Giana Sisters STE - New Game!

[Zamuel_a]

What about doing sound in play part complete with DMA ? Then would be no problems with sound levels, only 'small' problem of mixing diverse samples if they play at once. Or just using L-R channels ?

I'm not sure I understand what you mean? The music must be runned with the YM chip.

[AtariZoll]

Amiga has no YM or some other sound generator chip. All goes via 'her' DMA , as I know . So, need tune samples too. And possibly 4MB RAM

[junosix]

But then you'd need to have a .mod tracker playing in the background which will kill CPU, and will also take away the "ST"-ness away from the game, might as well just play the Amiga version then! I like the idea of keeping enhancements era-sympathetic, as in this is the version STE owners could have quite possibly had back in 1989 with some decent coding. Whacking a load of RAM and streaming music off a hard disk etc. takes away from that a bit I think.

[AtariZoll]

MOD tracker ? What kills not slower Amiga CPU ? So, you helped to add Amiga samples, and now talk about taking away ST-ness from game ?
If game worked on Amiga 512, and now needs 2MB RAM with YM audio (older versions), then it is not my fault for sure. Doing music 'streaming' was just demonstration of how to achieve really C-64 like music. And it was possible even in 1989 for people with hard disks ( no need for big speed) .
This is already taken away, if you ask me.

[junosix]

Thing is that the Amiga has the ability to play samples at arbitrary sample rates with little intervention from the CPU, and has 4 distinct sound channels. The STE only has 4 sample rates (6.25kHz, 12.5kHz, 25kHz, 50kHz) and two (interleaved) channels. To achieve a 4-channel stereo mod player takes a lot more effort in terms of coding and CPU usage than the Amiga. With the additional stuff that Zamuel_a has added that there may not be enough cycles left over.

To me the way I think of conversions like this is that it's a great example of going with the STEs strengths rather than fighting its weaknesses and it's a realistic demonstration of something that would have been more or less commercially possible back in the day.

[Eero Tamminen]

Yes it's a Devpac issue, but the files and configuration is the same. I use the same "harddrive" for both emulators. I'm using 1.7.0 and I'm close to filling up the entire memory. I'm using the same TOS file to. 1.62

What if you just set Hatari to have 14MB of RAM ("-s 14" option if you do it from command line)?

so there must be something that takes more RAM in Hatari than in Steem

If you're using harddisk image, I don't see how there could be less memory available under Hatari, as everything is done by TOS, in same way. If by "harddrive" you mean GEMDOS HD emulation (?), there could be some difference.

Hatari GEMDOS HD emulation calls on program Pexec() Hatari's m68k code located in the cartridge memory area. That m68k code does the program loading and relocation. It's possible that it differs from TOS code a bit, or from STeem code, if STeem does something similar for it's GEMDOS HD emulation... Thomas would know this best.

[Zamuel_a]

Some time ago I looked into adding a mod tracker to the game, but the cpu load is very different depending on how much sprites are on screen. The first levels could do it, but later on when there are rather much sprites at the same time, when the speed would go down. I do have something like 20-40% free cpu time left so it's not like I have reached the limit. I don't think the YM music is so bad (and in most reviews of giana sisters they say that the Atari version has better music than the Amiga).
If I had used a mod player, the sound effects would still run on the YM chip or else I need a 6 channel mod player!

What if you just set Hatari to have 14MB of RAM ("-s 14" option if you do it from command line)?

The problem is Devpac version 2 that won't let me use all 4Mb of RAM!

I got the same problem in Steem SSE, but not on the old Steem. Strange

[junosix]

I've attached the two samples that needed fixing, and have also included all the samples again but with a lower volume just in case you decide that the YM_Lower version of the music is easier/better to use. Can easily re-export again with different volume levels and sample rates if required, can only really tell whether I've got it right once it's up and running really.

[npomarede]

I've attached the two samples that needed fixing, and have also included all the samples again but with a lower volume just in case you decide that the YM_Lower version of the music is easier/better to use. Can easily re-export again with different volume levels and sample rates if required, can only really tell whether I've got it right once it's up and running really.

Note that both versions should be tried on a real STE first, especially the YM_Lower one with mixer=0, as it's possible the result is different from Hatari in that case.
If there's really no YM music with YM_Lower version on a real STE (as cyprian reported on his mega ste), then only remaining alternative is to use mixer=1 and either lower YM volume or boost sample volume (I think the target should be the result on a real STE, not the one under Hatari or Steem, even if nowadays many people are using emulators)

[troed]

(I think the target should be the result on a real STE, not the one under Hatari or Steem, even if nowadays many people are using emulators)

Agreed. Once we start tuning for emulators the scene could be said to have lost its roots and be something completely different from then on.

(Especially since emulators over time will get closer and closer to the real hardware anyway)

[hedning]

(I think the target should be the result on a real STE, not the one under Hatari or Steem, even if nowadays many people are using emulators)

Agreed. Once we start tuning for emulators the scene could be said to have lost its roots and be something completely different from then on.

(Especially since emulators over time will get closer and closer to the real hardware anyway)

Troed is of course completely correct, an no further discussion is necessary.

[junosix]

I agree wholeheartedly!

[dma]

Also part of the pleasure for a coder on such a "retro-"system, is, even if himself isn't coding on a real machine or maybe doesn't even own one, too see/imagine/know that some users run his production on their real machine.

[AtariZoll]

Indeed, only tests, work on real HW is what should be taken as 'reference' .
Despite all huge work done, and still going on, there are some things not properly emulated.

But, for developing, debugging Atari ST(E), and even Falcon, TT SW Steem Debugger is ideal IMHO . After it works in it - and that may mean work under some modded versions ( 68030 stackframe for instance ) coders, developers should test on real HW.
Unfortunately, as I see, there is number of developers who haves no real Atari HW at all - to not write here names .. Should buy at least some STE - it is still not expensive, but will be after some years, I'm sure, because there will be less and less working machines.

[Mug UK]

Whilst doing a batch of Kryoflux images, I finally got the time to play Giana Sisters STE on my Mega STE. Many thanks for this updated version Zamuel_A

[Cyprian]

Things to test on an STE : use mixing mode 0 and play an empty sample (bytes=0) + YM sound on voice A. Now change the YM volume (eg reg 8 for voice A) between 0 and 15 and check if any YM sound can be heard

thanks for the hint, I'll check that.

I've prepared small test application (attached) and I've checked it on 1040 STE and also on Mega STE.
Unfortunately there is no sound in case of LMC YM-12dB., therefore it seems that LMC documentation is wrong in this point.

When you try that app then:
Yellow Border means:

Code: Select all

move.w   #%10000000000,d0   ;mix DMA + YM-12dB

Blue Border:

Code: Select all

move.w   #%10000000001,d0   ;mix DMA+YM equally

You can close that program with "ESC" or with mouse movement.

[AtariZoll]

Yes, it silents PSG complete.
Here is what I use :

Code: Select all

  *Setting mixer, volumes:

   move.w $FFFF8922.w,old
   move.w $FFFF8924.w,om

   move.w #$7FF,$FFFF8924.w
   move.w #$401,$FFFF8922.w  *mixer - corrected PSG level !

w1
   move.w old(pc),d1
   cmp.w $FFFF8922.w,d1
   nop
   bne.s w1 

   move.w $FFFF8922.w,old
   
   move.w #$7FF,$FFFF8924.w
   move.w #$543,$FFFF8922.w  *master

w2
   move.w old(pc),d1
   cmp.w $FFFF8922.w,d1
   nop
   bne.s w2

   move.w $FFFF8922.w,old
   
   move.w #$7FF,$FFFF8924.w
   move.w #$545,$FFFF8922.w  *left

w3
   move.w old(pc),d1
   cmp.w $FFFF8922.w,d1  
   nop
   bne.s w3

   move.w $FFFF8922.w,old
    
   move.w #$7FF,$FFFF8924.w
   move.w #$544,$FFFF8922.w   *right

w4
   move.w old(pc),d1
   cmp.w $FFFF8922.w,d1
   nop
   bne.s w4

   move.w om(pc),$FFFF8924.w

[Zamuel_a]

I haven't tested it myself on a real machine yet, but if I understand it correctly, it is not working to use the YM -12dB bit on a real machine? Is there a way to get it to work or it's not possible?

[Cyprian]

I haven't tested it myself on a real machine yet, but if I understand it correctly, it is not working to use the YM -12dB bit on a real machine? Is there a way to get it to work or it's not possible?

Outcome of my tests is that there is no LMC mode "YM -12dB mixing". LMC has 4 stereo inputs (plus "Open"). I wrote another test app which checked all of them and YM sound is only present on "Input 1" (and this is default mode).
Interesting is that why in some STE documentations like Profibuch is mentioned that non-existed feature.

[Zamuel_a]

Ok, so there is no -12db option to use. When I don't know what to do get it correct without alot of extra work.

[npomarede]

I haven't tested it myself on a real machine yet, but if I understand it correctly, it is not working to use the YM -12dB bit on a real machine? Is there a way to get it to work or it's not possible?

Outcome of my tests is that there is no LMC mode "YM -12dB mixing". LMC has 4 stereo inputs (plus "Open"). I wrote another test app which checked all of them and YM sound is only present on "Input 1" (and this is default mode).
Interesting is that why in some STE documentations like Profibuch is mentioned that non-existed feature.

Hello
from your tests, what does mixer=0 sound like ? Is it the same as mixer=2 ? But from the doc it seems mixer=2 gives no sound output too.
So in the end, mixer=1 would be the only value that produces sound (mix ym and dma) ?

[npomarede]

Ok, so there is no -12db option to use. When I don't know what to do get it correct without alot of extra work.

Easiest solution would be to boost the sample volume of each .wav using a sound editor like Audacity (but this will alter the sound too)
Then it will be possible to hear if this degrades sample sound too much and if this allows to reach an output level similar to the YM music.

[Cyprian]

Hello
from your tests, what does mixer=0 sound like ? Is it the same as mixer=2 ? But from the doc it seems mixer=2 gives no sound output too.
So in the end, mixer=1 would be the only value that produces sound (mix ym and dma) ?

Attached you can find another test application.
With keys 1,2,3,4,0 you can change settings of LMC mixer, A/Z,S/X turn on/off SDMA/YM sound:

Code: Select all

0 - Open
1 - Input1
2 - Input2
3 - Input3
4 - Input4
A - Turn Off YM
Z - Turn On YM
S - Turn Off SDMA
X - Turn On SDMA

The outcome is that, whether you choose Input1/2/3/4 or Open - SMDA sound is always audible (with the same volume level). But YM is only present when you activate Input1.

[junosix]

I've just tried a quick boost by 6dB but it's making the sounds too distorted, I'll instead try compressing them tomorrow when I have some time. That will have the effect of not clipping the sounds but boosting the average volume.

Although the YM2149 volume levels aren't linear is it worth trying a quick subtract/shift before the data hits the soundchip registers?

[AtariZoll]

I'm not trying to be negative, but will write my thoughts: I think that it went in bad direction. Idea of using YM and DMA at once for sound seems for me as bad. I looked couple quality games: Magic Boy and Lethal Xcess , and both have only DMA audio on STE (Magic Boy will have YM on ST) . Both work with only 512KB RAM . So, I really don't think that background music in levels must eat lot of CPU power and/or RAM . To solve DMA music should not use some stand-alone MOD player code, which can use 100% CPU power, since it does nothing else, but preparing such music data, what can be played without need for mixing data and similar. So, music playback code should only control addressing and looping of DMA - what surely needs not more than 1 % of CPU power. The problem here is, as I see that STE DMA is with interleaved L-R channel, so you can not simply use music on 1, and effects on other channel. Some CPU time must go in mixing effects into music data, then restoring music data . Simple copying of 10 KB needs about 6mS. With some mixing more. Surely not easy to code.
Another way would be to silence complete YM music data - not on fly, but before assembling
But before all it, what about Falcon fixed version with YM audio ?