Homebrew RetroArch - A new multi-system emulator

[AntonioJosedeSou]

Audio samplerate frequency has absolutely nothing to do with the TV, the wii outputs an analog audio signal and the sampling is done internally.
48khz maybe takes more processing power since it needs to emulate more samples per seconds but the wii only supports 32 or 48 khz anyway so the 44.1 khz audio stream generated by the emulator has to be resampled by retroarch before sending it to audio hardware anyway and audio resampling isn't exactly without speed penalties.

By the way, stand-alone GenPlusGX always used 48 khz by default and does not have any issues. It's kinda odd to use a fixed samplerate for the libretro port since resampling is not only CPU taxing but will also affect audio quality. I might be wrong but I think retroarch always need to resample the emulator output by default because of that dynamic rate control thing though.

I have to admit I don't know what frequencies the Nintendo Wii supports. I thought it could support 44.1000 Hz as well.
Anyways, GenPlus GX does emulate the Genesis audio frequency at it original frequency that is 44.100 Hz.
That later on is resampled to 48.000 Hz using Blip Buffer is another story.

Now, if you read the change log of the latest RetroArch, it says that they changed the frequency from 44.100 Hz to 48.000 Hz.
So it's logic to assume that RetroArch GenPlus GX output 48.000 Hz directly without emulate the Genesis original frequency first.
At least, that is what I think he means with change from 44.100 Hz to 48.000 Hz.

And I didn't mean that change the frequency to 48.000 Hz had anything to do with the TV (actually I said HDTV but whatever).
What I meant is that's the standard that all the multimedia companies is using now.

Edit: I forgot to mention that RetroArch GenPlus GX 0.9.7.2 didn't have this "issue".

 

[the_randomizer]

Star Fox does not run full speed on Retroarch ( yet )

I don't know if it ever will, since the Wii is apparently too "weak" to handle emulating such a taxing game.

I have to admit I don't know what frequencies the Nintendo Wii supports. I thought it could support 44.1000 Hz as well.
Anyways, GenPlus GX does emulate the Genesis audio frequency at it original frequency that is 44.100 Hz.
That later on is resampled to 48.000 Hz using Blip Buffer is another story.

Now, if you read the change log of the latest RetroArch, it says that they changed the frequency from 44.100 Hz to 48.000 Hz.
So it's logic to assume that RetroArch GenPlus GX output 48.000 Hz directly without emulate the Genesis original frequency first.
At least, that is what I think he means with change from 44.100 Hz to 48.000 Hz.

And I didn't mean that change the frequency to 48.000 Hz had anything to do with the TV (actually I said HDTV but whatever).
What I meant is that's the standard that all the multimedia companies is using now.

Edit: I forgot to mention that RetroArch GenPlus GX 0.9.7.2 didn't have this "issue".

Well, like I said before, the standalone version of 1.7.3 doesn't have this issue, but when used as a core in RetroArch 0.98, it does. I don't know why they can't revert back to the 44100 Hz it was before. If 48,000 Hz is more CPU intensive, why use it? Surely, there must be a way to fix it. I use a 2.1 speaker system as TV speakers sound like garbage and have no bass.

Standalone 1.7.3 - No tinny audio
RetroArch 1.7.3 - Tinny audio

Seriously, what the crap? Surely, LibretroRetroArch can look at it and fix said issues, can he not?

Audio samplerate frequency has absolutely nothing to do with the TV, the wii outputs an analog audio signal and the sampling is done internally.
48khz maybe takes more processing power since it needs to emulate more samples per seconds but the wii only supports 32 or 48 khz anyway so the 44.1 khz audio stream generated by the emulator has to be resampled by retroarch before sending it to audio hardware anyway and audio resampling isn't exactly without speed penalties.

By the way, stand-alone GenPlusGX always used 48 khz by default and does not have any issues. It's kinda odd to use a fixed samplerate for the libretro port since resampling is not only CPU taxing but will also affect audio quality. I might be wrong but I think retroarch always need to resample the emulator output by default because of that dynamic rate control thing though.

Sorry to ruin the fun, but having it fixed at its current sample rate makes it sound worse than before; my ears are very sensitive to things like this and I am not imagining it. Genesis Plus GX sounds tinnier than usual in the latest version of RetroArch. And yet the separate 1.7.3 does NOT have this issue at all. It has to be how RetroArch samples/resamples the audio when it outputs, same applies with Nestopia and its tinny sound output. I'm not some nut who's hearing auditory hallucinations. I am convinced that these issues can be fixed. Going to 44 to 48 KHz for the sampling rate seems kind of a moot point to me. The real Genesis wasn't 48 KHz, it was set at 44.1 KHz, and the NES was set at variable rates, and the Snes was set at 32 KHz. I don't see why this can't be looked at or taken seriously.

 

[Jacobeian]

I have to admit I don't know what frequencies the Nintendo Wii supports. I thought it could support 44.1000 Hz as well.
Anyways, GenPlus GX does emulate the Genesis audio frequency at it original frequency that is 44.100 Hz.
That later on is resampled to 48.000 Hz using Blip Buffer is another story.

Now, if you read the change log of the latest RetroArch, it says that they changed the frequency from 44.100 Hz to 48.000 Hz.
So it's logic to assume that RetroArch GenPlus GX output 48.000 Hz directly without emulate the Genesis original frequency first.
At least, that is what I think he means with change from 44.100 Hz to 48.000 Hz.

And I didn't mean that change the frequency to 48.000 Hz had anything to do with the TV (actually I said HDTV but whatever).
What I meant is that's the standard that all the multimedia companies is using now.

Edit: I forgot to mention that RetroArch GenPlus GX 0.9.7.2 didn't have this "issue".

Where did you see the Genesis was outputting at 44100 hz ?
That does not make much sense, the console is mixing analog audio signals coming from multiple custom sound chips, one of them being a Yamaha FM synthetiser chip and those things aren't really comparable to modern digital audio cards. It's very unlikely they were using such a standard samplerate, actually the only thing being output at 44100 hz are probably Sega CD audio tracks.

 

[LibretroRetroArc]

Audio samplerate frequency has absolutely nothing to do with the TV, the wii outputs an analog audio signal and the sampling is done internally.
48khz maybe takes more processing power since it needs to emulate more samples per seconds but the wii only supports 32 or 48 khz anyway so the 44.1 khz audio stream generated by the emulator has to be resampled by retroarch before sending it to audio hardware anyway and audio resampling isn't exactly without speed penalties.

By the way, stand-alone GenPlusGX always used 48 khz by default and does not have any issues. It's kinda odd to use a fixed samplerate for the libretro port since resampling is not only CPU taxing but will also affect audio quality. I might be wrong but I think retroarch always need to resample the emulator output by default because of that dynamic rate control thing though.

Dynamic rate control's temporal samplerate adjustment is only done when we are in danger of underrunning and the variance in samplerate is unnoticeable to the human ear.

If you want math backing that up, here is a nice article -

https://github.com/libretro/libretro.github.com/blob/master/documents/ratecontrol.pdf

Seriously, if there is one thing maister knows well, it's audio. Nobody on this forum at least has any snowball's chance in hell matching him in that area.

Just to give a little example - Dborth/Tantric's sound in FCEU GX was a total sound pop fest until maister just mentioned within 2 minutes a way to get it sounding audio pop-less (ie. pushing slightly more than the sampling rate so that blocking is done on audio) - he managed to do within 2 minutes what that guy couldn't do within five years.

 

[LibretroRetroArc]

Now regarding Genesis Plus GX - it does resampling at two times - once in Genesis Plus GX's core - and the other time in RetroArch itself (through a Hermite resampler).

As for the samplerate change, it has already been changed back after I heard a guy in the PS3 scene complaining that it was screwing up the audio in certain Sega CD games. I don't know if the version that is in 0.9.8.2 right now has that at 44Khz or 48Khz, but it should be back to 44Khz in the next version.

 

[Jacobeian]

I do not deny that, he sure knows his stuff. I was only curious to know why retroarch was always resampling the audio output from emulator cores, even when the emulator output samplerate already matches the Wii output samplerate.

 

[LibretroRetroArc]

Sorry to ruin the fun, but having it fixed at its current sample rate makes it sound worse than before; my ears are very sensitive to things like this and I am not imagining it. Genesis Plus GX sounds tinnier than usual in the latest version of RetroArch. And yet the separate 1.7.3 does NOT have this issue at all. It has to be how RetroArch samples/resamples the audio when it outputs, same applies with Nestopia and its tinny sound output. I'm not some nut who's hearing auditory hallucinations. I am convinced that these issues can be fixed. Going to 44 to 48 KHz for the sampling rate seems kind of a moot point to me. The real Genesis wasn't 48 KHz, it was set at 44.1 KHz, and the NES was set at variable rates, and the Snes was set at 32 KHz. I don't see why this can't be looked at or taken seriously.

You want sinc resampling instead of hermite? It will be TWO TIMES SLOWER (conservative figure here), and the Wii CPU has no SIMD feature like Altivec to speed that up.

 

[the_randomizer]

You want sinc resampling instead of hermite? It will be TWO TIMES SLOWER (conservative figure here), and the Wii CPU has no SIMD feature like Altivec to speed that up.

Well, I didn't know about that resampling, my mistake .

Then can you please elaborate to what causes the tinny audio? What can be done to fix it? I know it exists in the Genesis Plus core, but not the standalone emulator, that doesn't make sense. It sounds worse when using RetroArch then using the separate emulator.

I'm not trying to come across as bitching but I feel I need to report this as a legit issue. Genesis Plus GX 1.7.3 sounds fine when using the standalone emulator, but when you load the core in RetroArch, it sounds worse. Surely, there's an explanation for that.

 

[megazig]

wii has paired singles for SIMD

 

[LibretroRetroArc]

wii has paired singles for SIMD

Thanks for letting me know - not sure if we'll write code for it though.

 

[LibretroRetroArc]

Regarding the audio quality debate concerning the_randomizer and other people - I'll let maister read this - he can decide whether he wants to reply then or not. However, he might ask you to do ABX tests - if you are not prepared or willing to do that for him, he might just disregard your issue altogether.

 

[AntonioJosedeSou]

Now regarding Genesis Plus GX - it does resampling at two times - once in Genesis Plus GX's core - and the other time in RetroArch itself (through a Hermite resampler).

As for the samplerate change, it has already been changed back after I heard a guy in the PS3 scene complaining that it was screwing up the audio in certain Sega CD games. I don't know if the version that is in 0.9.8.2 right now has that at 44Khz or 48Khz, but it should be back to 44Khz in the next version.

Nice to know that it will be back to 44.100 Hz in the next release. (It's 48.000 Hz in the latest RetroArch 0.9.8.2)
Anyways, performance should be the priority over accuracy at this point of development. And if you feel accuracy isn't important at all I'm fine with it too.
What is important for me is that the games look, sound and play nice on my 42" HDTV. ^^ (and yeah, input lag sucks in this new HD era)

 

[the_randomizer]

Nice to know that it will be back to 44.100 Hz in the next release. (It's 48.000 Hz in the latest RetroArch 0.9.8.2)
Anyways, performance should be the priority over accuracy at this point of development. And if you feel accuracy isn't important at all I'm fine with it too.
What is important for me is that the games look, sound and play nice on my 42" HDTV. ^^ (and yeah, input lag sucks in this new HD era)

That right there is probably the issue, the sampling frequency being 48 instead of 44 KHz, either that it's how RetroArch is sampling the audio twice. which doesn't make sense to me at all.

Regarding the audio quality debate concerning the_randomizer and other people - I'll let maister read this - he can decide whether he wants to reply then or not. However, he might ask you to do ABX tests - if you are not prepared or willing to do that for him, he might just disregard your issue altogether.

What the heck are ABX tests? Why would he disregard what clearly is a noticeable emulation glitch? I mean, if one was to compare both the standalone and emulator core of Genesis Plus GX side-by-side, one would irrefutably notice a degradation in sound quality when it's sampled two times (one via Genesis Plus and the other by RetroArch). I'm thinking that the sampling frequency change was a primary factor, glad it was reverted. Correct me if I'm wrong, but isn't one of the goals of RetroArch to be as accurate as possible without making too many sacrifices in performance? The issue doesn't exist in the Windows port of Genesis Plus GX using RetroArch, so there's something clearly wrong here. I'll record samples if I must.

Please don't get me wrong, I'm not trying to be antagonistic or contentious in the least, I just want to make sure that I'm not the only one who experiences this issue, because I know others can and will notice the audio degradation. If the goal is accuracy, then the Yamaha YM2162 emulation should be 1:1 to the standalone outputs, wouldn't you agree?

 

[mudlord]

What the heck are ABX tests?

A method to do double blind tests.

So, a means to tell if you are telling the truth or just spewing bullshit that you can hear any difference/issue in audio quality.

 

[the_randomizer]

A method to do double blind tests.

So, a means to tell if you are telling the truth or just spewing bullshit that you can hear any difference/issue in audio quality.

Makes sense. Hopefully he doesn't toss it aside as something trivial, well, because this particular issue detracts from what should be enjoying an emulator that focuses on accuracy and speed. Disregarding it is definitely is his prerogative, and yet at the same time, if he wants people give him feedback, he should at least take said feedback into consideration.

Seeing as I have hypersensitive hearing when it comes to perceive difference in emulation and how accurate something sounds (in relation to hardware vs emulation), I'm willing to do whatever it takes to neutralize this audio degradation.

 

[tacodaemon]

I've made some recordings that will hopefully put the sound issue to rest.

https://anonfiles.com/file/3b3ea0469b00d1d660b1a1c3566598cb

They're 44.1 Khz Mp3s. Let me know if it would be more helpful to choose a different sample rate and encode to a lossless format, but I don't think that's necessary to show the difference. I recorded the retroarch ports of FCE and Nestopia along with the real NES. All recorded at the same settings, no modifications to the files. The clip is of Super Mario Bros. 3 with the sliding sound effect, per Nintygaming's suggestion. Note that Nestopia is much louder. Perhaps this harshness is due to that difference? It sounds a lot like clipping on poorly mastered music. This isn't a huge issue for me, but I would like to see it recognized and addressed or set aside so that other things can be discussed. In the screenshot below, the top waveform is from the NES, the middle from FCE, and the bottom from Nestopia.

This files don't really lend themselves to ABX'ing since they differ in volume so dramatically and have different sound timing from my gameplay. Not to mention the NES is only in one channel, of course. Let me know if I can record any other examples for analysis.

Also, I would like to mention that the doublestrike modes are the main reason I use the Wii for emulation. I previously used a computer hooked up to a 15Khz production monitor via an Extron Emotia scan converter, and the Wii looks much better. I do hope that modern displays and filters improve to the point when I leave these old CRTs behind, but for now, they are the only thing that will do for me. Again, thanks for all your work on this fantastic project.

 

[nintygaming]

Man that is some wonderful news nintygaming and some great advice. Man I can't wait to get play some old school games tonight lol. Hmm.. But first some research I'm going to star looking into getting info on my service menu and hopefully it's possible to turn off digital processing (crossing fingers XD)

-_-.... My my what a sad state of affairs HD gaming has come too. Sad indeed. When I'm in the market for a new TV I'll have to reconsider my options. I'm an LG fan but the Sammy was a gift. However, I'll probably stick with LED, plasma's run way too hot for me, kills me in my room since its a bit cramped and I don't like running A/C all day lol.

I'll have to start doing research on best HDTV's for gaming.

Digital processing/post-processing is a demon when it comes to games due to it creating anywhere between small-to-massive amounts of input lag. Digital technology should (IMHO) have stayed far, FAR away from Televisions, especially when you consider that 95% of all programming and movies are shot in Analog Film or Analog Video, and whenever you do a "Conversion" from Analog to Digital, you loose alot of information in the process - degrading of sound/video quality - due to the way digital conversion works (taking a source that doesn't understand 1's and 0's and then forcing it into 1's and 0's, essentially its compressing it). Most people think that High Definition automatically means "Digital", but this is not entirely true. Take for instance that most movies have been recorded on Analog film, and some of the highest quality movies were shot on Film. Look at Wizard of Oz, a freakin OLD movie, and it was shot on 35mm film which has been "opinionated" to be roughly the equivalent to 8k (or higher) resolution (8k is 7680x4320p), but this is not true as Film has no "real" resolution as its not shot in pixels, so Film is definitely higher in resolution. For Blu Ray, Wizard of Oz was scanned in 8k due to Film not having a fixed resolution, and then compressed to 1080p. Not bad for "old outdated analog" right? So Wizard of Oz can look EVEN BETTER than 8k. While on the other hand, movies like Star Wars Episode II will NEVER look better than 1080p because it was Digitally shot IN 1080p. So much for "Digital is better" right?

So basically HDTV's could have been Analog all along if 2 things would have never happened.....

1. Digital TV transition would have never happened (this would mean that your HS500 would be analog instead)
and
2. SED or FED technology would have been the flat-panel tech instead of LCD and Plasma (I'm pretty sure that "Pixel" based displays like LCD/Plasma are digital regardless, although I could be wrong)

And as a result, we gamers would not even know anything about Input Lag (other than issues within Emulators), simply because it would not exist.

So we gamers have to suffer through the flaws of digital technology in order to play in HD nowadays.

If your serious about looking for another HDTV but with low input lag, then consider that Plasma's are the safer choice simply because they don't require nearly as much post-processing due to having much superior picture quality. Have you tried a Plasma to see if the heat is really that bad? I heard that heat output on those Plasma's have been significantly reduced over the past few years. Check out this thread right here - http://shoryuken.com/forum/index.php?threads/sub-1-frame-hdtv-monitor-input-lag-database.145141/ - and see what you may - or may not - find. Its REALLY hard to find a sub-1-frame HDTV, especially in 2013.

Otherwise, if you can deal with the size issue, then I'm willing to bet that your HS500 will make 1080i-capable consoles look amazing, along with Blu Ray's as well. Those Sony Hi-Scan TV's are truly like an old souped up muscle car (lots of bulk, yet lots of power) and while old retro games will never look "perfect" on it, they will still look far better than a pixel-based display with the CRT's anti-aliasing nature, and thus I'm willing to bet if you try Retroarch in 512x480p for NES, and 512x448p for SNES, and 640x448p for Genesis, etc that you will be pleased if you adjust sharpness to acceptable levels and turn off as much digital processing as possible.....and most importantly, turn off HDPT or the equivalent within the Service to get rid of Input Lag. You can find out about how to get into the Service Menu via looking at the Service Manual here - http://www.scribd.com/doc/79561778/sony-kv-32-42hs500-hv600-drc500-xbr800-drc800-v2-da4-ET - you can also do other great things like eliminating red-push (if your TV has it) and reducing overscan, etc. JUST BE SURE TO WRITE DOWN ALL ORIGINAL VALUES WITHIN YOUR SERVICE MENU BEFORE YOU CHANGE ANYTHING.

 

[the_randomizer]

I've made some recordings that will hopefully put the sound issue to rest.

https://anonfiles.com/file/3b3ea0469b00d1d660b1a1c3566598cb

They're 44.1 Khz Mp3s. Let me know if it would be more helpful to choose a different sample rate and encode to a lossless format, but I don't think that's necessary to show the difference. I recorded the retroarch ports of FCE and Nestopia along with the real NES. All recorded at the same settings, no modifications to the files. The clip is of Super Mario Bros. 3 with the sliding sound effect, per Nintygaming's suggestion. Note that Nestopia is much louder. Perhaps this harshness is due to that difference? It sounds a lot like clipping on poorly mastered music. This isn't a huge issue for me, but I would like to see it recognized and addressed or set aside so that other things can be discussed. In the screenshot below, the top waveform is from the NES, the middle from FCE, and the bottom from Nestopia.

This files don't really lend themselves to ABX'ing since they differ in volume so dramatically and have different sound timing from my gameplay. Not to mention the NES is only in one channel, of course. Let me know if I can record any other examples for analysis.

Also, I would like to mention that the doublestrike modes are the main reason I use the Wii for emulation. I previously used a computer hooked up to a 15Khz production monitor via an Extron Emotia scan converter, and the Wii looks much better. I do hope that modern displays and filters improve to the point when I leave these old CRTs behind, but for now, they are the only thing that will do for me. Again, thanks for all your work on this fantastic project.

Wow, a very nice comparison, how did you record directly from the emulator? If I could only find a way to test the difference between Genesis Plus GX 1.7.3 by itself and in the RetroArch core (same version). I'll let you know my view on your recordings.

 

[tacodaemon]

Wow, a very nice comparison, how did you record directly from the emulator? If I could only find a way to test the difference between Genesis Plus GX 1.7.3 by itself and in the RetroArch core (same version). I'll let you know my view on your recordings.

Thanks! Nothing fancy. RCA cables from the Wii and NES to the line in of my PC's sound card. Recorded with Audacity from there. I missed the posts on the difference in Genesis GX Plus versions. I'll try to get some recordings tomorrow. Any specific games I should record from?

 

[the_randomizer]

Thanks! Nothing fancy. RCA cables from the Wii and NES to the line in of my PC's sound card. Recorded with Audacity from there. I missed the posts on the difference in Genesis GX Plus versions. I'll try to get some recordings tomorrow. Any specific games I should record from?

Sonic 3 and Knuckles if you have it (or just Sonic and Knuckles), the title screen music and some sound effects. For accuracy, both Genesis Plus GX (separate emulator) and Kega Fusion get it 100% correct. I'm curious how it will sound using the same emulator in RetroArch Wii.

Oh, I just tested the samples and Nestopia on RetroArch is definitely off, and when I mean off, I mean off. There is apparent distortion and sounds like it was recorded at a lower sampling frequency, even though it was recorded at 44.1 KHz, so it's no fault of your work or Audacity; the emulator is definitely doing something weird when sampling the audio in RetroArch. I know for a fact that issue doesn't happen in the Windows version. That FCEUmm recording sounds very close to a real NES, while Nestopia does not. If this issue can be resolved, then that will be one down and two to go (Snes and Genesis).

Summary: FCEUmm and NES sound almost exactly the same.
Nestopia has some serious distortion and tinniness to the audio and the difference is quite perceptible. Genesis Plus GX and Snes9x Next have the exact same issue (though in Snes9x Next, many games do not have it).

When recording, you used a Y-cable (RCA-female cable to ) to plug in the headphone jack?