Huzak - Yet another music driver

[elmer]

Here's the thing, we certainly would want that, it would make it a lot easier to optimise the song to use less patterns ...

OK, we'll have to see if Delek can be persuded to change that behavior in a future version, then.

Using less patterns would be a big benefit! 

You'd do an assembly-language-source ".include", rather than a ".incbin" or a HuC "#incbin".
...
I suspect that the current data format (without banking) is going to be limited to 2 or 3 tunes (there is maximum of 255 "patterns").

This sounds super complicated. Also, would it be possible to have one bank have all the sound effects?

I'm sure that it can all be wrapped up into something that's reasonably-easy to use. 

And "yes", it is fairly easy to have separate banks for music/effects if I move to a "binary" format for the music data.

That format would just put all of the tables that point to the interesting bits of the music data in constant locations in a bank. It will waste some space, but allow more flexibility.

It's probably more useful for a HuCard game than a CD game.

I just wanted to pop in and thank you for doing all this hard work Elmer, and Michirin9801 thank you for providing him with such useful examples, awesome work everyone!

You're welcome!

It's really good to hear all of this PCE music that's being created these days. I love it! 

Now you caught me off-guard, I'm unsure on how 'exactly' it's supposed to work, but from what I can hear on the tracker itself, setting it to 04xF really has a difference of 1 semitone both upwards and downwards, but it seems to linger on the higher pitch of the vibrato effect a little longer than it does on the lower pitch, but maybe I'm mishearing things...

Just so that we're on the same-page with the language ...

http://en.wikipedia.org/wiki/Vibrato
http://en.wikipedia.org/wiki/Finger_vibrato

And this has a good picture of "depth" vs "speed" ...

http://www.listening-singing-teacher.com/Vibrato.html


Finally, this link from Wikipedia, is a good example of a simple sine-wave vibrato ...

http://upload.wikimedia.org/wikipedia/commons/e/e7/Frequenzmodulation.ogg



It's a very "regular" and, to-my-ears, "pleasing" sound.


Here is my DefleMask test, which is your Misty Blue, with the 1st Matrix Row changed to have empty patterns everywhere except for one test-channel ...

http://www.dropbox.com/s/s98ax5vk1o75xca/vibrato-test.dmf?dl=0

Playing that in DefleMask gives what I'd describe as a wah-wah sound.

It seems audibly less "regular" than the example from Wikipedia.

It's a C1 note, with an $041F effect (speed 1, depth 15).

That's a Vibrato Speed of 1/64th cycle per 1/60s -> 64/60s -> 1.07s-per-cycle, and a Vibrato Depth of, according to the docs, "Maximum depth is a full note.".

OK, if I choose to read "note" as "tone" instead, then we're talking about a range of +/- 1 semitone.

+1 ... -1 = 2 semitones = 1 tone.


Now, looking at the data that's output in the .hes file, the original "period" (not frequency) is $06ae.
And with the vibrato, the period varies between $0673 and $073C.

Here's a quick table of the period values that surround C1 ...

C1# $064F
C1  $06AE
B0  $0714
A0# $0780

Looking at that, you can see that the frequency varies between - 1.5 semitones and + 0.5 semitones.

But looking at the timing data as well shows that the pattern of the vibrato is like this ...

-----
|   |
|   |
    |   |
    |   |
    |   |
    |   |
    |   |
    |   |
    -----

... instead of ...

-----
|   |
|   |
|   |
|   |
    |   |
    |   |
    |   |
    |   |
    -----

Which is why it *sounds* different, to my ear.

BUT, and it's a huge "but", there is nothing "wrong" in absolute terms in having that unbalanced vibrato waveform.

AFAIK it is a valid, if very unusual, choice.

I'm trying to figure out if it's a deliberate "choice", or just a math-error in Delek's implementation. 

[Michirin9801]

AFAIK it is a valid, if very unusual, choice.

I'm trying to figure out if it's a deliberate "choice", or just a math-error in Delek's implementation. 

Which is one of the many reasons why you need contact with him...
Either way, I think you should try to implement it the way he does, at least for now, so that it's more accurate to what it sounds like on Deflemask...

[elmer]

Either way, I think you should try to implement it the way he does, at least for now, so that it's more accurate to what it sounds like on Deflemask...

Unfortunately, I believe that you're right.

Which means that now I've got to figure out a reasonably-efficient way to recreate the imbalance. 

[Michirin9801]

Either way, I think you should try to implement it the way he does, at least for now, so that it's more accurate to what it sounds like on Deflemask...

Unfortunately, I believe that you're right.

Which means that now I've got to figure out a reasonably-efficient way to recreate the imbalance. 

Or maybe not? To be honest, I don't think it would be such a big deal to implement it the most CPU-efficient way possible... I mean, yeah it would sound slightly different, but I wouldn't mind the difference, and maybe some other people would find that it sounds better that way? Maybe?
The way I use the sine vibrato isn't because I want the note to play those very specific frequencies, I don't really care what frequencies it plays, I use it in a subtle manner because I want long notes not to be flat you know, and sometimes I also do it to make the channel stand out a little, without necessarily worrying about how far up and down in pitch they go or how long do they take to go up and down in pitch...

[elmer]

OK, sorry for the delay! 

It's been a bit of a battle to get DefleMask's Vibrato and Detune effects implemented, and doing so has bloated up the size of the driver because I've needed to allocate 2KB for the fast-multiply tables.

I've split the ROMs into 3 banks (for a 24KB ROM) ...

1) Startup code & Font (approx 1700 bytes)
2) Huzak & Tables-of-Squares (approx 4900 bytes)
3) Music Data (varies, the largest is approx 7KB)

Anyway, adding those effects opens-up a whole bunch of tunes that couldn't play correctly before, so here's 5-at-once ...

Fighting Back - Thunder Force IV
Poka-Poka Island - Super Adventure Island 2
Pressure - Flame Zapper Kotsujin
Sidecrawler's Dance - Wonder Boy III
Stage 1 - Sega's Quartet

http://www.dropbox.com/s/qpx9iognjg27l80/Huzak3rdWIP.zip?dl=0


Michirin9801 ... could you please listen to these and let me know if you hear any problems???

My ears are a bit too weary to be really discerning after listening to the driver mess up or crash so many times.

I *think* that these are right now, but I may be missing something.

<edit>

Yeah ... I need a 2nd-opinion, but I think that I may need to look at the drum roll at the end of Poka Poka Island ... it doesn't sound quite right. 

[Michirin9801]

OK, sorry for the delay! 

It's been a bit of a battle to get DefleMask's Vibrato and Detune effects implemented, and doing so has bloated up the size of the driver because I've needed to allocate 2KB for the fast-multiply tables.

I've split the ROMs into 3 banks (for a 24KB ROM) ...

1) Startup code & Font (approx 1700 bytes)
2) Huzak & Tables-of-Squares (approx 4900 bytes)
3) Music Data (varies, the largest is approx 7KB)

Anyway, adding those effects opens-up a whole bunch of tunes that couldn't play correctly before, so here's 5-at-once ...

Fighting Back - Thunder Force IV
Poka-Poka Island - Super Adventure Island 2
Pressure - Flame Zapper Kotsujin
Sidecrawler's Dance - Wonder Boy III
Stage 1 - Sega's Quartet

http://www.dropbox.com/s/qpx9iognjg27l80/Huzak3rdWIP.zip?dl=0


Michirin9801 ... could you please listen to these and let me know if you hear any problems???

My ears are a bit too weary to be really discerning after listening to the driver mess up or crash so many times.

I *think* that these are right now, but I may be missing something.

<edit>

Yeah ... I need a 2nd-opinion, but I think that I may need to look at the drum roll at the end of Poka Poka Island ... it doesn't sound quite right. 

Oh don't worry about that ^^
Taking your time to make something means that said something will most likely be well polished!
But it sure is sounding like a lot of optimization needs to be made to Deflemask considering the amount of data these songs are eating up >w>

About the songs:
- Sidecrawler's Dance sounds A-OK
- Fighting Back sounds almost A-OK, just one very tiny part in the 2nd half of the song has a little problem, when that note that does the portamento up and goes really high-pitched, it should have a little bit of vibrato at the very end, but the vibrato seems to be missing, or isn't strong enough, but otherwise it's spot-on...
- Quartet sounds a little weird, the detuning has gone a bit too far up in spots, particularly when the song loops, also, you're using a bit of an outdated version of the .dmf, I don't remember if I sent you the newest version but this sounds like the version I've posted on the Deflemask forum, most of what I've posted there is outdated >w>
Here's my current version of it: http://sta.sh/01k7o5ak0iz8 (It has a better intro than what was posted on Youtube)
- Pressure is too fast and the snare drums sound weird, that should be fixed when I can make Huzak-specific drums...
You should be able to get it with the correct tempo with this DMF: http://sta.sh/01jfb21ayfx5
- Poka Poka Island is pretty much A-OK, the only issue I can hear is when the song loops one of the channels seems to be turned off for a good chunk of the song, but other than that the only real issue is that I'm using wave-phasing for one of the instruments playing in the end of the song, so there's a little bit of popping, but I can live with it, it's barely noticeable...

[elmer]

But it sure is sounding like a lot of optimization needs to be made to Deflemask considering the amount of data these songs are eating up >w>

Yep, they're coming out bigger than I expected, but that's because you're doing a *lot* of stuff in them. 

With all of the panning changes that you've got going on, even if the track were redone in MML/Squirrel, it would probably still be large.

Removing some of the redundancy inside each of the patterns themselves would help a bit, and I could try to do that ... but I don't really think that we have a problem. 

The music data is still smaller than one tiny sample, and we're nowhere near as tightly constrained on memory as developers were back in the original System Card 1 days.

Some of your tracks are downright frugal ... it's just the ones that add all of the extra ear-candy that folks like to hear, and that makes them *feel* like complex tracks.

Note that something like the Xanadu PSG tracks are very sparse in comparison, and they'd probably be tiny in either MML or DefleMask.

**********

About the songs:
- Sidecrawler's Dance sounds A-OK

Excellent! 

- Fighting Back sounds almost A-OK, just one very tiny part in the 2nd half of the song has a little problem, when that note that does the portamento up and goes really high-pitched, it should have a little bit of vibrato at the very end, but the vibrato seems to be missing, or isn't strong enough, but otherwise it's spot-on...

Could you please let me know what matrix row and channel that's on, and I'll investigate.

- Pressure is too fast and the snare drums sound weird, that should be fixed when I can make Huzak-specific drums...
You should be able to get it with the correct tempo with this DMF: http://sta.sh/01jfb21ayfx5
- Poka Poka Island is pretty much A-OK, the only issue I can hear is when the song loops one of the channels seems to be turned off for a good chunk of the song, but other than that the only real issue is that I'm using wave-phasing for one of the instruments playing in the end of the song, so there's a little bit of popping, but I can live with it, it's barely noticeable...

I'll take a look at these next, but first ...

- Quartet sounds a little weird, the detuning has gone a bit too far up in spots, particularly when the song loops, also, you're using a bit of an outdated version of the .dmf, I don't remember if I sent you the newest version but this sounds like the version I've posted on the Deflemask forum, most of what I've posted there is outdated >w>
Here's my current version of it: http://sta.sh/01k7o5ak0iz8 (It has a better intro than what was posted on Youtube)

OK, I've used you new .dmf, and there's one small problem.

The new intro uses Porta-To-Note with really high speed values ... much higher than I've seen in any other DefleMask track. 

I've got a limit  of +/- 127 on the speed in order to use simpler (faster) math, and so my converter refuses to handle the track as it is written.

I changed the speed values to stay within range, and it messes up the sound of the intro, but at least I could keep on looking for the problem.

Can you live with that +/- 127 range? Perhaps Arp Macros could be used for that 1/5s "thump"


Anyway, I reworked the math on the detune, and now it's a lot more accurate.

The particular notes that you're talking about at the loop point are a bit troublesome.

You've got a couple of high notes in there which have steps of 7 and 8 "period" values to the next semi-tone.

You're using a detune of that's basically 1/16th of a semi-tone.

Now my old math was a bit hokey, but the new math turns those into 7/16th of a step, and 8/16th of a step.

That (now) gets rounded to detune values of 0 and 1 steps.

That math seems right now, but it may not be what you're expecting to hear.

I don't know if Delek forces a minimum value of "1" whenever the math rounds to "0".

Anyway, take a listen, and let me know what you think ...

http://www.dropbox.com/s/ls96xar85eb0uha/Huzak3rdWIP2ndTry-QuartetStage1.pce?dl=0


Thanks!

[Michirin9801]

- Fighting Back sounds almost A-OK, just one very tiny part in the 2nd half of the song has a little problem, when that note that does the portamento up and goes really high-pitched, it should have a little bit of vibrato at the very end, but the vibrato seems to be missing, or isn't strong enough, but otherwise it's spot-on...

Could you please let me know what matrix row and channel that's on, and I'll investigate.

Matrix row 25, channel 2

- Quartet sounds a little weird, the detuning has gone a bit too far up in spots, particularly when the song loops, also, you're using a bit of an outdated version of the .dmf, I don't remember if I sent you the newest version but this sounds like the version I've posted on the Deflemask forum, most of what I've posted there is outdated >w>
Here's my current version of it: http://sta.sh/01k7o5ak0iz8 (It has a better intro than what was posted on Youtube)

OK, I've used you new .dmf, and there's one small problem.

The new intro uses Porta-To-Note with really high speed values ... much higher than I've seen in any other DefleMask track. 

I've got a limit  of +/- 127 on the speed in order to use simpler (faster) math, and so my converter refuses to handle the track as it is written.

I changed the speed values to stay within range, and it messes up the sound of the intro, but at least I could keep on looking for the problem.

Can you live with that +/- 127 range? Perhaps Arp Macros could be used for that 1/5s "thump"


Anyway, I reworked the math on the detune, and now it's a lot more accurate.

The particular notes that you're talking about at the loop point are a bit troublesome.

You've got a couple of high notes in there which have steps of 7 and 8 "period" values to the next semi-tone.

You're using a detune of that's basically 1/16th of a semi-tone.

Now my old math was a bit hokey, but the new math turns those into 7/16th of a step, and 8/16th of a step.

That (now) gets rounded to detune values of 0 and 1 steps.

That math seems right now, but it may not be what you're expecting to hear.

I don't know if Delek forces a minimum value of "1" whenever the math rounds to "0".

Anyway, take a listen, and let me know what you think ...

http://www.dropbox.com/s/ls96xar85eb0uha/Huzak3rdWIP2ndTry-QuartetStage1.pce?dl=0


Thanks!

I've listened to it and now it sounds pretty much A-OK, it's not 100% accurate, but it's about 98.01% accurate which is totally acceptable, if anything it no longer sounds weird so that's great!
As for the intro, It's just a matter of me reworking it, and I have an idea of how to do it... Just don't expect me to do it too soon because I don't feel like doing it right now >w>

Note that something like the Xanadu PSG tracks are very sparse in comparison, and they'd probably be tiny in either MML or DefleMask.

And yet they still sound better than anything I can put out =w=';
But that has more to do with music theory than actually pushing the limits of the soundchip, I'm just a self-taught amateur who loves detuned reverb...

[elmer]

Matrix row 25, channel 2

Thanks, I'll take a look!

I've listened to it and now it sounds pretty much A-OK, it's not 100% accurate, but it's about 98.01% accurate which is totally acceptable, if anything it no longer sounds weird so that's great!
As for the intro, It's just a matter of me reworking it, and I have an idea of how to do it... Just don't expect me to do it too soon because I don't feel like doing it right now >w>

I'm glad that it sounds OK now, but I'd like to get as-close-as-possible-within-reason.

I've posted on the DefleMask forum asking Delek for the exact math that he's using, and if he responds, we'll see if it is implementable.

As for the portamento ... no rush. I'll take another look at the code and see what the cost is of implementing the full 8-bit range.

[Michirin9801]

As for the portamento ... no rush. I'll take another look at the code and see what the cost is of implementing the full 8-bit range.

That would be very nice ^^

I'm also looking forward to any progress on the other songs, but no rush mate, take your time and make sure they're working well okay? ;3

[elmer]

That would be very nice ^^

I wanted to do some timing tests to see how much CPU time the driver is currently using before seeing what it'll take, and it looks like I don't need to be too paranoid, IF it can be done reasonably nicely.

The following are worst-case costs, because that's what a programmer needs to budget for.

The average cost looks like it will be rather lower.

"Pressure", "Quartet" and "Sidecrawler's Dance" are maxing out at 5.1% of CPU time.

"Poka Poka Island" is a bit higher.

"Fighting Back" is using the most time at 5.5% of CPU time.

That includes processing all the empty SFX channels, and allowing for the channel-overrides.

That's not at all bad, and makes me thing that we'll be well-within the 10% CPU budget that I'm used to leaving for music/SFX.

Sample playback will, of course, play havoc with that, but there's not much I can do about that.

I'm also looking forward to any progress on the other songs, but no rush mate, take your time and make sure they're working well okay? ;3

Thanks for letting me know where to look for that lack-of-vibrato in "Fighting Back".

It was actually playing, but it wasn't deep enough.

I've applied the same more-accurate-math fix from the Detune code to the Vibrato code, and Fighting Back sounds much better now.

"Pressure" seems to sound OK to me with the latest .dmf. I'll upload some new ROMs later on.

That last builds were using the same "undefined" $1F noise setting as DefleMask, but I've switched that back to the legal $1E setting again now.


One thing that I'm noticing in "Pressure", and a few of the other tracks, is that you're using Arp Macros that keep on looping back 1 step and repeating the very last note in the macro.

Is there a reason for this? Does DefleMask stop doing something when that Arp Macro ends?

For Huzak ... it costs time to keep on processing that Arp Macro change, even if it's not really a change, and it may also mess up the logic in the driver, because things like Portamento and Vibrato are skipped while  there is an active Arpeggio (either the effect, or the Macro).

That's because there doesn't seem to be any sane way of processing frequency shifts (Portamento/Vibrato) at the same time that you're forcing a different note anyway.

Am I missing something that you'd want to do as a musician? 

[Michirin9801]

One thing that I'm noticing in "Pressure", and a few of the other tracks, is that you're using Arp Macros that keep on looping back 1 step and repeating the very last note in the macro.

Is there a reason for this? Does DefleMask stop doing something when that Arp Macro ends?

For Huzak ... it costs time to keep on processing that Arp Macro change, even if it's not really a change, and it may also mess up the logic in the driver, because things like Portamento and Vibrato are skipped while  there is an active Arpeggio (either the effect, or the Macro).

That's because there doesn't seem to be any sane way of processing frequency shifts (Portamento/Vibrato) at the same time that you're forcing a different note anyway.

Am I missing something that you'd want to do as a musician? 

When you use an instrument that uses the 'Fixed Arpeggio' in Deflemask, as soon as that arpeggio effect ends the note you've put in the note column will play like normal unless you add a loop to the arpeggio macro, that loop is there in order to make drums easier to deal with on the tracker, so I can just put whatever note I feel like on the channel without a care in the world and the drum will play the same way every time, so I can just use C notes for everything except for when I need the drums to do different tones (like the Toms for example)

That last builds were using the same "undefined" $1F noise setting as DefleMask, but I've switched that back to the legal $1E setting again now.

If I may make a suggestion, whenever an instrument uses the $1F noise frequency, play back the $1E frequency but with its volume lowered by 4 okay?
I've already adjusted my Hi-Hats to not use the $1F frequency and I'll be using only the legal $1E for all my future songs, and from what I could hear, lowering the volume by 4 produces the closest/most acceptable results...

[elmer]

When you use an instrument that uses the 'Fixed Arpeggio' in Deflemask, as soon as that arpeggio effect ends the note you've put in the note column will play like normal unless you add a loop to the arpeggio macro.

Yeeks! That's not in the manual (I think). Then again, so much isn't.

Is that the same with the "normal" Arpeggio Macros ... i.e. it returns to the base note when it ends???

I can see that would make some sense, especially if that's how it works on the "fixed" Arpeggio Macro.


Do either of the other Volume or Wave Macros have similar features?

I think that I've seen you loop a Volume Macro once-or-twice.

Thanks! 

If I may make a suggestion, whenever an instrument uses the $1F noise frequency, play back the $1E frequency but with its volume lowered by 4 okay?

I could certainly do that, at least temporarily, just to get stuff working.

But I really, really, really don't like doing it long-term, because even testing for that value just wastes CPU time.

I'll probably implement it like that in a day-or-so when the my horror at the sheer-waste dies down a little.

It's a good idea ... thanks! But ... YUK! 

[Michirin9801]

When you use an instrument that uses the 'Fixed Arpeggio' in Deflemask, as soon as that arpeggio effect ends the note you've put in the note column will play like normal unless you add a loop to the arpeggio macro.

Yeeks! That's not in the manual (I think). Then again, so much isn't.

Is that the same with the "normal" Arpeggio Macros ... i.e. it returns to the base note when it ends???

I can see that would make some sense, especially if that's how it works on the "fixed" Arpeggio Macro.


Do either of the other Volume or Wave Macros have similar features?

I think that I've seen you loop a Volume Macro once-or-twice.

Thanks! 

Nope! That only happens with the Fixed Arpeggio, doesn't happen with the relative arpeggio or the volume macros, also, looping volume envelopes is a great way to get tremolo, MUCH better than doing it in the effect column... (Take note of that Fragmare! I remember you wanted to do a tremolo in Simon Belmont's theme but wanted more control of it, doing it this way is really useful!)

If I may make a suggestion, whenever an instrument uses the $1F noise frequency, play back the $1E frequency but with its volume lowered by 4 okay?

I could certainly do that, at least temporarily, just to get stuff working.

But I really, really, really don't like doing it long-term, because even testing for that value just wastes CPU time.

I'll probably implement it like that in a day-or-so when the my horror at the sheer-waste dies down a little.

It's a good idea ... thanks! But ... YUK! 

Oh you don't really have to do it long-term, when Deflemask has its PCE quirks fixed I'm sure you'll be able to optimise Huzak quite a bit to not waste nearly as much CPU time ^^
In the meantime though, it will be a useful thing to have...

[elmer]

Nope! That only happens with the Fixed Arpeggio, doesn't happen with the relative arpeggio or the volume macros, also, looping volume envelopes is a great way to get tremolo, MUCH better than doing it in the effect column...

What!!! That's crazy!!! Why do it in the Fixed Arp, but not in the Relative Arp??? 

OK, whatever, it's easy-enough to handle, now that I know about the behavior.

Using the Volume Macro for Tremolo makes a lot of sense, and it's a lot more controllable than the effect.

But I could have sworn that I'd seen you do a repeat of just the last value in a Volume Macro.

If there's no "special" meaning to that, then it might be temping to automatically remove it during conversion.

I *think* that it might be used by folks that want to fade up/down from a Volume Envelope, by using the Volume Slide of individual Volume settings.