the error.

Edited July 11, 2016 by artrag

[artrag]

In this version the data are packed volume+period, so now a channel takes a single word!

 

In intybasic terms we have:

 

DATA &VVVVPPPPPPPPPPPP,&VVVVPPPPPPPPPPPP,&VVVVPPPPPPPPPPPP, etc etc

 

In attach the results for the ECS demo, where we use 6 words per frame, i.e. 360 bytes per second

The ROM size passed from 31KB to about 18KB

Not bad considered the context.

 

 

 

PS

Oscar this is the result of the compiler for the /256/16 thing

	;[131] 			Sound 5,peek(#frames+0),(peek(#frames+0)/256)/16
	SRCFILE "C:\Users\Ragozzini\Documents\IntyBASIC SDK\Projects\intyLoFi\ecsLoFi_compact.bas",131
	MVI@ R1,R0
	MVO R0,240
	SWAP R0
	MVO R0,244
	MVI@ R1,R0
	SWAP R0
	ANDI #255,R0
	SLR R0,2
	SLR R0,2
	NOP
	MVO R0,251

Is that the optimization you was referring to ?

intyLoFi_compact.rar

Edited July 12, 2016 by artrag

[artrag]

The encoder able to generate intybasic DATA files is this one

 

Just take the files XXXX.wav_psg_cmpct.bas from the subdirectory \wavs that will be generated.

 

In order to run the voicenc.exe you need Matlab 2012 or its free runtime libraries (read the readme.txt to install them)

[+nanochess]

In this version the data are packed volume+period, so now a channel takes a single word!

 

In intybasic terms we have:

 

DATA &VVVVPPPPPPPPPPPP,&VVVVPPPPPPPPPPPP,&VVVVPPPPPPPPPPPP, etc etc

 

In attach the results for the ECS demo, where we use 6 words per frame, i.e. 360 bytes per second

The ROM size passed from 31KB to about 18KB

Not bad considered the context.

 

 

 

PS

Oscar this is the result of the compiler for the /256/16 thing

	;[131] 			Sound 5,peek(#frames+0),(peek(#frames+0)/256)/16
	SRCFILE "C:\Users\Ragozzini\Documents\IntyBASIC SDK\Projects\intyLoFi\ecsLoFi_compact.bas",131
	MVI@ R1,R0
	MVO R0,240
	SWAP R0
	MVO R0,244
	MVI@ R1,R0
	SWAP R0
	ANDI #255,R0
	SLR R0,2
	SLR R0,2
	NOP
	MVO R0,251

Is that the optimization you was referring to ?

 

That's right my friend