Upsamplig - downsampling

[filogab]

How many of you ME, when you have to deal with a mix file at 44.1K (or 48) wuld do an upsampling to 96K for processing it?
And how would you do this upsampling: the digital way or would you convert to analog and then resample it @ 96K?
The same question apply for the downsampling from 96K to 44.1K before going to cd premaster.
Thanks alot

[filogab]

Ok, no replies so far.
I know this is a subject discussed already. I would have expected the classic "use the search button a****le" but not even that.
I woul try to narrow the question. What I would like to know is: is it used to upsample something to analog convert it and then to reconvert it in digital at higher sampling rate? Or the upsampling process is mainly done in the digital realm?
Same for the downsampling.
Hope it's a little more moving now.
Thanks

[miditus]

no need for upsampling.

[acorneau]

Originally Posted by filogab

How many of you ME, when you have to deal with a mix file at 44.1K (or 48) wuld do an upsampling to 96K for processing it?

I tried that route for a while, but found the up&down SRC caused more degradation than mastering at the original sample rate with one SRC (if needed) towards the end.

[Jerry Tubb]

Originally Posted by acorneau

I tried that route for a while, but found the up&down SRC casued more degredation than mastering at the original sample rate with one SRC (if needed) towards the end.

+1 thumbsup

Same here Allan !

JT

[lucidwaves]

Degredation from increasing the sample rate? Can someone explain further?

[acorneau]

Originally Posted by lucidwaves

Degredation from increasing the sample rate? Can someone explain further?

It's usually not the trip up, but the trip back down that's does the damage.

[Riccardo]

Originally Posted by Jerry Tubb

+1 thumbsup

Same here Allan !

JT

and here +1

[Fretbored]

Originally Posted by Jerry Tubb

+1 thumbsup

Same here Allan !

JT

Here too.thumbsup

[MattGray]

Originally Posted by acorneau

It's usually not the trip up, but the trip back down that's does the damage.

But then like you say if the original file is 48kHz you would still have to do a trip down at the end anyway so what would be the difference to do an up & down in this case?

I agree that the downsample can degrade the signal if a quality SRC isn't used, but I find (particularly if you are using any downstream digital processing) that an upsample to 96kHz often sounds just that bit better even after the downsample to 44.1kHz. The trick is knowing where to do the downsample. If you do it post the limiter or clipping then you will get intersample peak grunge which sounds bad. If you do it pre limiter you will get a much cleaner result & if you use a high quality 64bit floating point SRC you will still keep a lot of the added advantages of the 96kHz processing.

Now this is where I change my mind... if I was to do a completely analog master I would playback at the existing sample rate 44.1, 48, 88.2 or 96k & then capture at 24bit 44.1khz if it's going to CD.

Matt

[Virtalahde]

I upsample now and then. Usually up to 88.2k, and only in cases where I know it's going to be loud. I find it makes the work a tiny bit easier for some reason.

But not on every project, and I always do it before anything else. Upsample, set up the session at 88.2k, process & capture, edit & limit and then to 16bit/44k1.

[djwaudio]

I use upsampling in the D/A before the analog chain. I don't like more than one resampling process in my mastering chain. Re-capturing is almost always 44.1 if it's a CD release.

I'm kind of an analog guy so I'm biased against any additional digital processing after EQ.

[filogab]

Sounds like someone is concerned about the downsampling post the up. Would it be different going to analog after the upsampling and then going back to digital recapturing at 44.1/16 or whatever is needed for the final medium?
Aren't the modern A/D converters capable of great performances also at the lowest sample rates?

[Robin Reumers]

I think it also depends on the kind of processes you're using. Most Weiss equipment I know, automatically upsample when passing signal at 44.1/48 K. So if you're using say 2 DS1-MK2 and 1 EQ1 in a row, which is better, first upsample to 96 with a seperate SRC, then do the processes at 96, and at the end go back to 44.1 or have the DSP in each box go up and down 3 times in a row? I think really it's all about knowing your equipment and knowing what sounds best in what place!

[HIGHENDONLY]

For listening to music, do any yous guys upsample your music to 88.2?

[af_analog]

Originally Posted by HIGHENDONLY

For listening to music, do any yous guys upsample your music to 88.2?

Are you expecting a change? You can't create samples that aren't already there...

[HIGHENDONLY]

Originally Posted by af_analog

Are you expecting a change? You can't create samples that aren't already there...

Yea but i thought by taking a song off a cd"44.1 16bit" than converting up to "88.2 24bit" will improve the the quality somewhat??

[UnderTow]

Originally Posted by HIGHENDONLY

Yea but i thought by taking a song off a cd"44.1 16bit" than converting up to "88.2 24bit" will improve the the quality somewhat??

Going from 16 to 24 bit will just add 8 zeros to each sample. Not useful. Going from 44.1 to 88.2 will just add interpolated values for the extra samples.
Don't forget that the vast majority of DA converters will up sample to much higher than 88.2Khz anyway.

So in short: No. That is a complete waste of time.


Alistair

[HIGHENDONLY]

Originally Posted by UnderTow

Going from 16 to 24 bit will just add 8 zeros to each sample. Not useful. Going from 44.1 to 88.2 will just add interpolated values for the extra samples.
Don't forget that the vast majority of DA converters will up sample to much higher than 88.2Khz anyway.

So in short: No. That is a complete waste of time.


Alistair

Thanks for clearing that up

[lucidwaves]

Originally Posted by HIGHENDONLY

Thanks for clearing that up

It's a common misconception, but yeah your not gaining sound quality, your just capturing the existing sound quality at a higher resolution which will sound the same.

[acorneau]

Originally Posted by MattGray

I agree that the downsample can degrade the signal if a quality SRC isn't used, but I find (particularly if you are using any downstream digital processing) that an upsample to 96kHz often sounds just that bit better even after the downsample to 44.1kHz.

The "best" SRC I have is the Z-sys 3src, which is good but it still changes the sound in a way that doesn't make it worth it.

Now this is where I change my mind... if I was to do a completely analog master I would playback at the existing sample rate 44.1, 48, 88.2 or 96k & then capture at 24bit 44.1khz if it's going to CD.

That's pretty much my situation; playback at native rate, capture at 44.1kHz. I don't do a lot of digital processing (for the most part) so the added benefit of upsampling is negligible in the long run.

Just my experience with the tools I have.

[aivoryuk]

for me the whole upsampling and downsampling thing really depends on what i'm doing. i've found that if im clipping and going for loudness then this does sound better at higher sample rates and the benefits of this outway that of doing upsampling and downsampling.

However it also depends on what genre it is as i've had experience with electro music that i've done where it sounded better the whole way through at source rate then it did with upsampling.

some plugins do sound better at the higher sample rates and the benefit of that can outway that of upsampling and downsampling.

its all swings and roundabouts really.

the one thing i do avoid is doing multiple upsampling and downsampling in the same pass for example, using psp mastercomp in fat mode followed by voxengo elephant in 4xsampling mode, it just doesn't sound right to my ears IMO

[Virtalahde]

Originally Posted by aivoryuk

i've found that if im clipping and going for loudness then this does sound better at higher sample rates and the benefits of this outway that of doing upsampling and downsampling.

This is exactly what I've found out. You just gotta experiment and know when to upsample and when not to..

the one thing i do avoid is doing multiple upsampling and downsampling in the same pass for example, using psp mastercomp in fat mode followed by voxengo elephant in 4xsampling mode, it just doesn't sound right to my ears IMO

I had a project where I was forced to use two Elephants in a row. First was clipping a little, the second one squeezing a tiny bit in AIGC-4/med (this combination works sometimes pretty well). The session was 88.2k, clipping Elephant upsampled 4x and I was shocked to see what a difference it made to use the 2nd Elephant in oversampling mode! It widened the sound in a way that really suited the particular project. It definately added something, wasn't neutral, but it was really good.

Always listen for different oversampling modes. Today I mastered something were the low end sounded way better in 2x than 4x. Otherwise it was pretty much the same.

[MattGray]

Originally Posted by acorneau

The "best" SRC I have is the Z-sys 3src, which is good but it still changes the sound in a way that doesn't make it worth it.

You should try either the Weiss Saracon software re-sampler or AudiofileEngineering's WaveEditor which has the 64bit iZotope SRC. Both are at the top of their game & sound very clean. I'd be interested if you did some tests using these SRC's whether you would not hear the benefits in your digital processing at 96kHz even after the downsample to 44.1kHz. Sure it's a subtle difference & some would argue whether the extra time & hassle is worth the effort. I have decided with jobs that I'm using a number of digital processors that 'yes' it is...

Matt

[MattGray]

Originally Posted by aivoryuk

for me the whole upsampling and downsampling thing really depends on what i'm doing. i've found that if im clipping and going for loudness then this does sound better at higher sample rates and the benefits of this outway that of doing upsampling and downsampling.

Really? you find that clipping sounds better at 96kHz? Are you talking about A/D clipping or digital clipping in your DAW? There is a flaw in your theory called "intersample peak overs". If you achieve your final levels at 96kHz whether it be through clipping, limiting or both & then you downsample to 44.1kHz you will generate intersample peak overs which can exceed 0.0dbfs & introduce further aliasing & grunge that doesn't sound as good to my ears as setting your final levels post SRC at 44.1kHz. These overs are often exaggerated even further downstream in cheaper consumer CD players or broadcast chains & are best avoided where possible.

Matt

[aivoryuk]

Originally Posted by MattGray

Really? you find that clipping sounds better at 96kHz? Are you talking about A/D clipping or digital clipping in your DAW? There is a flaw in your theory called "intersample peak overs". If you achieve your final levels at 96kHz whether it be through clipping, limiting or both & then you downsample to 44.1kHz you will generate intersample peak overs which can exceed 0.0dbfs & introduce further aliasing & grunge that doesn't sound as good to my ears as setting your final levels post SRC at 44.1kHz. These overs are often exaggerated even further downstream in cheaper consumer CD players or broadcast chains & are best avoided where possible.

Matt

generally yes, i think limiting-clipping does sound better at higher sample rates but as i said in my post i don't automatically do it. I use my ears and if it benefits the track then i process or capture at higher rates and then downsample. While then limiting after the SRC to set the output to normally about -0.3. Intersample peak over is not a flaw in my theory, as i stated in my post i'm more interested in what it sounds like rather than worrying what my meters show.

I could always use the ozone limiter with the intersample box ticked if i want to be anal about it

[MattGray]

Originally Posted by aivoryuk

generally yes, i think limiting-clipping does sound better at higher sample rates but as i said in my post i don't automatically do it. I use my ears and if it benefits the track then i process or capture at higher rates and then downsample. While then limiting after the SRC to set the output to normally about -0.3. Intersample peak over is not a flaw in my theory, as i stated in my post i'm more interested in what it sounds like rather than worrying what my meters show.

I could always use the ozone limiter with the intersample box ticked if i want to be anal about it

Setting your ceiling to -0.3dbfs alone isn't enough to catch these sorts of overs & they won't show on normal peak metering so looking at meters won't tell you the whole story. It may also sound fine through your high quality monitoring DAC but it's the downstream processing on radio & consumer CD players where these overs will reveal themselves in a distorted manner.

P.S. - Isn't it our job to be 'anal'?

Matt

[aivoryuk]

Originally Posted by MattGray

Setting your ceiling to -0.3dbfs alone isn't enough to catch these sorts of overs & they won't show on normal peak metering so looking at meters won't tell you the whole story. It may also sound fine through your high quality monitoring DAC but it's the downstream processing on radio & consumer CD players where these overs will reveal themselves in a distorted manner.

P.S. - Isn't it our job to be 'anal'?

Matt

with ref to the recent WUMP, a lot of engineers including the more experienced ones had intersample peak overs even when capturing at source rate. So either they are not aware of of intersample or are not bothered. It would be interesting to hear what other peoples opinon on it are.

a further point of interest that still a lot of current of cd's are still going to 0dBfs, so on less then adequate equipment you may hear distortion or even when converted to mp3 (the current trend) will produce overs and can distort.

This diesn't seem to be a issue for some of the great mastering engineers

[meloco_go]

before writhing off SRC check how your DAW performs it - SRC Comparisons
there can be DRASTIC difference

[Nordenstam]

Originally Posted by MattGray

Really? you find that clipping sounds better at 96kHz? Are you talking about A/D clipping or digital clipping in your DAW? There is a flaw in your theory called "intersample peak overs". If you achieve your final levels at 96kHz whether it be through clipping, limiting or both & then you downsample to 44.1kHz you will generate intersample peak overs which can exceed 0.0dbfs & introduce further aliasing & grunge that doesn't sound as good to my ears as setting your final levels post SRC at 44.1kHz.

There's something to be said for both ways with regards to intersample peaks. Limiting and clipping at higher rates produce less intersample peaking in the resulting material. Clipping at say 96kHz and SRC'ing to 44.1 will produce less IS peaks than say clipping at 48kHz before going to 44.1. Clipping at 44.1 removes the potential problem for clipping when downsampling, but it will, generally speaking, produce more intersample peaks than clipping at higher rates.