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

[welwynnick]

Plus most of the very best speakers dont reproduce anything meaningful above 20KHz.
Look at the speakers of KEF, Genelec, Neumann, Perlisten, Arendal et al. => Falling of like a cliff above 20KHz.

 

[Steakhouse]

And which speakers are this?

 

[kifeep]

Hopefully this is true but I can't find that in the manual currently, rather just the bit I shared above explicitly about an ethernet cable. It would seem reasonable that it could also operate over WiFi but just cautioning here in case this is a dealbreaker for anyone.

Side-note: I hope the same functionality can be added for my older miniDSP SHD. This has an ethernet connection for music streaming already.

Looks like the first batch is shipping without working wifi:

1. Wi-Fi antenna. Screw on the supplied Wi-Fi antenna here.
   
   Info
   
   The Wi-Fi interface is not currently enabled. It will be enabled in a future firmware update.

Front and rear panels - miniDSP Tide16 User Manual

docs.minidsp.com

I would imagine this will be a high priority fix for later firmware.

 

[Ultrasonic]

Looks like the first batch is shipping without working wifi:

Front and rear panels - miniDSP Tide16 User Manual

docs.minidsp.com

I would imagine this will be a high priority fix for later firmware.

Well spotted!

 

[welwynnick]

And which speakers are this?

Look at the speakers of KEF, Genelec, Neumann, Perlisten, Arendal

If you think transparency can be achieved with a 20kHz bandwidth, you should see this ABX DBT by googlebot on HydrogenAudio, who compared 16/44 vs 24/96 audio.

I do not hear a difference over my regular Canton speakers, which top at about 20kHz! Neither over my Grado headphones.

My hearing tops at about 17 kHz.

Still I can hear pretty clear difference over a pair of Elac FS 607 X-Jet speakers, which are rated for 28-50000Hz (IEC 268-5):

He got 16/16 on foobar2000, but only with the wideband speakers.
Rhetorically, how can he hear the difference above 20kHz if he can only hear 17kHz?
I think the answer is that you can't base fidelity requirements on the audibility of sinusoidal tones.
They do this sort of test a lot on HA, and there are other examples.

 

[peng]

If you think transparency can be achieved with a 20kHz bandwidth, you should see this ABX DBT by googlebot on HydrogenAudio, who compared 16/44 vs 24/96 audio.





He got 16/16 on foobar2000, but only with the wideband speakers.
Rhetorically, how can he hear the difference above 20kHz if he can only hear 17kHz?
I think the answer is that you can't base fidelity requirements on the audibility of sinusoidal tones.
They do this sort of test a lot on HA, and there are other examples.

I don't doubt some people, under some conditions, can score well enough to prove there are audible differences regarding 16/44 vs 24/96, probably same for 24/48 but to lesser degree. What I consider hyperbole are the claims of obvious, even night and day differences heard, to me, in those cases, either there were other factors at work, or they just exaggerated, or imagined.. miniDSP has been in this game long enough, and appeared to be successful enough to not know the potentially audible bad effects from not using sampling rate higher than 48 kHz, based on their track record in choosing decent DAC ICs, and often implemented in schemes that resulted in SINAD >105 dB, I am not concerned about the effects would be audible to most humans with normal hearing, let alone those over 40 who most likely couldn't hear that well frequencies >15 kHz to begin with. Any fold back effects, such as those from IMD would most likely be at very low level, such as -80 dB or lower, as could be seen in some of those ASR measurements.

Regardless, just to save the forever debates on multiple pages of posts on various forums, I really do wish the likes of miniDSP would soon add the option to use higher sampling rate, even 96 kHz will go a long way to please more potential customers. I am hopeful they would do it for the Tide 16 without hardware changes, but likely limited to certain use conditions, such as for two channel stereo mode, and/or when Dirac Live is disabled, that's based on their subtle (very subtle) hint in the FAQ section.

 

[101geo]

I don't doubt some people, under some conditions, can score well enough to prove there are audible differences regarding 16/44 vs 24/96, probably same for 24/48 but to lesser degree. What I consider hyperbole are the claims of obvious, even night and day differences heard, to me, in those cases, either there were other factors at work, or they just exaggerated, or imagined.. miniDSP has been in this game long enough, and appeared to be successful enough to not know the potentially audible bad effects from not using sampling rate higher than 48 kHz, based on their track record in choosing decent DAC ICs, and often implemented in schemes that resulted in SINAD >105 dB, I am not concerned about the effects would be audible to most humans with normal hearing, let alone those over 40 who most likely couldn't hear that well frequencies >15 kHz to begin with. Any fold back effects, such as those from IMD would most likely be at very low level, such as -80 dB or lower, as could be seen in some of those ASR measurements.

Regardless, just to save the forever debates on multiple pages of posts on various forums, I really do wish the likes of miniDSP would soon add the option to use higher sampling rate, even 96 kHz will go a long way to please more potential customers. I am hopeful they would do it for the Tide 16 without hardware changes, but likely limited to certain use conditions, such as for two channel stereo mode, and/or when Dirac Live is disabled, that's based on their subtle (very subtle) hint in the FAQ section.

What's needed is a clear definition of "Night and day"
If they mean something like 17:59 in the day and 18:00 at night, they might have a point.

 

[RichB]

I don't doubt some people, under some conditions, can score well enough to prove there are audible differences regarding 16/44 vs 24/96, probably same for 24/48 but to lesser degree. What I consider hyperbole are the claims of obvious, even night and day differences heard, to me, in those cases, either there were other factors at work, or they just exaggerated, or imagined.. miniDSP has been in this game long enough, and appeared to be successful enough to not know the potentially audible bad effects from not using sampling rate higher than 48 kHz, based on their track record in choosing decent DAC ICs, and often implemented in schemes that resulted in SINAD >105 dB, I am not concerned about the effects would be audible to most humans with normal hearing, let alone those over 40 who most likely couldn't hear that well frequencies >15 kHz to begin with. Any fold back effects, such as those from IMD would most likely be at very low level, such as -80 dB or lower, as could be seen in some of those ASR measurements.

Regardless, just to save the forever debates on multiple pages of posts on various forums, I really do wish the likes of miniDSP would soon add the option to use higher sampling rate, even 96 kHz will go a long way to please more potential customers. I am hopeful they would do it for the Tide 16 without hardware changes, but likely limited to certain use conditions, such as for two channel stereo mode, and/or when Dirac Live is disabled, that's based on their subtle (very subtle) hint in the FAQ section.

If folks can reliably identify differences between 44.1 and >48 kHz, then foldback is the most likely cause.
The filtering is less severe and timing affects impacting < 20 kHz could be measured.

IMO, measuring sidebands in a lab is not going to expose all behavior or misbehavior of a given product.

I have dome experiements at home with the Topping D900 and D90 III Discrete DACs and the diffenences in the sound from various inputs are easily discernible.

What I find strange, is the lack of recognition that not everything is measured with lab equipment, sinewaves and multitone.
Some devices may have better power supplies and isolation that may perform worse in home environments.

Here are some other differences between inputs on my Topping DACs.

• PEQ circuitry may not be fully bypassed for some inputs
• Clocking can be internal of USB but external for other inputs.

There may also be bugs. These things get firmware updates...

- Rich

 

[welwynnick]

I don't doubt some people, under some conditions, can score well enough to prove there are audible differences regarding 16/44 vs 24/96, probably same for 24/48 but to lesser degree.
What I consider hyperbole are the claims of obvious, even night and day differences heard, to me, in those cases, either there were other factors at work, or they just exaggerated, or imagined..

The hyperbole that I object to are the sweeping statements that 16/44 audio is unequivocally, demonstrably, provably, and without any shred of doubt 100.00 % transparent.

Done right, lossless 16/44 audio offers very high quality, and is difficult to improve on.
The improvements with higher resolution formats are inevitably vanishingly small, even inaudible in many cases, and often not worth pursuing.
They're still detectable, but if anyone suggests there's a night and day difference, it must be because they're doing a sighted test.

The example I gave from HydrogenAudio highlights the fact that tester himself could not hear 20kHz, but that he could nonetheless hear the difference with HRA.
So it's not about who does or doesn't have golden ears, nor whether digital audio does what it's supposed to, nor whether measurements are right or wrong.

The point is that establishing the broadest requirements for transparency shouldn't be based on what sine waves we can hear - it should be based on what we can hear.

 

[peng]

What's needed is a clear definition of "Night and day"
If they mean something like 17:59 in the day and 18:00 at night, they might have a point.

Haha, to me, I would just ignore those "..night and day..." claims/comments..

 

[peng]

Done right, lossless 16/44 audio offers very high quality, and is difficult to improve on.

IMO that should be a given, of course it has to be "done right". If not, then even 24/192 may not be transparent.

The improvements with higher resolution formats are inevitably vanishingly small, even inaudible in many cases, and often not worth pursuing.
They're still detectable, but if anyone suggests there's a night and day difference, it must be because they're doing a sighted test.

Or there are other factors, variables that they failed to disclose, such as comparing samples of different masters, etc...

The example I gave from HydrogenAudio highlights the fact that tester himself could not hear 20kHz, but that he could nonetheless hear the difference with HRA.
So it's not about who does or doesn't have golden ears, nor whether digital audio does what it's supposed to, nor whether measurements are right or wrong.

We already know it is not the HF itself, but the fold back to the lower frequencies, so may be the tester heard those lower frequencies from the IMDs and for some reasons the magnitudes of those were high enough for his listening level, in his room, don't really know and don't really care, because to me, one could easily avoid those use/listening conditions it there's in fact audible issues.

The point is that establishing the broadest requirements for transparency shouldn't be based on what sine waves we can hear - it should be based on what we can hear.

Not sure what you mean..

 

[john61ct]

Proper double-blind, volume matched randomized ABX testing designed to ensure statistical validity with enough participants of normal hearing, would settle the question, even with untrained listeners.

@Floyd Toole

 

[Steakhouse]

If you think transparency can be achieved with a 20kHz bandwidth, you should see this ABX DBT by googlebot on HydrogenAudio, who compared 16/44 vs 24/96 audio.





He got 16/16 on foobar2000, but only with the wideband speakers.
Rhetorically, how can he hear the difference above 20kHz if he can only hear 17kHz?
I think the answer is that you can't base fidelity requirements on the audibility of sinusoidal tones.
They do this sort of test a lot on HA, and there are other examples.

I was referring to measurements taken by Amirm. I dont really care about the measurements you delivered here. Concerning the blindtest in the forum: Needs to be replicated by scientists. Not convinced.

 

[gnarly]

The improvements with higher resolution formats are inevitably vanishingly small, even inaudible in many cases, and often not worth pursuing.

Agreed. Although my opinion is always not worth pursuing.
I say put every penny we have into better speakers and room treatments, before higher resolution source material & hardware capability.

 

[Oddball]

Agreed. Although my opinion is always not worth pursuing.
I say put every penny we have into better speakers and room treatments, before higher resolution source material & hardware capability.

Or better EQ like Tide offers - albeit at lowely 48hHz. Not sure what actually does decent EQ with the higher sampling rates - I don't really use that so might be my blindspot.

 

[welwynnick]

Not sure what you mean..

Its rather fundamental, but difficult to explain because it's counter-intuitive. From a systems engineering perspective you have to start with the right requirements. If your objective is to develop a transparent audio system, it's natural to analyse frequency range, dynamic range, amplitude linearity, noise floor, etc.

Breaking that down, we establish the individual requirement for frequency range by measuring the audibility of low and high frequencies. That's typically done with sinusoidal tones - you listen to tones of increasing frequency until you can't hear any more, and you specify that as the upper limit of audibility.

However, the reasoning for that approach makes the key assumption that tones will give you the right bandwidth requirement, but this is an indirect solution. What we actually need to know is what's the bandwidth requirement for transparency, and that's slightly different. There can be lots of ways to do it - by listening to tones or by listening to jingling keys or by listening to cymbals or by listening to music or whatever. It doesn't matter what the method is, we just need to find out what bandwidth is needed so that ANY of those sounds can be heard transparently.

In the case of music or cymbals or keys, the sounds are more complicated than tones because there's more than one frequency being reproduced, and the sounds are time-varying. In practice that's what we listen to, and transparency of ALL of those sounds is what we need, not just transparency of ONE of those sounds. If all those sounds give you the same answer for the bandwidth requirement, that's fine, but if they give you different answers, then you need to accept the highest requirement, not the lowest.

There are several things that I've found in the last year or so that make me think the requirements aren't aligned, but that will have to wait for tomorrow.

EDIT: sorry it was late, here are some paragraphs....

 

[Oddball]

Its rather fundamental, but difficult to explain because it's counter-intuitive. From a systems engineering perspective you have to start with the right requirements. If your objective is to develop a transparent audio system, it's natural to analyse frequency range, dynamic range, amplitude linearity, noise floor, etc. Breaking that down, we establish the individual requirement for frequency range by measuring the audibility of low and high frequencies. That's typically done with sinusoidal tones - you listen to tones of increasing frequency until you can't hear any more, and you specify that as the upper limit of audibility. However, the reasoning for that approach makes the key assumption that tones will give you the right bandwidth requirement, but this is an indirect solution. What we actually need to know is what's the bandwidth requirement for transparency, and that's slightly different. There can be lots of ways to do it - by listening to tones or by listening to jingling keys or by listening to cymbals or by listening to music or whatever. It doesn't matter what the method is, we just need to find out what bandwidth is needed so that ANY of those sounds can be heard transparently. In the case of music or cymbals or keys, the sounds are more complicated than tones because there's more than one frequency being reproduced, and the sounds are time-varying. In practice that's what we listen to, and transparency of ALL of those sounds is what we need, not just transparency of ONE of those sounds. If all those sounds give you the same answer for the bandwidth requirement, that's fine, but if they give you different answers, then you need to accept the highest requirement, not the lowest. There are several things that I've found in the last year or so that make me think the requirements aren't aligned, but that will have to wait for tomorrow.

I do have a problem with reading posts in one paragraph. Perhaps just me but it is good to actually break out the thoughts in paragraphs sp other can follow with more ease.

Just my 2 c...

 

[welwynnick]

There are several things that I've found in the last year or so that make me think the requirements aren't aligned, but that will have to wait for tomorrow.

The first thing that made me sit up and think was listening to someone playing sine waves and square waves at different frequencies. At the highest frequencies they sounded the same of course, but drop down a bit and I could hear the difference between sines and squares at 6 or 7 kHz. A square wave is composed of the fundamental plus just the odd order harmonics - 3rd, 5th, 7th etc. That meant I was hearing the difference that 18 and 21 kHz harmonics were making. But I can't hear single tones that high.

I knew that infinite baffle speakers sounded better because they were phase linear to lower frequencies than other subs. I was convinced that Storm and Trinnov prices were exorbitant until I realised they corrected phase as well as amplitude. This improved time domain performance, especially in the mid-range. Wondering if high frequency time domain performance was being compromised by amplifiers, I looked at the phase measurements performed by SoundStageNetwork, and I found nothing. However, they also tested DACs, and despite flat amplitude response, there was a clear correlation between sample rate and pass-band phase linearity, like this RME ADI-2.

Some research at Kyoto University by Shibasaki & others was published by the Journal of Neurophysiology where music was played to a large number of listeners with different bandwidth filtering applied. The 50kHz bandwidth was broken into two halves - audible and ultrasonic. By various means, the listeners could distinguish the audible sound from the audible + ultrasonic combined. The interesting bit is the ultrasonic sound was completely inaudible, when it was played on it's own, without the audible sound.

The other was the DBT by googlebot on hydrogenaudio mentioned above. He freely admitted that he couldn't hear (sinusoidal tones) above 17 kHz, but he could hear the difference between a 20 kHz speaker and a 50 kHz speaker. That test was indeed done with robust scientific principles - level matched, double blind, and randomised. Those are the rules on hydrogenaudio. Yes, the test material was taken from the same master, and even Arnold Kruger himself (RIP) said of the test methodology: "I don't see anything obvioiusly wrong with the samples." Googlebot performed multiple tests and got a foobar2000 test result with a high degree of confidence, and other people also got positive tests with over 95% confidence.

People often try to answer positive results like this by attributing it to intermodulation distortion, and foldback of high frequencies into the audible band. In googlebot's case, he was using the same DAC and amplifier, and if they were causing the IMD, then that would have been audible with the 20 kHz speakers. Of course the speakers themselves could have been causing the IMD, but it would probably have been even more audible with the inferior speakers, working outside their normal operating range.

Some people have said there were two variables at play - both different bandwidth and different depth of modulation (16 vs 24 bits). If that was the explanation for the results, then it wouldn't have made any difference which speakers were being used.

 

[welwynnick]

Agreed. Although my opinion is always not worth pursuing.

I say put every penny we have into better speakers and room treatments, before higher resolution source material & hardware capability.

Or better EQ like Tide offers - albeit at lowely 48hHz.

No argument with any of that. MiniDSP and Storm and others have almost certainly done the right thing for their individual cases.

Not sure what actually does decent EQ with the higher sampling rates - I don't really use that so might be my blindspot.

However there are products that do process DRC at higher rates.
The Datasat LS10 and RS20 process Dirac at 96 kHz using hardware overkill - they used NINE Analogue Devices SHARC DSP chips.
The Trinnov Nova and Altitude 16 perform processing at 96 kHz.
The Trinnov Altitude CI and Altitude 32 process at up 192 kHz.
For full disclosure, Theta claim that the Casablanca IV can perform Dirac processing at 96 kHz, but I understand this is the DSP I/O sample rate, not the processing rate.

48 kHz processing is probably good enough, BUT Trinnov use modest i3 12100 and i5 12400 CPUs for their processing. They only cost about $100. Yes that's too much for any AVP costing $ hundreds, and more than the ADSP-21593, 21489 or SC598 used in many audio processors. But it's only about $60 - $70 more, hardly prohibitive for any AVP costing $ thousands.

 

[Ultrasonic]

miniDSP's SHD series models run at 96 kHz with the stereo version of Dirac Live. Rather different DSP overhead for multichannel ART though.