Covered lots of times. Try searching for threads.

Well I'm still stuck with the good old Digi001, so 48k is my maximum.... but two friends who run big Digi HD3 systems still only record at 24/48 for bands, but one had a string quartet in last year and he used 192khz for that.

Superb sound quality doesn't come from what sampling rate/bit depth your working at--it is determined by the quality of the instrumentalist, then the instrument, then the room, then the engineer, then the mic, then the preamp, then the conversion quality (not determined by what sampling rate your at). 44.1K/24-bit is as high as I'll go if I know the final product will be a CD- Of course that changes if your doing A/V work, or DVD-A work. Dithering is at a place where working at 24 then going to 16 is not a bad thing---but SRC is not (at least to me) beneficial to the final project--if anything going from 48k to 44.1 hurts a project more than recording at 44.1 and staying there. Of course there are millions of opinions on this topc.

Here is the somewhat 'end-all' thread on the subject.

http://recforums.prosoundweb.com/index.php/t/288/0

Unfortunately I dont think the link really answered the question. I have been wondering this myself for quite sometime, but I havent found anything that definitively gives me any answer. 96khz takes up more space than 48, and 48 takes up more space than 44.1. Now I dont know about the difference in processing, but I think it has been settled that 24bit dithered to 16 sounds better because of a number of reasons... however, does 24/96 dithered down, sound better than 24/48 dithered down, or 24/44.1? Is there any benefit to using 96 instead of 48, or using 48 instead of 44.1?

Its not about the signal chain... its a specific question designed to clear up a particular dilemma. Hopefully this clears things up a little and maybe we can draw some conclusions.

Sweet

You know - Dan makes a valid point (or alludes to it at least) that there's the science involved and then there's the pure live testing (listening).

Many people claim that there is no information to be had above 20kHz - many people claim otherwise.

The interesting fact is, in blind tests, people have been able to identify the same program material recorded at different sample rates and identify whether they were 44.1 or high sample rates. Obviously, something is going on and whether it's the distortions that Dan references or some actual program material, it is discernable.

Of course, the converse is also true - some people actually preferred the sound of 44.1 in numerous of these listening tests - so - what to do???

Personally, the majority of the projects I record and mix are in 24bit / 44.1 kHz. If I do a major orchestral project, I will bring up the sampling rate to 88.2 or 96k - sometimes 192 or DSD.

Ultimately, is there a difference? Sure, it takes up more of my harddrive space and I can hear a subtle difference. But, I can't say that it's such an improvement that it makes me die to use higher sampling rates. Now, I'm a firm believer in DSD however. I know a lot of people who crap all over DSD because of its theoretical limitations, but frankly, it just sounds better and there are plenty of scientific reasons why. (What I find to be funny is that the opponents of DSD are often the proponents of 24/44.1 and their argument for 44.1 is that there is nothing to be had outside of 20k. Then they turn around in the very next breath and say that one of the serious drawbacks to DSD is the excess of "out-of-band" noise centered above 60 kHz. Wow, now that's a contradiction in philosophy!)

J.

It's weird, I've heard "definite" differences when notching out 22kHz by only a half dB on a final mix. I didn't think I could hear 22kHz but was proven wrong.

Ya gotta hear it to believe it.

24/96 on good converters sounds great, but I end up going 24/44.1 about 95% of the time. It just saves so much disk space and processing power. I also feel like some of the mediocre converters do not necessarily sound better at 96k than they do at 44.1.

If I'm doing only a few tracks and it's classical or acoustic music and I have some decent converter's then I go 96k...but most of the time on my digi002r I will just go 44.1.

Cucco wrote: (What I find to be funny is that the opponents of DSD are often the proponents of 24/44.1 and their argument for 44.1 is that there is nothing to be had outside of 20k. Then they turn around in the very next breath and say that one of the serious drawbacks to DSD is the excess of "out-of-band" noise centered above 60 kHz. Wow, now that's a contradiction in philosophy!)

J.

J,

We meet again. I hope you're well. To clarify the perspective, the issue with the out of band noise in DSD is three fold:

1. many proponents of DSD tout its high frequency response. I don't bring up the out of band response of DSD in order to say that I believe we can hear it, but rather to discount that posit on their behalf that that is why it is better.

2. the out of band noise at high levels contributes to distortion in amps and speakers that are ill-designed to handle that kind of signal. Tweeters become particularly non-linear in that range, and the non-linear behavior causes distortion in our audible frequency range.

3. in order to prevent this excess out of band noise Sony recommends putting a low pass filter on. This filter, however, does not get the benefits of a digital, FIR, linear-phase filter as it needs to be in the analog world. Instead, this filter causes phase shift of audible frequencies as it inherently must.

Yes, the out of band noise is an issue and continues to be a source of concern regarding DSD, but not because we speculate it can be directly heard.

Nika

Wow! I thought this was a simple question. Turns out it's a topic of much discussion and many opinions.

I guess I'll do some recordings at 24/96, 24/44.1, and 16/44.1 and then burn some CD's and let my ears decide.

If this thread isn't dead yet I'll report my findings here, at least as far as my ears and equipment can tell anyway.

Thanks guys
Keith

took-the-red-pill wrote: Wow! I thought this was a simple question. Turns out it's a topic of much discussion and many opinions.

Keith

Simple question? I've seen flame wars, fist fights, couple of stabbings, and at least one shooting over this very topic.... :wink:

Seriously though, the general consensus that I've been able to gather is that 5 years ago most people HAD to have 96K or 192K capabilities...and those exact same people are now reporting they stick to 24bit/44.1K as the cost/benefit ratio is drive space is just too high...even with the higher rates, there are those who argue that the better conversion systems running at 44.1 exponentially sound better than lesser conversion at 96K....

I tend to use 96K ONLY when recording string ensembles or acoustic jazz...for the typical rock/country production, its 24/44.1 all the way....for radio ads, PSA voiceovers, etc, I don't even think twice - its the lowest I can get away with and still provide a good sounding mix....

Really long story short: you are correct. Let your own ears decide...IF you think the higher rates are worth the investment in drive space/lowered processing abilities, then go for it.

Yeah, in truth, I would never refer to this as a simple question.

If you plan on doing comparisons, make sure that you are taking the signal from the exact same source. I have a setup which has allowed me to do just that.

The RAMSA WZ-AD96M uses ADAT Lightpipe and AES outputs. With 2 of those hooked up to my True Systems Precision 8 (which has 2 identical and active sets of balanced output) has allowed me to perform these comparisons with the identical signal present at both ADs and thus at mix down. In my experience, I've heard some differences, but bear in mind, all of the differences have been subtle and noticeable only on my high-end system or monitoring set-up, not with standard playback equipment.

Nika-
How do you do? All settled in to the new life yet?

For anyone who's interested, Nika and I have pretty vastly differing opinions on some of the finer points regarding high sample rates and specifically DSD. The difference is, he's done a bit more scientific research into it than I have. It's an interesting read to say the least if you can find the threads.

You make some good points here, but I know you are aware that, with the proper playback system, many of these limitations cease to exist. True, many tweeters do become quite unpredictable at these higher frequencies - a problem solved one of two ways: a simple analog filter analogous to a crossover with no hand-off to a higher frequency band or specific materials used for their more linear production of UHFs. Of course, the Bose paper cone tweeters aren't good at this, but several of the aluminum, titanium, ceramic/metal hybrids do a fantastic job of resisting distortion and break up well into the octaves above theoretical audible limit.

Many amplifiers are now being designed and built to handle this kind of bandwidth as well. Of course, all of this suggests that one has taken the time, effort, and money to piece together a system capable of these feats.

Of course, much of the debate re: DSD (quasi-PSM) vs PCM are akin to those debates of higher vs lower sample rates. It really boils down to the single bit modulation and the theoretical advantages and problems inherit in it. Experts have come out on both sides of the fence on this one with facts to prove that DSD both sucks and reigns supreme. Personally, due to the conflict and my lack of all the information, I rely entirely on my ears. Thus, I enjoy DSD and SACDs.

So, I guess the moral of the story is - get all the facts you can (which should be easy, there are several good books on the matter) and avoid all the hype/opinion on the matter. Then, you be the judge - you should be the toughest person to convince. And ultimately, if you ever lose money on a gig b/c a client demands 24/192 - then RUN, don't walk to your nearest store - drop $3000 on the newest converter and regain your lost business.

J.

Midlandmorgan wrote: Simple question? I've seen flame wars, fist fights, couple of stabbings, and at least one shooting over this very topic....

There's proof you're from Texas.... we've had that problem in Oklahoma too...

Anyway, I've probably already made it clear that I don't record digitally, but that's besides the points I am about to attempt to make based on my knowledge. First of all, it is a standard procedure to broadcast time signals via radio at 60 kHz. These signals are typically used to update the date and time automatically on many electronic devices (cell phones, for example, usually have a "feature" to automatically update the time itself). This can be a possible explaination the out-of-band noise mentioned in an earlier post. A time code is synchronized with the 60 kHz carrier and is broadcast continuously at a rate of 1 bit per second using pulse width modulation. The carrier power is reduced and restored to produce the time code bits. The carrier power is reduced 10 dB at the start of each second, so that the leading edge of every negative going pulse is on time. Full power is restored 0.2 s later for a binary “0”, 0.5 s later for a binary “1”, or 0.8 s later to convey a position marker. The binary coded decimal (BCD) format is used so that binary digits are combined to represent decimal numbers. The time code contains the year, day of year, hour, minute, second, and flags that indicate the status of Daylight Saving Time, leap years, and leap seconds.

Also, the changes that can be "heard" by adjusting, say, 22 kHz aren't necessarily heard, but rather perceived much like subharmonic frequencies.

These are just things that I learned about when I should've been trying to pass high school. It's just a hunch, though, considering that I nearly dropped out of high school so I can't possibly know what I am talking about. (Sorry I'm venting there, I just had a rough night at work about how my supervisor is convinced that hydraulic equipment doesn't need a fluid and filter change after over 5 years....he mentioned that I couldn't know what I was talking about and how he had a degree....) Anyway, the following link should help diffuse a little information about interference. http://

Notice how much falls between 9 kHz and ~192 kHz.

Anyway....

Cool, 8)

Seems to me the best way to test this specific sample rate issue is to record the same signal in 24/96(or similar) and 16/44.1. Cut them to a regular CD-A and then A/B the results on a system like what Joe Sixpack might have in his house, or his car.

Isn't that 'real life' situation the one that matters most? I hate to be a poop, but who cares if we can hear the difference through our squeaky clean, high end signal chains, since ultimately that isn't a 'real life' situation.

It only really matters if Joe's less than perfect D/A converters, average monitors, running a 16/44.1 CD-A can spit out the difference in such a way that it's discernable to the naked ear. All else appears, to me at least, to be pretty much academic.

Anyway, many thanks for the replies. I'll add that 24/44.1 to the list of things to test out, as some of you have suggested.

By the way, as downloading explodes, and music cheapens, I think the importance of purity of signal is going to get pushed further and further down the list. This whole discussion starts to look a lot more like intellectual masturbation in that light.

...but that's a topic for another day...

Keith

If you plan to be using a lot of plugins, then definitely stay at 48Khz or you'll run out of CPU/RAM quickly.

If you're recording more than 8 tracks at once its probably best to record at 48Khz (unless you have an FW800 bus supporting audio interface like RME's Fireface800 which can handle the bandwidth).

But going from 48Khz to 96Khz should increase high end clarity and tighten bass somewhat, so whenever possible use it.

Disk space really shouldn't be an issue, if it is get some external firewire hard disks. Lacie/Porsche makes some great ones, but make sure it doesn't have any USB ports - just 2 FW ports so you can daisy chain them. They've got a 160GB, 7200 RPM, 2 FW ports for less than $130.

When you archive sessions made at 96khz, definitely remove all the out-takes and un-used audio (SX has a nice command for this "Save Project To New Folder").

Arrowfan -

You've stated some of the untrue myths about higher sample rates --

High SRs will not improve bass frequencies. For awesome bass, you could use a sampling rate of 400 Hz and get the EXACT same bass (from 0 Hz up to 199 Hz) as you would if you used 384 kHz sampling rate. Remember, you only need 2 points on the graph to identify a frequency. No additional points will help - they will only help identify higher frequencies.

As well, it takes no significant additional processing/RAM power to process 96 vs 48 vs 44.1.

Keith -

Ultimately, yes it is advisable to test anyway that you can. However, NOT testing it on the highest end system that is available is a mistake. I assume that every disc I make will be played back through:
The finest CD transport
Awesome D/A converters
Krell or Mark Levinson Amplifiers
B&W or Revel Loudspeakers
over $10,000 cables

Why - some of my clients (and probably yours too) have this kind of set up. If you want them (the people closest to being "in the know") to think that your disc sounds like shit b/c all you did was make it sound good on your Ford's stereo system, then by all means - ignore this demographic.

Moral of the story - never skimp on the monitoring chain - EVER. :D

J.

Seems to me the best way to test this specific sample rate issue is to record the same signal in 24/96(or similar) and 16/44.1. Cut them to a regular CD-A and then A/B the results on a system like what Joe Sixpack might have in his house, or his car.

Well, that's a start, but only part of the process. You also open up the "can of worms" that includes sample rate conversion (which one is best?) and dithering down from 24 to 16 bit. (I know what "I" like, but....)

All of these steps "corrupt" the scientific method in determining which one is "best". By the time it's all boiled down to Redbook CDs, who knows what's happened to the sound, and where/how/why?

In a perfect world (not the way people listen, anyway), you could go with a DVD-A or DVD-ROM or HD wav file playback to first find out if there's an appreciable difference for you at the top of the listening chain. (let someone other than the listener keep track of which track is which.) To be scientific about it, you'd have to have recorded all different versions at the same time or under strictly identical conditions, with similar gear - 16/44, 24/44, 16/96, 24/96, etc.

Keep your results noted there, and then continue with the CD making process, noting carefully what steps were taken.

Make your CDs with the SRC and dithering of you choice, compare the same tracks again (also with the "control" person aware of which one's which) and see which one you like, what the differences are, etc. No one ever REALLY takes those steps when making divine "pronouncements" and comparisons, and IMHO, by the time anyone's done any real listening, it's way too far down the chain to be scientific enough to matter.

I like to record at the highest level practical for the project at hand, take the best care of it throughout the DSP and mixing, and then master it with all the TLC possible down to 16/44. (Always keeping the hi-res master on file for "someday"....whatever that might bring.)

Jeremy, point taken.

Actually I don't necessarily agree. I don't know anyone in my world who possesses a stereo system even remotely approaching what you are describing. I know guys with money, and I know guys who really care how music sounds, and they are still listening through the stuff you can get at Future Shop for a few grand. So to create music to the standard you describe would not be a 'real world' situation for me.

It would be kind of like the government saying "We need to assume when we build all our roads that a formula one car can go screaming down it at 200MPH and not hit a hole or bump."

That would be nice, but hardly practical, since in reality that road will never see a formula one car, so they let the odd bump and pothole go because Joe and Jane Sixpack can drive it in their Civic and it will get them where they're going.

I'm not advocating a weak signal chain on recording, mixing and mastering, merely stating that we need to consider the average stereo's capacity instead of overkilling, or adding redundancies. I'll just create recordings of the highest quality I reasonably can and if there are 3 people out there who are offended, because they have 25K in listening gear and my CD has holes in it due to my D/A converters or sample rate...well I guess I'll include that in my refund policy. Currently I don't have a 'list of clients' who have expectations, so I have more room on that one than some.

Well anyway, this is beginning to wander off topic, and I did say I wasn't trying to start a war, so I'll shut up until I've done some testing.

Cheers
Keith

Hey Keith -

I don't think you're starting a war or straying off topic - don't worry.

My logic behind aiming for the highest quality is simple -
True, you may not have clients with that esoteric stuff now, but you might and hopefully will. You don't want to alienate them because you're designing for the lo-fi or mid-fi crowd. If you provide the absolute best sound possible, everyone will be satisfied from the dude with the RadioShack boom box all the way up to $100K invested in audio equipment.

My point is, don't exclude and of the market segment.

J. :D

I don't mean to be a bother (if I am, let me know), but does anyone else see what I was getting into with my last post? About interference?

Cucco, you will be able to run about 2x as many plugins with 48khz audio than with 96khz audio. Period. No mythology there.

So if you know you will be running many plugins during mix stages you should just record at 48Khz from the start and avoid the downsample process from 96khz to 48khz.

OK, I'm back for another round

Jeremy,

I want to meet this guy with the $100K in audio gear. I know of a small country he could sponsor.:shock:

Joe H

Your methodology seems very sound. YOU'RE HIRED! E-mail Jeremy for the name of that guy with the extra 100K, he's a...friend of mine...yeah, that's it...and he will front the fees for the test :lol:

Brian

I doubt you're a bother, though it's hard for me to tell. The last thing you said that I understood was something about fist fights in Texas :?

Arrowfan

Okay, I know nothing, but I'm with you. I would think that it would take more processing power to do 96K than 44.1. I would think that the brain of the computer would have to plow through a lot more zeros and ones per second and it would eat up juice, though I wouldn't know if it sucked more from your CPU, your RAM, or both :-?

Not trying to hog the airwaves, but somebody in the history of the universe MUST have done tests like what Joe H described, and then published the results of such tests, no?

Any links out there, maybe written in such a manner that a monosyllabic soul such as myself might understand them?

Cheers
Keith

Well, I work with 96khz and 48Khz audio + plugins galore every day - and even with a 3.8Ghz CPU and 4GB RAM the 96Khz audio tracks will limit the plugin count. A workaround is to 'freeze' tracks of course.

Actually, if your destination format is CD audio (44.1Khz) then 88.2Khz is the optimal sample rate (not all audio interfaces can run at this rate). The dithering should have less artifacts, mythologically speaking, since its a straight division by 2.

OK - here is the document!

Took-the-red-pill,

Are you ready for some blue pills?!?! Its pretty in technical but I think you'll like it. Of course our aim is to convert you to an Audio-borg (half-analog/half digital, best of both worlds!)

SAMPLING THEORY FOR DIGITAL AUDIO, by Dan Lavry
here's a link...
Details and pictures of the setup.

It deals mainly with 192Khz vs 96Khz but the concepts are there.

Great debate everyone!!! :D

Congrats to all! 8-)

Now if only the trade publications would address this...

I think this whole thread can be summed up in a few bullet points:

• Use 24/96 if A) you need it, or B) You have the HD space and horsepower to deal with it. If you're just going to CD, and don't have the HD space, don't worry about it. Always do 24 bit though. Just do it.

• Mix all your projects the best you can, no matter what the bit depth or sample rate. If you're mixing it so that it sounds decent on a boom box, but don't care what it sounds like on a real hi-fi, go look for another career/hobby, like counting the requisite number of McNuggets for each box.

• Monkeys taste good with sauce.

took-the-red-pill wrote: Not trying to hog the airwaves, but somebody in the history of the universe MUST have done tests like what Joe H described, and then published the results of such tests, no?

Any links out there, maybe written in such a manner that a monosyllabic soul such as myself might understand them?

Cheers
Keith

I don't think so. I tried (albeit not very hard) to find someone around here that had two redundant identical DAWs to record a session on 44.1k on one and 96K on the other for a test like this, and I don't think it is very feasible to do. Would be cool though.

Thanks for the link, Arrowfan

I understood a bit of it. Unfortunately on page 3 my head exploded. As you can imagine, growing a new one takes a while so I probably won't get to my tests as soon as I like.

If I understood Dan Levry correctly, then it would almost seem that 44.1 is fine, since there may be information going on up past that, but for practical purposes, not much that anyone would care about.

So back to the event that sparked this lively debate, When I mixed down from the 24/96 to 16/44.1 there was definitely a difference in the signals, and the original sounded clearer. It was almost indescernable, but it was there when flipping back and forth between the two. It also was across the board, not just the highs.

So as I see it, this minute loss of clarity could only be caused by one of 4 things:

1)the switch from 96 to 44.1, though the Levry paper would indicate that this shouldn't do it.

2)Something in my program. I run Cubase SL. I've heard Steinberg's stuff panned here, though I don't have enough personal experience to say if it could be related to this issue.

3) The switch from 24 to 16 bit. This is starting to look like the most promising contender to me anyway.

4) Not enough sauce when I cook monkey...

Any input on that?

Reggie,

Yeah, I suppose people with two computers sitting around waiting to do tests on things that probably don't matter a hill of beans anyway are probably few and far between.

Sounds like Cucco may have the required gear, the knowhow, and the wherewithall, but he may have better things to do with his life.

In my next life I'm going to set up a test lab, to test real world examples and go around dispelling myths about mics, pres, DAWS, etc. The only equipment I'll allow for analysing signals will be human ears...

Cheers
Keith

Reggie wrote: ...find someone around here that had two redundant identical DAWs to record a session on 44.1k on one and 96K on the other for a test like this...

I am the recording engineer for Straight Ahead Records, a label newly formed by Bernie Grundman & Stewart Levine. We record and mix direct to 2track onto simultaneous recorders: 1/4" 30ips Analog, DAW@1644, and another identical DAW@2496. We do this so that we have native recording formats for Vinyl, CD, and DVD.

What this allows is a direct comparison between the three formats. As one might expect, every person who has heard these recordings prefers the Analog Tape, hands down. But more to the point of this thread, we can compare the 2496 against 1644. This only proved the obvious: doubling the sampling rate and increasing the word length provides a higher resolution recording that is audible and preferable. Unfortunately, the necessary and invasive dvd authoring process steals some of that fidelity, imo.

This is a dang hard topic to get my point across clearly. I have no doubt that good analog tape sounds better than 96k, which in turn (fairly obviously) sounds better than 44.1k. The thing I want someone to try is two redundant daws recording a multi-track session, one at 24/96k and one at 24/44.1k.
THEN mix them as identically as possible to separate stereo tracks.
LASTLY dither the 24/44.1 to 16/44.1, and also SRC & dither the 24/96 down to 16/44.1 for CD. This would prove whether it was worth recording at the higher rate if you only intend your final product to appear on CD. Of course this whole process would kind of be a pain.
Sounds like you may have a way to do this experiment in at least one set of circumstances?

Right, my last example bypassed your specific point of interest. I can't address the multitrack scenario, but I can say that as a general statement, 2496 original masters do make for better 1644 production masters.

But trying to determine whether 2496 recording/mixing is "all around the best way" is kind of pointless. What really matters is how well the equipment you are using handles the different formats. In other words, some equipment running at 44.1 can sound 'better' than others at 96. For example, it is not uncommon for us to receive original masters in various formats, and occasionally the 44.1 DAT version is preferred over the 96k file! Not because it is 44.1, but rather because of all the other circumstances and influences that went into making it.

Zilla, thanks for coming on board. Now we are getting somewhere.

So what I'm hearing from you, as it relates to the original question, is that from your experience, if you had a choice, you would record to a higher word length and higher sample rate EVEN IF it were going to end up as a run of the mill CD, and that the reason for doing this would be quality of the final product? (I'm assuming all other factors, like A/D/A converters and such are equal)

Am I putting words in your mouth or is that the crux what you're saying?

Keith

That is the crux.

Cucco,

Sorry, I lost track of this thread.

Conceptually of course I like the idea of limiting the presence of extraneous high frequency content (especially artificial HF content such as is produced in noise-shaped systems such as DSD) so that it doesn't drive tweeters into a non-linear state and create audible distortion. The problem is that such filters will inherently then cause audible phase distortion. An analog filter that rolls off even up around 50kHz or higher will still provide phase non-linearity in the audible range below 20kHz.

It seems that if we don't want that HF content the best way to get rid of it is to use a linear-phase filter - which can't be done in the analog world, of course.

Nika