Super NES Digital Audio Mod

[zedrein]

I've posted a handful of threads over at the gamesx mod forum (http://www.gamesx.com) about this mod but I wanted to get more opinions on how best to accomplish this project as well as the potential pitfalls. This mod (http://www.alpha-ii.com/Info/snes-spdif.html) is definitely going to require some finesse as it calls for soldering onto a surface-mount IC that has extremely small traces to work with. Actual engineers have told me that this is going to be problematic and could potentially ruin my console because when you subject any IC -especially such a small one- to extreme heat and electrostatic energy, you are prone to electronic failures that manifest themselves in many ways. I just would like to know how best to approach this project in a way that's going to keep me from ruining my SNES while tapping the best possible audio signal from this machine. Thanks for your time!

[APE]

The hard part is finding a SNES that has the appropriate board inside as well as the fact the clock isn't 100% compatible with what is needed.

[Druid II]

So, all that is required to add s-pdif is a CS8406, a soic to dip pack, a toslink connector, and some wire to solder to the correct parts?

That sounds awesome and pretty easy to do, I wonder why no one did it before. It would boost up the audio quality soooo much, and allow for 1:1 accurate recordings from the console.

[zedrein]

So, all that is required to add s-pdif is a CS8406, a soic to dip pack, a toslink connector, and some wire to solder to the correct parts?

That sounds awesome and pretty easy to do, I wonder why no one did it before. It would boost up the audio quality soooo much, and allow for 1:1 accurate recordings from the console.

That's what I'm saying! Some guys over at the gamesx forum have done this mod and said that it's absolutely well worth it. The 60 Hz hum and other noise issues are completely eliminated and the signal-to-noise ratio is drastically improved. I am looking for more people that have done this so I can get some good advice.

[APE]

Every SNES I've ever opened has had a different board configuration than the one he used for the mod, this is the sole thing that has stopped me from attempting this mod.

That and my only coax input based speaker system died a few months ago.

[Druid II]

That's what I'm saying! Some guys over at the gamesx forum have done this mod and said that it's absolutely well worth it. The 60 Hz hum and other noise issues are completely eliminated and the signal-to-noise ratio is drastically improved. I am looking for more people that have done this so I can get some good advice.

I'm more interested in doing this for my Saturn than a SNES, mostly because the worthless shit play-asia rgb cable adds an insane amount of noise on brighter screens. Also, my current amp has no digital input, but my PC does, so I could just use that as a crossover and have 1 less thing hooked up.

[Ly-Colizer]

It would boost up the audio quality soooo much, and allow for 1:1 accurate recordings from the console.

But you need a soundcard then that can sync to 32khz (and can record at that rate)... my M-audio "Fast Track Ultra" can't record 32khz but can sync properly so you can at least listen to the noise free sounds:-)
To record 32khz do i have an older PCI soundcard "Delta Audiophile 24/96".

[Alchy]

my M-audio "Fast Track Ultra" can't record 32khz but can sync properly so you can at least listen to the noise free sounds:-)Have you tried capturing from the Microsoft Sound Mapper/MME-WDM/Wave Mapper (it may be called any one of these things in your audio app)? If your out-of-spec audio is playing fine in the background that should work.

[Ly-Colizer]

I can only record the s/pdif input by choosing "windows directsound" and record 8 channels and the 2 last is the s/pdif, so i need to delete the other 6 channels when the recording is done and make the 2 remaining a stereo track. This does not work for 32khz even if i hear the sound perfect... it just gives a distorted sound when trying to record it.

[la-li-lu-le-lo]

That sounds like a really cool idea.

On a related matter, there's something I've been wondering about for a while. Is there any difference between coax-based and optical (toslink) based S/PDIF? They both carry the same signal, right? Is there any reason to use one over the other?

[Alchy]

I can only record the s/pdif input by choosing "windows directsound" and record 8 channels and the 2 last is the s/pdif, so i need to delete the other 6 channels when the recording is done and make the 2 remaining a stereo track. This does not work for 32khz even if i hear the sound perfect... it just gives a distorted sound when trying to record it.What are you using to record?

[Druid II]

I can only record the s/pdif input by choosing "windows directsound" and record 8 channels and the 2 last is the s/pdif, so i need to delete the other 6 channels when the recording is done and make the 2 remaining a stereo track. This does not work for 32khz even if i hear the sound perfect... it just gives a distorted sound when trying to record it.
... You get whitenoise *during* recording, or after playing back the recording? It could be playback issue in the recorded file too, it makes no sense that you can play 32khz but can't record.

[Ly-Colizer]

What are you using to record?

In that example so did i mean "Fast Track Ultra".

Druid II: this is how FTU sound when i try and record when i have the S/PDIF (32khz signal) connected.. http://www.ym2149.com/ftu_spdif_32khz.flac

the audiophile 24/96 can record at several rates like 8khz, 9,600khz, 24khz, 32khz etc... but not at the sample rate it is synced to thru S/PDIF, only the preset rates, it does not sound as bad as the FTU, only playing at faster or slower speed depending on what rate you select to record at... example recording a 44.1khz sound at 32khz will make it sound slow and recording it at 48khz will make it sound too fast.

BTW: i am "Stefan_L" at the gamesx forum that zedrein linked to in the first post, i have done several S/PDIF mods on arcade games (and SNES) :-)

[Alchy]

In that example so did i mean "Fast Track Ultra".I meant what software package? Anyway, change it up and see if you get any further. If you're hearing it, Windows is coping with it, and that means one way or another you can record it.

[Calpis]

SPDIF hardware needs to be able to capture at any arbitrary rate, it does this by greatly oversampling the signal. If the control software outputs a 44.1kHz file just change the rate in the file header to whatever it truly is. Since the data comes out of a buffer it should still be lossless.

[Ly-Colizer]

If the control software outputs a 44.1kHz file just change the rate in the file header to whatever it truly is.

Yes i have done some tests of that earlier.
Here is an example recording from Taito F3 arcade PCB (Top Ranking Stars) that was recorded at 44.1khz and later i changed the sample rate of the recording to 29.762khz.
http://www.ym2149.com/spdif/top_ranking_stars_spdif_ingame.flac

[Druid II]

Oh, thats a Taito F3.... I was listening to it and thinking "DAMN best quality SNES sound I've ever heard".

[Calpis]

On a related matter, there's something I've been wondering about for a while. Is there any difference between coax-based and optical (toslink) based S/PDIF? They both carry the same signal, right? Is there any reason to use one over the other?
The only difference is in the physical transmission, they are the exact same signal. Personally I think optical is stupid and a waste of money. Coax can easily handle data rates in the MEGABITS lol. If SPDIF supported 7.1 24-bit PCM @ 192 kHz coax could handle it without much worry as it's ~10% of the cable's bandwidth.

[Druid II]

Optical has more bandwith (doesn't matter cause s/pdif is limited) and less signal degradation (you can use longer cables).

[Calpis]

But you can damage optical cables by bending them too far or stepping on them and they are unitaskers. More equipment will only have coax input/output so that's a good reason to use it too. I think Toslink was just a novelty of 1999.

[mr. newbie]

... Coax can easily handle data rates in the MEGABITS lol. If SPDIF supported 7.1 24-bit PCM @ 192 kHz coax could handle it without much worry as it's ~10% of the cable's bandwidth.

the cable may physically be able to transmit that much data, but even 5.1 lossless is beyond coax/spdif in actual application

we already have spc's. i can see someone doing the mod for fun, or for actual game playing, but you're gonna have a hard time matching an spc (which is ripped directly from the cart)

[Alchy]

we already have spc's. i can see someone doing the mod for fun, or for actual game playing, but you're gonna have a hard time matching an spc (which is ripped directly from the cart)You've got that backwards. SPCs have a hard time matching this, since they're emulations and this is the real deal.

[retro]

Absolutely. Where S/PDIF is used in a professional environment, it'll more than likely be via phono jacks, not optical. Actually, S/PDIF via co-ax is limited to 10m, whereas optical is only meant to be used with cables up to about 6m.

The clue is in the name TOSLINK. Here in the UK, toss means rubbish! :P

[mr. newbie]

You've got that backwards. SPCs have a hard time matching this, since they're emulations and this is the real deal.

spc's are sound information ripped directly from a cart. in fact, they can be played back on real hardware if you have a flash cart because the files are just instructions for the sound chip.

just because something is emulated does not automatically make it inferior to the original. i'm willing to bet in a blind test an spc is either going to be equal or better than a wav form an actual console. remember you're dealing with w/e noise or interference (jitter maybe?) is coming from the console, then it's a question of your sound card's quality in making the recording.

[Alchy]

just because something is emulated does not automatically make it inferior to the original. i'm willing to bet in a blind test an spc is either going to be equal or better than a wav form an actual console.Define "better". Better, for me, is more accurate. You can't get more accurate than the real hardware. If I wanted my SNES music to be tarted up there's no shortage of covers and remixes out there.


remember you're dealing with w/e noise or interferenceIt's a digital signal, if interference is making any difference you're doing something very wrong. I can't speak to jitter, that's outside of my experience really.


then it's a question of your sound card's quality in making the recordingEmulated SPC playback has the same issue, i.e. it's only as good as your playback hardware.

[Druid II]

You mean playback software for SPCs, not playback hardware.

[Alchy]

I mean both. At some point you've got to hear it and the hardware that facilitates that is pretty important ;)

[Calpis]

the cable may physically be able to transmit that much data, but even 5.1 lossless is beyond coax/spdif in actual applicationSPDIF is meant to be expanded as everything but STEREO audio is unsupported. Kludging on 7.1 wouldn't be that different than anything else not on the spec.

OT: the cable isn't just able to transmit vastly oversampled 7.1 (<30 Mbit = 30 MHz uncompressed digital signal), everyday coax is good for ~300 MHz which with QAM256 (52 Mbit/chan) is good for 200+ MiB/s of data.


we already have spc's. i can see someone doing the mod for fun, or for actual game playing, but you're gonna have a hard time matching an spc (which is ripped directly from the cart)If you're playing back via hardware the SPDIF rip WILL match the DSP's output. And you don't have to spend hours-days hacking game code or weeks-years writing an emulator for an unemulated/poorly emulated system (such as most arcade games). As for emulated SPC playback, it has a hard time matching the actual DSP output since even with a bit perfect DSP (I'm not sure emulation is there yet) the poor SPC emulation will cause phase errors in the audio. Obviously these errors are inaudible but they create bitstreams far less accurate than randomly cycling a real console. If you were only going by what's audible formats like VGM which do not emulate a CPU at all (they just contain periodic register writes) would be entirely sufficient.


You've got that backwards. SPCs have a hard time matching this, since they're emulations and this is the real deal.I'm sure you know in theory a SPC and DSP can be for all intents and purposes perfectly emulated (the jitter will be in the order of nanoseconds from propagation delay inherent in all circuits and there will be unpredictable phase differences between two different oscillators as there is in real hardware). In that respect every time you start up the console you could have a slightly different bitstream of the same music track. Also the analog filtering can be emulated to the component's ideal value. All this is not a reality as historically emulators have poor instruction timing and drop extraneous read/write cycles (only one or two emulators emulate a system to the clock cycle much less to the clock edge, much less logic gate propagation on an asynchronous output signal XD).


Where S/PDIF is used in a professional environment, it'll more than likely be via phono jacks, not optical.Thrifty and very technical pros will be using SPDIF, big pros should be using AES (XLR connector or something) since it can do a lot more apparently with sub data and probably all the studio software is for it.


just because something is emulated does not automatically make it inferior to the original. i'm willing to bet in a blind test an spc is either going to be equal or better than a wav form an actual console.Of course. It won't have unintended analog filtering or sample aliasing and it will have perfect dynamic range.


remember you're dealing with w/e noise or interference (jitter maybe?) is coming from the console, then it's a question of your sound card's quality in making the recording.
Jitter sounds like audiophile bullshit. Clock recovery should be very accurate on either interface. Any receiver should be able to handle 32 kHz since it's below the spec.

[mr. newbie]

actraiser filmoa (http://cid-b0252154262ad541.office.live.com/self.aspx/.Public/actraiser%5E_fillmoa%20%5E51%5E6.flac)

[/URL][URL="http://cid-b0252154262ad541.office.live.com/self.aspx/.Public/actraiser%5E_fillmoa%20%5E52%5E6.flac"]actraiser filmoa
(http://iegrrq.blu.livefilestore.com/y1po5eqV28BZ1NwHhKSg8C_OFvfCPebvctjZDqFLtE3Abw46zU 8rt3ki2Hqqi3Kjhz8hdZvR2e2Ivo3VZ8JkcbAfOTTULNzbUgk/actraiser_fillmoa%20%282%29.flac?download&psid=1)
one of these was captured from snes via spdif. the other is an spc.

[retro]

Thrifty and very technical pros will be using SPDIF, big pros should be using AES (XLR connector or something) since it can do a lot more apparently with sub data and probably all the studio software is for it.

Oh absolutely, AES/EBU is better, which is why I was careful to say WHERE it is used by pros ;) However, there are some items (usually semi-pro or older equipment) that only use S/PDIF.

mr. newbie, your links don't work.

[Calpis]

I listened to them and neither was spectacular. (1) is clearly the emulated SPC since it's 44.1 kHz which means the DSP output was interpolated, where (2) is 32 kHz. (2) was quieter, but I suspect that's the actual waveform and (1) has inaccurate DSP mixer emulation or a gain stage. Both of them would sound better if they were sent (via SPDIF lol) to a receiver instead of being (FURTHER) interpolated and attenuated/amplified (and possibly mixed with any other faint audio or microphone noise) by the operating system's sound mixer for playback.

[mr. newbie]

my bad about the links, they should work now

1. was actually the spdif output from this site (http://board.kohina.net/viewtopic.php?p=6358&sid=531abb210ad32bcc014467dd9d82ca71)

2. was converted from spc to flac via foobar.

my computer is connected to my receiver via spdif and i cannot hear any difference whatsoever. that's actually the reason i posted them.

[Ly-Colizer]

The S/PDIF Actraiser recording was done by me and compared to the emulated recording so does it sound sharper and i don't think it is because the recordings is at 44.1khz... it's more because of that the emulator is emulating/simulating the analog filtering that happens after tha DAC when the sound passes thru caps and resistors.

Maybe a new thread should be created as Zedrein might not be interested in "VS" discussions? :-)

[Calpis]

So the actual 32 kHz SPDIF rip was interpolated into 44.1 kHz during compression while the emulator output simulates analog filtering and is interpolated is still 32 kHz? XD this is not a very good test.


my computer is connected to my receiver via spdif and i cannot hear any difference whatsoever. that's actually the reason i posted them.How about with headphones? I can notice the difference with extremely poor netbook speakers. That wasn't what I meant though about being hooked up to a receiver; unless you have custom drivers for your sound card you aren't actually listening to the raw PCM, there is a single sample rate and bit depth which the OS resamples all audio to. So with a computer it doesn't really make a difference if you use a SPDIF receiver because it's not being clocked at 32 kHz anyway, the only improvement is that it's a digital signal longer.

[la-li-lu-le-lo]

The clue is in the name TOSLINK. Here in the UK, toss means rubbish! :P

I always assumed toslink was supposed to be pronounced tozz-link.

[Druid II]

Both recordings sound identical to me, outside the volume of course. (on my better-than-logitech-but-not-quite-home-theater 5.1 set)


How about with headphones? I can notice the difference with extremely poor netbook speakers. That wasn't what I meant though about being hooked up to a receiver; unless you have custom drivers for your sound card you aren't actually listening to the raw PCM, there is a single sample rate and bit depth which the OS resamples all audio to. So with a computer it doesn't really make a difference if you use a SPDIF receiver because it's not being clocked at 32 kHz anyway, the only improvement is that it's a digital signal longer.

The OS resampling depends on what OS you are using. XP uses kmixer, and I forgot how that works exactly. Vista and Win7 use some ridiculously high floating point format (64bit I think) for the internal mixer, and downmixes that to whatever format your output device supports (you can pick this value under sound properties, and even onboard can do 24bit 192khz nowadays!). Or rather, what the device drivers tell that it supports, cue in the SB Live and its fixed 48khz line where you could still select 44.1khz spdif output, that was twice resampled.

However, this quality is in far excess of what would make a Snes recording sound bad. Something like a FM sine waveform might sound bad with the crap hardware resampling of a Soundblaster Live (but not with any kind of software resampling), on PCM recordings you'd need extremely specific samples and equipment to notice issues. Or you could claim "this sounds bad!" because the spectograph of the recording is an ugly picture, this is what most audiophiles do anyway.

So to make my point: OS resampling means fuckall. The real issue is whether SPC files are emulated correctly or not, and as far as I know, the emulation itself is as close to 1:1 as possible thanks to byuu & co. So if you are using an up-to-date SPC player (foo_gep is one such thing I think), the only thing to degrade quality is improper SPC files - they are sound ram dumps from the SNES, done by emulators, so if an incorrect emulator sends junk to the sound chip, the saved SPC will be junk as well. Snesmusic has a huge selection of presumably correctly saved SPCs, but I'm not quite sure how they ripped them. Either way, they are better than the junk at Zophars Domain.

[Calpis]

OS resampling means fuckall? How do you know Win 7 doesn't use nearest-neighbor or linear interpolation? Or that your driver doesn't have an auto-equalizer to restore dynamic range compression / other crap you can't turn off? I'm not being an audiophile, just real; for a good listening test raw 32 kHz emulated SPC output should be compared to the real thing.

And as far as I was aware we WERE talking about nitpicking lossless waveforms, not what is necessary for casual listening. With this example however there is a difference, not just in dynamic range but tonally.

As things are now, both the SPC and DSP were RE'd through black boxing so yes, we know SPC are emulated well, but as emulator DSP output doesn't match the hardware DSP output barring output resampling, amplification and filtering so they aren't emulated as correctly as possible. The difference is for the most part inaudible, but the output is freaking digital, it should be correct*.

*except for any slight nondeterministic differences due to out of phase, unideal clock signals

[Druid II]

OS resampling means fuckall? How do you know Win 7 doesn't use nearest-neighbor or linear interpolation? Or that your driver doesn't have an auto-equalizer to restore dynamic range compression / other crap you can't turn off? I'm not being an audiophile, just real; for a good listening test raw 32 kHz emulated SPC output should be compared to the real thing.

A 32khz spc recording and a 32khz snes spdif recording would be both subjected to whatever your OS does to play back audio, so they are on equal footing if you play them back on your PC. Like I said, unless you are using some old soundblaster to play back sine waves, the audible difference caused by crap resampling is minimal (also, even if Win7 uses low quality linear interpolation, which it does not, you can still do HQ interpolation on your audio player).


And as far as I was aware we WERE talking about nitpicking lossless waveforms, not what is necessary for casual listening. With this example however there is a difference, not just in dynamic range but tonally.

As things are now, both the SPC and DSP were RE'd through black boxing so yes, we know SPC are emulated well, but as emulator DSP output doesn't match the hardware DSP output barring output resampling, amplification and filtering so they aren't emulated as correctly as possible. The difference is for the most part inaudible, but the output is freaking digital, it should be correct*.

*except for any slight nondeterministic differences due to out of phase, unideal clock signals

Last time I checked, they had matched the DSP filters and other stuff bit to bit, and recently got hires die shots as well. But I didn't pay attention to that work much, someone with more knowledge (byuu?) should confirm how things stand regarding SPC emulation.

[Ly-Colizer]

SPDIF hardware needs to be able to capture at any arbitrary rate, it does this by greatly oversampling the signal. If the control software outputs a 44.1kHz file just change the rate in the file header to whatever it truly is. Since the data comes out of a buffer it should still be lossless.

Now i don't remember exactly how i recorded the Actraiser (S/PDIF) recording, but i think it was at 44.1khz and then i changed it to 32khz using Audacity... but it still says it play at 44.1khz? well, anyway at least the speed was corrected.

[mr. newbie]

my concern is really whether one could differentiate an spc vs actual output in a blind test. i understand that we could legitimately have pages of fascinating talk about the intricacies of audio recording and playback equipment. I think the only measurement that matters is how similar the songs sound without being able to look at sample rates or anything else.

is this possible? can we contact anyone who has done the mod?

are you sure the audio re sampling is done in every case? i can set foobar to send my 48khz files to my receiver at 48khz (confirmed on my receiver's screen) and my 96khz files at 96khz etc.

i'm not trying to to question question your ears but are you sure your netbook speakers are the best devices to use? i used an onkyo ht-r340. the only headphones i have that work are koss portapros :(.

[Druid II]

are you sure the audio re sampling is done in every case? i can set foobar to send my 48khz files to my receiver at 48khz (confirmed on my receiver's screen) and my 96khz files at 96khz etc.
SPCs play back at 32khz in foobar, so yeah, they are resampled when you play them back.