Sounds Authoritative…But Wrong!
I see it everywhere. Marketing departments, music reviewers, writers, and audio pundits all writing things that at best stretch the truth and at worst outright falsify things. I don’t want to become the online, high-resolution audio fact checker, but it seems every time I pick up a brochure or read an online post there’s inaccurate information being presented as fact.
Here’s a quote from the front page of a brochure I picked up in Washington DC during the Capitol Audio Fest:
“24-bit, 192 kHz digital music processed during studio mastering retains 6.5 times the information that is lost when transcribed to a standard audio CD. 24-bit audio results in a finer, more precise and enhanced sound. We refer to 24-bit audio as Mastering Quality Sound (MQS).”
This is an example of “blinding people with science”. It works by taking some math or big numbers and using them in some sort of comparison in the hopes that readers will be impressed. If you follow the logic of the sentences quoted above, you might assume that having a digital container that is 6.5 times bigger than a CD would result in “finer, more precise and enhanced sound”. Not true. It’s only true if the source recording was made using 192 kHz 24-bit PCM…not just because the mastering engineer converted it to 192/24.
Here’s another example from a recent webpage that is focused on high-resolution audio. Under the “LEARN” tab, the author wrote the following:
“On a CD, each audio sample is stored using 16 digital bits, which means a CD can produce 65,536 different volume gradations. That may seem like a lot, until you calculate that the 24-bit high-resolution downloads can produce 16,777,216 volume gradations! That’s more detail, more depth, more subtlety. It’s also less noise: With CDs, the noise is –96 dB below the highest sound level. With high-resolution downloads, it’s –144 dB below.”
There are some chunks of truth in the preceding paragraph but most of it is completely wrong. CDs do have 16-bits to store the amplitude of an analog signal but this doesn’t mean that they have 65,536 different volume gradations! And moving to 24-bits and its 16 millionish discrete values aren’t volume gradations either…the increase moves the potential dynamic range for 93 dB (accounting for dither which is a must have process) to something around 132 dB. Both 16 and 24-bits deliver dynamic range…the 24-bit flavor can handle higher input and output levels not more resolution within the amplitude range.
Decibels are not linear and the increase is substantial (although not routinely heard in the real world or in recordings) but the numbers aren’t actual volume levels. And this was written by a respected audio writer…I actually know him and consider him a friend.
My last example today is a blatant sales pitch and completely misrepresents the facts. I’ve been tempted to write a comments on the blog page and have actually started a couple of times but resisted. Here’s the pitch:
“MP3 is fine for headphones but it distorts when played loud on a stereo system. That’s what compression does, it squashes the dynamic range (life) out of your music so it takes up less space.
To restore the dynamic range, you need a ‘digital hub’, a device that can ‘up-sample’ your low res music to 24 bit/192K Hi-Res so you can crank it up at home when you want to rock. The sound quality is stunning, the best fidelity in audio history. That’s why Neil Young is behind it with his own Pono Hi Res Music service. Once you hear it, you’ll understand.”
This reminds me of the confusion about compression in the “Distortion of Sound” film mentioned a couple of weeks a go in these posts (click here to read my comments).
MP3 files use of data compression (not audio compression) and do not affect the dynamic range of music…it reduces the size (and the bandwidth) of a track. MP3s can sound bad at very low bandwidths (but actually pretty good at 256 or 320 kbps)…they can be distorted. But the distortion is the same in a good set of phones as it is in loudspeakers.
And thinking that you can “restore dynamic range” through upsampling a low resolution MP3 to 192 kHz/24-bits is pure nonsense. You’ll get a perfect (and very large) version of the file with the same fidelity you had with the MP3. To state that this audio alchemy outputs the “best fidelity in audio history” is laughable.
Furthermore, Neil Young is not talking about the same thing in his Pono effort. He’s trying to capture the magic, the “soul” of analog tape recording using higher sampling rates, longer words and FLAC compression (but just got in bed with people that are ripping CDs for delivery through PonoMusic…go figure).
We all know that we shouldn’t believe everything that we read. I do my best to present accurate information with minimal spin…but we all have an agenda.
Thanks for keeping the honesty flame burning.
It just drives me crazy…how is a person supposed to know? You read these things from seemingly reliable and smart people and it turns out that they shouldn’t be writing the stuff.
Thanks for all the newsletters.
I do not have the time to delve into the technicalities and debates about sound but I do struggle to find HD recordings that I like. Up to now I have mostly been dependent on importing SACD’s disk.
Please add -some audio clips to your HD ads. I do not want to buy material I cannot per-audition.
Thx
There should be audio samples on iTrax.com for all of the tracks I offer.
“…the 24-bit flavor can handle higher input and output levels not more resolution within the amplitude range.” I’ve read a number of statements like this in your blog,, and they are wrong, or at least represent a limited viewpoint. Consider a sound source with a maximum amplitude of 120 dB SPL — it doesn’t matter whether it’s an orchestra, rock band, or anything else. You can capture that sound source with either a 16-bit or 24-bit system. The 16-bit system cannot resolve anything below -93 dB, or 27 dB SPL in this example. The 24-bit system may capture the goings-on approaching 0 dB SPL — limited by the noise of the microphones, microphone preamplifiers, and analog to digital converters. At those low levels, the ventilation system in the hall will be competing with the sounds produced by the instruments, but that’s still information not resolved by the 16-bit system.
Looking at it another way, an analog to digital converter might accept a maximum signal of +24 dBu. If the ADC is set to 24-bit operation, it will record smaller fluctuations in the voltage signal than if it was set to 16-bit. The maximum input level does not change, but 24-bit operation captures more resolution.
Andrea…you and I have discussed this previously. The addition of 24-bits does not increase the number of “loudness levels” as claimed in the quoted piece. 16-bits are able to capture and deliver a theoretical dynamic range of 96 dB, with the noise floor actually 93 dB below the peaks as you state…and that’s a very long way down. Adding 24-bits doesn’t increase the number of “loudness levels” within that same range, which I believe is consistent with your comment. This the problem with writers stating that 24-bits increases the resolution…it does not. It adds additional capability with regards to capturing more dynamics. Perhaps you and I appreciate the term resolution differently. I don’t agree that 24-bits “slices” the available dynamics into more discrete values or that it adds resolution to system.
Apparently I haven’t explained well enough.
An alternating voltage representing a music signal comes into an ADC. Presumably you’ve adjusted the microphone preamplifiers etc. so that the maximum SPL generates a voltage swing that is close to the maximum input of the ADC — e.g. an ADC with maximum input of +24 dBu excepts voltages from -12.28 V to 12.28 V. You can record that signal with either 16 or 24 bits. The maximum amplitude is the same. As the voltage fluctuations get closer to zero, they can no longer be captured by the 16-bit system. If using 24 bits, smaller amplitude variations can be recorded. For example, a signal with amplitude 110 dB below that of the loudest signal you set your levels to capture is lost by the 16-bit system, but preserved by the 24-bit system. The ability to discriminate smaller signals in the same amplitude range seems a good working definition for more resolution.
I can’t comment directly on the article you are quoting, because you haven’t provided the link. I would not use the words “volume levels,” but increasing the system’s bit depth does increase the number of discrete voltage levels that can be recorded in a given range.
Stating that a 24-bit system can handle higher input and output levels, as you do, is backward to the way in which we use increased bit depth/dynamic range. Whether recording with 24 bits, 16 bits, or analog tape, you set your levels not to clip at the maximum volume from the performance. The tape loses all of the detail 60 or 70 dB below the maximum input, the 16-bit system 93 dB below, and a state-of-the-art 24-bit system better than 120 dB below peak amplitude. It’s the small signals we gain with greater dynamic range in the recording medium. (I acknowledge that having more dynamic range than is needed in the 24-bit system allows for greater safety margins when setting levels, but that does not greatly change the argument.)
Lastly, the rest of the world does seem to regard bit depth as a measure of audio resolution:
http://en.wikipedia.org/wiki/Audio_bit_depth
Alternatively, you can Google “analog to digital converter resolution” to read many articles on this basic concept of digital systems.
Mark, you’ve got a lot of good information on your website. That’s why I don’t like to see this error persist.
Andrea, I really think we’re saying pretty much the same thing. It’s just the terminology that is getting in the way. The Wiki article and the general concept that 24-bits offers greater resolution than 16-bits is absolutely true…but it’s the details that get in the way. Some writers want to think of the same amplitude range being covered by the additional levels and therefore we get greater resolution…like the additional density of pixels when the dots per inch goes up. I look at it based on the capabilities offered by each additional bit added to the digital words. Yes, we get more discrete values into which we can approximate a sample value (quantization noise) but we also expand the dynamic range by approximately 6 dB. That’s why I insist that a 24-bit system can handle more “input and output” levels than a 16-bit system. 24-bits can deliver a wider dynamic range by virtue of the lower noise floor than 16-bits…theoretically 144 dB vs. 96 dB.
It’s absolutely correct to say that the system with greater bit depth and greater dynamic range can handle a wider difference in amplitude between the largest and smallest signals it can record. It’s also true that the total available range is broken up into more discrete levels.
A 16-bit system divides up the voltage range — e.g. the -12.28 V to 12.28 V of my previous example — into 65,536 discrete voltage levels. Moving to a 24-bit system adds 255 discrete voltage levels between each available value in the 16-bit system for a total of 16,777,216 discrete levels over the same voltage range.
A simple analogy is that of moving from a system that rounds values between -10 and 10 to the nearest integer to one that rounds to the nearest 0.1. The increased resolution isn’t only between -1 and 1, but also between 7 and 8, for example.
I read your comment in this article as disputing that fact, but perhaps that wasn’t your intention.
When thinking about resolution and dynamic range, it’s important to remember that small signals don’t exist only in isolation. A signal 110 dB below full-scale may be present when nothing else is going on. More likely, it’s overlaid on a far larger signal. In either case whether or not it can be accurately recorded depends both on the number of quantization levels available and on the analog noise floor of the system.
Although it is correct to state that increasing bit depth increases the number of discrete voltage levels into which the overall voltage range can be divided — as it seems the other writer was pointing out –, and that a greater bit depth allows for greater dynamic range — as you have written many times –, neither statement gives a very complete picture of what is gained by increasing the number of bits. Writing good educational articles is hard, and trying to summarize a topic like this in a sentence or two tends to lead to misinterpretation.
Andrea, I reached out to my guru of all things digital and sent him your comments questions. Check out today’s post and this link.
Andrea, Mark is right. You are confusing the terms ‘range’ and ‘resolution’. A dictionary will confirm this. Regards.
Thanks Mark. It’s a nice article by John. I’ll follow up on your new post.
I want to thank you for the exchange…without it, I would not have reached out to John for his input. I think he’s one of only a few experts that can write and explain this stuff well.
Top CD players of yesteryear used to pay special attention to digiatl and analogue filtering of player of output. Not so much today.
192 kHz oversampling often makes previous CD recordings sound better because of getting rid of filter artifacts that can often be heard on modern players.
However when it comes to recovering missing information, I was remindened of a terrible saying that I once heard:.
“You can’t make a pig from a pork sausage”
I am glad that you did not write a negative comment on your friend’s blog. I do not believe that some of these HRA explanations are intended to mislead. HRA is difficult to explain in a short and easy way so that general consumers can understand it. Most people would never read the full explanation.
I do hope that we will begin to see the recently adopted source labels on HRA downloads soon though. For those of us who do understand that it is the source that matters, I would like to save some money buying up-sampled downloads.
Donald, the sales pitch about upconverting MP3s to 192/24-bits is misleading. Either the author doesn’t know the facts or is deliberately writing lies…I’m not content with either choice.
Mark,
Thanks again for the daily updates. They really do round out my day!
Facts are not really important to most people. Most folks just want to believe. They have all sorts of constructs built around themselves. Most of them are as flimsy as fairy tales.
Take a solid piece of cabinetry. You know what good joinery is (judging by your dovetail remarks in the past). You know what constitutes even a simple, but elegant piece. Solid materials. A sound material understanding of how different pieces work and fit together. And most importantly how expansion and contraction over time will affect the work.
You explain this in intimate detail to all of your friends and acquaintances. You even show and demonstrate the simplicity, durability and beauty of a well-made piece. Don’t waste your money on flimsy nailed/fastener connected junk, you say.
But the Ikea – or pick your woodpile – retailer continues to grow.
The Austrian economist Ludwig von Mises understood this apparent cognitive dissonance a long time ago. Everyone has different valuations. Value has nothing to do with price, or quality, or time preferences – short term or long term. Those are only flags or markers. Values are always subjective and are as unique as people are individual.
But I still appreciate your dogged determination! Keep it up.
Thanks. The woodworking analogy is appropriate and I miss not being able to work physically with wood. There was a real joy in feeling a couple of pieces of wood come together in a dovetail.
The gentleman with the up-sampling digital hub should be arrested and taken out of the gene pool.
Yes, he’s definitely been drinking the cool aid. And yet he vigorously defends his hardware. But I guess that’s what salespeople are supposed to do.
Interesting — Author X has been spreading this misinformation since 2012 when he first wrote this “advertisement” for a Sound & Vision and HDTracks collaboration back in 2012.
This “new” article on HighResAudioCentral is just a regurgitation of that original advertisement.
It’s unfortunate that this “expert” continues to spew this nonsense.
I was taught that if you don’t know something…do some research and make sure you get it right. The amount of bad information out there is astounding…including the stuff that passes for “standards”. The high resolution audio definition strikes me as a prime example.