Short white noise bursts in AAC audio that were not there before

[Foebane72]

This is yet another chapter in my issues with Shadowplay and Handbrake:

When I record 8-Bit audio (from the Atari 800XL) via Shadowplay and then play it back, very rarely I will hear these nasty loud short bursts of white noise in the audio. It only seems to happen with the distinctive nature of such audio, and it is NOT there when I recorded it in the first place.

However, I can't be sure if the bursts are there before or after I encode with Handbrake.

Is there a tool that can visualise the audio from an MP4 file as a waveform so I can see these spike bursts for myself? If so, then I could simply re-record them and hope the problem doesn't appear again. However, I am baffled as to why they appear in the first place.

[Foebane72]

I found a tool for viewing my MP4 audio files, Audacity with Ffmpeg, and it seems only three out of the 62 files I recorded from this Atari emulator were affected. What's more, the spikes (which are only in the right-hand side) are not audible in the emulator and it seems they're introduced with Shadowplay. In any case, there's nothing I can do if they appear, and I doubt I can edit the audio wave without re-encoding.

Anyone know if I can edit AAC audio without re-encoding?

[Foebane72]

Since no-one here has answered, I've decided to not use Shadowplay at all any more and I'm trying something better, like Bandicam.

[LigH.de]

AAC is a heavily compressed audio format. Usually it will require decoding to allow editing, so you would have to re-encode it afterwards. In theory, it may be possible to manipulate single "frames" in encoded format, but I don't know any software doing that with AAC (for MP3, there are tools which can change a few attributes per audio "frame").

If the audio of your Atari 800XL (emulator or real hardware?) is really recorded with 8 bit resolution per sample, I could imagine that there might be mistakes converting them to 16 bit in a way that sometimes a byte is missing or inserted, swapping the meaning of the two bytes of a word (big endian / little endian). The analogy in video processing would be a ranged phase shift, resulting in pseudo interlacing and combed frames whereas the source used to be progressive.

Using a different recording software is probably a good decision. At least it adds another chance to use more reliable technology.

[Foebane72]

I've tried Bandicam for a while now, and whilst it has greater options for custom recording settings, it too is prone to this same white noise spike problem when it comes to recording Atari 800XL and even Amstrad CPC audio (both 8-bit computers) and it seems there's nothing I can do about it.

I'm using emulators, BTW.

[LigH.de]

Do these emulators have their own screen recording feature? Which emulators are they, maybe there are alternatives?

[Foebane72]

The emulators are Altirra, WinAPE and Spectaculator in the case of the ZX Spectrum.

Yes, they all have their own recorders but the only one I found good enough was Spectaculator's, as I simply could not record it with Shadowplay OR Bandicam.

[LigH.de]

ZX Spectrum emulators ... pro: there are many; con: there are too many to know good ones immediately.

One was made by Christian Hackbart ... do I know this name already? ... Ah, yes: DVBViewer. Same guy? Well possible.

[Foebane72]

I tried several Spectrum emulators when I wanted to run Demoscene productions - Spectaculator (which I paid for) is the only one good enough for all of them.

[raffriff42]

Experiment with lowering audio sample rates -- both in the emulator and on the capture side. I've seen ultrasonic noise leaking into the audible band, especially with very high sample rates. Maybe raise the emulator sample rate, I don't know.

How to fix Fraps High pitch/Whistling sound from input [short answer: lower sample rate]
https://www.youtube.com/watch?v=KT-0GfwZxZU

The Science of Sample Rates (When Higher Is Better — And When It Isn’t)
http://www.trustmeimascientist.com/2013/02/04/the-science-of-sample-rates-when-higher-...-when-it-isnt/

two super-sonic frequencies that cannot be heard, say 22 kHz and 32 kHz, can create an intermodulation distortion down in the audible range, in this case at the “difference frequency” of 10kHz.​

[Foebane72]

Shadowplay is very limited in recording settings, there is only the option to set vertical resolution (480p, 720p and all that), FPS and Mbps. I doubt I'll buy Bandicam.

[raffriff42]

Shadowplay should follow Windows audio capture settings. (Control Panel, Sound, Recording, <your sound card>, Properties, Advanced)

If you're looking for an alternative to Shadowplay (good idea), try OBS Studio.

[Foebane72]

Tried OBS Studio, couldn't figure out how to get it to record video.

What if I provided an example of these audio spikes? Maybe if you were to analyse them yourselves?

[Foebane72]

Here is an example of what I've been talking about: https://drive.google.com/open?id=0B5CK-_D7y_M_Y2tHQTFvcjlldlE

If you examine the waveform, you'll see three or four prominent spikes in the middle and towards the end of the audio which are not meant to be there, and when you play it back, you'll hear just how bad they sound in general listening. Maybe you can extrapolate what the cause is from this?

[raffriff42]

It's not what I thought it was; it's a software glitch of some kind. You need to ask someone who knows about this emulator. Do you have any choice of sound generator? Any sound options at all?

If all else fails, Audacity low pass filtering (5000 Hz, 48 dB/octave) kills the first glitch in your sample, and most of other two. Going to a lower cutoff (3500 Hz) kills the glitches almost completely, but dulls the sound too much.

EDIT be sure to capture and do any processing using uncompressed, PCM audio. Compression "smears" any large, sharp glitches and makes them more difficult to filter out.

EDIT this smearing effect may be why the glitches are audible now, after encoding, and not before.

[Foebane72]

Well, Shadowplay encodes the audio as AAC on the fly, and I've recorded this particular demo a few times and the glitches never occur in the same place twice.

The emulator's sound options are stereo or mono, non-linear mixing (which I turn off), volume, drive volume (for the floppy drive sounds), audio API (WaveOut or DirectSound), show debug info, latency and extra buffer.

However, most of the demos have recorded fine, just this one has issues. Also, I use an Amiga emulator which has issues with a few demos out of well over one hundred, so I cannot figure out why this issue singles out certain demos only. But since they are so few and far between, I've decided to just delete the offending audio spike files until I can get this situation resolved.

[El Heggunte]

Well, Shadowplay encodes the audio as AAC on the fly

That's a serious design flaw

[raffriff42]

>non-linear mixing
>audio API (WaveOut or DirectSound)
>latency
>extra buffer
Hmmm, all of these might possibly cause glitching. Experiment with different settings:
try flipping nonlinear and API, and doubling latency and buffer. One setting at a time, of course.

>drive volume (for the floppy drive sounds)
Whoah, like singing floppy drives?
https://www.youtube.com/watch?v=cM_sAxrAu7Q

[Foebane72]

The sounds of a floppy drive, not as many as that, mind!

I won't be investigating this issue on the audio spikes just yet, as I want a break from all of that. Maybe in a few months I'll go back to it, but not now.

Thanks for everything, anyway!