Why Bother with Higher Bit Rates?

[HCMarkus]

In a recent thread, the subject of bitrates in recording arose. As someone who is happy to work at 44.1 or 48k, I am curious of others' experiences and opinions in this regard. A few arguments against pushing bitrates beyond 48k:

•••Many plug ins don't work at high sample rates. Some do, but provide reduced performance/signal to noise ratio when operated beyond 48k.

•••Many DA/AD converters also suffer from reduced signal to noise ratios when operated at increased sample rates.

•••Most microphones' response falls off rapidly above 16kHz, which is well within the frequency repsonse provided at 44.1k.

•••Few speakers reproduce source material that exceeds 20kHz.

•••Only exceptionaly gifted listeners can hear sound beyond 15kHz - 20kHz.

•••With a doubled sample rate, the computer works twice as hard for nominal gain in sound quality, at least as compared to the big boost in headroom and corresponding freedom from the tyranny of 16 bit maximum word length we got when we moved to 24 bits.

•••Files take up twice as much space on our hard drives. Admittedly, this point is rendered almost moot by the proliferation of big, cheap hard drives we now enjoy.

Personally, I am unconvinced going to 88.2 or 96k sampling rates is worth the trouble, especially if the end product is to be downsampled to 44.1 or 48k, or distributed as MP3. I do believe if a project is destined for one of the high bitrate audiophile formats, it should be tracked at a higher rate as a matter of professional intergrity. But I wonder how many "96k" recordings are merely upsampled at mastering to qualify?

On a related note, I question why product that is to be delivered at 44.1k is recorded at 48k or a multiple of this frequency. As I understand it, the reasons the "consumer" 44.1k standard and "pro" 48k standards were adopted is becuse they are numbers that are difficult to convert... not easily divisable. The rational for the "Pro" standard was not sound quality, it was piracy control. Recording at the same, or double (88.2k) sample rate in projects destined for CD release makes a lot of sense to me. DVDs use, I believe, the 48k rate, so music for picture should be cut at 48 or 96k to allow an easy, distortion-free division to 48k.

I post this message in the hopes that we can explore whether the high sample rate bandwagon is driven by what we hear, or by what marketing departments want us to buy.

Finally, while we're on the subject, do all the MOTU-provided DP plugs work beyond 48k? If so, how well do they function?

[Frodo]

HC--

What caught my eye about your topic was that it mentioned "bits". I thought it was bit depth, ie: 16b/24b.

Sample rates!!

I know that MachFive is 192k capable. Specs for the other MAS instruments are not readily listed-- however, it doesn't make sense to include VIs and plugins that would not be compatible with a company's own best interfaces. It's a very good question, so I won't assume so easliy that the interface determines the rate until I hear otherwise.

As for sample rates in general, engineers I've spoken to always point to the rate divisibilty issue. Most recently, one engineer I know has always believed from day 1 that once sample rates got near 100k that most of his concerns over analog vs digital would be much less of an issue-- and that for his taste he'd still perfer to sum via high end analog and re-encode at whatever lower rate rather than to deal with whatever damage dithering can do.

This sort of chimes in with something Al Schimidt discussed in an article on Mixoline.com. He spoke of recording for film at higher rates in the digital realm, then sending the stems out to the encoders and summing AD at that time; again claiming that to his ears it produced a better result. I can't speak to what he's done most recently, but he brings an old school wisdom to a new school process. Of course, to someone who's only worked with the digital realm, some of the discipline of analog signal processing may sound foreign or seem pointless. I like Al's thinking-- and his work.

Of course, the whole analog argument bypasses the rate division issue, but since these days 192k is more easily had, it just may be time to reconsider the benefits of digital summing and dithering.

You've been so helpful with my studio quandaries that I hope I can contribute something useful here. This post is not at all complete-- and I hope to return with worthy articles on the pros and cons for you.

[Frodo]

HC:

You might find this PDF interesting. It doesn't always speak in favor of 192k as a sensible rate, but it is interesting notwithstanding.

http://www.lavryengineering.com/documen ... Theory.pdf

[HCMarkus]

Interesting article Frodo. Not having a lot of time this AM, I scanned it, noting:

The determination of sample rate must be decided by bandwidth of the ear.

and

Sampling audio signals at 192KHz is about 3 times faster than the optimal rate. It compromises the accuracy which ends up as audio distortions.

While there is no up side to operation at excessive speeds, there are further disadvantages:

1. The increased speed causes larger amount of data (impacting data storage and data transmission speed requirements).

2. Operating at 192KHz causes a very significant increase in the required processing power, resulting in very costly gear and/or further compromise in audio quality.

The optimal sample rate should be largely based on the required signal bandwidth. Audio industry salesman have been promoting faster than optimal rates. The promotion of such ideas is based on the fallacy that faster rates yield more accuracy and/or more detail. Weather
motivated by profit or ignorance, the promoters, leading the industry in the wrong direction, are stating the opposite of what is true.

I'll be looking closer at the data generated. Thanks for sharing it Frodo.

[Frodo]

Yeah, some of it defies common sense, altough I need to explore the numbers further.

Complaints about the ear's limitations concern me a bit. Lots can happen with LFO sub-audio which are "felt" more than heard. I do acknowledge that certain undesirable anomalies must be avoided, but I'm still weighing in on the absolutes of this report-- to opt for a wholesale elimination of the extremes without the facility to understand exactly what needs to be elimiated. To only work within a certain frequency range is to ignore both the good and bad sonic activity at the extremes.

Now, the statements regarding the 192k genie being released from the "bottle" and then squeezed back into a 44.1 or mp3 bottle all within the digital realm is something to consider seriously. In this case, analog summing starts to make more sense.

Write speeds, files sizes-- these all make sense to me.

The other argument of what additional details are present and noticeable at 192k which are not there at 96k is something else to explore. What, indeed, is "overkill" in this context? When do we reach the point of diminishing returns?

[sdemott]

If I remember correctly, Dan went on to say that the optimal sampling rate was 60k (translating to a top reproducible freq. of 30kHz). Now if we could just specify CD v2.0 be 24bit/60kHz instead of 16/44.1, we'd have something.

Until then I keep all my CD projects at 44.1kHz and all my broadcast projects at 48kHz.

[stephentayler]

From what I understand, the issue with extended high end is how the frequencies we can't hear intermodulate with the frequencies we can hear. Digital always has a finite ceiling, and so prevents that happening to some extent.

If the final delivery is going to be a digital format, I guess it is always going to defeat that, but it has always been an issue for makers of high-end analog equipment.

Stephen

[PedroCruz]

The higher sample rates cover in fact a higher frequency spetrum, but its not this frequencies that we should have in mind

In 44.1KHz your higher frequency is 22.05KHz right?
(why is the higher frequency always half of the sample rate? because you need at least two values to build a wave, one positive and one negative, this way you will end with a triangle wave)
If you take a better look, a 22.05KHz wave will be represented by only "2 dots" so you will get a triangle wave and this also happens in lower than 22.05KHz,
Upgrading to higher sample rate this triangle wave will have higher definition, more than "2 dots" to represent the 22.05KHz.

Cheers
PedroCruz

[Frodo]

Obrigado, Pedro!

I wonder, however, if the article is not suggesting that this definition in the higher end at 192k is a waste of time? Your explanation makes more sense to me in a way.

[sdemott]

Exactly, but, as Dan Lavry contests in his white paper, there is no benefit from sampling rates above 60kHz - 88.2/96/192 are all marketing hype.

Also, if you read anything Bob Katz has written on the subject you will understand that the original reasoning that higher sampling rates were better was because you brought the Nyquist filter (the steep roll off at the point of 1/2 the sampling rate) out of the range of our hearing. In early digital the filter was very harsh. In todays digital world there have been so many improvements that it has become a moot point (in his words).

[HCMarkus]

I admit it is counter-intuitive that one can create a perfectly accurate digital reproduction of a band-limited wave at a sample rate just higher than 2x the highest frequncy to be reproduced, but the science is real. Remember our hearing is band-limited, too!

Yes, the filter roll off vs phase issue was huge at the dawn of the age of digital audio. The typlical complaint was brittle high end, and the problem was severe enough to drive a lot of folks back to vinyl. As SD notes, these issues have been rendered moot by oversampling. As Mr. Lavry's paper points out, the audio sample rate to which we typically refer is the rate at which sample are stored, not the rate at which AD converters sample an analog source. Modern converters are actually oversampling at huge bitrates in the conversion process, allowing steep yet phase accurate filtering to take place.

Lavry also notes that inaccuracies creep into conversion as rates go up... another factor to add to the list of reasons to keep things simple.

Thanks again for sharign this resource Frodo. This thread ties in with what I have been reading here: http://www.unicornation.com/phpBB2/view ... hp?t=15613

[Frodo]

Thanks again for sharign this resource Frodo. This thread ties in with what I have been reading here: http://www.unicornation.com/phpBB2/view ... hp?t=15613

Sure thing, HC. Glad the doc came in handy for you. Yes, I've been following that thread, too. To me, it's helpful to revisit the old arguments in an effort to examine how they are impacted by today's improvements and developments. As SD mentioned, so many issues have been rendered moot as a result that it's really worth denoting how the point of diminishing returns has moved.

As you know, I'm beginning a re-do of my system and have discovered why certain things are necessary and others are a waste of time and money. Comparing these assessments with similar ones over the years, I find the changes significant and invaluable.

Thanks, guys!

[David Polich]

I think higher-end converters and clocks have more to due with sound quality than higher sample rates.

With high-end converters and clocks, and of course good engineering skills, I see no reason to record at anything other than 24 bit/44.1khz. We've been through this discussion quite a few times, I know there are many who would "beg to disagree" and that they can hear the difference between 48khz and 44.1khz. I never have. I can tell a little bit of difference at 96khz, but it is simply not enough for me to warrant the extra hard drive space required to do a project at that rate.

There is one area of concern for projects at 48khz - I believe that DVD audio has to be at 48khz, correct me if I'm wrong on that. That's a "standards" argument, not a valid audio/musical argument.

The only plausible reason I've ever encountered for doing something at the higher sample rates is - sampling itself. If you're developing a sample library for consumer release or for some hardware product, I can see where higher sample rates are beneficial.

[Jim]

The only plausible reason I've ever encountered for doing something at the higher sample rates is - sampling itself. If you're developing a sample library for consumer release or for some hardware product, I can see where higher sample rates are beneficial.

What about future-proofing your projects? Aren't the super-audio CDs using (or going to use) higher bit depths and sample rates?

[HCMarkus]

24 bit word length gets us beyond the highest possible SN ratio possible in the analog world, so there is no advantage to going beyond 24 bits. And while some may claim better sound quality at higher sample rates, moving to higher rates may actually degrade the sound quality of recorded material as has been noted in this thread.

In time, technologies should improve, and there may be an audibly superior recording technology devloped. Until then, it is a simple process to upsample 44.1k or 48k projects to 88.2k, 96k and beyond. Anyway, everybody is buying 44.1k CDs or MP3s these days. The audiophile group is a nich market.

If a project is destined for sale in a high-end stereo store, it should probably be recorded at higher bit rates, whether or not the end result is actually better, for marketing purposes. Better yet, invent some arcance terminology to describe your recording process and sell the music on that basis.

I think I'll be sticking to 44.1k and 48k samples rates for the forseeable future, unless a client wants to pay extra for what he feels is an essential service, of course.