Hextracker (updated to 0.849B + BPM preview)

[ljbk]

Hi !

Thanks for testing it.

Well, from the point where it fails, my best guess is the following:
- Your PAK board is fooling the STE detection mechanism that is based on the colour hardware registers: if i set 16 bits, on a normal ST / Stacy, only 9 bits are really stored and so that is why we have only 512 colours as for an STE 12 bits are really stored but apparently you PAK board detectes the acess to a known hardware register where probably the full 16 bits are stored and remembered instead of the "normal" STacy hardware;
- The program assumes then that it is dealing with an STE and in the Tracker Init phase where a T should appear on your screen, it tries to init the "STE" DMA accessing addresses $FF89xx and this should be causing the bus error due to the protection of that memory space on STFs and STacys and also aparently with the PAK;

So what to do ?
Well, first we need to check that this is the problem. I hardcoded the HMIN844 to work always as an STF (you can check for a DarkLord comment in the source).
If this works, then i can provide you with an hardcoded version of the program that will assume the hardware to be always an STF.
The correct solution would be of course to have a new STE detection mechanism that is not fooled by the PAK board.

So please have a try at the new zip with program HMIN844D.PRG.


Bye,
Paulo.

[DarkLord]

Okay, just downloaded and tried the "D2" version.

It does NOT two bomb crash, makes the background light blue, then freezes
up right after:

CMDVFTL

The audio also pops one time. I think we're getting closer?

Thanks.

PS Oh, just to rule things out, I disabled all ACC's, CPX's and the
AUTO folder and tried it "clean" like that. Same result.

[ljbk]

Hi !

So your STacy 030 was really detected as an STE ...
But now we have a crash during play. That can be caused by the shadow addressing. May be the PAK as the Falcon does not like that.
The zip attached has a new HMIN844D with a MSYN844D to test the basic funcionality removing the shadow addressing.
It also has a new HEXT844D and DSYN844D that forced a STF recognition on top of removing the shadow addressing.

Bye the way, does the original package work on your Falcon ?


Bye,
Paulo.

[DarkLord]

Woohoo!!!!!!! Success!!!!

It's working! It loaded up just fine, and I played the included mod, no problems.

I also loaded up the original 844 release on my Falcon and it worked fine there,
even in 060 mode.

Thank you, thank you, thank you! <bows low in awe> This is one of the things
I love about the Atari community - so many great people!

PS One quick question - <space> stops the current song and returns the mouse
pointer, but how do you exit back to the desktop? Probably something obvious
but I didn't see it...

[ljbk]

Woohoo!!!!!!! Success!!!!

It's working! It loaded up just fine, and I played the included mod, no problems.

I also loaded up the original 844 release on my Falcon and it worked fine there,
even in 060 mode.

Thank you, thank you, thank you! <bows low in awe> This is one of the things
I love about the Atari community - so many great people!

PS One quick question - <space> stops the current song and returns the mouse
pointer, but how do you exit back to the desktop? Probably something obvious
but I didn't see it...

I am glad it works.
To exit, just left click on the top left corner.
Please take a minute to read the small manual included in the main package.

Enjoy it.
Paulo.

[DarkLord]

I am glad it works.
To exit, just left click on the top left corner.
Please take a minute to read the small manual included in the main package.

Enjoy it.
Paulo.

Already enjoying it, and I promise I will go back and read the manual now.

Thanks again!

[ljbk]

This procedure just waits for the vertical blank

Thanks. It seems that calculating this register value on reads (like ST/STE/TT code does) is TODO in Hatari's Falcon/Videl emulation for some reason, it just returns value last written to this register by emulated code. I mailed hatari-devel about that.

Hi !

As you can read from the previous posts, you can fully test Hextracker 0.844 with Hatari in Falcon mode.
If it works on the real machine, then it must work on the emulator.

Bye,
Paulo.

[ljbk]

Hi Yerzmyey !

I found back the calculations for the FSAMPL value.
It depends directly on the Amiga clock.
For D#2, the divider in the Amiga pitch table is $168 = 360 decimal.
Paula, the Amiga sound chip uses that value to divide the computer master clock.
Amiga NTSC version has a master clock of: 28.63636 MHz
Amiga PAL version has a master clock of: 28.37516 MHz
If you divide those master clock values by 4 you will get the CPU frequency: 7.15909 MHz and 7.09379 MHz respectively.
Now dividing by 2 and by the divider (360), one gets: 9943 Hz and 9852 Hz.
If one use C-2 as reference with a divider of $1AC (428) and we then multiply by 2^(3/12), we get: 9946 Hz and 9855 Hz.
So a you can see the American Amigas will replay things faster.

So here you get the why i have the value 9855.
I assumed the European (PAL) value as most of the MODs were done in Europe and the MOD movement started in Europe.
One could also assume an average value of 9900 to get the average Amiga behaviour.
Hextracker lets you test and save the one your prefer as the FSAMPL value is stored in the Hextracker format as the "base frequency" for the set of samples used.
With its samples lenght up to 16 MB and a variable sample base frequency, Hextracker allows you better quality samples (with up to 6 times more resolution than the normal MOD samples) and so better mixing results.
You can for example set FSAMPL to 59130 and sample your instruments at 6 times the rate you would do it for "normal" Amiga MODs.
This requires of course huge quantities of memory, but most of the interested ST community has now a STE with 4MB ...


Bye,
Paulo.

PS: Bye the way, Lance´s player uses a reference Amiga clock of 7.09 MHz.

[ljbk]

Hi !

An update to version 0.845 is now available.
STE oversampled and multiplexed modes are now available.
Octalyser tags CD42, CD43, CD62, CD63, CD82 and CD83 are now recognized.
This means that any module using 20KHz or 25 KHz sample base frequency saved in Octalyser can be played or reworked in Hextracker.
In the zip there is an example: spicey.mod from Alan F, using 25 KHz base for C-3 note.
That mod was adapted to have volume effects using 9 changed to C. But the original file is also there.

This is not a "full" package as i expect to do some more updates soon.
The multiplexer implemented strategy is the one that fitted the strategy for the other parts of the code like DIST calculation and Osciloscope. But it is not the fastest. A new one is in preparation that will need a specific STE codepath for both above features. It is much faster. But this will require some time.

Enjoy,
Paulo.

[FedePede04]

Hi
I can't find the download link.
in the first post, its the Hextracker_0.844.zip that you can download.

/Peter

Hi !

An update to version 0.845 is now available.
STE oversampled and multiplexed modes are now available.
Octalyser tags CD42, CD43, CD62, CD63, CD82 and CD83 are now recognized.
This means that any module using 20KHz or 25 KHz sample base frequency saved in Octalyser can be played or reworked in Hextracker.
In the zip there is an example: spicey.mod from Alan F, using 25 KHz base for C-3 note.
That mod was adapted to have volume effects using 9 changed to C. But the original file is also there.

This is not a "full" package as i expect to do some more updates soon.
The multiplexer implemented strategy is the one that fitted the strategy for the other parts of the code like DIST calculation and Osciloscope. But it is not the fastest. A new one is in preparation that will need a specific STE codepath for both above features. It is much faster. But this will require some time.

Enjoy,
Paulo.

[ljbk]

Sorry !

The file was not there for some moments has i add to modify a small detail.
After that, i uploaded the file but forgot to submit ...

But now it is there ...

Paulo.

PS:
DarkLord, this version will not work with your "special" STacy. At the next major version, i will provide you with a special case.

[nativ]

Kool stuff. nice to see you take on octalyser

[DarkLord]

PS:
DarkLord, this version will not work with your "special" STacy. At the next major version, i will provide you with a special case.

Well, hate to see you go through all that trouble, but I sure do appreciate it!

Been jammin' all the time, it's great software.

[FedePede04]

thanks
i will test it this weekend

/Peter

[ljbk]

Hi !

I wish to inform that version 0.845 does not work correctly with STE DMA output at 12.5 KHz both in mono and stereo. This is only a small bug when setting the STE DMA as there were some changes due to the multiplex and oversample modes.
I am sorry about this. I will correct that with a new version soon. Meanwhile, if you need 12.5 KHz you can use version 0.844.

Bye,
Paulo.

[ljbk]

Hi !

An update to version 0.846 is now available.
What's new ?
- the 12.5 KHz STE replay problem is solved;
- a new DMIN mode (minimum DATA display = 1 line) is available;
- Hextracker file format was slightly changed to include 4 VQ values;
- DarkLord special Stacy board should now be handled correctly without special version;

The most important change is the VQ one. We will have a different VQ for each type of mix.
If we have 8 voices, we can have:
- 8 to 1 mix in Mono;
- 4 to 1 mix in Stereo;
- another 4 to 1 mix in Mono with Multiplex;
- 2 to 1 mix in Stereo with Multiplex;
Even if you just work with one of these cases and define just one VQ, during the save, the 3 others will be included.
If your mod is loaded with another situation one of the 3 other VQs will be used.
If you change output type, the other VQs will be displayed automaticaly.
You can check that with the example ART_HXT.mod.

Very important: this version can load old Hextracker files but the files saved from now on can not be loaded with previous Hextracker versions.
The YM47K and YM50K players have also been adapted to load the new files.

Bye the way, there is a bug remaining with the Mutiplex mode and frequencies out of the 36 half tones that are not played correctly. I will not bother to correct it as i want to change the Multiplex implementation strategy to a faster one.
Finally, if you notice some high pitch noise with Multiplex, this is no bug. This is a cause of the Multiplex option that is better for some samples and worse with others.


Enjoy,
Paulo.

PS: DarkLord, please tell me if this does not work with your STacy ...

[bullis1]

Wow, this is a very nice update. I'll transfer it to my ST tomorrow.

Thanks for the continued support.

[ljbk]

Hi !

After all, even for the new multiplex strategy, i had to solve the remaining bug because i will have a similar case later on ...
So, here it is, a minor update to version 0.846A.
Just overwrite the existing files from version 0.846 with the two new ones.
You can check the difference with the module SPICEY_H.MOD, in STE Stereo mode, at 25X50 (with STeem at 16 MHz , Hatari at 16 MHz or a Mega STE at 16 MHz) at position $23 for example.

Bye,
Paulo.

[DarkLord]

Hi. Thanks for including me and my STacy in your updates!

Been working last couple of days, won't get a chance to check the
new and improved version until Wed/Thursday. I'll report back then.

Thanks!

[DarkLord]

Okay, just some real quick questions...

The 0.846 update is "stand alone", right? I mean just download the archive, unpack
it to a folder and run it - no other files needed, correct?

If so, then it does not work on my modified STacy. It crashes with something like
15 bombs, then returns to the desktop.

I tried 0.846, and I tried adding the 2 files from the 0.846a update, but it made no
difference. Same result.

Thanks, and sorry for all the trouble...

PS Special version 0.844 continues to work just fine. .

[ljbk]

Hi !

@DarkLord, thanks for the test. Don't worry about the 15 plus bombs appearing: an exception was generated on top of another one and so on ... I am allowing the CPU go to Bus Error where i included my routine. But something goes wrong when leaving it because the data piled on the stack is probably different on the CPU included on your special board compared to the standard 68000. So probably this generates some more exceptions i did not modify like: illegal instruction or address error. Anyway, i made an update to this new bus error routine to go back to the correct address in the code after the exception processing has ended in this new 0.847 version.

Well, here is a new version 0.847 that includes:
- multiplex bug correction already provided with version 0.846A;
- ALT-P or CONTROL-P command to save MOD as Protracker / Fast Tracker / Octalyser (lossy save);
- new text information when dealing with TOS operations like file selection;
- new STE MFP outputs to use the STE as a DAC and all the MFP replay frequencies;

The STE MFP outputs are of course slower compared to the STE DMA outputs at the same frequency.
The Protracker save is lossy. This means that all the data that does not fit in its format will be lost like: VQs, samples above 128K, pattern list after 128, odd values for sample length, restart position, loop length ... If the samples frequency is 11892 (10000 * 2^0.25) or 14865 (12500 * 2^0.25) then the Octalyser tags CDx2 and CDx3 will be used to indicate 20 KHz and 25 KHz base C-3 frequency. If we have 4 voices, M.K. tag will be included. For all other cases, xCHN or xxCH tags will be included.
- SPICEY.MOD is the original ALANF mod file;
- SPICEY_H.MOD is the Hextracker version of it where volume commands $9 have been changed to standard $C;
- SPICEY_O.MOD is the direct output of the ALT-P or CONTROL-P option for SPICEY_H.MOD;
- SPICEY_P.MOD is identical to SPICEY_O.MOD except the tag was changed to 8CHN because standard trackers like MIlky Tracker do not recognize Octalyser CD83 tag used when the samples have a 25 KHz C-3 base => this means that the MOD will be replayed off key ...


Enjoy it,
Paulo.

[DarkLord]

Hi. Sorry to report, but 0.847 doesn't work either. It simply freezes up with a mouse
pointer icon on a blank screen.

Listen, it's okay though - 0.844 is working fine so if you don't want to invest a lot
of time trying to patch for what is admittedly one single oddball machine, I do
understand.

Thanks for all your hard work on this.

[ljbk]

Hi. Sorry to report, but 0.847 doesn't work either. It simply freezes up with a mouse
pointer icon on a blank screen.

Listen, it's okay though - 0.844 is working fine so if you don't want to invest a lot
of time trying to patch for what is admittedly one single oddball machine, I do
understand.

Thanks for all your hard work on this.

Ok, thanks for the testing.
That is odd ...
Can you please reload the complete zip and try it in a separate directory as the .BLK files were not in the first zip available for download ?
If it still does not work, then the attached 847D version, copied to the same directory as above, should as it uses the same stuff applied to 844D.

Paulo.

[DarkLord]

And we're back working again.

I tested the complete archive first but it showed the same
behavior - locked up with a "busy bee" icon froze on the screen.

I copied the "D" version of the program into that directory and
it took right off, no problems.

Thanks much!

[ljbk]

And we're back working again.

I tested the complete archive first but it showed the same
behavior - locked up with a "busy bee" icon froze on the screen.

I copied the "D" version of the program into that directory and
it took right off, no problems.

Thanks much!

Again, thanks for the testing.
If it would not be asking too much, could you please test the attached E and F versions (to be tested like the D one).
I read some stuff about the high end 68K CPUs and found some references to other stack pointers like MSP and ISP instead of the 68000 SSP and also diferent stack frames for the exceptions.
This E and F cases tests 2 other possible common points to all models that might be our solution.

The idea is of course to have a single file working for all. But producing a D version takes less than a minute: it is just hardcoding that a STF was found and that shadow addressing is disabled: 3 lines at all.

Thanks again,
Paulo.