sample rates

[Kris]

What I don't get about "upsampling".... is that if it isn't there to begin with, you aren't going to get it no matter what.

I spend a good deal of my time "upsampling" standard definition video to high definition for clients who want to future proof their catalogue. And you know what? It looks like crap compared to the original SD version. The only way to truly upsample is to re-scan all the film and re-mix all the audio. It ain't gonna happen any other way, so I am a little wary of plug ins or boxes that claim they can do this. Where do they get the data from if it was never there in the first place????

[David W]

[quote="Kris"]What I don't get about "upsampling".... is that if it isn't there to begin with, you aren't going to get it no matter what.

I spend a good deal of my time "upsampling" standard definition video to high definition for clients who want to future proof their catalogue. And you know what? It looks like crap compared to the original SD version. The only way to truly upsample is to re-scan all the film and re-mix all the audio. It ain't gonna happen any other way, so I am a little wary of plug ins or boxes that claim they can do this. Where do they get the data from if it was never there in the first place????[/quote]

I think you are talking about up-sampling for a different reason. I'm sure UA know what there doing as they're making great products.

I agree though I would never upsample audio just for the sake of it.

[Mark Bassett]

You're still talking technical......

Alert readers will recall that last month I began a discussion of how we try to determine what it is that us humans can really hear. I noted that people often report hearing “awesomeâ€

[Howard Jones]

Mark - a great post and thanks for putting it up.

I'm all for promoting scientific method but it's not apparent to me that that's what the audio industry wants. In fact, I'd say that interest in quantifiable and verifiable techniques in audio peaked somewhere in the 80s and has been heading downhill since then. I can't remember the last time I saw a pro audio magazine equipment review that carried any lab testing whatsoever. Certainly not in the last decade and certainly not in Australian publications, although I think that some of the consumer audio mags still manage this.

Laboratory testing of audio hardare & software isn't the be-all and end-all, but it makes an excellent starting point and serves as a counterweight against the in-built brand prejudices you talk about. Mind you, even if you took a colleague through a double-blind test in the course of which she definitively decided that the Shure sounded better than the Neumann, would she be willing to follow through on this when purchasing? Would she even admit that she thought it to others?

I also have a slightly different take on the bit depth/sample rate argument. I figure that 24 bit and 96kHz is as far as the industry will ever push, at least in terms of practical, everyday products. There will be a slow drift towards either or both 24 bit/96k as new products come on to the market and older 16 bit/44.1k machines die or become otherwise unrepairable. Therefore, there will come a day when 24/96k is the norm and we will have wasted a great deal of breath over 16/44.1k. We'll likely end up not having the choice to worry about, so why expend too much energy on it now?

If you've got 16/44.1k and are happy with it/can't afford 24/96k, I'd forget about it and get on with making music. You can shoot the breeze all day long about this stuff but you'd be better off making music with what you've got. You can advance yourself by being a better player or producer, but not by owning a fancier bit of gear. Good way to drive yourself nuts.

[heathen]

I can remember a very valid point by Bob Katz on this subject and from memory said he believed the reason 96k sounds/perceived better is due to the shape of the anti-alias filters in converters. With such a steep curve between 20k and 20.5k harmonic distortion (i think) artifacts can be introduced and become more obvious. 96k has a milder filter curve up to the nyquist and will most likely capture a more natural signal.

With bit depth I can almost certainly pick a 24 bit track to a 16 bit track compared side by side. 24 bit just sounds more realistic and natural to me every time. I remember Rick saying when he gets an extra 3% quality there is no going back, dithering down to 16 bit feels like losing about 1/3, well it is really.

96k 24 bit would be nice but in the age of mp3 44.1 16bit is still ok with me.

Build quality also, I guess $10 000 prisms will sound better at a lower sample rate than say #$&ringer at 192k.

[David W]

[quote="Howard Jones"]

I'm all for promoting scientific method but it's not apparent to me that that's what the audio industry wants. In fact, I'd say that interest in quantifiable and verifiable techniques in audio peaked somewhere in the 80s and has been heading downhill since then. [/quote]

I think this is spot on.

There is definitly a cool factor in using gear that my not be considdered industry standard or technically superior.

This is clearly demonstrated on countless occasions when "XYZ" producer claims they used some budget microphone on the latest chart topping hit. It tends to them be the next hot item in Pro Audio Forums for the next month.

I guess in Music Industry is all about cool but can't we leave this to the rockstars and as Audio Engineers just stick to recording as best we can.

As to why we argue differences in say advantages in audio resolution with no clear answer, yet many default to the arguably lowest quality option can only be explained as either an emotional decision (similar to what Mark Andrew Bassett has described in above post) or lack of resources.

[tonymite]

check the dan lavry whitepaper - 60kHz seems optimum anything higher is overkill ... and it probably wont mean jack with the new generation 1-bit converters - correct me if im wrong ....

[Howard Jones]

For years and years Rupert Neve has maintained that analogue mic front ends need to have usable response out to around 200kHz.

If anyone's interested in what Rupert thinks on the matter.

[rick]

i am all for wide bandwidths , except i get very confused when i ponder on the technical realities of what we do
if no microphones anybody uses on any recording responds to anything above 35k ( most dont go above 20 k )
just what is it that we use the ultra bands for ..? what musical noise is up there ..?

there are a bunch of audiophile dweebs that are currently using bandwidth limited cables ( limited at 100k) and the claims are they sound more neutral then the normal full bandwidth cables

before you right this off as crackpot magical nonsense
factor in that gateway mastering in the usa recently rewired completely with this stuff
so the point ...?
well the science and the reality of sounds dont agree ...what else is new ?

[Howard Jones]

There is no reality of sound. Not anymore... and maybe there never was. See previous post.

[tonymite]

[quote="Howard Jones"]For years and years Rupert Neve has maintained that analogue mic front ends need to have usable response out to around 200kHz.

If anyone's interested in what Rupert thinks on the matter.[/quote]


that is all about phase linearity ... if your equipment is phase linear octaves above and below the audiable spectrum ... u know the rest ...

[mal stanley]

The 'upsampling' I referred to earlier is usually used to move any artifacts (aliasing etc ) from further digital processing out of the lower audible range....not so much for wider audio bandwith...
and to take advanatge of being able to implement digital filters more gently...

[Mark Bassett]

I've been ponderng the idea that the numbers are just that. Sure you can do the math and say that theoretically we can use smoother anit-aliasing filters in our A/D conversion if we use higher sample rates but the advantage of this ends way below 96kHz......

Is there any tangible link between the sample rate numbers and what you hear?

How about bit depth?

If you can hear the difference between 16-bit and 24-bit take a second to comprehend the idea that you're comparing two electrical/chemical-mechanical-acoustical-mechanical-electrical-mechanical-acoustical-mechanical-electrical/chemical-perceptual transduction processes in terms of symbols and then stand firm and tell me that 96 is better than 44.1 even though your guitar sound is thinner than paris hilton and you're record has (to quote a board member) the dynamic range of a puddle cause you you over did it with a L6 uber-ultra-supermaximizer that 'sounds the same as the hardware version these days at 96kHz'.

So why not use some other arbitrary symbol (instead of a number) to represent the sample rate, like a geometric shape as rick has once suggested. Express all sample rates and bit-depths as simple geometric shapes (that are still as mathmatically related as 48/96) and then tell me that the square sounds better than the rhombus.

The amount of records that I hear, no scrap that, sounds, that I hear that others subjectively judge as sounding good that I perceive as rubbish/unsuitable is a constant eye opener - i'm not right - they're not right - that's why it's subjective - but it certainly sheds light on the whole 'I can hear the difference between a tomato and an orange' rant, throw in the fact that no 2 humans hear the same, and we're off to a great start.

Now, I'm all for the math, without it we would not even be at the stage of digital, let alone analogue recording.

But talk math if you talk math, make records if you make records, and write songs if you write songs.

Tautology (rhetoric), use of redundant language that adds no information.

[heathen]

[quote="Mark Andrew Bassett"]I've been ponderng the idea that the numbers are just that. Sure you can do the math and say that theoretically we can use smoother anit-aliasing filters in our A/D conversion if we use higher sample rates but the advantage of this ends way below 96kHz......

Is there any tangible link between the sample rate numbers and what you hear?

How about bit depth?

If you can hear the difference between 16-bit and 24-bit take a second to comprehend the idea that you're comparing two electrical/chemical-mechanical-acoustical-mechanical-electrical-mechanical-acoustical-mechanical-electrical/chemical-perceptual transduction processes in terms of symbols and then stand firm and tell me that 96 is better than 44.1 even though your guitar sound is thinner than paris hilton and you're record has (to quote a board member) the dynamic range of a puddle cause you you over did it with a L6 uber-ultra-supermaximizer that 'sounds the same as the hardware version these days at 96kHz'.

So why not use some other arbitrary symbol (instead of a number) to represent the sample rate, like a geometric shape as rick has once suggested. Express all sample rates and bit-depths as simple geometric shapes (that are still as mathmatically related as 48/96) and then tell me that the square sounds better than the rhombus.

The amount of records that I hear, no scrap that, sounds, that I hear that others subjectively judge as sounding good that I perceive as rubbish/unsuitable is a constant eye opener - i'm not right - they're not right - that's why it's subjective - but it certainly sheds light on the whole 'I can hear the difference between a tomato and an orange' rant, throw in the fact that no 2 humans hear the same, and we're off to a great start.

Now, I'm all for the math, without it we would not even be at the stage of digital, let alone analogue recording.

But talk math if you talk math, make records if you make records, and write songs if you write songs.

[b][i]Tautology[/i][/b] (rhetoric), use of redundant language that adds no information.[/quote]

I think most people here have pondered the thought of electro/mechanical/acoustic/neurological transduction processes.

To me a bit depth reduction is more noticeable than sample rate increases. This I did not judge by reading facts or figures, I used my ears to determine the results, disclaimer:"that was a subjective and perceptual deduction".

The designers of these pieces of gear are still writing white papers on this stuff usually not even agreeing amongst themselves, so who is right. We should probably just ask our dogs what sounds better.

"L6 uber-ultra-supermaximizer", I thought this tool was designed for some mastering engineers to creatively destroy a reasonable mix to unintelligable garble, actually that was before clipping became fashionable.

We should all end the rhetoric now and vote Labour number 1, I agree.

[jkhuri44]

i've been reading this thread....for a while...after my original post...andbeen learning quite a bit, which has been great..

but one question, and I do not mean to backtrack any arguments being made here..

but i have a few comments to make, regarding sampling rates and what not...

1) sample rate increases sound better, during the course of a project, with the culmination of more and more tracks...

i could be wrong in saying this, but this has been something that i have observed over the course of working on my current project...

and in reference to these blind tests that are talked about, i would be interested to know what is being blind tested.....because there are so many variables in this game that its not funny...


i would probably say that with good convertors, it is tricky to tell the difference between 1 instrument with 1 mic recorded in 44.1 or 96khz....

BUT, surely the differences would present themselves listening to a whole mix done at 44.1 vs 96K ??

these are all observations i have made, that arent particularly solid, but nevertheless, do some people here agree??

[rick]

what about this to ponder on
my feeling ( based on spending more money on converters in the last twenty years then i have cars and computers combined)
is simple - good sounding converters have good sounding analog audio paths , the right use of power supplies, amplifier circuits and in general good ...no make that excellent sonic electrical audio analog house keeping.
this is hard to do , a bit more expensive and really not actually currently a "known" issue with converters
and the bit rates and such we all talk about can only be experienced by hooking up analog converters to hear what all those numbers are really doing.

we all seem to fall over ourselves to get the cleanest "class a high voltage all discrete mic pre " but as soon as we start talking digital then its all subminature surface mounted low voltage switchmode driven works of computerised heaven that nobody questions or even wants to question , but jeez those marketing numbers get a real going over. ( thus this thread goes on and on )

hmmm...not sure exactly what a good line level amplifier should sound like ....?
well the difference is ACTUALLY a sonic issue for me
sample rates and bit depths are simply not something for me to worry about
i dont understand the maths but i do understand the sounds

not sure about your converters analog audio paths ?
well its not the only test to do but a good place to start is
is it heavy and does it get hot ?
if it is there is a fair chance somebody in the design stage understands what excellent sounding audio has always needed
".....over engineering ! "

i recorded a project on a digi 01 several years ago and i wired some neve line amps ( on a 30cm patch lead) from the analog outs that digidesign kindly put on the output of the box
for the first time ever that box sounded great , it could drive a 5 metre cable without sounding dull :)

so maybe get some quality line amps cheap on ebay to stick on your current converters to do all the hard analog work
then lets all talk about bit rates and computers dsp crunching

it will be about $12k cheaper then a new set of "special 400 bit" converters or that new ua card and i am sure it will quiet down even the most worried theoretical bit number crunching audio engineer.

[rick]

oh yes
i think blind ab testing is for nerds
i use the am i still using the box after i have finished paying for it test
its much more reliable

[heathen]

Absolutely I agree 100% about the analog audio path in converters being one of the main culprits, the crappier the analog in or out and yep for sure the harder the levels are pushed the worse things will sound, well things just wont sound too good no matter what.

Quite a few people in the USA have been getting low end converters analog circuits modded with reportedly better results afterward, as a lot of the actual digital components are the same as some of the higher end ones anyway. Though as you said clean power supply and other internals are also a big part, so improvements would most likely be marginal. I was going to follow this route but just thought screw it and went for reasonably good converters and sold the others.

Also agree 100% about the test of time, some units are warm every day and others sort of just get forgotten.

[rob]

interesting post fellas

the psycoacoustics rate as probably the single biggest factor in our perception of what things sound like. Often when i complete the latest "black box" and phone the client up to say it's ready, i invariably get asked "so, what's it sound like?". I really try to avoid answering this question with subjective descriptions, i might say " well the noise floor is nice and low at -100dBu and the frequency response is good out to 200kHz" but to say it sounds " fat " or "warm" or "airy" or whatever is going to predudice the clients perception of what they think it sounds like. " you tell me what you reckon it sounds like!" is far more productive.
So, Mark's point about numbers affecting our perception is spot on.

As to "BUT, surely the differences would present themselves listening to a whole mix done at 44.1 vs 96K ??" Maybe, maybe not. Who here has done an entire project from start to finish simulataneously tracking and mixing at 44.1 and 96K with every other variable identical, and to top it off listen back to the result without knowing which one is the 44.1 and which is the 96K?

Rick's point about the analog path in converters is also of huge importance and as he indicates usually not mentioned by manufacturers, they'd prefer to quote big ( or little ) numbers. We all get a hard on when Neve is mentioned because it is supposedly the rolls royce of audio and we all want our audio to pass on it's merry way through a Neve. But what happens when the audio hits the 50 cent IC through 5 cent capacitors in our converters? Does it now matter that it is sampled at 44.1 or 96K? Interestingly whilst converter manufacturers have a huge range of analog devices and whole legacy of analog circuit topology to draw upon, there are only a handful of converter IC's on the market. Lots of different converters covering a wide price range can have the same converter IC inside them, what differs is the analog path, the power supply and the case, front panel and eye candy.

any one come across a good song lately?

Rob

[graemeh]

Great post fellas.

And though I can't contribute anything meaningful I really appreciate your various points of view.

and now my 2 cents worth.

I dunno that I could hear the difference between 44.1k and 96k in blind listening tests, but I know that I could tell the difference between Sting and Pavarotti over the phone...

[heathen]

[quote="rob"]
any one come across a good song lately?

Rob[/quote]

Silver, from the new Ozzy album.

[rick]

"I dunno that I could hear the difference between 44.1k and 96k in blind listening tests, but I know that I could tell the difference between Sting and Pavarotti over the phone..."


you mean sting still rings you ? :)

[Damien]

we have 2 new converters. they both get really hot, now i know they must sound good. the hottest one may be the best. ill measure the temp tomorrow amd it will be my new bestest one. now which of my pres cooks the most........

seriously the new converters really cook. and i like em.

[graemeh]

I dunno that I could hear the difference between 44.1k and 96k in blind listening tests, but I know that I could tell the difference between Sting and Pavarotti over the phone..."


you mean sting still rings you ? :)

never rings never calls round anymore - musta been something I said :)