32bit recordings only have 24bits of audio

WaveLab
42

Multi-path makes a perceptual difference by reducing distortion at lower listening levels. Multi-path THD reduction can be an order of magnitude, or far more, compared with legacy single-path architecture. This is easily measured on THD+N test equipment (see Bob Katz review). But most importantly – and something Bob experienced – reduced THD is immediately perceived in improved imaging, texture, and atmospherics.

Good audio gear is now achieving -120dB THD+N at max levels. Multi-path THD reduction is achieved when the signal drops into the “low path” (say -35dBFS). As any classical/jazz/acoustic recording engineer knows, the “detail and space” of an acoustic recording in a good room is perceived in the quiet passages, the reverb tails, and the immediate / initial texture of percussive notes.

So, multi-path isn’t really about “more bits” or “sample rate” — it’s mostly about a dramatic reduction in THD+N. Compared with today’s best D-to-A converters, multi-path improves dynamic range and linearity by 100X, and reduces THD+N by 10X to 100X or more. This is the greatest single audio improvement in the 140-year history of audio (LP to CD was a net 30X improvement in dynamic range and linearity).

1 Like
43

Call me ‘skeptical’ but 10x t0 100x better than this …

Let’s just wait and see shall we. Hopefully, you guys turn out to be correct.

All the best

Paul

44

Most people today listen on really crappy playback equipment and 44.1KHz and 16 bits is more than they need or want…FWIW

This quest for better and better specs is OK but the average consumer could care less. Most people today, IMHO, listen to MP3s on earbuds and could care less about 32 bits or 384KHz sampling rates IMHO…

45

Was it ever different?

46

Yes in the years I was growing up almost everyone had a “nice” home music setup and most cared about their music reproduction. Of course there are always those that played their 45 records on a 3.5" speaker enclosed in the changer…

1 Like
47

Hmm. Noise level in a very quiet room is extremely seldom below 25dB. Don’t confuse this with mics saying they have self-noise in the single digit A-weighted.

Max level any instrument makes is about 130dB.

Even if we add quite a bit of padding on both ends, our microphones would not need more than 24 bits (144 dB range) to cover the whole. Going above 24 bits turns into marketing hype, which of course is interesting but not really of much relevance.

49

ghellquist,

What you say is true, for most microphones, but misses the main point. Which is: multi-path architecture reduces THD+N by orders of magnitude at perceptual levels. Read the Katz review to learn how it achieves this. This new standard of THD+N performance is immediately audible in improved timbre, imaging, and atmospherics.

That said, let’s revisit real-world dynamic range. Today most people are listening on headphones, which attenuate room noise by at least 25dB, often much more at mid-freqs. If we use your “25dB room noise” number, headphones reduce ambient noise to below 0dB SPL – in many cases down to the threshold of hearing (-8dB SPL - edit). That’s the new normal for our low side audio performance.

On the loud side, high-end home entertainment installs are using +148dB SPL subwoofers – for car crashes, explosions, SFX, and other sub-50Hz information (ears are not harmed at these brief VLF levels). That’s the new normal for our high side experience. Which gives us a real world dynamic range of 156dB, or 26-bits.

As an aside, many microphones do perform at 155dB SPL, even low cost Shure SM57s. We stick these on snare drums every day. Snares have a peak output of around 155dB SPL.

50

What could be below absolute quietness?

1 Like
52

I don’t know. That’s why I ask.
0dB SPL is the possible minimum.

Really?

dBu is not SPL. And there are no negative SPL values possible.

Sound Pressure Level!!!

53

OK, I googled a bit, and there are some issues with my explanation (because the translation to German). I got confused with pressure and level… :wink:

So, to correct it, sound pressure can’t be negative, but for sound pressure level there is a reference used of p 0dB = 20 μPa = 2·10−5 [Pa]
So it can be negative.
But it is not measured in dBu. That is a voltage level.

1 Like
54

Most people are listening to the music today for the music and not for how quiet it can get. 44.1 KHz and 16 bits is more than most people need or want. It is nice to have GREAT specs but the average consumer could care less about a 384 KHz sample rate or 32 bit audio…IMHO

2 Likes
55

This is basically what I was saying earlier … it’s about the music … and of course that Topping DAC linked above has a THD+N ratio of 0.000065% which isn’t too shabby for a non multi-path architecture device

56

I think the idea is not that we go below that but rather that the noise added to the signal we receive is significantly lower relative to the part of the signal we want to hear higher when our overall level is higher than where our normal noise floor is.

I’m not sure if there was a link to the article or if I searched my way to it, but in the article it is explained just where we get less SNR+THD relative to desired signal as this ‘dual design’ ‘switches over’ from one ‘path’ to another, and that happens before we get down to the noise floor of a decent environment.

I don’t think anyone is arguing that the average person listening to crap music compressed to hell 20 seconds at a time because they have attention problems, listening on crap earbuds or a laptop or TV or tiny speaker… is going to hear a difference.

57

Correct. Multi-path architecture is a new professional model, for working engineers who need to hear more precisely into the program at lower perceptual levels (-35dBFS, etc). Think mastering engineers and post production mixers. The imersiv company preview page has a number of professional comments on it. The difference is not subtle, but it depends on program. Metal with 10dB dynamic range probably isn’t a good candidate. :upside_down_face:

Sorry, 0dBu was a typo. I fixed it.

Note: the lowest acoustic energy we can perceive is -8dB SPL @ 3kHz, not 0dB SPL. The 0dB number is from 1920s testing. See graph.

A lot! I think the lowest sound we’ve actually measured is something like -30dB SPL. But that’s far below the human threshold of hearing. Microsoft has an anechoic room with self-noise of around -30dB SPL.

Well designed single-path and multi-path devices have equivalent performance at full-scale levels. When program drops to -35dBFS (or so), that’s when multi-path dramatically outperforms single-path. In multi-path architecture, this is called the “cross-fade” point. At the moment of cross-fade, THD+N drops by more than an order of magnitude, often much more. This is audible, and is what Bob Katz heard in his beta-test review.

So, for instance, on the Topping, at -50dBFS, its THD ratio is probably in the 0.1% range (every -6dB level change = a doubling of THD ratio), while a multi-path DAC will be in the 0.001% range, or probably less. This is because, at the moment of cross-fade, multi-path transfers to a different DAC core, and starts processing at that new DAC’s 32nd-bit (full scale), which gives the highest performance THD numbers. It’s counterintuitive, but that’s the basic idea.

Katz actually ran tests on all the numbers, so this isn’t just “made up”. It’s a dramatically new audio architecture, and if I had my choice, I would use audio devices that exhibited 0.001% at low levels rather than 0.1%. But that’s just me.

Ear Perceptual Graph
Ear Perceptual Graph1920×1521 186 KB

58

You may like this microphone:

https://www.grasacoustics.com/products/special-microphone/low-noise-measuring-systems/product/180-40hf

1 Like
59

Yes! When designing multi-path microphones, you need at least 2 transducers. There are a bunch of patents on this, but the basic idea is you use something similar to the GRAS mic (or DPA, etc) for ultra-low level information. Then you add a second transducer in tight proximity that’s optimized for ultra-high level information, like snare hits, gunshots, and trumpet blasts.

Then you feed each transducer into its own processing chain (preamp => ADC => DSP). The DSP is programmed to sum (“cross-fade”) the incoming signals as required to create a final high dynamic range output, which now represents the self-noise of the low mic to the clip point of the high mic, creating a microphone with potentially 170dB of dynamic range (or more).

This same concept can be applied to power amplifiers. Effectively what multi-path does is dramatically reduce the self-noise of every element of the audio path, while adding a bit of extra headroom on the top. Multi-path from mic to power amp, and everything in between. I do think it’s the next paradigm of audio architecture. Once it’s miniaturized into an IC, it becomes the standard way of doing things.

2 Likes
60

Well, the very low noise mics are not very good at high decibels. They might get damaged.

61

Precisely! Multi-path is a wickedly hard engineering puzzle. Many hurdles yet to overcome. An interesting parallel in the video industry is Sony Video, who took many years to develop a multi-path CCD sensor, which had to overcome this very same problem, except with bright light. They used a low-level CCD for low light and a high-level CCD for bright light. But the bright light saturated the low-level CCD. They had to reduce the “recovery time” of the low-level CCD to within one frame rate. They did it! This is the next paradigm of “HDR” video performance. Google “Sony 2-Layer Transistor Pixel”

2 Likes
62

Thank you for this detail. I guess where I am coming from is that in my mastering world, I (and my DACs) pretty well never see any program material anywhere near as low as -35dBFS. In context, I regularly get mixes at -8 to -6 LUFS from some of the top mixers in my country. But I get what you are talking about and I am not saying it’s not valid.

I have nothing but respect for Bob Katz, but this is an example of where non Audio Science type reviews (with literally USD100,000+ equipment) fall short: audiophile subjective assertions that something can be ‘heard’.

Anyway … all of this is getting outside the scope of the WL forum so I’m happy to wait to see how the released product goes and tests. Look forward to it.

All the best

1 Like
63

Yes, most music today has a dynamic range of 10-25dB, or so. It’s all compressed into the very top. The types of music Bob was using for the multi-path review (jazz, classical, ambient – he printed a list) has far higher dynamic resolution, which is why the performance difference (single vs. multi) became obvious. Then again, a number of rock-pop producers and engineers have been beta testing multi-path D/A prototypes, and have had similar praise. Larry Crane at Tape Op Magazine, for one. Paul Blakemore at Concord is another.

What’s interesting about Bob’s review is that he uses test equipment to objectively characterize and compare performance (like Audio Science Review). The differences in noise + distortion are profound. I’m sure once the ASR people test multi-path, their numbers will be similarly impressive, and will validate Bob’s results. Bob knows what he’s doing.

A paradigm shift is defined as a 10X breakthrough.Multi-path isn’t a 10X improvement. It’s closer to 100X improvement of every key parameter: dynamic range, linearity, self-noise, and lower-level THD. I think this is why some very important audio gurus have recently joined the imersiv advisory board. We’re seeing the next architecture of audio capture and delivery.

1 Like