I shoudn't have tried 96 Khz !

decimator · Post by **decimator** » Sat Oct 02, 2004 12:25 pm

Well on a Pulsar II, you don't have much room left for fantasies and since 96 Khz eats 2 X DSPs compared to 44.1 ...

I tried looong ago when I had way less synths, nowadays I got more and also became more and more picky about sound clarity and quality.
And now I've got to buy a Scope SRB I guess because the results were quite striking, unfortunately one voice only on not hungry devices so the need for more DSP.

I run sonograms at 96 Khz : lot of things happening above 22050 Hz and I realize 192 Khz is not superfluous either since in most cases I don't see the " top of the tree ".

In fact by recording at 44.1 you've got sounds * terribly * altered, muddied and polluted by all those reflected frequencies and no need to go very high in the keyboard at all.

Run tests yourself and you'll see this is not audiophils matters !

To be noted that downsampling can produce interesting sounds though at times, it's a case by case situation ...

Now in my head it's " 96 Khz or bust !! " I'll only do presets and such waiting to be able to buy some SRB ... long wait ... sigh

Shayne White · Post by **Shayne White** » Sat Oct 02, 2004 1:37 pm

I know; 96KHz sounds way better than 44.1 for synths. But since it takes up so much DSP, I've arranged projects using 44.1 and then separately recorded one or two synths at 96 and then downconverted to 44.1 using high-quality algorithms. It's a bit cumbersome, but it works.

decimator · Post by **decimator** » Sat Oct 02, 2004 2:41 pm

Hey, you stole my idea Shayne !

I ran other 96 / 44.1 comparisons and more WOWWW !!

For instance on a given preset : it sounds good but not excellent, dull and muffled ( at 44.1 Khz ) on 96 Khz : sparkling and well defined !

Sonogram at 96 Khz : a lot of noisy background above 22050 Hz so when I listen at 44.1 all this crap is mirrored and the sound is muddy !

A more drastic case : sometimes quite noticeable ( in intensity ) and not noisy frequency content lies above 22050 Hz so when you play at 44.1 you throw in a structured pack of frequencies that really modify the sound and at times 44.1 and 96 Khz gives quite different sounds on a given preset !

At other times, there's not that much over 22050 Hz so the difference is negligeable.

Late revelations ! tomorrow the benefits of the wheel !

blazesboylan · Post by **blazesboylan** » Sat Oct 02, 2004 2:53 pm

I know I'm going to get flamed for this. That's OK, this is the document someone else took me to school with on digital sampling:

http://www.dbtechno.com/documents/Sampling_Theory.pdf

Flame away!

nitty gritty · Post by **nitty gritty** » Sat Oct 02, 2004 4:54 pm

Nice one blazesboylan!!

Hee hee....

decimator · Post by **decimator** » Sat Oct 02, 2004 4:59 pm

No flame if you save me money ...
I put up to 8 stock high pass filters in chain at 20 Khz, though it seems to terminate a lot above I prefer unfiltered 96 Khz by a very large margin.
I can't dream of 192 Khz, we don't have it ( yet ) 96 Khz will be well enough for me unless you have better filters to propose.

Liquid Len · Post by **Liquid Len** » Sat Oct 02, 2004 5:21 pm

Not a flame, but I've ran Lightwave at 44KHz, 48Khz, and tried running it at 96KHz, and as far as I'm concerned, at 96Khz it sounds clearer and more detailed. I don't know that this contradicts the article you posted - making something do its internal calculations at 96khz is different than listening to the final output at 96khz. Generally, when you do calculations, it's recommended to hang onto all the precision you can, and only round the final answer. If there was a way to make Lightwave run internally at 96khz but output at 44khz or 48khz (I believe Minimax does this), p'raps there would be no discernable difference whether you were running the card at 44, 48, 96. I don't know for sure what the answer is, but I know I hear a noticeable difference in quality in Lightwave running at 44/96. If it's a placebo effect, it's a pretty stubborn one.

I have considered schemes similar to Shayne White's, the only question is whether it's worth the bother, to me it just isn't. I can use Minimax and Proodysey (and Profit 5!) running the Creamware card at 44 or 48, rather than jump through hoops to get a 96khz-calculated Lightwave recorded at lower frequencies.

astroman · Post by **astroman** » Sat Oct 02, 2004 7:17 pm

I also did some comparisons, but not to deep.
While the Profit5 seems almost immune to sample rate settings, the UKnow benefits in the described way by less artifacts, as does Uberplastic.
Python sounds more defined and OP8 seems play the same sound at a higher rate.

Imho it's difficult to judge if the perceived improvement is in fact due to an increased precision or a side effect from different loading of the DSPs.
The phase shifting error (by crossing DSP boundary) may even result in a more pleasing sound

The Python actually revealed the most significant differences, some patches definetely sounded more defined while others s*cked.

Anyway, with 8 DSPs I wouldn't trade the 'improvement' for the lower load.
Imho it's a task for a new hardware layout, if that's ever to come...

cheers, Tom

[ This Message was edited by: astroman on 2004-10-02 20:18 ]

Shayne White · Post by **Shayne White** » Sat Oct 02, 2004 9:14 pm

Well hey, it'd be nice if we could convince to CW to reprogram Lightwave (and Vectron, for that matter) to run internally at 96Khz all the time!

Post by **garyb** » Sat Oct 02, 2004 9:57 pm

this is where hubird suggests making music.

the article said it all, i think. a better process at 96 or 192k would be just as much( or mo') better at 44.1, 48 or maybe 60k. maybe we like the distortion of 192k...still, this is a problem with what is being sold rather than actual samplerate. i'm sure that some processes might be better done at the higher rates and then the audio outputed at a lower rate and that's a job for the engineers(

),but for the rest of us, high sample rates are a trap for the obvious reasons(stated in the article) of storage space and processing power(already overtaxed by poor programming as pointed out by astroman repeatedly).

it's important to see that while using the imagination is wonderful, it's good to live in the real world. only SOME things benefit from the higher rate and some things suffer. in the end, you can have the same quality at the lower rate if the material(hardware) and programming are equal. it's clear that the best processes are not always being used, and that expensive, silly ones often are because we like glamour and romance and because we are not interested in simple, elegant solutions(if we were, the these companies wouldn't be making all these endless, unessessary products, but would be making a few products as well as possible...).

so my machine sounds better at 96k. it sounds like the radio already and it is half useless at that rate. what's the point?

Post by **hubird** » Sat Oct 02, 2004 10:42 pm

I love this way of thinking, Gary

I don't wanne be blind for new devellopments, better recording technics, but I don't wanne try to focus on the max if it leads to less 'production'.
Working with today's standard values in recording technics should bring me good (sounding) songs, otherwise I'm not good enough (to sell anything).

Making the right choises bastling together a song is thausend times more important than trying to get that last 2 procent more sound quality.
When I start a song by choosing the right 2 bar drumloop which inspires me to a new song, I've made a very important decision.
Before it there was silence, now I hear something musical, by choice.
Looking for that impuls is where it all starts for me.
If I then can't get the sound right, the song isn't good enough

[ This Message was edited by: hubird on 2004-10-02 23:42 ]

krizrox · Post by **krizrox** » Sat Oct 02, 2004 11:25 pm

I know what I'm about to say is very unscientific but I agree there is a difference at 96kHz. However, I record everything here at 24/44.1 because I felt I got most of the advantages of 24 bit recording without the extra baggage. Most of my clients don't even know how to tune their guitars. Dealing with those giant files is just a big ol' waste of storage space and time for me.

But I think you all should definitely go 96k!

blazesboylan · Post by **blazesboylan** » Sun Oct 03, 2004 1:57 am

On 2004-10-03 00:25, krizrox wrote:
Most of my clients don't even know how to tune their guitars.

I feel your pain Krizrox!

wolf · Post by **wolf** » Sun Oct 03, 2004 2:25 am

hi,

from a developers point of view it is just, that it is far easier to make filters for 96kHh than for 44.1kHz, as you don't need to program a steep curve, when it comes to the aliasing frequency. In a turnaround you can also say, synths, that sound "better" in 96kHz, are in need of better coded filters.
That said it is very easy to make a filter, but not easy to make a good and musical sounding filter.
On the other hand sometimes aliasing is wanted from a sound designers point of view ..

best
Wolfgang

ernest@303.nu · Post by **ernest@303.nu** » Sun Oct 03, 2004 5:25 am

Here's another believer (because I hear it!) of higher samplerates.
One thing thats's always left out when a scientist 'proves' that higher sampling rates are useless, is that the human brain is extremely sensitive to phase information. Very small timing differences between the left and right speaker are essential for our brain to determine the position of a soundsource, as well as the exact nature of a (natural or artificial) ambience. Altough even at 44.1kHz sampling rate the phase accuracy of the sampling equals little more than one centimeter (due to the speed of sound in relation to the sampling rate), the brain seems to be even more sensitive than this 1 sample/1cm quantisation when it comes to spatial experience.
Listening reports of Super Audio CD proves this to be true, especially with live classical recordings where the spatial position of the individual musicians can literally be drawn out from listening, even during complex parts which tend to get muddy on a 44.1kHz CD (I have heard this personally, and please don't say that normal 44.1kHz classical CD's are equally spatial sounding; because the are not!)

any thoughts on this?

[ This Message was edited by: ernest@303.nu on 2004-10-03 06:27 ]

astroman · Post by **astroman** » Sun Oct 03, 2004 7:21 am

On 2004-10-03 06:25, ernest@303.nu wrote:
...Listening reports of Super Audio CD proves this to be true, especially with live classical recordings where the spatial position of the individual musicians can literally be drawn out from listening, even during complex parts which tend to get muddy on a 44.1kHz CD ...

right Ernest, and you've just prooved that statement wrong in regard to frequency.
SACD is a 4 channel format, hence you have a more detailed spatial resolution...

I can confirm your sentence about the phase sensitivity of the ear (off topic).

On a bicycle in regular traffic it becomes a question of survival (sometimes) to know exactly what's going on behind you.

With some experience you'll be able to do the following trick:
a car comes closer from behind - unnoticed, because it's too far away yet - another vehicle approaches in the opposite direction, emitting it's own sound and the reflection of the noise from the car approaching you from behind.
Your brain immediately splits up the 2 components of the sound, telling you something comes from behind - and to pay attention because of something from ahead, too.
Even with each component in the 'system' moving at a different speed, you'll make stunningly accurate guesses about this situation.

cheers, Tom

decimator · Post by **decimator** » Sun Oct 03, 2004 8:12 am

Here's a small sonogram of the same preset in Prodissey, 44.1 then 96.

So when there's nearly nothing above 22050, the 96 Khz does mainly bring fatter files now when there's something ... hmmm ... and this is a mild example !

I prefer pristine sounds to work with at first then degrade if I choose to.

The filtering solutions I tried weren't convincing, I even put ISON 48 dB HP filter at 24 Khz but still I prefer 96 Khz unfiltered.

[ This Message was edited by: decimator on 2004-10-03 09:20 ]

ernest@303.nu · Post by **ernest@303.nu** » Sun Oct 03, 2004 9:38 am

On 2004-10-03 08:21, astroman wrote:
On 2004-10-03 06:25, ernest@303.nu wrote:
...Listening reports of Super Audio CD proves this to be true, especially with live classical recordings where the spatial position of the individual musicians can literally be drawn out from listening, even during complex parts which tend to get muddy on a 44.1kHz CD ...
right Ernest, and you've just prooved that statement wrong in regard to frequency.
SACD is a 4 channel format, hence you have a more detailed spatial resolution...

I hope you're kidding me, or are you really that misinformed?
Super Audio CD can accomodate for 4 channels of audio (or even 6 channels, according to the specifications), but it's main feature is the so-called 'Direct Stream Digital'-technology, allowing for much higher resolution and stereo coherence (2 channels!) than the traditional 44.1kHz recordings.
This 'Direct Stream Digital' differs somewhat from 'traditional' quantized sampling techniques, but both 96kHz/192kHz recordings and SACD-recordings benefit from higher phase-accuracy, which allows for a more accurate stereo image than possible with a 44.1kHz recording.

All I wanted to point out is there's much more to the quality of a recording than just frequency response (which can or cannot be perceived by the human listener); the critics of higher sampling rates never seem to take these issues into account!

I can confirm your sentence about the phase sensitivity of the ear (off topic).

I don't think this example is off-topic; it perfectly sketches why recordings (especially live recordings in a delicate ambience) can benefit from higher sampling rates to reproduce a highly accurate binaural image of the recording environment.

nprime · Post by **nprime** » Sun Oct 03, 2004 12:26 pm

On 2004-10-03 06:25, ernest@303.nu wrote:

One thing thats's always left out when a scientist 'proves' that higher sampling rates are useless, is that the human brain is extremely sensitive to phase information. Very small timing differences between the left and right speaker are essential for our brain to determine the position of a soundsource, as well as the exact nature of a (natural or artificial) ambience. Altough even at 44.1kHz sampling rate the phase accuracy of the sampling equals little more than one centimeter (due to the speed of sound in relation to the sampling rate), the brain seems to be even more sensitive than this 1 sample/1cm quantisation when it comes to spatial experience.

any thoughts on this?

[ This Message was edited by: ernest@303.nu on 2004-10-03 06:27 ]

In the high end audio world the consensus is that the quality of the clock signal is absolutely critical. A great deal of the design effort goes in to providing ultra stable power to the clock. We are extremely sensitive to jitter, which as stated above casues phase issues in the signal. This is what I would call the "smearing" of the image.

In other words, I think the information is there at 16 bit 44.1, but it is still very expensive to allow you to hear it all properly.

I think the issue of ultimate high frequencies is a red herring. Everyone of us would be shocked and appalled at the highest frequency we can "hear". I know the last time my ears were tested I was told that I wasn't hearing much above 16 KHz.

The reason that this is not such a big deal is that we are only talking about the last octave of musical information. Muaic spans about ten octaves of perceivable information, give or take from 20 Hz to 20,000 Hz. The stuff that we are debating here (16 Khz and above) is all in the top quarter of the last octave, (the last octave spanning from 10,000 to 20,000 Hz). In other words the top quarter of the top one tenth of the range.

I do believe that it is all about phase accuracy. Perhaps the 96Khz rendered signal you are hearing either has less or different phase issues.

R

[ This Message was edited by: nprime on 2004-10-03 13:28 ]

[ This Message was edited by: nprime on 2004-10-03 13:31 ]

Immanuel · Post by **Immanuel** » Sun Oct 03, 2004 12:58 pm

I don't know *hit about, what is actually important here, but in these discussions my mind always go to the 15kHz sine wave topic. At 44,1kHz sampling, you have ~3 samples to create a sine wave. At 96kHz sampling, you have more than 6 samples for the same task. We usually do not listen much to sinewaves at this frequecy, but I mypothesise, that if we are able to hear the difference between a saw tooth and a sine wave at 15kHz, then we will benefit from higher sample rates, simply because the complex sound of music will be less chunky.