QTM and QTM_PlayRawSample

Hiya,
I’m trying to use the QTM module for sound sample playback in another program.

First Question:

I’m trying to figure out the sound format I need to convert the data too so QTM can play it, the documentation suggests “8-bit logarithmic format” but !SoundCon offers many more options than that for RAW sample data

Stereo/Mono?
Sample Rate, 11,22,44,48 KHz?
Logarithmic format, u-law, a-law, Arc VIDC

I’ve assumed 8-bit, little endian.

I’ve tried a few different formats without success

8-bit u-law mono 44100Hz
8-bit a-law mono 44100Hz
8-bit Arc VIDC mono 44100Hz
8-bit Arc VIDC mono 20833Hz

These all playback with enough distortion that I can barely discern the original sound.

Second Question:

The R5 param for QTM_PlayRawSample is the note to play it back with (I presume as in a tracker, samples are pitch bent up or down to play the specific notes), what would be the value for ‘playback without bending the pitch’, the natural speed of the sample? (Some value in the middle?)

Afraid I can’t help as I haven’t ever been interested in 8-bit sound. However it may help others advise if you say what hardware and OS version you’re using.

Jim

Doesn’t QTM play Amiga IFF? This is a billion year old memory, so maybe wrong. ..
https://en.m.wikipedia.org/wiki/8SVX

The hardware is a Pi B+ running 5.21.

QTM itself will play back many tracker formats (and the many formats of the samples inside them), but QTM_PlayRawSample is likely to need a specific type, given you just give a pointer to the data in memory and a length (assuming no autodetection).

From the Technical Docs

SWI “QTM_PlayRawSample”                                               &47E57

-————————————————————————————————————-

On entry: R0 = channel number (1-8) or -1 for automatic (new in QTM v1.28)

          R1 = sample address (logarithmic sample data)

                            – if address is 0, the channel is silenced

          R2 = sample length  (in bytes)

          R3 = repeat offset  (bytes from start of sample)

          R4 = repeat length  (in bytes)

          R5 = note and flags (note in bits 0-27, flags in bits 28-31):

                If bits 28-31 = 0, use 3-octave notes or Amiga periods

                   bits  0-27    = 1 to 36, 1=‘C-1’, 2=‘C#1’… 36=‘B-3’

                              or = 37 to 1999 use Amiga period values
If bits 28-31 = 1, use extended 6-octave note values
bits  0-27 = 1 to 72, 1=‘C-0’, 2=‘C#0’… 72=‘B-5’
R6 = linear volume  (0-64)
If R0=-1, R7 = 0    }_ When using R0=-1, R7 and R8 must be set as
R8 = 255  }  shown here for future compatibility!
On exit : If R0=-1 on entry, R0=channel used (1-8) or -1 if sample was not

                                played due all channels currently in use.

                   else R0 preserved

          R1-R8 preserved
V flags set if the contents of the registers contain illegal values.
This SWI allows you to play a sample stored in any area of memory. The

sample should be in 8-bit logarithmic format.
If R0=-1, the automatic channel finding system is used. This does not

need a channel specified in R0, but instead searches for a free channel (with

no music or samples playing through it) and plays the new sample through that

channel, returning the channel number in R0. Note R7 and R8 should be set

as shown above when R0=-1, this is to enable your program to work with the

next version of QTM – which will use R7 and R8 for further information. (R7

and R8 remain unused and unaltered if you are using R0=1-8.)
This SWI will also start up the QTM sound system if it is called while the

song is stopped or paused, though it will not switch the system off after

the sample finishes, use SWI QTM_SoundControl to do this.
See file ‘NoteFormat’ for more information on the 6-octave 72 note format.

“The sample should be in 8-bit logarithmic format”

This is what I’m struggling with, that’s not enough information. Alternatively SoundCon could be converting things slightly incorrectly. Are there any other RISC OS Wav file to RAW convertors?

“See file ‘NoteFormat’ for more information on the 6-octave 72 note format.”

This has had enough information in for me to determine that my samples recorded at 44100Hz should be played back at note F-3 (value of 80 is closest at 44336Hz).

It expects the sample to be in standard VIDC 8-bit audio format (mu-law). Strictly speaking it’s not a pure logarithm as the chords are logarithmic but the offset within the chord is linear, where you’re possibly going wrong is with the sign which is in bit 0:

bit Value
7..4 chord
3..1 offset
0 sign

The correct way to do the encoding is via the lookup table provided by the Channel Handler looked up via Sound_Configure

MOV R0, #0
MOV R1, #0
MOV R2, #0
MOV R3, #0
MOV R4, #0
SWI Sound_Configure
LDR R0, [R3, #8] ; get linear to mu-law conversion table address
STR R0, log_to_lin_table

The table is 8Kb in length, so you should scale your linear value accordingly via LSR #x

If you’re converting the sample outside of fill code, you could simply call Sound_SoundLog with a 32bit signed sample and it will return the mu-law equivalent.

Replying to my own question incase anyone searches for this in future.

Ta Jon, with that information I was able to use !SoundCon to output 8bit, mono, ‘vidc’ unsigned format raw sound (at 44100Hz)

At that frequency the correct code is

DIM somespace% 100000
SYS"OS_File", 16, “SampleFileName”, somespace% TO ,,,,len%
SYS"QTM_PlayRawSample", -1, somespace%, len%, 0, 0, 80, 64, 0, 255
REM note don’t quit your program until the sample finishes playing else sound breaks up.

There is an issue on the Pi B+ with slightly distorted sound, but this is due to the presence of the disable_audio_dither flag in config.txt, remove that from config.txt and the playback is fine, but there is the constant background hiss reported in the other threads.

https://www.riscosopen.org/forum/forums/11/topics/3384