upsampling from 44khz to 96khz higher amplitude?

[Reptil]

hmmm it could be that the process introduced spikes, or emphasised faults in the original.
so...... this is reason to NOT push it to the maximum ceiling
go for a -10 dB (digital scale) or lower!

it will sound MUCH better (try it?!)

[-tc-]

No it's not strange at all for that to happen ... it's entirely expected but has nothing to do with adding harmonics (unless you use a low-quality src, and then it would be inharmonic distortion not harmonics) or emphasising faults

Your Ableton meters display sample values.

Upsampling audio can essentially be thought of as a process of interpolating a curve using the existing sample values, and then resampling this "implied curve" at the higher sample rate.

If the sample values of your original 44.1kHz file are not at the peak amplitude of the waveform* (and the vast majority of the time they're not), then the new sample points at the higher rate will be very likely to contain higher values ...

A visual aid ... imagine that it is your original 44.1kHz file ... new sample values taken on the curve interpolated from the old sample values are very likely to be higher in amplitude ...


So you can see that, because your meters display sample values and not an actual "waveform", it's not at all surprising for an upsampled signal to have a higher peak value than its source.

*this applies not only to recorded audio, but also to synthesised waveforms ... the sample points are "descriptors" of what the waveform will be when it is converted to a continuous signal at the output of your d/a conversion.

Hopefully that makes sense ... if not, research "intersample peaks" and if you end up back on g-slutz, avoid the posts of a certain oldanalogueguy because his relentless and repetitive pedantry might just confuse you

[Augsy]

It was a mastered House track...the kick peaked right at 0db.

[Augsy]

Originally Posted by Reptil

hmmm it could be that the process introduced spikes, or emphasised faults in the original.
so...... this is reason to NOT push it to the maximum ceiling
go for a -10 dB (digital scale) or lower!

it will sound MUCH better (try it?!)

well it does sound better (obviously) anything over O db on the meter and I can usually hear some sort of distortion.

[oldeanalogueguy]

Originally Posted by Augsy

I noticed when i load a track that is an .mp3 44khz quality into ableton it will peak at 0 db, as you would expect, but when I run ableton at 96khz sample rate the 44khz .mp3 will peak slightly higher around a db or so.

Does anyone care to shed any light on this?

Im assuming that the upsampling is adding some of the upper harmonics back into the audio. Ive always thought 96khz sounded "brighter".

artifacts?...

how the upsampler worked and how the meters work...

upsampling accurately is not easy but shoudl not add highs just more resolution of what was there

[-tc-]

Originally Posted by oldeanalogueguy

upsampling accurately is not easy but shoudl not add highs just more resolution of what was there

In the interest of clarity, the above is potentially misleading ... upsampling an already digital signal simply increases the potential bandwidth of that signal. If it is done well, no information will be added above the upper bandlimit of the original signal*, and the increased bandwidth will have no effect on the "accuracy" of the content of the original signal ...

*in practice, distortion that peaks at around -180dBFS is essentially nonexistent for our purposes.

[golden beers]

this is all a guess. i've got no idea but here's my thoughts.

the original transients are most likely clipped

the upsample algorithm makes curves where previously there was none.

it does this because it has to fill the gaps in the information. youre going from less to more information. this info has to come from somewhere: the algorithm inserts it.

the resolution of the original may have for example, at the transient of the kick; 2 sample points next to each other at 0db (a clip). during the upsample process these 2 points are moved apart and the gap in between is filled by an appropriate curve, which may go over 0db.

[filipv]

Originally Posted by Augsy

I noticed when i load a track that is an .mp3 44khz quality into ableton it will peak at 0 db, as you would expect, but when I run ableton at 96khz sample rate the 44khz .mp3 will peak slightly higher around a db or so.

Does anyone care to shed any light on this?

This one is easy: intersample peaks. Upsampling created a data point which didn't exist previously.

If the original material didn't peak at 0 db, this wouldn't have happened.

[golden beers]

Originally Posted by filipv

This one is easy: intersample peaks. Upsampling created a data point which didn't exist previously.

If the original material didn't peak at 0 db, this wouldn't have happened.

oh snap

[nms]

Originally Posted by filipv

This one is easy: intersample peaks. Upsampling created a data point which didn't exist previously.

If the original material didn't peak at 0 db, this wouldn't have happened.

Tangible evidence of intersample peaks for those who scoff at the idea. Also, another benefit of working at higher sample rates