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

[EWL5]

I wanted to follow up after I started a bit of a quarrel about the lack of HDMI 2.1 a few weeks ago.

The trigger was an annoying HDMI handshake issue between my Denon X3700 AVR and the LG G5: the Denon injects ALLM, which forces the G5 into PC Mode and disables a lot of processing features. So I took a closer look at HDFury's portfolio and came across the VRROOM 8K.

At first glance, nearly €600 for an HDMI 2.1 I/O processor feels steep. But after two or three showers thinking it over, I actually came around on it — because it lets you bypass any wonky HDMI processing and use any AVR as a pure pre-processor. Anything that handles ARC/eARC can be reduced to audio-only duty, and the VRROOM effectively becomes the AVR itself. Considering what AVRs usually cost, €579 suddenly looks pretty reasonable for a device that solves basically every HDMI setup problem for the next 5–10 years (HDMI 2.1 should carry us a while at 4K120 / 8K60).

So yeah — I'm happy to backtrack here: the Tide16 is actually a pretty neat device, if you factor in that gamers and PC-couch users will probably need a VRROOM alongside it.

Just wanted to close the loop on this. Have a nice weekend!

FWIW, no handshake issues b/w my 3800 and a 65" G5.

 

[lozoyad]

That follows by definition just by expressing a desire for higher bitrates in a listening context as opposed to music production.

> if I feed a processor 192kHz

That is a mistake if you don't trust the device doing the downsampling.

If there is any audible degradation, it is truly broken.

> I do know when not using Dirac I get 96kHz

The fault is miniDSP skimping on processor hardware, it simply cannot handle complex high tap count filtering at higher rates.

I wish that were different, but to get more taps not higher rates.

Interesting. So I had no idea what you were referring to with "taps". Seems to be related to FIR as well.

So I asked an AI engine -
"MiniDSP’s basic 48 kHz products were designed with a fixed internal clock and limited MIPS, so they deliberately cap sample rate to keep enough taps and features available at 48 kHz rather than offering higher rates with dramatically reduced processing headroom".

So yeah they cooked in a hard stop, as suspected they skimped a bit to conserve processor cycles and memory. I'm sure it can be done - feed 192kHz and get 192kHz out but cost prohibitive. Look I get it that 48kHz is good enough for most folks (me included) as we can't really hear the difference. I was mostly curious why miniDSP chose the design path....now I know.

 

[anphex]

FWIW, no handshake issues b/w my 3800 and a 65" G5.

The 3800 is fixed in this regard, only the 3700x had this rushed chipset. So bad that the first batch of these AVR back then needed a special connector that Denon provided for free.

 

[Harryharryharry]

Where have you seen this? Asking as the manual explicitly refers to an ethernet connection and doesn't mention WiFi?

Power and control - miniDSP Tide16 User Manual

docs.minidsp.com

From memory I think the manual said laptop with DC on can control the tide over WiFi/local network rather than a usb cable. Could have misread though

 

[john61ct]

So I had no idea what you were referring to with "taps". Seems to be related to FIR as well.

Yes.

@Keith_W is a great learning resource here, https://www.audiosciencereview.com/...asr-acourate-users.47057/page-10#post-2566504

will at least give you good keywords for searching to help you climb the learning curve

 

[Ultrasonic]

Really the biggest gripe I have with my MiniDSP Flex Digital is the sampling rate of 48kHz. Looks like this one falls in line as well - pretty disappointing, is it that they cannot find a processor(s) to handle the processing? Can someone explain this to me?

The SHD Studio already exists as an all-digital unit that uses a 96 kHz internal sample rate (and pre-dates the Flex digital).

As others have said though it's not obviois you're really missing out.

 

[NTK]

This is what miniDSP says.

Why is Tide16 only 48k? Why not 192kHz?

As much as we’d love to crank the sample rate as high as possible, it’s just not that simple. Most decoding and upmixing stages run at 48 kHz because that’s what the decoding libraries expect. It’s also the standard rate used by studios when th...

support.minidsp.com

 

[welwynnick]

I for one would love to see AVPs die.
Or rather, the proprietary-algorithm format decoding moved into specialist preamp units that DSP ONLY that function, let them plan churned obsolescence on that unit only, make it as cheap as possible, even a loss leader.
The systemic DSP functions - per-speaker EQ, crossovers, phase / time tuning, all bass management, room compensation functions in a separate box designed to have technological longevity.
Accommodate all filter types, imported from REW, rePhase, Acourate, MSO, Audiolens, the VST world, LADSPA / ACDf / asoundrc formats...
Connect the two different type boxen via multiple ADAT - or something else just as "open" NOT proprietary.
Sell 8- 16- 24- etc channel count units, allow for coordination between them so users can transparently stack them.

I think we see things from a similar perspective. I used to be outraged by the 5 digit prices of the high end AVPs, and wondered if you could achieve the same thing with a big pile of individual hardware and software elements, properly integrated together. I concluded it was possible but difficult. The components would add up to about £4000.

This line of thinking led to two things:
Firstly, an audio system could be considered to have a layered architecture, and separating the functions could be more efficient than an integrated solution.
Secondly, those high end AVPs actually did do more than cheaper processors. The DRC filtering was better, and they could also correct phase and speaker errors.

An AV system typically has the following key processes. These can all be performed in one box, but they can also be broken out:

• Dolby / DTS decoding
• Immersive rendering
• Amplitude correction
• Phase correction
• Reconstruction Filtering
• D A Conversion
• Gain control
• Buffering
• Amplification

I figured that an ideal audio system architecture would have one layer and one unit for each process or set of similar processes, each connected by non-proprietary, industry-standard interfaces, and each unit would be the optimum performance/cost solution for each function. Each layer could be exchanged or upgraded without upsetting the architecture, which would look something like this, and have the following advantages:

I started out looking for a cheaper solution to the megabucks processors, but learned this:

• Expensive processors do sound better than cheaper processors
• Digital decoder outputs do allow better amplitude linearity
• Both amplitude-frequency and phase-frequency linearity need equalisation
• A good system needs an open, scalable, non-proprietary architecture
• Layered architectures work effectively with complex systems
• One layer - one function - one unit - one input - one output
• Every unit is high performance, moderate cost and available off the shelf
• Each layer interfaces using open standards and connections
• A good preamp can help avoid the volume control bottleneck
• A good preamp can help improve DAC and power amp performance
• Good performance is needed, across the whole envelope, across every layer

My preferred solution is separate layers for playback, decode, process, convert, levels and amplification, and has the following advantages:

• High processing power
• Relatively low cost
• Compact and quiet
• Easy upgrading route
• Open, flexible, scalable, non-proprietary architecture
• Physical separation of playback & processing & conversion & amplification
• Opportunity to use DSP for speaker cross-overs
• Opportunity to use any source
• Opportunity to use any interfaces
• Opportunity to use any hardware
• Opportunity to use any software
• Opportunity to use any speakers

The open, flexible, scalable architecture matters, because it allows you to start small, add channels, upgrade or downgrade, and mitigate obsolescence.
You can see what works for you, without any heavy commitment to a single proprietary solution, which is a leap of faith if the sonic benefits are in any doubt.

EDIT 1: I think our difference is that I plan to do the DSP in a PC or Mac, and you plan to use an RPi (about which I have got no further than keen curiosity).
EDIT 2: The Tide16 does all of the above (except I believe digital domain volume control) in one box - but should achieve similar cost & performance.

 

[peng]

Really the biggest gripe I have with my MiniDSP Flex Digital is the sampling rate of 48kHz. Looks like this one falls in line as well - pretty disappointing, is it that they cannot find a processor(s) to handle the processing? Can someone explain this to me?

I suspect it doesn't matter if a thousand of those "someone" would explain that to you, if you don't believe the explanations.

 

[peng]

I think miniDSP is just using the status quo that we won't hear any differences beyond 48 kHz to inform their AVP design.

There was some indication that miniDSP might up it to 96 kHz at some point, and I've the impressions that they would do so just to please believers like @lozoyad and there is nothing wrong with that, it should cost that much more to do it at 96 kHz (not sure about 192 khz) and if they could sell more units then I guess it could be a win win..

 

[peng]

I think we see things from a similar perspective. I used to be outraged by the 5 digit prices of the high end AVPs, and wondered if you could achieve the same thing with a big pile of individual hardware and software elements, properly integrated together. I concluded it was possible but difficult. The components would add up to about £4000.

This line of thinking led to two things:
Firstly, an audio system could be considered to have a layered architecture, and separating the functions could be more efficient than an integrated solution.
Secondly, those high end AVPs actually did do more than cheaper processors. The DRC filtering was better, and they could also correct phase and speaker errors.

An AV system typically has the following key processes. These can all be performed in one box, but they can also be broken out:

• Dolby / DTS decoding
• Immersive rendering
• Amplitude correction
• Phase correction
• Reconstruction Filtering
• D A Conversion
• Gain control
• Buffering
• Amplification

I figured that an ideal audio system architecture would have one layer and one unit for each process or set of similar processes, each connected by non-proprietary, industry-standard interfaces, and each unit would be the optimum performance/cost solution for each function. Each layer could be exchanged or upgraded without upsetting the architecture, which would look something like this, and have the following advantages:

I started out looking for a cheaper solution to the megabucks processors, but learned this:

• Expensive processors do sound better than cheaper processors
• Digital decoder outputs do allow better amplitude linearity
• Both amplitude-frequency and phase-frequency linearity need equalisation
• A good system needs an open, scalable, non-proprietary architecture
• Layered architectures work effectively with complex systems
• One layer - one function - one unit - one input - one output
• Every unit is high performance, moderate cost and available off the shelf
• Each layer interfaces using open standards and connections
• A good preamp can help avoid the volume control bottleneck
• A good preamp can help improve DAC and power amp performance
• Good performance is needed, across the whole envelope, across every layer

My preferred solution is separate layers for playback, decode, process, convert, levels and amplification, and has the following advantages:

View attachment 525617

• High processing power
• Relatively low cost
• Compact and quiet
• Easy upgrading route
• Open, flexible, scalable, non-proprietary architecture
• Physical separation of playback & processing & conversion & amplification
• Opportunity to use DSP for speaker cross-overs
• Opportunity to use any source
• Opportunity to use any interfaces
• Opportunity to use any hardware
• Opportunity to use any software
• Opportunity to use any speakers

The open, flexible, scalable architecture matters, because it allows you to start small, add channels, upgrade or downgrade, and mitigate obsolescence.
You can see what works for you, without any heavy commitment to a single proprietary solution, which is a leap of faith if the sonic benefits are in any doubt.

EDIT 1: I think our difference is that I plan to do the DSP in a PC or Mac, and you plan to use an RPi (about which I have got no further than keen curiosity).
EDIT 2: The Tide16 does all of the above (except I believe digital domain volume control) in one box - but should achieve similar cost & performance.

All these, made me think, since long time ago, that people often referred to AVP + Preamp/poweramp vs AVRs as separates vs AVRs when in fact neither are "separates". I guess there is no such thing as "separates" but if we follow the preamp/poweramp convention that I might consider ADC+DAC+DSE+preamp+poweramp as truly, but loosely/literally "separates".

 

[EWL5]

There was some indication that miniDSP might up it to 96 kHz at some point, and I've the impressions that they would do so just to please believers like @lozoyad and there is nothing wrong with that, it should cost that much more to do it at 96 kHz (not sure about 192 khz) and if they could sell more units then I guess it could be a win win..

The question is once the hardware is sufficient, what will happen to the processing time for Dirac? Is it linear, exponential, etc.? How big will these filters be and what storage space is required on the PC? Folks on ART right now are complaining about GB of data when they don't regularly clean-up files so are we talking about TB for 96 kHz and beyond?

It just doesn't become practical for most of us like a 100MP camera unless we're professionals!

 

[peng]

The question is once the hardware is sufficient, what will happen to the processing time for Dirac? Is it linear, exponential, etc.? How big will these filters be and what storage space is required on the PC? Folks on ART right now are complaining about GB of data when they don't regularly clean-up files so are we talking about TB for 96 kHz and beyond?

It just doesn't become practical for most of us like a 100MP camera unless we're professionals!

I would speculate that if they follow through that thought, they could make it conditional, such that 96 kHz or even 192 kHz are enabled/hopefully selectable or selected automatically per input, only when DL is disabled. That's of course only a half way approach but it would still likely please the two channel audiophiles who probably value higher sampling rated more so than room correction anyway. Again, that's for the believers, who just trust their ears/brains. To me, I wouldn't want to pay for such extra features/capabilities if I can't hear the differences anyway. Also, reality is, there are AVPs that don't down sample at all, have they been selling more units than the likes of Marantz, and others that do down sample? Still, I do hope miniDSP to go ahead with that on the Tide, as long as they don't increase the price.

 

[Ultrasonic]

From memory I think the manual said laptop with DC on can control the tide over WiFi/local network rather than a usb cable. Could have misread though

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.

 

[welwynnick]

The question is once the hardware is sufficient, what will happen to the processing time for Dirac? Is it linear, exponential, etc.? How big will these filters be and what storage space is required on the PC? Folks on ART right now are complaining about GB of data when they don't regularly clean-up files so are we talking about TB for 96 kHz and beyond?

I think the processing delay is more a function of the length of the filter rather than the sample rate.
Suppose you have a typical 1000 tap filter running on 48kHz audio data, that equate to about 21ms.
A 1000 tap filter running on 96kHz data equates to 10ms, but the filter will be shorter, so it won't have the same low frequency resolution.
To get the resolution back, the filter needs 2000 taps, which brings the delay back to 21ms, so the delay won't be any longer.
The downside is that those 2000 taps now have to be processed in half the time, so the processing power needed has increased four-fold.

(I think I might have missed out a factor of two somewhere in there - someone put me right - but it doesn't change the comparison.)

 

[Oddball]

I think the processing delay is more a function of the length of the filter rather than the sample rate.
Suppose you have a typical 1000 tap filter running on 48kHz audio data, that equate to about 21ms.
A 1000 tap filter running on 96kHz data equates to 10ms, but the filter will be shorter, so it won't have the same low frequency resolution.
To get the resolution back, the filter needs 2000 taps, which brings the delay back to 21ms, so the delay won't be any longer.
The downside is that those 2000 taps now have to be processed in half the time, so the processing power needed has increased four-fold.

(I think I might have missed out a factor of two somewhere in there - someone put me right - but it doesn't change the comparison.)

I think Dirac has bigger fish to fry than catering to the 96kHz crowd, or should we rather say a "bunch" to describe them in more proper words having in mind the scale.

Dirac's ART breakthrough is immense, but now they need to do more and better for Dirac Live. Improving the handover and response in Dirac Live region would be the priority for most users, at least IMO. While the best consumer grade product at this point, if they don't keep developing it might just slip behind. We are in the area of incredibly fast and disruptive technology transition.

 

[peng]

I think Dirac has bigger fish to fry than catering to the 96kHz crowd, or should we rather say a "bunch" to describe them in more proper words having in mind the scale.

If one browse through their website and read their comments on sampling rate, one might conclude that they really don't believe in the benefits of higher sampling rate than 48 kHz (they have enough PhDs in related fields to know the pros and cons lol) anyway so yes, they naturally would aim for and set their bait accordingly for the bigger fish, and fry those caught.

 

[john61ct]

Yes I agree with your separates argument, but not the final conclusion applied to long-term longevity.

The Tide16 does all of the above (except I believe digital domain volume control) in one box - but should achieve similar cost & performance

I like the Layers enumeration but each of us will have different principles and goals.

Other than speakers, a few hundred per component is my limit - no way could I spend many thousands, even on the whole system after years of hunting.

To me it would be silly to limit any one component to a single function. Loudness contouring is integrated with volume control, and IMO ad-hoc tone controls / PEQ, monitor loops, to me there's a reason "a preamp" bundles those. But source selection in my case may need to be a separate box.

For me, power amps and speakers are always separate.

Crossovers / bass management / speaker EQ must be kept separate from room compensation.

As you noted, I do not want my system to depend on a PC staying on, ideally just use for measurements and filter creation, and watching an occasional film.

If it turns out a RPI5 **simply cannot** do the necessary convolving I'll likely just scale back but maybe, if I find a cheap / silent / efficient enough on power miniPC...

 

[welwynnick]

Yes I agree with your separates argument, but not the final conclusion applied to long-term longevity.
Other than speakers, a few hundred per component is my limit - no way could I spend many thousands, even on the whole system after years of hunting.

It seemed like we were thinking along similar lines, and I thought you might be interested in what I'd been exploring for the last year. But maybe not.
I said $4000 because I was thinking in terms of a multi-channel audio interface like a MOTU 16A or Antelope Orion Studio, but there are lots of good options, like a used Lynx Aurora or this Apogee DA16X which would work fine.

To me it would be silly to limit any one component to a single function. .........
Crossovers / bass management / speaker EQ must be kept separate from room compensation.

I didn't understand that bit, it sounds like a contradiction. I think a PC could do both just fine.

If it turns out a RPI5 **simply cannot** do the necessary convolving I'll likely just scale back but maybe, if I find a cheap / silent / efficient enough on power miniPC...

I've read many examples of where people have used an RPi or NUC. Mdsimon2 is always worth reading, and he's very keen on using CamillaDSP:

RPi + CamillaDSP Tutorial

As of May 22, 2024, this tutorial has been migrated to GitHub to allow for better revision management and a more universal location, please see link below. https://github.com/mdsimon2/RPi-CamillaDSP In general the GitHub tutorial is the same as previous ASR tutorial but now covers the RPi5...

www.audiosciencereview.com

Did you see this piece where Archimago used Accurate Sound's Hang Loose Convolver Multichannel on low power Mini PC?

Accurate Sound's Hang Loose Convolver Multichannel on low power Mini PC. The disgraceful AQVOX ethernet switch. And some E1DA upcoming evaluations.

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

I'll probably duck out for now to get back on thread. Long live 96kHz.

 

[Steakhouse]

The explanation is nothing higher adds anything real, just higher costs for the sake of marketing to audiphools

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.
For a reason: The tweeters a designed so that their stiffness and weight lets them operate efficiently withing the range they need. Break-up modes are pushed outside just of 20KHz and filtered out. No need to compromise the range below 20KHz just to reproduce some high frequency noise of some recordings that no one will hear anyway.