Is it possible that what you guys are hearing at 96k+ sample rates and up isn't really increased accuracy, and is actually artifacts of the analog stage ahead of the A/D's trying to drive a fast converter chip?

The reason I'm asking is that I was recently involved in the design of a data acquisition system, which featured what I thought was a pretty good analog front end (differential amp and a simple 2 pole filter....we were heavily oversampling). We didn't pay enough attention to slew rate, and soon found that there were errors from one channel to the next as the analog stage couldn't drive the (slightly) capacitive load of the A/D fast enough while switching channels. We tried some ridiculously fast IC's (1000V/us slew rate), but they would ring. Eventually we found some IC's with good settling times, that worked out well for the application, but the big lesson learned is that the analog part of an A/D converter is not a trivial matter, it can play a huge (dominant even) factor in how a converter sounds.

I'm of the belief that 44.1 kHz is an adequate sample rate for audio...If done properly. It's just more of a challenge (for designers) to get the filters to sound good and have sufficient attenuation. I think really the key thing to note is that jsut because a converter is faster (in sample rate) does not necessarily mean it's better than a (different) 44.1kHz converter. We need to use our ears....

Cheers,

Kris

Scientific proof first requires understanding of science. It is very easy for Dan to preach to a semi informed internet crowd.

I wish we could hear some educated opinions about the substance of this discussion. For example opinions about ten other papers on the subject that I quoted. I'm wondering if anybody read these. It's unfortunately a requirement to understanding what are we even talking about here.

Michal www.mytekdigital.com

I'm sorry to hit back, in the previous post but I'm not sure if you understand that I may have felt offended with your particular choice of words.

Is this another right question? No, I quoted them randomly just picking up titles and I couldn't even understand these titles because I'm a dumb idiot who just happened to design great sounding converters by pure freak luck.

I'm sorry too. I wish we were capable of real conversation. You are indeed very strange- I almost feel like you haven't read or you forgot my posts. There was a lot of details there , concrete things that readers here understood but you simply ignored. You said nothing about filters, DSD, concrete chip devices (5394, 5396,5381 etc) and the fact you completely misunderstood me as in your previous to this post. This is very weird.

I don't know if real discussion is possible. Looking at your posts at prosoundweb you tend to turn to dust anybody with different opinion. I'm getting tired too.

I think this discussion could only proceed with a moderator who really understands the subject.

I'm off, thanks

Michal www.mytekdigital.com

You gotta be kidding, who is this lot?

Interesting point and I heard similar from experienced engineers and producers , some of them even prefer and use older converter designs with 44 and 48 kHz with 20-bit depth for tracking (although they use only analogue processing and mixing) THAT SOUND DEFINITELY CLOSER TO ANALOGUE than 24/192 presently.
I don't fully share their opinion, but finished projects I heard were astonishing.
If somebody is striving to reach sonic nirvana, focusing on 96 or 192 sample rates only would not do a lot.
It would bring real improvement to consider more appropriate mixing solution and choice of effect-processors than to scream how 96k or 192k dramatically improved 'flat, two-dimensional, thin, plastic, harsh etc.' sound of 48k.
Such audible differences, if exist on particular material, resulted to greatest extent due to lesser quality converters, questionable processing and mixing algorithms than principal improvement that 96k gives over 48k (that is practically not audible when good conversion and appropriate mixing is applied).

So, returning to original question
96k vs. 48 will most likely not provide or guarantee better mix, if everything else is done correctly and proper routine is used during the whole path. In same time some marginal differences that could be perceived as improvement are noticeable on some materials, so without doubt when such occurs there is no reason to exclude 96k as preferred choice.
To calm minds of our 'hi-tech conscious' clients we have 88k-192k always ready to be switched on and often this is the main improvement we can really do with that.

GYang

Not science here, but projects using my Digi 192 i/o & PT HD sound better to me and others at 96k than 44.1, even after mixing down to Redbook.

I sit in front of the machine.
There is a little pull down menu that allows me to choose the sample rate.
I record a bit at one rate and listen.
I record a bit at the other and listen.
Everyone in the room thinks B sound better than A.

We proceed to make a record.
later we order dinner.

this requires no paperwork.

Mr. Wittman surely you jest ! There simply isnt enough angst and bravado in that scenario.You straight thinking enjoyer of life get thee to a nunnery and thank you for the normality.

I apologise for not understanding 95 % of what you two are talking about.

priceless.

"Originally Posted by wwittman

I sit in front of the machine.
There is a little pull down menu that allows me to choose the sample rate.
I record a bit at one rate and listen.
I record a bit at the other and listen.
Everyone in the room thinks B sound better than A.

We proceed to make a record.
later we order dinner.

this requires no paperwork."


Later, everyone listens to my choice via mp3, on an iPOD.



TH

I have read all the papers on the dCS website, and find them to be lucid and easily understandable if you have some backfround in electro acoustics.
The paper on higher sampling rates is excellent, explaining that digital filters ring before and after the signal contributing a low level a-musical
hash that obfuscates inner detail, dynamic nuance, reverbrant decay, etc,
I own dCS upsamplers and dacs, and invite anyone to come and listen to different filter topologies at different sampling rates. YES, there is a difference!

im still wondering if it is actually a step down from the $50 bookshelf systems people buy to listen on or not.

*sigh*

All of ya, go make some music will ya? My head hurts....



I will say this.. It is human nature to think the grass is greener on the other side of the fence.

I distinctly remember when CD's first came out and everyone was talking about how much better they sounded than albums. Could have been that they were buying marketing hype? You bet yet I remember that everyone I knew in the hifi audio game wanted to be the first one to go down to the music shop to hear one.

Now CD's are the norm and the same 44.1 CD's are crap, we need 48, then 88.2 and 96 and... Could this be marketing hype as well?? Ask yourself.

I do agree with William Wittman, choose what sounds good. That is a given as it should be but it is not the real argument here. The real argument is should designers spend time reaching for pie in the sky sample rates or should they spend there time making 44.1 or 48 sound better.

The ears might say 96 sounds better (not on some converters go back and reread this thread) and the math says 44.1 and 48 can sound good.

I am going to go make some music now.. have fun.....

Kinda depends on whether they're using apple's $0.30 earbuds or not...

I've got an iPod...pretty cool peice of technology, but those earbuds are absolute crap. Relaced 'em with a good set of headphones, and I'm now I'm getting re-aquinted with items from my CD collection that I haven't listened to for quite some time. With good heaphones, and a high enough bit rate AAC, the iPod sounds pretty damn good to me!

Cheers,

Kris

my fireface sounds a lot better @ 96

but due to bandwidth i stay at 44.1

he ipod headphone amp is possible the worse peice of technology to ever grace the market. yeah..and if you use some decent cans, say the seinnehiser hs280s..u only get an hour or so of battery life!

LOL...you almost make me die here!!!

_________
I

The pre and post "ring" of which you speak is the energy in the transition band that remains when the high frequency information is removed. Most analog filters only have a post "ring," and this is evident in, say, the human ear. The human ear itself has this "ring" because the absence of the high frequencies exposes the lower frequency (transition band) content that remains. Digital filters can have this "ring" on both sides (pre and post). I keep putting the word "ring" in quotations because it is not actually a ring - it is merely the remaining low frequency content and sounds nothing like a ring if designed properly. Further, because this content is in the transition band only, if the filter can be restricted to having a transition band above the human frequency range then this frequency content isn't audible anyway - whether before or after the waveform. The content is only present in the audible range if the filter is poorly designed.

Nika

Nikka,
Thank you for the clarification.
I'm not sure my conclusion does justice to te dCS paper Mike Story wrote. What do you think of it's observations?Thanks,
Mike Chafee

It has been a long time since I read this paper, but if I recall it correctly its points are valid on poorly designed filters OR if we concede that humans can hear above 20kHz. Otherwise this issue is irrelevant. Unfortunately, many filters are indeed poorly designed.

Nika

Sampling at 44.1KHz often yields great results, when done well. But 44.1KHz sampling (22.05KHz audio) is a bit tight. The ACCUMULATED limitations on high audio frequencies can degrade both high frequency flatness response and high frequency phase linearity.
Most mics are limited to 20KHz. What does it mean? It means there is already a 3dB loss of response at 20KHz (3dB is the half power point). So it is possible to have say 1-2dB loss at some 18KHz…
The same is true for speakers, so a mic at the record end and a speaker at playback can yield 6dB accumulated loss at 20KHz, so what happens at 17-18KHz?
Adding yet another device, an AD, with yet another 3dB loss at 20KHz, may become very audible, depending on the music.

That is were some sensible increase in sampling rate may help. How much is sensible? It helps to do the sampling at 48KHz, and better yet at 60KHz. Unfortunately, there is no 60KHz standard, so 88.2-96KHz may be a reasonable alternative. A bit faster then the optimum rate of 50-70KHz, but not outrageously so.

It is NOT about processing signals that the mic can not capture, the speaker can not play and the ear can not hear. It is about having a good audio range, that range that ear can hear. If we want flat audio response to 20 KHz, we need the mics and speakers and AD converters to be flat to 20KHz. Most mics and speakers are not, but at least we can get the AD flatness out of the way. It only takes a few KHZ increase. 88.2KHz is more then enough.

It is not about unexplained mystery. It is, in fact, “nuts and bolts”, easy to understand stuff…

Regards
Dan Lavry
www.lavryengineering.com

well put, that makes sense.
thanks Dan.