Subject:Where's 0dB in SF 5.0? Attn: Engineering
Posted by: Rednroll
Date:2/28/2001 4:02:30 PM
I was doing some testing on 1Khz 0dB Sinewaves today for engineering test purposes. 1 ripped from a Abex test CD (TCD-782 track#2), 1 ripped from a Ford Audio System Test CD, and then 1 generated in SF 4.5 simple synthesis generator. I wanted to verify that all these sinewaves were truly at a 0dB peak. I did a normalize on each using Sound Forge 4.5 using peak level set at 0dB. When SF 4.5 did a scan on all 3 waveforms it detected that they all where at 0dB and did no further processing. I then reopened all 3 of the same waveforms in SF 5.0 and did the same test. All three of them did processing and raised the levels. So why would v5.0 detect that there was available headroom and 4.5 wouldn't? Which program is correct? Seeing I did a digital rip of 2 engineering test CD's, I would suspect 4.5 to be correct. Why is there this discrepancy? regards, Brian Franz |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: Itna
Date:3/2/2001 9:26:54 AM
Hey Ted, Seems to be a very important topic here ! Any answer ? Itna |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: SonyEPM
Date:3/2/2001 5:12:03 PM
Quote from the SF5 lead developer: If he was working on a 16 bit source file as a float temp file, then there actually would be a bit of extra headroom in the temp file. Of course, when you convert back to 16 bit on the save, that change would get truncated back to the original waveform. (in theory). |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: Rednroll
Date:3/2/2001 8:10:16 PM
Yikes, did anyone here understand what that meant? You just introduced me to a new term, "float temp file". I don't know what that means,I'm sure it has something to do with floating point integers. I opened the file as I normally would, which I know does create a new temporary file and it was a 16 bit file. In any case, as everyone knows this breaks all laws of audio conservation (ie don't do any unnecessary processing of audio if it is not needed). This is what the scanning process is suppose to do for you, (ie scan audio.....if a peak is at 0dB, no processing is necessary...cancel operation). This also waste time by processing unnecessary files, that have been already normalized. Well anyways, I'm sure there's a lot of down sides to this, I hope someone fixes it correctly though. FYI I found this by mistake. I was just trying to compare the two test CD's to verify that 1Khz @ 0dB was just that, because some engineers at ford found different RMS output levels from the 2 test CD's, on CD mechanisms I design. So to verify this on the two CD's I did a normalization test and then measured the frequency. I know this is anal, but this is the kind of precision engineering I have to deal with, I guess I just won't use the v5.0 downgrade to do this kind of test. :-) I'll do some further testing Monday also. I'll try opening these same files in 5.0, do the normalization and then save them, then close them. I'll re-open and then try normalization again. If the developer is correct, then it should actually process it and normalize it again when I re- open it. Thanks for the reply, Brian Franz |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: sreams
Date:3/2/2001 10:08:13 PM
Go to Options/Preferences and click on the File tab. There should be a checkbox labelled "Use 32-bit (IEEE Float) temporary files". Be sure it is unchecked. -S |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: Rednroll
Date:3/3/2001 6:54:10 PM
Thanks for the info, I will check this out. Brian Franz |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: Rednroll
Date:3/5/2001 11:58:55 AM
Update: First I opened the Abex sinewave and Normalized it in 5.0. I then saved it as another file. I reopened this file and normalized it again. SF 5.0 did no further processing to it, (IE IT DETECTED THAT IT WAS AT PEAK LEVEL). So the statement of re-truncating it didn't hold up under this test, because if it disregarded this information, then it should have therectically normalized it again. Next, I went to option-properties-File. I checked "use same bit depth for temp files as the original file". I unchecked "enable 64bit internal processing". Again I opened the Abex sinewave and normalized, again it raised the level. Here's the format of the file under the file menu-properties menu. SOURCE: 44,100 HZ,16 BIT ,STEREO TEMPORARY: 44,100 HZ,16 BIT ,STEREO PROCESS/DSP: 44,100 HZ, 32 BIT (IEEE FLOAT),STEREO PLAYBACK: 44,100 HZ,16 BIT ,STEREO So where is 0dB in v5.0? Is there something I'm missing here or is it, "back to the drawing board?" regards, Brian Franz |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: SonicCaleb
Date:3/6/2001 11:47:39 AM
This may be a silly question, but exactly how are you determining that Sound Forge is raising the level of the audio? Thanks, Caleb |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: Rednroll
Date:3/6/2001 1:48:37 PM
Just do to the fact that in the bottom left hand corner it shows the "normalize" status as it is processing and in the normalize setup menu it is set to "peak level" and is set to 0dB. So from this I would have to conclude that Sound Forge detected that the waveform wasn't at 0dB and raised the level. When I open this same file in v4.5 with the same settings, it does the scan of the waveform and detects that it's at 0dB and does no further processing, as expected. This is a 10meg file, if you would like me to upload it for you to further investigate I will happily do that. My email address is REDNROLL@HOTMAIL.COM. Or SF could purchase an Abex audio test CD (TCD-782). Regards, Brian Franz |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: SonicCaleb
Date:3/7/2001 2:43:04 PM
I can't seem to reproduce the problem with 0 dB files generated by 4.5. Now, i am using the soon-to-be-released 5.0a update version, but nothing to my knowledge has changed with regards to the normalize function. There may have been differences between the beta version and the final retail version. So, if you are using the beta, i'd suggest downloading the demo version of the final and seeing if that does the same thing. Let me know. Thanks, Caleb |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: Rednroll
Date:3/7/2001 3:32:31 PM
Yes, I am using the Beta version and FYI it is installed on a Win2k Pro OS. I will download the demo and retry the tests. Thanks for the response, Brian Franz |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: CDM
Date:3/7/2001 5:48:37 PM
I am also able to create a 0dB sine wave in Forge 4.5 and normalize it in Forge 5.0. It doesn't seem to really change anything but it does process. Forge 5.0 Retail |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: SonicCaleb
Date:3/8/2001 12:21:37 PM
Ok, so this is all figured out. It is, in fact, the difference between processing in floating point math (SF 5.0) and processing in integer math (SF 4.5) for normalization as posted earlier. The reason it keeps normalizing after saves is because everytime the file goes back to 16-bit, the same problem exists. A max level of 32767 in a 16-bit integer when converted to 32-bit floating point is 0.9999694. That means that there is an incredibly small amount of headroom left to get to 1.0. It is inpercievable, but it still exists. So, we are going to add a little fudge in the normalize code to detect this situation and not do anything. You won't notice any sonic differences at all. Hopefully that is all cleared up now. |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: Rednroll
Date:3/8/2001 12:29:16 PM
Thanks for the explanation and the good work Sonic Foundry. I think most of us realized this was not a big difference, just wondered where the discrepencancy was and where the extra headroom came from. Thank you, Brian Franz |
Subject:RE: Where's 0dB in SF 5.0? Attn: Engineering
Reply by: CDM
Date:3/8/2001 2:17:33 PM
Thanks Caleb. I wasn't particularly worried. :) |