miniDSP Tide16 - Holy Grail with 16 Channel Atmos/DTS:X, high SINAD

[peng]

Over the years I've built up a favourites folder full of evidence about the audibility of high resolution audio. Here's another one from AES Convention 128 Paper 172 by Pras & Guastavino at the Centre for Interdisciplinary Research in Music Media and Technology, Multimodal Interaction Laboratory, McGill University, Montréal

https://www.researchgate.net/publication/257068631_Sampling_Rate_Discrimination_441_kHz_vs_882_kHz

Are you talking about cases where recording was done at 88.2, that's very different than playing back at 88.2 vs 44.1.

 

[Curvature]

Yes you can hypothetically represent any delay as if it is one due to distance, but not all delays are due to propagation delays through the air. Accurate spatial information misses other causes of delays and so is not sufficient for optimal time-alignment.

I think I understand the confusion. You are making a distinction between an object map, where the speakers and listeners are in a room, and an acoustic map, how and where sound travels, taking into account the electronic signal path as well.

There is a somewhat bewildering method in AVPs of time alignment where you enter the physical distance between the listening position and each speaker. And if that fails because the speakers or other downstream devices cause a per channel difference in timing, you increase/decrease the "distance" in the AVP as an approximation.

A microphone array does not measure distance. It measures timing differences that are converted, using the speed of sound as a basis, into approximate locations. It is completely agnostic about the source of delay. In this sense it produces an acoustic map foremost.

 

[welwynnick]

Where you have contradictory evidence like that, my view is that you should take the result with the highest positive result, rather than the lowest negative result.

That's an example of a negative outcome, where subjects were unable to discern a small difference. There are lots of those.
My point is that the limit of audibility should be set at the smallest discernible difference that CAN be detected, rather than the largest difference that can't be detected.

My reason for focussing on positive results is because a negative result doesn't prove anything. You can't prove that something doesn't exist - only that it does exist.

Here's a disgraceful example of pseudo-science by Ashihara et al that concluded "the detection threshold for random jitter was several hundreds ns for well-trained listeners under their preferable listening conditions". Audiophile bashers all over the World cheered, because it was seen as scientific proof that a high level of jitter is inaudible.

What they actually did was simulate the effects of jitter by modifying the audio data itself (see their other papers) and get their listeners to play the samples on a PC, DAC and audio system. Firstly, if you know anything about jitter, that's not jitter. Second, their transport was a PC, which is full of noise and jitter. Maybe if the test was done today it would be different, but this was 20-odd years ago. The researchers simply assumed they had a good source, and didn't do anything to quantify the performance of their baseline. This is analogous to testing a linear amplifier using a tone generator with 1% THD, and concluding that the amplifier has 1% THD (this is why the APx555 is so good - because its signal generator is so good). Ashihara concluded that their subjects couldn't hear hundreds of ns of jitter, because the source had hundreds of ns of jitter. It's an (extreme) example of a negative result obtained because the test wasn't good enough

If you get a negative test result, it could be for two reasons - because the sonic difference you're looking to detect is too small to be audible, or because the test isn't good enough. The problem is, if you have a negative result, you don't know which one it is.

If you get a positive result (achieved with proper scientific rigour of course) then you know BOTH that the test is good enough AND the sonic difference is audible.

EDIT: I don't want to go any more off-thread, so I'm just going to dump all my audibility links here. The 2014 Meridian AES paper is also available in full, now.
It's not just one or two positive test results. Whenever I probed deeper, I kept finding more, so there's lots out there, and probably lots that I've never seen.

Audibility of "typical" Digital Filters in a Hi-Fi Playback - Page 4
Audibility of 20kHz brick wall filtering
Audibility of Group-Delay Equalization | IEEE Journals & Magazine | IEEE Xplore
Detection threshold for distortions due to jitter on digital audio
Double Blind tests *did* show amplifiers to sound different | Audio Science Review (ASR) Forum
E-library page - AES
HD high noon
High-resolution music with inaudible high-frequency components produces a lagged effect on human electroencephalographic activities - PubMed
https://ieeexplore.ieee.org/document/9450008/
Inaudible high-frequency sounds affect brain activity: hypersonic effect - PubMed
Inaudible High-Frequency Sounds Affect Brain Activity: Hypersonic Effect | Journal of Neurophysiology | American Physiological Society
Mastering Captured Vinyl For CD
Modulatory effect of inaudible high-frequency sounds on human acoustic perception - PubMed
Proof that DACs CAN make a difference! - Blind ABX Testing - YouTube
Sampling Rate Discrimination: 44.1 kHz vs. 88.2 kHz
Successful ABX of 24/96 vs. 16/44.1
The human ear detects half a millisecond delay in sound | Aalto University

My own conclusions:
This is the 21st century, and we don't use CRT TVs, VCRs, film cameras, typewriters and fax machines any more.
The bottleneck in digital audio should lie in storage & distribution, and not in expensive reproduction equipment.
High Resolution audio definitely makes a difference, I'm even more confident of this now than I ever was.
The difference between CD and HR is tiny, often inaudible, and isn't worth pursuing in most cases.
24/48 audio is fine for film soundtracks, it's only well-recorded music that shows the benefit of HR.
MiniDSP and Storm process at 48kHz, and I'm sure they will stay that way. Better filters outweigh higher sample rates.

 

[welwynnick]

The first step removes information. The second step adds information that's not there?!!!

Yes the first step removes information, but shouldn't that be inaudible? The second step is interpolation, like a digital reconstruction filter, and cannot add information.
I assume the second step was done so the D to A conversion process was the same in both cases.

Are you talking about cases where recording was done at 88.2, that's very different than playing back at 88.2 vs 44.1.

Both scenarios were tested - comparing an 88.2 recording with both a 44.1 recording, and an 88.2 recording down-sampled to 44.1.

 

[Curvature]

The only argument for high resolution audio that makes sense to me, for audio production and reproduction and not scientific or engineering purposes (e.g., using 1/10th scale models of buildings and using ultrasonic signals to understand sound propagation), is enforcing exporting the master at 24 bits and 44.1 or 48kHz.

Based on how a lot of musicians work, and use of badly designed plug-ins, and unclear AD/DA chains, and questionable recordings, it is really unlikely that all of the stems are going to meet that standard, but some will. At least the master will be consistent.

Based on how audio delivery works these days, the vast majority of content involves compression in one format or another, which will have no very high or ultrasonic frequency information anyway.

If we limit this discussion to playback of special film or audio releases meant for small audiences, then, assuming the signal actually has high frequency information and isn't just 44.1/16 dumped into a larger container, this topic becomes a highly niche. In my mind, equivalent to asking whether or not the Storm device is worth it compared to this Tide16 and how willing or unwilling you are to compromise, and what you actually gain by through the latter.

 

[EWL5]

Yes the first step removes information, but shouldn't that be inaudible? The second step is interpolation, like a digital reconstruction filter, and cannot add information.
I assume the second step was done so the D to A conversion process was the same in both cases.

Both scenarios were tested - comparing an 88.2 recording with both a 44.1 recording, and an 88.2 recording down-sampled to 44.1.

That's not how downconverting works as there's always something lost. My music teacher told me that when you take something analog like a record and convert to something digital like a CD, you have to shave off quite a bit of info for the bits to fit on the CD (otherwise, the info would be nearly infinite). A CD will ever only be an approximation of a record. In your example, you are going from digital to digital so less info is shaved but shaved nevertheless! You can't restore what isn't there!

 

[masterhw]

My music teacher told me that when you take something analog like a record and convert to something digital like a CD, you have to shave off quite a bit of info for the bits to fit on the CD (otherwise, the info would be nearly infinite).

I understand the sentiment, but this is also kind of implying a record is an infinite data source. There are practical limits (usually because of the materials and production methods) to where the record can faithfully store and reproduce the original source, and that limit is generally at best something like 12 bits of depth before you hit noise, and 18-22kHz max (doesn’t exceed 44.1kHz sampling).
Thus a standard CD can capture all of the signal plus some of the noise. Just not an infinite amount of noise.

 

[kyuu]

My music teacher told me that when you take something analog like a record and convert to something digital like a CD, you have to shave off quite a bit of info for the bits to fit on the CD (otherwise, the info would be nearly infinite). A CD will ever only be an approximation of a record.

This is nonsense. Please look up Shannon-Nyquist and/or just get educated by Monty:

 

[Mort]

MiniDSP confirms there will NOT be a version sold without Dirac for any reason, tariffs or otherwise.

 

[pat355]

MiniDSP confirms there will NOT be a version sold without Dirac for any reason, tariffs or otherwise.



View attachment 506407

Glad we’re getting back to the Tide16 at last

 

[TurtlePaul]

My music teacher told me that when you take something analog like a record and convert to something digital like a CD, you have to shave off quite a bit of info for the bits to fit on the CD (otherwise, the info would be nearly infinite). A CD will ever only be an approximation of a record.

Yes, downconverting losses data.

But on the record side, you can compare imperfection is the record (microscopic flaws in the vinyl or macroscopic errors with the angle of the tone arm or warped vinyl) with the rounding that is used in the smallest bits of digital audio. The result is that 16-bit audio has 20-30 dB better dynamic range capability than records.

Also, 99% of records were mastered using digital delay equipment, so it is missing that information too. Not a lot of people talk about how the last step before the mastering laythe is a analog -> digital (delay) digital -> abalog step.

 

[Mort]

Glad we’re getting back to the Tide16 at last

I'm gonna keep trying....


MiniDSP also confirms that they will neither confirm nor deny the possibility of Digital AES3 / Dante in the future.

 

[pat355]

I'm gonna keep trying....


MiniDSP also confirms that they will neither confirm nor deny the possibility of Digital AES3 / Dante in the future.

View attachment 506408

I for one would be thrilled if they did at some point. Since they do read comments might be marginally helpful to express my interest

 

[Oddball]

MiniDSP confirms there will NOT be a version sold without Dirac for any reason, tariffs or otherwise.



View attachment 506407

That is something that Monoprice and especially D&M should do. For D&M Audy is basically free as they have a long term arrangement that either costed them money or is costing them money without the way to get out. However, they would need to negotiate better Dirac terms for their users.

One thing that only D&M could probably afford is to include whole Dirac pack up to ART as free 30 day trial that disappears if not activated, but if activated, that would be at the lower price than current options.

 

[Curvature]

Some years ago I read this in Hydrogen Audio, where you need evidence of a Foobar2000 ABX DBT or similar in order to even discuss subjective differences.
Like AVS but draconian. Googlebot took a 24/96 recording, down-sampled it to 16/44.1, and up-sampled it again to 24/96.
He compared the original and re-sampled recordings, and correctly identified them 16 / 16.



IgorC and Wombat also got 5/5 ABX DBT results.
Googlebot said he couldn't hear anything above 17 kHz, and he couldn't hear the difference in the HF content of the 2 samples.
He was only able to hear the difference with the Elec speakers, and not the Canton speakers, which extend beyond his own hearing.
This was some time ago, and I'm sure there are lots more DBTs on HydrogenAudio.

There are two points.
Firstly, there is evidence out there.
Secondly, the audibility of high frequency tones isn't enough to determine the requirements for transparency.

If the original files are available (link fails to load for me), we can compare them to understand the differences in the signal. If the files don't show anything, then we have to raise questions about the signal path and if anything, devices, drivers, DSP, contributed to incorrect playback.

Otherwise this doesn't rise above the level of anecdote about the "time smearing of Redbook".

 

[EWL5]

I understand the sentiment, but this is also kind of implying a record is an infinite data source. There are practical limits (usually because of the materials and production methods) to where the record can faithfully store and reproduce the original source, and that limit is generally at best something like 12 bits of depth before you hit noise, and 18-22kHz max (doesn’t exceed 44.1kHz sampling).
Thus a standard CD can capture all of the signal plus some of the noise. Just not an infinite amount of noise.

I should have included the caveat that despite something analog embodying nearly infinite information, when it comes to our ears we only need the amount of info where it would be considered lossless to the studio master. Unless I'm mistaken, this only came about w/Dolby TrueHD and DTS-HD MA during the early Blu-ray days. So my question to the group (and continued apologies for OT comments):

1) Is lossless audio a scam?
2) Is it possible to downconvert something that was (1) TrueHD to (2) CD quality and then upconvert back to (3) TrueHD and still tell the difference w/o losing information in the process? I would think there's no point to lossless audio if this were true as (1) and (3) would sound the same!

 

[Curvature]

I should have included the caveat that despite something analog embodying nearly infinite information, when it comes to our ears we only need the amount of info where it would be considered lossless to the studio master. Unless I'm mistaken, this only came about w/Dolby TrueHD and DTS-HD MA during the early Blu-ray days. So my question to the group (and continued apologies for OT comments):

1) Is lossless audio a scam?
2) Is it possible to downconvert something that was (1) TrueHD to (2) CD quality and then upconvert back to (3) TrueHD and still tell the difference w/o losing information in the process? I would think there's no point to lossless audio if this were true as (1) and (3) would sound the same!

1. Lossless compression retains all information. E.g. wav to flac. This is the only context in which this concept is valid.
2. Downconverting, not the same as compression even though file sizes decrease, if done correctly, loses information only if the container is has information to lose. If you downconvert from 192kHz to 96kHz when your audio signal contains nothing over 13kHz, then nothing is lost. Upconversion only has a theoretical benefit when trying to avoid the effects of the reconstruction filter in a delta-sigma DAC.

Some people claim stuff about increasing the sample rate being important because time resolution (rather than frequency resolution) increases. This is sorely mistaken. Here's the math: https://troll-audio.com/articles/time-resolution-of-digital-audio/

 

[Oddball]

I should have included the caveat that despite something analog embodying nearly infinite information, when it comes to our ears we only need the amount of info where it would be considered lossless to the studio master. Unless I'm mistaken, this only came about w/Dolby TrueHD and DTS-HD MA during the early Blu-ray days. So my question to the group (and continued apologies for OT comments):

1) Is lossless audio a scam?
2) Is it possible to downconvert something that was (1) TrueHD to (2) CD quality and then upconvert back to (3) TrueHD and still tell the difference w/o losing information in the process? I would think there's no point to lossless audio if this were true as (1) and (3) would sound the same!

Digital is just different than analogue. As people pointed out, records have their own limitations, and tapes even more. While analogue gradation of the signal might be more granular, overall these formats are - let's just say old.

Lossless audio is not a scam. Same titles on Blu Ray do sound better than on streaming where they get clipped for the bit rate. In my case, I prefer the convenience of streaming vs ordering, changing and then returning the discs to the library. Lazy dog and penguin is even worse.

As to your second point if you had the same algo downsampling and then upsampling, the upsampled version could be close to original, but not quite there. The downsampled version would likely not ever be the same as the True HD.

How much it matters? Little to me, but then splitting hairs is the favourite hobby of this forum. A strand of hair is a strand of hair, but then when you split it 100 ways it is hundred strands of hair.

 

[EWL5]

1. Lossless compression retains all information. E.g. wav to flac. This is the only context in which this concept is valid.
2. Downconverting, not the same as compression even though file sizes decrease, if done correctly, loses information only if the container is has information to lose. If you downconvert from 192kHz to 96kHz when your audio signal contains nothing over 13kHz, then nothing is lost. Upconversion only has a theoretical benefit when trying to avoid the effects of the reconstruction filter in a delta-sigma DAC.

Some people claim stuff about increasing the sample rate being important because time resolution (rather than frequency resolution) increases. This is sorely mistaken. Here's the math: https://troll-audio.com/articles/time-resolution-of-digital-audio/

Yes, I think this is where I got confused. To summarize, there should be a greater audible difference b/w compression types of the same track (i.e. Dolby TrueHD vs DD 5.1) and fewer differences b/w sampling rates (the 24/96 recording down-sampled to 16/44.1 in @welwynnick 's example).

 

[Curvature]

Yes, I think this is where I got confused. To summarize, there should be a greater audible difference b/w compression types of the same track (i.e. Dolby TrueHD vs DD 5.1) and fewer differences b/w sampling rates (the 24/96 recording down-sampled to 16/44.1 in @welwynnick 's example).

If compression is done correctly, even when lossy there should be little to no audible differences. https://ccrma.stanford.edu/~malcolm/13dB_Miracle/ In practice there will be edges cases where glitches emerge, but thise are rare and in my experience not the fault of the container but incorrect settings.

In the end, most containers are transparent and the reason compression works is that it is shaped to the audio signal. I went out of my way to acquire limited CD runs (the only original source) in the 2000s of some music because I had variable bit rate mp3s from file sharing. The VBRs averaged 60-250kbps, while the ripped wavs were the usual 1.4mbps. Couldn't tell the difference, sadly.

The reason that 44.1/48kHz at 24bits is enough for audio is that you it captures everything and more of human hearing. Most people can't hear over 15-17kHz, and most soundsystems won't deliver more than 16bits (96dB) of dynamic range, though in practice this is more like 12-14bits at home (72-84dB).

It's like the SINAD discussion. Generally 70dB of SNR and 40dB of THD is transparent (the SINAD value is set by whatever is higher, noise or distortion). But there are many cases where that's not true, and many cases where a fraction of that value works without audible consequence, so to meet all circumstances, devices are made with 100-110dB SINAD at high output (or should be, at least, like the Tide16).