V-DAT Quality

voidar · Post by **voidar** » Sun Jan 02, 2005 12:09 pm

Found an editor which supposedly loads/saves 32-bit integer/signed format. http://www.ngwave.com
I won't be able to test this before a week or so.

[ This Message was edited by: voidar on 2005-01-02 12:09 ]

decimator · Post by **decimator** » Mon Jan 03, 2005 5:29 am

Adobe Audition 1.5 ( formerly Cool Edit Pro ) supports the 32 bit integer but it was already mentioned.

astroman · Post by **astroman** » Mon Jan 03, 2005 5:42 am

the above mentioned NGWave converts 32-bit integer files to it's own 32-bit float format when opening files (according to the web page)

cheers, tom

voidar · Post by **voidar** » Mon Jan 03, 2005 10:37 am

On 2005-01-03 05:42, astroman wrote:
the above mentioned NGWave converts 32-bit integer files to it's own 32-bit float format when opening files (according to the web page)

cheers, tom

This might be true for Audition too, or most editors in general.

Is it possible to compare two audiofiles on the sample-level? I mean, does any tools for this exist?

spoimala · Post by **spoimala** » Tue Jan 04, 2005 10:43 am

Is it possible to compare two audiofiles on the sample-level? I mean, does any tools for this exist?

Any hex-editor? (Like http://www.chmaas.handshake.de/delphi/f ... /xvi32.htm )

voidar · Post by **voidar** » Tue Jan 04, 2005 1:30 pm

On 2005-01-04 10:43, spoimala wrote:
Is it possible to compare two audiofiles on the sample-level? I mean, does any tools for this exist?
Any hex-editor? (Like http://www.chmaas.handshake.de/delphi/f ... /xvi32.htm )

Right,

Wonder why I did not think of that.. It would be compared at the byte-level though, but it should work.

I want to compare an original 32-bit integer file to one that has been loaded (converted to 32-bit float) and saved with NGWave.

astroman · Post by **astroman** » Tue Jan 04, 2005 3:09 pm

does not compute...

even if the files would have the same numerical content the different representation results in 2 completely different files. Matching bits would be due to chance only.

aside from that the internal structure isn't necessarily a stream of datawords. Header, checksums, block structure, byte order - there's a lot that can happen in a file

cheers, Tom

voidar · Post by **voidar** » Wed Jan 05, 2005 9:11 am

On 2005-01-04 15:09, astroman wrote:
does not compute...

even if the files would have the same numerical content the different representation results in 2 completely different files. Matching bits would be due to chance only.

aside from that the internal structure isn't necessarily a stream of datawords. Header, checksums, block structure, byte order - there's a lot that can happen in a file

cheers, Tom

No, no, no..

I am going to compare two files of the same format/representation, but one of them will have been passed through a integer/float/integer conversion. I want to see what such a conversion will do to a file.

Wave-files are linear and not compressed. You can basically read and understand them using notepad.exe.

[ This Message was edited by: voidar on 2005-01-05 09:12 ]

Music Manic · Post by **Music Manic** » Wed Jan 05, 2005 10:03 am

Could you also record the same signal through different drivers and audio programs to see how signals differ when recorded.Still believe that Cubase sound differs.
Happy with VDAT

voidar · Post by **voidar** » Wed Jan 05, 2005 3:46 pm

On 2005-01-05 10:03, Music Manic wrote:
Could you also record the same signal through different drivers and audio programs to see how signals differ when recorded.Still believe that Cubase sound differs.
Happy with VDAT

Somewhat more work, though I could record a 32 bit file in VDAT, play it back and record it in some ASIO app and see where I would need to go from there.

My idea however was to check whether NGWave would lessen the quality of a file when automatically converting to 32-bit float. I hope not.

voidar · Post by **voidar** » Tue Jan 11, 2005 7:18 am

Just tried this. The conversion changes the file in drastic ways. So, search goes on.

astroman · Post by **astroman** » Tue Jan 11, 2005 8:08 am

... just as predicted

and you have no way to corelate numeric values (their binary representation as shown by the file editor) to their audible result

honestly, it will not sound any better because it's 32 bit whatsoever.
I currently stick with VDAT 16 bit as I don't have any source with enough dynamic to take advantage of an increased resolution (but it's nice to know I could immediately...).
The 'less transparency' of one or the other audio engine results from processing, not from numerical encoding.

If you record a file by VDAT (arbitrary resolution) and pass it to Cubase just to immediately save it (without any channel or master operations) you will not be able to hear any difference.

cheers, tom

voidar · Post by **voidar** » Tue Jan 11, 2005 9:09 am

On 2005-01-11 08:08, astroman wrote:
... just as predicted
and you have no way to corelate numeric values (their binary representation as shown by the file editor) to their audible result

honestly, it will not sound any better because it's 32 bit whatsoever.
I currently stick with VDAT 16 bit as I don't have any source with enough dynamic to take advantage of an increased resolution (but it's nice to know I could immediately...).
The 'less transparency' of one or the other audio engine results from processing, not from numerical encoding.

If you record a file by VDAT (arbitrary resolution) and pass it to Cubase just to immediately save it (without any channel or master operations) you will not be able to hear any difference.

cheers, tom

It is a linear format so yes, I can. I wanted to check for alterations, and that is what I found. I would probably not hear any difference though.

Two files saved with NGwave from the same source yielded a 100% match.

A new VDAT-"tape" differs in the header only, from file to file.

The original 32-bit signed file was drastically different to the one that had passed through NGwave's signed/float/signed conversion. It could make an audiable difference, but probably not.

24-bit signed would probably work just as well.

Immanuel · Post by **Immanuel** » Tue Jan 11, 2005 9:15 am

My guess it, that the difference would be in the -144dB and below region. You may want to take the 2 files. Time align them, so they start at the same bit. Then play them thru 2 mono channels on a CWA mixer (with phase compensation. Invert the phase of one signal. Check the master meter for activity. This ought to tell you, how significant this conversion thing is.

voidar · Post by **voidar** » Wed Jan 12, 2005 12:32 pm

On 2005-01-11 09:15, Immanuel wrote:
My guess it, that the difference would be in the -144dB and below region. You may want to take the 2 files. Time align them, so they start at the same bit. Then play them thru 2 mono channels on a CWA mixer (with phase compensation. Invert the phase of one signal. Check the master meter for activity. This ought to tell you, how significant this conversion thing is.

I did the test. It turns out that I would need to amplify the out-phased signal about 119dB for it to peak. Result is of course very noisy but I can still hear the recording.

voidar · Post by **voidar** » Tue Sep 19, 2006 3:37 pm

I just figured out that using the ASIOx-32 drivers I can record and edit .AIFF's at 32-bit integer in Tracktion 2 that will load in VDAT or STS's, as long as I rename the files to .aif in the end.

However there seems to be a problem with recording .aiff's on my system at all via Tracktion 2 right now. But the little I get to record works.

voidar · Post by **voidar** » Wed Sep 20, 2006 4:08 am

It works!

I can now record VDAT/STS-compatible 32-bit integer 44.1KHz .AIF's in Tracktion 2 flawlessly via ASIO at ULLI set to max (block-size 1024).

It also works at at 512- and 256-sizes, but with an irritating post-record load-time of some seconds (5+).

For people looking for a audio/midi-sequencer that can record our buff SFP-resolution

in an easy and cheap fashion I say look to T2.

Files can later be edited in T2 and exportet to the same 32-bit .AIF-format, loaded up in VDAT/STS and played back, if you would rather do a wholy integer mix in SFP.

Or you could mix using T2's 64-bit floating point engine.

On a technical level 64-bit float > 32-bit integer > 32-bit float.
But I don't know about that.

It at least seams like what I get in is what I get out on a 1:1-level.
A wave recorded in T2, played back and recored again, will cancel out the initially recorded wave when phase-reversed.

My 0.02

iskra · Post by **iskra** » Wed Sep 20, 2006 10:23 pm

just came across this long thread on VDAT, I have two questions

I use cubase to sequence creamware synths, some outboard synths and samples (mostly drums) - if I record the output of all this into VDAT instead of cubase it will sound better?

also I dont really need the multirecording feature, is there a simple recording device for scope which just records a stereo mixdown, which would give the same quality of VDAT?

iskra · Post by **iskra** » Wed Sep 20, 2006 11:06 pm

one more thing, occasionally when I have cubase close to maxxed out performance wise (& scope similarly almost out of dsp) playback will be fine, but when I record in cubase I will get the odd audio glitch or stutter which is annoying, would VDAT likely solve this (as there is no problem with playback)?

also is anyone able to post a short example of the sound quality difference between vdat & cubase?

[ This Message was edited by: iskra on 2006-09-21 00:09 ]

voidar · Post by **voidar** » Thu Sep 21, 2006 4:38 am

iskra wrote:

I use cubase to sequence creamware synths, some outboard synths and samples (mostly drums) - if I record the output of all this into VDAT instead of cubase it will sound better?

VDAT records in the native SFP-environment, so if you record at 32-bit resolution in VDAT you will actually record what you hear.
Some complain about Cubase sounding dull in contrast. I don't use Cubase.

iskra wrote:

also I dont really need the multirecording feature, is there a simple recording device for scope which just records a stereo mixdown, which would give the same quality of VDAT?

The STS3000 and onward, can record native 32-bit resolution as VDAT does, however the STS sampler records to memory which might be a problem if you have little of it.

One VDAT-instance creates no less than eight pre-formated mono tracks. If you were to mix down to VDAT you would later have to splice two mono tracks into one stereo track in some wave-editor (which reads the 32-bit linear format).

iskra wrote:

one more thing, occasionally when I have cubase close to maxxed out performance wise (& scope similarly almost out of dsp) playback will be fine, but when I record in cubase I will get the odd audio glitch or stutter which is annoying, would VDAT likely solve this (as there is no problem with playback)?

VDAT might solve this as using VDAT you will pre-format/create a set of sound-files of a given resolution and lenght prior to recording. Think of it as a custom length tape.

You could also try to set your latency bigger (ULLI). That should help with glitches. Gives your system more space to breeth.

iskra wrote:

also is anyone able to post a short example of the sound quality difference between vdat & cubase?

I am not sure something like this exists. I did slave Tracktion 2 to VDAT via MTC, and recorded the same input to both. The Tracktion 2 recording would drift over time, but this we suspected was due to MTC not being sample-accurate.
I might try again without syncing, but sample-aligning two files is hard.

In my opinion, storing audio-data as floating point numbers is a horrible thing to do. For me it is only 16-, 24- or 32-bit integer format.
I choose the last one as I want to record at my DSP's native resolution. This might not be very important when recording inputs from A/D-C's, but when creating sounds within SFP you will actually be able to record what is.

[ This Message was edited by: voidar on 2006-09-21 05:38 ]