Topic: upsampling prior to performing DSP in post

[Brian]

Hey all,

For all the people who like to perform post production DSP on their recordings(for recordings with sample rates of 48 or 44.1),  do you upsample to 96k prior to performing any DSP?  I had an interesting discussion the other day with the engineer i'm interning with.  We talked about DSP and how it can "crunch" the sound of your recording at lower sample rates.  He then mentioned that when he masters, he'll upsample the recordings to 96k, perform the necessary DSP, and then resample back down to 44.1.  He claimed that when you do this, all of the "processing noise" will be way up in the unaudible spectrum so when you downsample to 44.1, it won't affect the sound of the recording(or as much as it would when performing DSP at lower sample rates).  When I asked him about adding the two sample rate conversions into the process(two extra DSP's that some might deem necessary) he reinforced my thoughts that sample rate converters are so much better than they were say 3 years ago and most.  So programs like WaveLab, Peak, nuendo, pro tools, and other similar quality programs all have decent sr converters.  I had a similar conversation some time ago with a teacher but i never really tested it out that much.

So i was messing around with it last night and I must say I could hear a difference.  The difference resulted in a smoother, slightly clearer recording.  It especially helped when using heavy EQ on the hi-mids to high frequencies as well as compression(mastering limiter in my case).

Does anybody else do this?  I think it's going to be standard procedure for me from now on for any work I do with recordings with sample rates of 48k and 44.1k.

thoughts?

edit: fat fingers

[Karl]

It sounds like a good idea.  Though, if you were working with 44.1, I would think it wiser to upsample to 88.2, then do DSP, then go back to 44.1.

I guess the only drawbacks would be time and harddrive space if you are working on unusually large projects.

[Brian]

yes, karl...correct.  thanks for that bringing that up.  +T i failed to make that distinction.

[BobW]

To really test this you need to compare:

<orig file>  -- <DSP filters>  -- <upsample> -- <downsample>

      to

<orig file>  --  <upsample>  --  <DSP filters>  -- <downsample>

This would ensure that the second upsampling and downsampling (first example) isn't making the changes that you find pleasant.
 
I am beginning to believe that a  very  small amount of hiss and hum is sonically pleasing and improves soundstage.
So double dither may cause this among many other sonic changes.

Beyond that, it should be added that in a good workflow, bit-depth conversion, then resampling and finally tracking (if DAO)  are the completing steps.

[Brian]

bob -  what's the point of upsampling if you are only going to downsample right after?  there isn't one IMHO.  down and upsampling is NOT dithering.

edit: misread post and re-replied.

edit2: some engineers say it's a good rule of thumb to dither after major changes with DSP.  my next test wili be <orig file> -- <upsample> -- <DSP> -- <dither w/24bit output) > downsample. and then repeat without the dither.  Dither most definitely affects your files soncically. 

edit3: i've been working with only 24bit files.

[Sloan Simpson]

edit3: i've been working with only 24bit files.

Do you think (realizing you haven't tested yet) this process would apply to 16-bit recordings as well?

[Karl]

Awhile back I did a few tests for fun.  I had better results (in 16 bit) when upsampling, doing dsp, then downsampling.  The high end seemed to be more pure.

When you are working in 44.1, higher frequencies can not represent a pure sine wave.  So it would make sense that upsampling would allow dsp to work cleaner.  Then the downsample will as a result be cleaner (esp if you are doing multiple dsp stages).

[BobW]

bob -  what's the point of upsampling if you are only going to downsample right after?  there isn't one IMHO.  down and upsampling is NOT dithering.

It will ensure the the Upsample and Downsample does not change the sound (I believe that it it does, YMMV)
Makes the playing field more even from a scientific perspective when you compare the files.
Also, why would upsampling affect sound ?  (rhetorical question, no need to answer)

edit2: some engineers say it's a good rule of thumb to dither after major changes with DSP.  my next test wili be <orig file> -- <upsample> -- <DSP> -- <dither w/24bit output) > downsample. and then repeat without the dither.  Dither most definitely affects your files soncically.

edit3: i've been working with only 24bit files.

Any filter affects should automatically dither, if necessary.
Dithering is a necessary evil, usually not a sonic benefit.

The workflow was just added in, not totally relevant to your question, sorry about the confusion.

Let us know what you find !
I'm always open for new ideas.

What DAW are you using ?

[Brian]

sorry....i didn't catch up with this thread after i got back home.

it will most certianly apply to 16 bit files.

I've been messing around with these tests in both Wavelab 5 and Peak 4.