24-Bits Makes Sense!
There is a lot of information and a lot of misinformation about audio out there on the World Wide Web. Who and what to believe is sometimes challenging for a layman. Heck, it’s hard for an experienced engineer to make sense out of it. I fully admit that there are people writing papers about the value of DSD and the fallacies of 192 kHz/24-bits that might know more than I do. But I’m pretty good at researching and discovering the truth…or at least a truth uncolored by hype and self-interest. So when I got an email from Rob the other day pointing me to an article by Monty at xiph.org entitled, “24/192 Music Downloads…and why they make no sense”, I knew I had to respond.
First, let me state that I like Monty’s overall approach. He’s unapologetic, rationale and has plenty of technical knowledge to speak with some authority on the topic. He gets a lot of things right but overlooks other things. I’d like to discuss some of the points that he makes in the rather lengthy article over several posts. Today, the main topic is his claim that 24-bits makes no sense. They do and here’s why.
He states, “OK, so 192kHz music files make no sense. Covered, done.
What about 16 bit vs. 24-bit audio? It’s true that 16 bit linear PCM audio does not quite cover the entire theoretical dynamic range of the human ear in ideal conditions. Also, there are (and always will be) reasons to use more than 16 bits in recording and production.
None of that is relevant to playback; here 24-bit audio is as useless as 192kHz sampling. The good news is that at least 24-bit depth doesn’t harm fidelity. It just doesn’t help, and also wastes space.”
The focus of his discussion regarding 24-bits for downloaded files is based on the recognition of the status quo and the limits that have been accepted by (or forced on) recording engineers and producers. The music business has long used radio as a primary means of promoting new songs and artists. If the record companies want their music to he heard over the background noise of your automobile, then they instruct the mastering guys and gals to push the volume and reduce or even eliminate dynamic range. It’s called the “loudness wars” and it has essentially dictated the number of bits that are actually required to completely capture and deliver a highly compressed commercial record.
And pop/rock music isn’t the only genre afflicted by this contagion. Classical music and jazz are subjected to the same processing although usually to a lesser degree.
The article states that most records don’t have sufficient dynamic range to utilize the potential range of even 16-bits. I would generally agree…but some of us have been successful in recording and releasing projects using 24-bits that actually do eclipse 16-bits or the theoretical limits of 96 dB SNR. It’s certainly not every piece of music that has more than 96 dB of dynamic range.
Figure 1 – A Chart Showing the Dynamic/Electrical Range of Sound, Music and Audio Systems
However, the instantly recognizable Ravel Bolero that I recorded in Bucharest in the summer of 2001 does. The background level of the Atheneum in downtown Bucharest is a very quiet place late at night. The opening snare drum rhythmic motif of the piece is played very quietly. I can distinctly recall how many times we had the percussionist play the opening to get the least volume with good articulation and detail.
And we all know how big the piece gets at the final modulation section. The whole 120 piece orchestra is playing at maximum volume and pings an SPL meter beyond 125 db for brief moments. This simply cannot be captured within 16-bits without some sort of dynamic compression (as is usually the case for vinyl and CDs) but we can deliver all of that range with 24-bits!
A rather funny side note: the review that I got of that recording in The Absolute Sound magazine stated that the recording had “too much dynamic range” in the reviewer’s opinion! So what am I supposed to do? The conductor and the players do their thing and I captured it without modifying the dynamics…that’s what I’m supposed to do. What I didn’t do was screw with the dynamics during the postproduction phase of the project. The reviewer needs to get a better and quieter listening environment.
So finally we have technology that can capture AND deliver the full potential dynamic range of human hearing. Analog tape can’t do it. Vinyl LPs can’t do it and 16-bit CDs can’t do it either. The dynamic potential of a 24-bit recording system can AND a 24-bit file can be delivered to consumers on a Blu-ray Disc or a DVD-Audio disc or as a download can too. If you have a current state-of-the-art DAC like the new DAC2 from Benchmark (which has a dynamic range of over 130 dB!) then you can reproduce the entire dynamic range of any piece of music. BUT THE RECORDING HAS TO HAVE BEEN MADE WITH THAT MUCH RANGE!
Admittedly, virtually all commercial music doesn’t even come close to exceeding 16-bits worth of dynamic range. So does that mean I compromise my efforts to live within the constraints of 16-bits. No it doesn’t. And whether your room and system can handle that much dynamic range is another possible constraint. But I don’t have to worry about that…you do.
Monty describes the current state of affairs but ignores those of us that are striving and delivering more. He didn’t factor that into his article.
1) The Benchmark DAC2 has not over 130 dB dynamic range but 123 dB unweighted (see http://benchmarkmedia.com/products/benchmark-dac2-hgc-digital-to-analog-audio-converter)
2) SPL of 125 dB? Hmmm… may be at the stage, very close to drum section or very close to a brass bell; at a reasonable distance from the orchestra, you have very seldom values in excess of 110 dB (more on this below,). See http://www.soundadvice.info/thewholestory/san12.htm
3) Background noise in any theater, even in a very quiet one as the one you have recorded at, is typically in excess of 30 dB (unweighted), even if they may measure 15 dBA (A-weighted). Theater noise comprises mainly low-frequency, inaudible noise, but a kind of noise that counts when discussing bit depth, since the ADC / DAC doesn’t know anything about frequencies. Beranek, in his book on Concert & Opera Halls (page 496 in the 1996 edition) recommends a noise level in octave bands below the NCB-10 graph, which allows 30 dB at 125 Hz and 40+ dB at 63 Hz.
4) Combining 2) and 3), at an audience location, you’ll have typically a maximum dynamic range of about 80 dB, considering a very loud passage at a very quiet theater.
5) This does not mean that for an individual tone you cannot get a better figure. Indeed, perceived dynamic range may be larger than the 96 dB (actually 16*6.02 + 1.76 = 98.08 dB) that allows a 16 bit representation, since the actual masking noise for, say, 1 kHz is what is contained in a critical band centered at 1 kHz, which has a bandwidth of 160 Hz, and has a level about 21 dB lower. So, believe it or not, you could perceive a 1 kHz tone at a level of -(98 + 21) dB = -119 dB below full scale (and even less)… with only 16 bit… The pianissimo snare drum, being a wide band signal, may require several dB more to stand out from the digitization noise.
6) Here comes also the important aspect of hearing health care. If in order to be able to listen to the snare drum at the beginning of the Bolero you have to raise the volume to the point that you’ll be subject to hearing loss risk at the 125 dB passages, there is something wrong. Besides, there is the fact that it gets quite uncomfortable for most people.Classical music enthusiasts will suffer from high levels, and surely that’s why the reviewer wrote that the recording had too much dynamic range. Rock or metal lovers, cannot even perceive sounds a few LSB in amplitude because of noise related hearing loss. I think sound engineers have some responsibility on this particular question. As a side note, even 24 bit may be not enough to represent all that is audible (let aside the fact that the loudest sounds can destroy at once the hearing system). At 3 kHz there are people capable of listening up to -8 dB, and a loud but short detonation may easily reach 150 dB, so you get a dynamic range of 158 dB. This doesn’t mean we should be using 27 bit to encode music.
7) Last but not least, one thing is the human hearing range capability and a different thing is hearing quality. Everyone’s ear starts distorting sound at a level typically ranging from 90 dB to 100 dB. This may be another reason of your negative review. In order to be able to listen to the snare drums pianissimo you are forcing listeners to exceed by far the distortion threshold of the ear.
8) I think the 24 bit technology is valuable for scientific research, since it allows to study acoustic phenomena that may prove interesting, such as the exact (?) waveform of a sonic boom wavefront, or the effect of turbulence. But if it is true that there is any perceivable difference (demonstrated through a truly scientific and well planned double blind test involving no known clues) several explanations should be explored before attributing any difference to the extra bits. May be better equipment is used when processing 24 bit than 16 bit, may be better processors are applied or also that more skilled engineers work at the mastering stage. See, for instance this scientifically conducted study: http://www.aes.org/e-lib/browse.cfm?elib=14195, also available for free here http://drewdaniels.com/audible.pdf.
Federico Miyara
Thanks for your comments.
I want to point out that 16 bit CD is capable of storing 120 db of dynamic range by dithering the 24 bit record, any noise produced by the dithering is pushing into the inaudible spectrums.
This is point out in Monty’s article. I believe that 24 bits for playback does have benefits regarding fewer quantization errors improving timing and reproduction of the sinusoidal analogue wave especially in the higher frequencies which I’ll explain later. I don’t believe you have read his article through, and by dismissing and ignoring his evidence without considering it, that is confirmation bias and ignorance.
I’ve watched Monty’s videos and generally agree with his positions. And he right about 16-bits being capable of 120 dB of dynamic range…but its’ also true that non of the CD that are commercially released have that much range. 24-bits is critical for recording engineers but not important at all for reproduction systems since the mastering houses dumb down recordings to much less than 24-bits. It’s a rare recording that can benefit from 24-bits on the delivery side of the equation…some of mine do. Thanks.
why not invent 18 or 20 bit?
Because processors works in units of bytes and each bytes is 8 bit. You can make music file that uses 18 bit per sample but it would either require extra processing and/or custom sound card, which is not practical
I really think that you guys, producers and equipment users should stop saying about technical things that you guys do not understand well. Signal processing has nothing about what you think or feel. Everything is measurable. What I see at most in the audio community are engineers that understands nothing about music and musicians / technicians that understands nothing about electronics engineering.
simple math … your hearing can not differentiate for REPRODUCTION a file of 16 … or N bits. NO MUSIC, NO ENVIRONMENT, has 140db of dynamic range … also NOBODY considers the ANALOG part of the equipment, where, man, NO-WAY you’ll get this dynamic range. a simple resistor has a lot more noise than this … not talking about the rest of circuit, analog potentiometers, operational amplifiers, filters … yes, resistors also, if you do not know what I’m talking about, it’s a good proof that you have to study electronics a little bit more before discussing the numbers that the manufacturers use to fool ordinary people. Please, do not get me wrong.
RECORDING is different … the music is uncompressed and will be edited later, with accumuled quantization noise of every processing stage, being common to find DSP’s with internal calculation down to 80 bis of precision … after getting out of this stage .. forget about the so-called hi-res.
Also interesting is that a lot of people have money to buy expensive gear of audio are over 30-35 y.o. and barely hear at 15KHz sine tone, with luck.
I’ll talk about the 96-192khz and so on … same bullshit … ok, maybe you’re producing bat, douph, or dog music with ultrasonic audio. Now, it’s time for the old audiophools which does not understand newer ADC / DACs technologies to start talking about how they feel the “air” of ultrasound with their ears cutting off at 14Khz. LOL.
Thanks for your input. I can’t say that I agree with everything in your comment…there is music that exceeds the dynamic range available on analog tape, vinyl LPs, and even CDs. And some audiophile equipment can produce dynamic range above 120 dB! Yes, it does take a rare playback environment to appreciate it but if it’e easy to provide high-resolution recordings to end users, why shouldn’t I? Making a recording using 24-bits at 95 kHz PCM provides a number of production advantages…whether you “hear” the results of those advantages is another thing.