Playing a long running sample...

[mcgeezer]

I have (currently) a 2 minute sound byte that I to sample and convert to be played as part of a scene on the A1200.

Now using this site https://www.colincrawley.com/audio-f...ze-calculator/
i have put in a sample rate of 22Khz, 8 bit depth and 1 mono channel and the file size uncompressed comes out at 2.64mb.

Ideally I have about 750Kb available to the sample so I'm looking at the best ways that I can get it to play on the Amiga.

As I know absolutely nothing about Paula I'm looking for suggestions on the best way to achieve this... my understanding is that Paula will only play a sample of upto 64kb in size.

The first obstacle would be compressing the 2.64mb of data so I was going to slice it up into 64kb packed buckets. About half the clip would be loaded and uncrunched in chip ram, once the first samples start playing I can then stream in the from floppy the remaining buckets and play those.

Is this the right way to go about this or is there a simpler way? If needed I could lower the quality but i'd like to keep it as clear as possible because it is speech over music so clearer the better.

Any help/advice is appreciated.

Geezer

[roondar]

One small detail: Paula can play up to 128KB in one sample, not 64KB.

As for the method, it may be possible to use simple audio compression forms and still end up real-time. Though AFAIK fast audio compression tends to not be very good in terms of gains (or quality if you use lossy).

[DanScott]

Might want to look at Kalms ADPCM stuff

https://github.com/Kalmalyzer/adpcm-68k

[mcgeezer]

Originally Posted by roondar

One small detail: Paula can play up to 128KB in one sample, not 64KB.

.

Nice one Roondar, thank's for pointing that out

Originally Posted by DanScott

Might want to look at Kalms ADPCM stuff

https://github.com/Kalmalyzer/adpcm-68k

That looks very good, cheers Dan.

Always best to ask the EAB first on stuff like this.

Geezer

[mcgeezer]

OK so I had a play around with this and I guess it's a bit challenging...

What I'm trying to do is get the audio sample here:
https://www.amigagamedev.com/Downloads/newblack_raw.mp3

which is 2 minutes long to be able to play with decent quality on the Amiga, BUT i need to get the file size down to about 700kb.

So I transcoded the file using sox to both 8Khz and 12Khz...

The 12Khz I would go with but it would mean I have to fit it across two disks, although it would fit in the A1200 chip ram.

the 8Khz version is reduced quality... but still quite large and I don't think they'll compress down to one floppy.

https://www.amigagamedev.com/Downloa...lack_8khz.8svx

https://www.amigagamedev.com/Downloa...ack_12khz.8svx

I tried Kalms utils but all i got was "Fail" when trying to convert the wav's...

Can this be taken to a 1 disk floppy with audio akin to the 12Khz???

Possibly a madman task.
Geezer

[Don_Adan]

you dont need to play long sample from chip ram.
for handling long sample players on Amiga you can check source players for:
ADPCMmono
NTSP
onEscapee
and from custom modules from Dragons Lair and Space Ace.
All available on the Wanted Team page.

[mcgeezer]

Originally Posted by Don_Adan

you dont need to play long sample from chip ram.
for handling long sample players on Amiga you can check source players for:
ADPCMmono
NTSP
onEscapee
and from custom modules from Dragons Lair and Space Ace.
All available on the Wanted Team page.

This implies I would be using another ram type...?

[Don_Adan]

Originally Posted by mcgeezer

This implies I would be using another ram type...?

for very long samples, you must/can use small buff in chip ram. samples are copied/depacked from fast memory to this buffer. or you can use Ready Soft style player, blitter decoding, 4bit quality.

[mcgeezer]

Originally Posted by Don_Adan

for very long samples, you must/can use small buff in chip ram. samples are copied/depacked from fast memory to this buffer. or you can use Ready Soft style player, blitter decoding, 4bit quality.

I don't have any fast ram though.

[ross]

A noisy test

[Don_Adan]

Originally Posted by mcgeezer

I don't have any fast ram though.

Decoding from chip ram to chip ram can be not very good for 68020+. But i dont know how many CPU time you have. If I remember right for pure chip ram sample playing you can use replay from AfterBurner custom module.over 400kB sample, if I remember right. Or maybe from CarrierCommand? I forget, it was Dave Lowe music. He used two versions for sample type playing. Pure one sample only and splitted for more samples version.

[Don_Adan]

I dont know how many chip ram free you have for sampled music, but you can use ADPCM (rate 1:2, if I remember right) or NTSP (rate 5:9). Or simple pack sample with PackFire (LZMA) or Arj7, and depack to chip ram. If you want to use only one disk for sampled music, your packed (ADPCM, NTSP, LZMA,ARJ7) file can be maximum 901120 bytes long or ~1.1MB if you used longtrack MFM format.

[mcgeezer]

Originally Posted by ross

A noisy test

Any chance you could let me in on how you did that buddy?

I wasn't able to get adpcm to work that Dan linked above.

[ross]

Originally Posted by mcgeezer

Any chance you could let me in on how you did that buddy?

I wasn't able to get adpcm to work that Dan linked above.

It was thrown in a few minutes last night, so I don't even know if I did everything right

For the encoding phase: sox.
I've set a frequency for an Amiga PER=300, with a low pass filter at 6Khz and dither.
Sure some better filter and tests be needed.

Then I've assembled PlayAdpcm from Kalms's package.
Probably the exe from Kalms's encoder need a recompilation because do not work on today Win machines.


EDIT:
Ok, now I'm in the laptop where I launched the encoding stage.
Note: it's absolutely not a series of recommended filters, they are simply the first thing that came to my mind and that could give 'decent' results, certainly it can be improved, we can discuss of this.

sox command line used:

sox.exe -v 0.92 -S -V -D "newblack_raw.mp3" "newblack_raw.ima" remix - lowpass 6000 rate -v -s -I 11822 dither -S

Results:

Level adjust   : 0.92 (linear gain)
Channels       : 1
Sample Rate    : 11822
Precision      : 13-bit
Sample Encoding: 4-bit IMA ADPCM
effects chain:
 input        44100Hz  2 channels
 remix        44100Hz  1 channels
 lowpass      44100Hz  1 channels
 rate         11822Hz  1 channels
 dither       11822Hz  1 channels
 output       11822Hz  1 channels

Done.