Using blitter for sound mixing?

[idrougge]

I found a post by Gibs on Atari-forum where he has been experimenting with mixing and modulating samples using the STe blitter. Is this applicable on the Amiga, and does it have any useful applications?
http://cyber.savina.net/sound.htm

[Galahad/FLT]

Originally Posted by idrougge

I found a post by Gibs on Atari-forum where he has been experimenting with mixing and modulating samples using the STe blitter. Is this applicable on the Amiga, and does it have any useful applications?
http://cyber.savina.net/sound.htm

Pretty sure this is how all more than 4 channel audio replayers work by using the blitter.

TFMX 7 channel, Star Trekker and others

[idrougge]

Oh, I always thought they used the CPU for mixing.

[frank_b]

Originally Posted by idrougge

I found a post by Gibs on Atari-forum where he has been experimenting with mixing and modulating samples using the STe blitter. Is this applicable on the Amiga, and does it have any useful applications?
http://cyber.savina.net/sound.htm

Bear in mind the Atari blitter has an indirect addressing mode and 16 word look up table built in. That can be combined that with first/middle/last word masks protecting selected bits in the destination during translation. That allows 0-4 bit to any 16 bit value translations on the fly. That's how he's able to do volume variation and other translation effects. The same feature can be used for graphics operations. 1-2x-4x-8x-16x scaling on the x or hflipping a raster in four passes. I'm not sure some of it is applicable to the Amiga blitter. It's a major difference between them. Paula has built in mixing of voices, frequency and volume variation per channel anyway

[meynaf]

Originally Posted by Galahad/FLT

Pretty sure this is how all more than 4 channel audio replayers work by using the blitter.

TFMX 7 channel, Star Trekker and others

All the mixing code i've seen so far used nothing but the cpu.

[ross]

I do not think it's possible to create a fast mixer with the blitter (if you want it anyway a really slow one can be made..).
At least not the classic (fast) one made with linear ADD in the same frequency domain.
And I can not think of a quick way to use blitter for a linear interpolation oversample usage..

At least you could create a blitter interlaced mixer (one byte of a sample followed temporally from that of another sample), but qualitywise is a no-no.
Or prepare a downward shift of the sample in real time and then cpu add without clip.

[frank_b]

He's using the halftone/smudge mode for some free effects in real time. Looks like it maxes out at 16% CPU time. Impressive!!!
[ Show youtube player ]

[meynaf]

Well, so apparently it's not for sound mixing but for applying quick'n'dirty sound effects. I'm not impressed.

[Don_Adan]

Originally Posted by meynaf

Well, so apparently it's not for sound mixing but for applying quick'n'dirty sound effects. I'm not impressed.

This is the only one known me Amiga module, which used blitter for synth samples manipulations:
http://aminet.net/package/mods/misc/CriticalsPromo
Second known me Amiga music, used blitter for sample depacking only.
http://aminet.net/package/mods/cust/SpaceAcedemo

[Photon]

Manipulation and unpacking isn't mixing, tho. The Blitter ALU can be used in passes for arithmetic operations, the problem is the carry. It would likely also need the instruments to change from sampled waves to something prepared for mixing. Typically, a proof of concept used in a demo is not usable in the general case. Finally, it must also sound almost-as-good as the one wave per hardware channel version. That's the real trick

[metalages]

Originally Posted by meynaf

Well, so apparently it's not for sound mixing but for applying quick'n'dirty sound effects. I'm not impressed.

I use the Blitter for mixing Bit decimation is an optional bonus stuff provided by halftone registers while copy is done.
Of course the STe hardware is basic and the routine is very constraining as explained here https://github.com/jhubrt/demOS/tree...r/DOCS/BLITSnd


Retrocomputing is about playing with contraints, and probably programming on the Atari ST has always been about that too


Here is what we have managed to release using this routine thanks to Dok talent [ Show youtube player ]

[Photon]

Originally Posted by metalages

Here is what we have managed to release

ST(e) can sort of do samples, and under some circumstances, it can do scrolling, too.

Your readme puts some tighter requirements for running it at all, though. E.g. beyond 1040 STe, 520 STe can't run it as I understand it and you have to set it up just so. Please correct me with exact platform target.

In this topic, the question was asked if the more capable Blitter of the Amiga could even substitute for the CPU in proper software mixing. This however requires arithmetic and carry. (For any Blitter including the simplistic ones on some PC graphics cards, and they didn't really get any less simplistic with time either).

Thus far objectivity.

If you are posing this against Amiga, know that you are competing with 1990 CPU 8-channel mixing, and as we know the 7-channel SMFX, launched not as a limited special-case proof of concept, but to solve the problem of music+SFX with a game using the hardware fully at the same time. [ Show youtube player ]

[saimon69]

I thought the first turrican did not have the 7-channel mixing routine, the second one for sure do

[TCD]

Originally Posted by saimon69

I thought the first turrican did not have the 7-channel mixing routine, the second one for sure do

It does not: https://www.exotica.org.uk/wiki/TFMX_7V

[alexh]

The 7-voice mixing routine used in TFMX 7V was written by Jochen Hippel of Thalion who did the arrangements for the Atari ST version of Turrican I & II. It was also used in Lethal Xcess and Amberstar.

[metalages]

Originally Posted by Photon

If you are posing this against Amiga, know that you are competing with 1990 CPU 8-channel mixing, and as we know the 7-channel SMFX, launched not as a limited special-case proof of concept, but to solve the problem of music+SFX with a game using the hardware fully at the same time. [ Show youtube player ]

Nope i was just answering to the comment saying the blitter was just used as a post process for bit decimation which is not the case. Also with compliant composition it is not so specific as that.



I am curious on how the more than 4 voices are usuallly implemented on amiga ? From what I understand Paula can get its data from 4 places in memory at variable pace. In case of more than 4 voice does it mean some of the voice run at fixed freq + cpu mixing ?

A long time ago i have heard amiga blitter has a bresenham engine. Is it the case ?

[metalages]

Originally Posted by alexh

The 7-voice mixing routine used in TFMX 7V was written by Jochen Hippel of Thalion who did the arrangements for the Atari ST version of Turrican I & II. It was also used in Lethal Xcess and Amberstar.

Whaou i was not aware mad max wrote this kind of routine on amiga

[ross]

Originally Posted by metalages

In case of more than 4 voice does it mean some of the voice run at fixed freq + cpu mixing ?

Yes, in case of 7-voices TFMX one channel with fixed frequency gets four cpu mixed sources and 3 channels are free for normal playback.
But there are other players who behave differently.

A long time ago i have heard amiga blitter has a bresenham engine. Is it the case ?

Yes, but the engine is not exposed. So only a destination pointer is calculated and pre-defined data is used for line mode.

[roondar]

Originally Posted by Photon

If you are posing this against Amiga, know that you are competing with 1990 CPU 8-channel mixing, and as we know the 7-channel SMFX, launched not as a limited special-case proof of concept, but to solve the problem of music+SFX with a game using the hardware fully at the same time. [ Show youtube player ]

The 7-channel TFMX routine is not used during gameplay (only for the title and perhaps the outro/high score - but not sure about those), it just plays 4 voices as normal. The TFMX modules mostly only use 3 channels (there's one tiny bit that uses 4 channels, but that is played when no SFX can play so it works out). SFX gets played on the free channel. Some other music channels get cut out when there's more than one simultaneous SFX playing back.

[no9]

To add to what Roondar said we can clearly hear in the Turrican 2 credits how mixing quality is downgraded every time the player character appears and gets back to higher mixing samplerate when there is only text displayed on the screen. During full gameplay software mixing was to much [ Show youtube player ]