Playing samples

[Lonewolf10]

Hi,

I am trying to add a few sound samples to my demo, but I am having trouble calculating the period. I believe I have the correct formula (below), but my results (also below) give me the result of 30 - this gives a short high pitched noise. A period value around 200 gives a better sounding result, but doesn't help me to calculate the proper value for this sample (and other samples).

Formula for calculating the period

period = frequency (Hz) / (1/sample length (bytes) )
        --------------------------------------------
          0.281937 (PAL) microseconds per interval
          0.279365 (NTSC)

My calculations

period = ( 11025 / 0.0001303 ) / 0.281937
       = ( 11025 / 1303 ) / 0.281937
       = 30

Am I using the correct formula for calculating the period?

The sample should be played at 11025Hz and is 7673 bytes long.


Regards,
Lonewolf10

[Leffmann]

For PAL the period is 3546895 / frequency in Hz / length in bytes, and for NTSC you use 3579545.

Samples have to be an even number of bytes in length, and there's a lower limit of about 125 for the period value (for NTSC and PAL at least).

[Codetapper]

Here's a little from Mapping the Amiga:

Most sound-digitizing software specifies the waveform frequency by the number of samples per second, not by the period value that the Amiga's hardware expects. The equation for converting the samples-per-second to period value is:

Period = 3579564 / Samples-Per-Second

So, if you want to play a sound that was sampled at 12 kHz, use a period value of 3579546 / 12000 = 298

To calculate the period value for a note, use the equation

Period = 3579546 / (Length * Frequency)

where Length is the size of the waveform in bytes and Frequency is the note's frequency value obtained from the table:

A = 440
A# or Bb = 466
B = 494
C = 523
C# of Db = 554
D = 587
D# or Eb = 622
E = 659
F = 699
F# or Gb = 740
G = 784
G# or Ab = 831

To raise a note by an octive, divide the period value by 2.

It also states the lowest value you should ever use is 124, since this is about as fast as the Amiga can go. If you use a value less than 124, you run the risk of having the same 2 bytes of data output twice in a row.

[Lonewolf10]

Thanks guys, that makes sense

Page 138 of the hardware manual is a little confusing and I ended up using the clock interval instead of the clock constant!


Edit:

Do games and demo's calculate the period in realtime before playing samples, or do they just detect whether it's a PAL or NTSC Amiga at startup and use the appropriate lookup table?


Regards,
Lonewolf10

[Leffmann]

I don't know of any games nor music replay routines that adjust periods automatically, there are probably some out there, but the vast majority will just assume PAL. With music replay routines it's complicated further because typically it's all synchronized to the frame rate.

[Lonewolf10]

When I wrote my demo back in 2011 I only had the one sound sample (dog barking) and just waited a specific length of time before switching off the DMA audio channels. For the game I'm working on right now I plan on having sound effects playing at random intervals in the background, so waiting a fixed length of time would ruin the game and complicate things.

I have written a routine for playing samples of different lengths and frequency's. It works fine, but it keeps playing the sound sample over and over. How do I work out when to turn off the DMA for the audio channel without cutting off the samples halfway through?

[daxb]

You may have a look at phx (I guess he was it) released play-mod-and-samples source. I can`t remember more, but you should find it on eab.

Edit: http://eab.abime.net/showthread.php?t=65430

[Toni Wilen]

Originally Posted by Lonewolf10

I have written a routine for playing samples of different lengths and frequency's. It works fine, but it keeps playing the sound sample over and over. How do I work out when to turn off the DMA for the audio channel without cutting off the samples halfway through?

No need to turn of the DMA. Simply point audio pointer to short empty sample (update both pointer and length) after sample playing has started. Paula will automatically switch to second sample after first sample ends.

Just make sure sample has started playing (Paula has fetched first 16-bit word), you can also wait for audio interrupt, it comes just after playing starts.

[Lonewolf10]

Thanks for the help guys