Genesis - SNES audio comparison

[sheath]

So, in the YM2612 case we can calculate it's output samplerate. Here I'd done some calculations, based on oscilloscope measurement.

Yes, this is what I'm looking for. What my ears tell me is that anything we might measure from both sounds outputs will end up being very similar. I don't have an oscilloscope, so it'd be great if we could get some sort of technical comparison posted. Even if I have to save the pictures and post the data myself, it'd be better than what we have (or don't have as it is).

[sheath]

Okay, so if the oscilloscope reads 20Khz, but audio doesn't approach sounding right in emulation until its set to 44.1 Khz it looks like we have a discrepancy. I also noticed when I was making digital recordings of music and sound from the actual console that if I recorded at anything lower than 44.1 it was noticeably lower quality and flatter than the actual output sounded. I know that the conversion from analog to digital has never been truly mastered, but 44.1 Khz was the point at which I was happy with the precision of the sound quality.

[tomaitheous]

I believe the rule is, whatever the approximation of the analog signal is in digital/quantitative resolution - you need to sample at twice the rate when digitizing to capture all of the signal.

[sheath]

That makes sense, and I think I found an article online last week that said the same thing about the analog to digital conversion needing a double sample rate. Unfortunately, I don't think I could use this to prove any differences in original sample rates.

[tomaitheous]

you can't really make 8KHz or less sample sound crisp and clean.

I'm not sure I understand what you're saying. Amiga MOD instruments are based on 8khz samples and they sound great. And that's 8bit PCM vs the DSP 16bit output. The DSP also interpolates in between sample values to give a much smoother sound - eliminating artifacts for 'nearest neighbor' resampling. The frequency is fixed output (iirc 32khz driven from a 1mhz input), not 8khz. That would just be the rate as which you are playing the sample for pitch bending it into a note.

[Near]

I'm not sure I understand what you're saying. Amiga MOD instruments are based on 8khz samples and they sound great. And that's 8bit PCM vs the DSP 16bit output.

I would posit that 8-bit PCM sounds better than 16-bit PCM encoded as 4-bit ADPCM, but I haven't listened to the two side-by-side before.

Real SNES sounds nothing like (most) emulators do

Fixed that for you
blargg's S-DSP emulator is bit-perfect except for one single edge case -- toggling mute on and off quickly makes the real chip go batty for a few samples.

[Chilly Willy]

Early Amiga music used 8 kHz sample rates because the Amiga came with only 512 KB of memory. You really didn't have space for higher frequency (larger) instruments until later. Because of that, the sound channel had a low-pass filter at about 8 kHz to help cut the noise. When the A2000 and A500 were released, MODs (and game music in general) were often using higher frequencies, so the A2000 and A500 both had a way to switch off the sound filter (you turned off the power LED).

The Original ChipSet was capable of sample rates up to 27 kHz, and many games used 22 kHz samples for sounds and music. MODs also started doing so as well. With the Enhanced ChipSet (and AGA chipset), you could get sample rates up to 54 kHz as long as the chipset was set for a VGA mode instead of the default NTSC/PAL modes. So by the end of the Amiga era, people were doing 48 kHz sounds and music on the Amiga.

While the samples could only be 8 bit, there was a way to chain two channels together so that one did the samples, and the other did the volume. The volume was a 6 bit log envelop, so you couldn't get exactly 14 bits out of it (8+6). After correcting for the log function, it was closer to 12 bits. However, it sounded great and was pretty easy to generate. So in the end you saw ~12 bit, 48 kHz sound and music for the built in audio.

[Snake]

I would posit that 8-bit PCM sounds better than 16-bit PCM encoded as 4-bit ADPCM

I would agree with you.

I don't buy this "low quality genesis DAC" crap. The only distortion you get from the YM2612 is from the FM channels, where the output to the DAC is more than 8 bits. It's fine for 8 bit, and the samplerate is way higher than the SNES can do.

Whereas SNES is pretty much limited to it's 64KB RAM for samples, the Genesis is only limited by how much space you can spare on the cartridge. The main reason for low quality samples on the Genesis is because the sound driver needs to do other things as well as write samples. But the system is capable of excellent quality, and there have been some good examples of this.

However you can't compare SNES and Genesis sound. At all. They are very different. Personally, I much prefer a system that allows for decent synthesis and not just samples. I think most musicians would agree.

[Huge]

I would posit that 8-bit PCM sounds better than 16-bit PCM encoded as 4-bit ADPCM, but I haven't listened to the two side-by-side before.

I can only rely on PC modplayers, but for the few tunes that exist both on Amiga and SNES, the Amiga versions are ALWAYS much clearer. Higher quality samples, most likely.

Pinball Fantasies is one such game I'm thinking of. Chuck Rock is another, but it has the music rewritten so its not such a great comparison. Pinball Fantasies has the theme song (Steelchambers2) intact on the Snes, in its original form, only with much worse samples, like if I was playing the original mod tune with 8khz output.

[Shiru]

Personally, I much prefer a system that allows for decent synthesis and not just samples. I think most musicians would agree.

That very depends from type of music you want to make - techno or symphonic, for example. Both SMD and SNES solutions are not best, they are too polar. I think, Neo Geo, Capcom Play System and Sega 16-bit arcade machines had better compromise - they used other Yamaha FM synth chips with hardware ADPCM support (Neo Geo - YM2610, 4 FM, 7 ADPCM, 3 PSG channels). Such setup is much more flexible.

[TmEE co.(TM)]

Early Amiga music used 8 kHz sample rates because the Amiga came with only 512 KB of memory. You really didn't have space for higher frequency (larger) instruments until later. Because of that, the sound channel had a low-pass filter at about 8 kHz to help cut the noise. When the A2000 and A500 were released, MODs (and game music in general) were often using higher frequencies, so the A2000 and A500 both had a way to switch off the sound filter (you turned off the power LED).

Those filters cut out a lot of good sound too, not just noise, and you don't have a possibility to turn them off on most devices... I'm rather positive than an Amiga without that filter at all sounds much nicer than with the filter... when I do get an Amiga sometime, the that filter is the first thing that's going to fly out from the system. Crisp and slightly noisy sound is better than dark and muffled sound. SNES and MODs sound better with interpolation and filtering turned off.

Personally, I much prefer a system that allows for decent synthesis and not just samples. I think most musicians would agree.

I like mixture of both, but the samples do have to sound good (especially in case of percussion).
An 8ch FM chip + ADPCM chip along with PSG would have been sweet to have in MD. The ADPCM chip would have needed its own RAM which would have driven the cost of MD quite a bit higher... more than 64K of RAM for sound... or something like NES and NeoGeo use, ROMs for each part of the system.

[Chilly Willy]

Early Amiga music used 8 kHz sample rates because the Amiga came with only 512 KB of memory. You really didn't have space for higher frequency (larger) instruments until later. Because of that, the sound channel had a low-pass filter at about 8 kHz to help cut the noise. When the A2000 and A500 were released, MODs (and game music in general) were often using higher frequencies, so the A2000 and A500 both had a way to switch off the sound filter (you turned off the power LED).

Those filters cut out a lot of good sound too, not just noise, and you don't have a possibility to turn them off on most devices... I'm rather positive than an Amiga without that filter at all sounds much nicer than with the filter... when I do get an Amiga sometime, the that filter is the first thing that's going to fly out from the system. Crisp and slightly noisy sound is better than dark and muffled sound. SNES and MODs sound better with interpolation and filtering turned off.

Yes, I liked the sound with the filter off as well. I wrote a kicktag module that turned the filter off every time you rebooted the system. You can also find instructions on aminet on how to permanently disable the filters on various models. I preferred the software approach rather than hacking up the Amiga... I did enough of that installing a fatter Agnus, a 68030 accelerator, 2.5 MB of slow memory, and a dual kickstart board.

[Near]

Whereas SNES is pretty much limited to it's 64KB RAM for samples, the Genesis is only limited by how much space you can spare on the cartridge.

This is not entirely true. While the SMP and DSP share only 64kb of RAM, the CPU and SMP can communicate with each other to transfer more sound data while current data is playing back.

In fact, it has a perfect mechanism for it ... HDMA. Have it send a few samples every scanline, and time your SMP code, have them sync between frames ... and you can stream voice-quality audio from the cart, without having to tie up the CPU to handshaking.

Given, it's not particularly amazing ... 8-12KHz or so quality, but it gets the job done. See Breath of Fire II - German by d4s, Tales of Phantasia opening, Star Ocean prologue or Earthworm Jim 2 sound effects for examples.

[TmEE co.(TM)]

its still using another resource to achieve something... I love how Z80 has a lot of control over outside its own area... too bad accessing RAM does not work !!!

[tomaitheous]

The SPC method looks to be less restrictive/complicated than the z80 driver. I mean, you don't have to worry about stalling the z80 with VDMA or such.

Even with the slow communication/transfer to the SPC, copying over ~134 bytes per frame to the SPC, even if it's a once per vblank bulk copy, can't be that taxing on the main CPU for 16khz ADPCM sample streaming. That hardly comes across as resource hogging. Probably around 1-2% cpu resource, let alone the faster HDMA method byuu mentioned (I wasn't aware that HDMA could transfer to the SPC ports).

I'm not sure what the talk about the Amiga pass filter has to do with the SPC DSPs sample point to point interpolation per channel/waveform. Unless I'm missing something here?

While the samples could only be 8 bit, there was a way to chain two channels together so that one did the samples, and the other did the volume. The volume was a 6 bit log envelop, so you couldn't get exactly 14 bits out of it (8+6). After correcting for the log function, it was closer to 12 bits. However, it sounded great and was pretty easy to generate. So in the end you saw ~12 bit, 48 kHz sound and music for the built in audio.

Wait! You're talking about the mode where you can turn one channel into amplitude modulation mode and the data in that channel modulation the volume of it's target channel. So that's how it's done for the 14bit mode? That's cool

I can only rely on PC modplayers, but for the few tunes that exist both on Amiga and SNES, the Amiga versions are ALWAYS much clearer.

A 4 channel original MOD format!? I mean, it doesn't even have volume envelopes and such that SPC and other formats like XM have. Check out Satellite One by Purple Motion XM format auto-converted to the SPC. Sounds awesome.

Pinball Fantasies has the theme song (Steelchambers2) intact on the Snes, in its original form, only with much worse samples, like if I was playing the original mod tune with 8khz output.

Sounds like 8khz fixed output? You're either exaggerating ( ) or that was just a good example of a really poor conversion.