I get your point, but - to use your analogy - do we know of some song/sound that’s the equivalent of ASP/Desert Fighter’s drop shadow appearing in bsnes-accuracy/higan but not in snes9x?
Some people are mentioning Comix Zone and Scooby-doo mystery as problematic games.
edit: Tested it and I can hear the difference on those games.
Drums and percussions are notably more defined, stereo separation is wide.
Castle of Illusion is a game where it’s really easy to spot I think. Lot of different instruments there.
It’s really good, great addition.
El Viento intro sounds so nice too now. (and the whole game as well )
Should this new YM2612/YM3438 nuked opn2 setting be used with or without the Lowpass audio filter ?
Thanks for this excelent work!!! By the way, Castlevania Bloodlines have missing vibrato in the music with the Nuked-OPN2 audio core in Genesis Plus GX. Someone can confirm if this is the same in the real hardware or is an issue in the Genesis Plux GX (or an issue in Nuked-OPN2)? Thanks!!!
I personally don’t, but Nuked-OPN2 is very young, maybe someone sometime will find differences.
In DOSBox with Nuked-OPL3 is was easy to compare to the hardware because I had full control of the YMF262 chip on my Sound Blaster cards with a dos sequencer called Adlib Tracker 2 so I could create custom sounds and compare the two directly, with Nuked-OPN2 vs MAME cores it would be not so easy because I don’t have any custom access to the YM2612 chip on a hardware Genesis except games.
With Lowpass.
Nuked-OPN2 has the same output, volume, and instrument balance as the Mame OPN2 core, but it accomplishes the result by more accurate code.
What BGM? Are you sure?
In game go into Options screen and test BGMs.
Most game have music/sound test in their options.
Hmm, the MAME core is very good, I don’t think the differences should be that obvious.
Thanks, that’s what I was using but wanted be certain it was correct.
In all BGM’s. In the BGM 06 the difference is more noticeable.
Great catch!
Indeed the vibrato has less range with Nuked-OPN2.
In this case Mame core sounds like the hardware, while Nuked has a bug.
I will report to nukeykt about this bug ASAP.
I think it is fixed now, but still not merged.
Meanwhile here is a comparison of Model1 VA6 vs Model2 VA4, both considered best in their model number.
Model2 is MUCH more filtered!
Hopefully one day GPGX will have the option to choose between first and second order lowpass filtering so that the user can choose freely… it wouldn’t hurt if I ask ekeeke to add second order filtering.
Streets of Rage 2 is about half way, but all the video provides good example of the filtering order and how it compares.
There is another option YM2612 DAC quantization. What is it and should it be on or off for maximum accuracy?
“YM2612 DAC quantization” should be ON to quantize from 14bit to 9bit like the real chip does.
The result is ADDed dithering noise while the FM channels are playing.
Without it, the FM emulation is crystal clear and some games might not sound correctly which rely on that added noise from the chip.
To hear what this option does;
Streets of Rage 2 - BGM 10, intro.
Listen with headphones and switch the DAC option On/Off.
Here is BGM 10 on a real Model1 VA1: https://www.youtube.com/watch?v=QwMchuT4SsE&list=PLEeAvoMXmmr8NBw_2tnb4FjHlrTED4LJY&index=16
Note the emulation is not perfect and still is much cleaner and less distorted than the hardware YM2612.
The DAC noise is the biggest difference between the YM2612 and YM3438, with YM3438 being the newer and quieter variant in the Model2 with much less noise than the original discrete YM2612 in Model1, but the quantization to 9bit is the same on all models, so the DAC option in GPGX should be ON.
This difference in noise is one of the differences in audio between Model1 and Model2, but the Model2 had badly engineered amplification circuitry after the clear YM3438 which made it sound distorted, that’s why many people modify the Model2 with “Mega Amp” circuitry which replaces the miss-engineers circuitry with a proper one.
More info here: http://www.sega-16.com/forum/showthread.php?25531-Discrete-Vs-Integrated-YM3438-Sega-Genesis-music-blind-test-results
Okay so to make sure I got this all correct, to get that Model 1 Genesis sound the settings should be
YM2612 DAC quantization - On
YM2612/YM3438 core - nuked opn2
Audio filter - low-pass
Sound output - mono?
Actually, you should wait for the next build, we’re all in for a big treat!
NukeYKT has fixed all the bugs and added accurate DAC quantization and “Ladder Effect” emulation to his cores.
Basically, it is the first OPN2 emulation that sounds exactly like the hardware including all the noise and distortion of the YM2612 DAC.
“DAC quantization” option applies only to the Mame core, Nuked cores have everything already included.
He included three new options for the Nuked-OPN2 varitions:
- YM2612
- Asic YM3438
- Discrete YM3438
Here are some examples where there is clear difference.
In the archive I included my Model1 VA6 recording and all new OPN2 core variants.
The clips have been professionally recorded and normalized to be same volume.
Aladdin: listen for intro noise and 16sec note stops tail quantization.
Konami Logo: listen for music end tail quantization noise.
Can’t wait, so glad this stuff is being taken care of.
Here’s test build compiled by nukeykt:
https://drive.google.com/file/d/0B4n8cpbZ9qicQ1JQNHc5LW9odEk/view?usp=drivesdk
Rename the proper version to genesis_plus_gx_libretro.dll
Great improvement!!! The DAC emulation of nukeykt is very impressive!!!
By the way, I have 2 questions:
- The lowpass filter is present in the Model 2 with ASIC YM3438 or only in the Model 1?
- The lowpass filter affects the CDDA audio of Sega CD in the real console?
Thanks!!
- Yes, it is even more filtered than Model 1.
- No.
so the bug fixed core still isn’t merged, is it? Just tried the latest Genesis Plus GX core, and the vibrato in Castlevania is still incorrect.
What is the best setting to have the Model 1 sound?
Thanks!