fhgaac

2015-03-04 20:39:38

Hi.

Due to the response to my question here a while back, I assumed it was okay to set the bit depth for fhgaacenc.exe (the version in the foobar2000 encoder pack) to 32 in the encoder configuration. So I did, and it's been fine, only today I was getting an error and realised a bit depth of 32 is fine for stereo audio, but not for multi-channel audio (at least not for 5.1ch).

Conversion failed: The encoder has terminated prematurely with code -1073741819 (0xC0000005); please re-check parameters

Changing the bit depth to 24 fixes that, so it's no big deal, but I thought I'd report it.... just in case somebody cares......

While I'm here.....
The way I understand it the fdkaac encoder is ultimately 16 bit. A higher bit depth input is down-sampled to 16 bit internally. Is that correct?
If so, is there a reason why setting a bit depth of 16 for the fdk encoder configuration produces a slightly different result than a bit depth of 24 or 32? It's only a tiny, tiny difference in bitrate and nothing I'd be concerned about but I'm a tad curious as to why there'd be any difference. Is it a dithering thing?

Thanks a lot!

fhgaac

Reply #1 – 2015-03-05 16:29:41

Thanks for the report. There was a buffer overflow bug in the encoder frontend and it was crashing behind the scenes.

I can't answer your FDK question but I found another bug while checking bitness behavior with FhG. An unfortunate integer overflow in the float conversion code has inverted all audio encoded as 32-bit. Fortunately the problem is inaudible.

Fixed fhgaacenc will be included in the next Free Encoder Pack and you can download it immediately here: [attachment=8196:fhgaacenc.zip]

fhgaac

Reply #2 – 2015-03-05 19:03:24

Awesome! Thank you!