The best way possible to build an active system.

[Blizzard]

Here's some interesting FAQ's on the Merging NADAC and HAPI:


https://confluence.merging.com/plugins/servlet/mobile#content/view/9338933

 

[Rodney Gold]

Much cheaper option but no "best of the best" parts is the miniDSP 4x10hd can handle 3 way actives and subs

Digital and analog in , digital and 8 chans (l/r = 2 chans) output (So it has an ADC and a DAC)
x over , limiting , delay , 12 chans PEQ . 24/96 , sharc , infinite mapping per channel
all for $499... Add a raft of power amps and you are done.
No , its not SOTA , but im pretty sure it wont be that far behind
I actually had a set of meridian DSP6000's whose DSP had died , I replaced the drivers with much better ones and used the 4x10 to replace the meridian DSP... after much fiddling and a ton of measurements I did get it to sound good , but boy o boy , was it a mission...

 

[Blizzard]

Much cheaper option but no "best of the best" parts is the miniDSP 4x10hd can handle 3 way actives and subs

Digital and analog in , digital and 8 chans (l/r = 2 chans) output (So it has an ADC and a DAC)
x over , limiting , delay , 12 chans PEQ . 24/96 , sharc , infinite mapping per channel
all for $599... Add a raft of power amps and you are done.

Yes those components have their place at the price level they are at decent value. A great experiment to try with their flagship product the "Minisharc" is to put it in unity bypass mode and just do A/B testing with it in and out of the chain in a 2 channel passive system. You may decide to take advantage of their return policy after this experiment. That's when you find out all sales are final.

 

[Rodney Gold]

I have done that , digital in to the sharc boxen (openDRC Di and my DDRC22 ) from a Squeezebox touch to my Devialet digital in, and to compare , optical out of my Squeezebox to the amp directly...
Sounded more or less the same to me when the box was in bypass.. what was I supposed to hear?

 

[Blizzard]

I have done that , digital in to the sharc boxen (openDRC Di and my DDRC22 ) from a Squeezebox touch to my Devialet digital in, and to compare , optical out of my Squeezebox to the amp directly...
Sounded more or less the same to me when the box was in bypass.. what was I supposed to hear?

A lot of people say there's a haze and genericness to the sound. But maybe you can't notice it because your comparing it to the optical out from the squeezebox.

 

[Groucho]

It is relevant if it involves putting extra sound degrading components in the signal chain. Also part of the hardware is the DAC's, A/D, I/V stages, power supplies, low phase noise clocks, board layouts, etc.

I was talking specifically about the DSP, where it matters not one jot if the calculations are done on a PC or dedicated DSP box (or boxes) - or an iPhone. However, the capacity for FLOPS is an issue. The more FLOPS the unit can perform, the larger the filters you can use, which has some implications for accuracy of response.

 

[Rodney Gold]

A lot of people say there's a haze and genericness to the sound. But maybe you can't notice it because your comparing it to the optical out from the squeezebox.

I have compared optical to spdif out with the SB touch and they sound identical to me.

I will try another way .. cue up 2 squeezeboxes (I have a few) , sych them , one plays thru the DSP box , the other into the amp .. both would use SPDIF...

 

[Blizzard]

I was talking specifically about the DSP, where it matters not one jot if the calculations are done on a PC or dedicated DSP box (or boxes) - or an iPhone. However, the capacity for FLOPS is an issue. The more FLOPS the unit can perform, the larger the filters you can use, which has some implications for accuracy of response.

But the problem is finding a dedicated DSP box solution that has the same grade of components as a high end stand alone DAC such as the NADAC. I don't know of any DSP chips with power like Intel processors. Also using Moore's law, Intel chip power grows much faster than the SHARC chips. I've talked to a lot of pros about this. I'm going to be using a real time operating system running on 12 core Xeon processors to handle my front end. SHARC chips might be at this level by 2025 if we are lucky.

 

[Blizzard]

I have compared optical to spdif out with the SB touch and they sound identical to me.

I will try another way .. cue up 2 squeezeboxes (I have a few) , sych them , one plays thru the DSP box , the other into the amp .. both would use SPDIF...

The guys I'm talking about were just using the raw MiniSHARC boards powered by ultra noise linear supplies connected via I2S. This is the best possible way to pass the audio through these boards.They were comparing it with just the USB-I2S (Amanero) connected direct.

The miniSHARC SPDIF-AES/EBU/Toslink boards (digi-fp-io) only just add more noise and jitter.

 

[Groucho]

I don't think this anymore. I totally agree that to fully realize the ultimate potential for a digitally active system, one needs a speaker designed from the ground up to be digitally active. The reason for this is the measurement process. IMO, the best measurements can only be taken in very large quasi-anechoic spaces or real anechoic spaces.

An excellent example would be a speaker like the JBL M2.

I agree that a digitally-active speaker can, or should, be designed from the ground up, because it gives the designer more freedom in the choice of which, and how many, drivers to use. More 'ways' means that dispersion is much more controllable, the job of each driver is easier; doppler distortion is reduced; the job of each amplifier is easier and so on. As such, the need for sophisticated measurements in anechoic chambers is actually reduced: you would only have acoustic problems if you were trying to extract quality from a beaming mid-woofer handing over to a tweeter, with both drivers in 'break up'.

Measurements are yesterday's technology, anyway. The modern way of doing things is to simulate the system in a computer. If you know what you want (but do people really know what they want? I am not convinced), you can design your system entirely by simulation - particularly if you know that the drivers are not being stressed and won't be running in break up modes.

The JBL's achille's heel is that it is ported - a defect that cannot be corrected even with DSP.

 

[Groucho]

But the problem is finding a dedicated DSP box solution that has the same grade of components as a high end stand alone DAC such as the NADAC. I don't know of any DSP chips with power like Intel processors. Also using Moore's law, Intel chip power grows much faster than the SHARC chips. I've talked to a lot of pros about this. I'm going to be using a real time operating system running on 12 core Xeon processors to handle my front end. SHARC chips might be at this level by 2025 if we are lucky.

I run my system on Linux and a small, fanless Intel-based PC - yes the power of general purpose Intel PC chips is astounding. My speakers are three way, and I find 32 bit floating point maths and huge filters are comfortably handled (an unfortunate consequence of the large filters is latency - about a second). I haven't tried 64 bit maths, but I'm fairly sure I could convert the code if I thought it would make much difference. However, I think I'd rather have a slightly cooler PC.

 

[Blizzard]

I run my system on Linux and a small, fanless Intel-based PC. My speakers are three way, and I find 32 bit floating point maths and huge filters are comfortably handled (an unfortunate consequence of the large filters is latency - about a second). I haven't tried 64 bit maths, but I'm fairly sure I could convert the code if I thought it would make much difference. However, I think I'd rather have a slightly cooler PC.

All depends on what kind of algorithms you are processing. 64 bit floating point, and this kind of horsepower is required for upsampling 8 channels to DSD 256, and applying the FIR and IIR filtering for Xovers and room correction. 4 cores are going to run the real time OS, and each channel will have its own dedicated core for the DSP. And latency is around 1ms with the real time OS.

 

[Groucho]

All depends on what kind of algorithms you are processing. 64 bit floating point, and this kind of horsepower is required for upsampling 8 channels to DSD 256, and applying the FIR and IIR filtering for Xovers and room correction. 4 cores are going to run the real time OS, and each channel will have its own dedicated core for the DSP. And latency is around 1ms with the real time OS.

Is it worth upsampling from PCM to DSD? If DSD has any merit (I'm not convinced!) has it not been lost as soon as you convert (and process) it as PCM? Starting with PCM as a source, a DSD-ised version won't be true 'DSD'.

The latency of the raw hardware may be 1ms, but if you are going to be using FIR linear phase filters or similar to flatten the phase (this, I would suggest, is the outstanding benefit of this technology), you will need to allow latency which scales with the size of the filters you are using. I don't find it a problem at all, but if I were using my system with video I would need to delay the video to match.

 

[microstrip]

(...) The best speaker I have heard (Lotus Granada) was active but used pretty basic DSP (parametric EQ, delays, etc) and A/D, D/A converters. I guess the message is that the digital side is much lower priority than the electro-acoustical speaker engineering part.

Are you referring to the speakers using the ultra-expensive artisanally built Feastrex drivers? A good friend of mine dreams about building a speaker around these drivers!

 

[Tam Lin]

I'm going to be using a real time operating system running on 12 core Xeon processors to handle my front end.

4 cores are going to run the real time OS, and each channel will have its own dedicated core for the DSP. And latency is around 1ms with the real time OS.

That's absurd. 1ms latancy? And since when does a real time OS require 1/3 of the available CPU capacity?

In my experience, you get better performance when you let the OS manage its resources dynamically rather then trying to out-guess it and manually assign specific processes to specific cores.

 

[Blizzard]

Is it worth upsampling from PCM to DSD? If DSD has any merit (I'm not convinced!) has it not been lost as soon as you convert (and process) it as PCM? Starting with PCM as a source, a DSD-ised version won't be true 'DSD'.

The latency of the raw hardware may be 1ms, but if you are going to be using FIR linear phase filters or similar to flatten the phase (this, I would suggest, is the outstanding benefit of this technology), you will need to allow latency which scales with the size of the filters you are using. I don't find it a problem at all, but if I were using my system with video I would need to delay the video to match.

Most modern DAC's convert PCM to SDM internally within the chip anyways. By doing it on the computer end, much better algorithms can be used. Look into HQplayer and why everyone is raving about it.

As for DSD in general, I find that it takes sitting down and listening to well recorded DSD through a good DAC to appreciate DSD. If just reading third party opinions about it, it's not so impressive.

 

[Blizzard]

That's absurd. 1ms latancy? And since when does a real time OS require 1/3 of the available CPU capacity?

In my experience, you get better performance when you let the OS manage its resources dynamically rather then trying to out-guess it and manually assign specific processes to specific cores.

1ms latency isn't absurd. Only 1 core is required, but I'll only need 8 cores for the DSP, so the other 4 can go to the OS. The guys building this system know what their doing.

Check out Merging Masscore for an example of low latency DSP software running on realtime OS. They claim 1.33ms round trip latency. This is going from mic to amp. latency isn't an issue with an audio playback system anyways. Every DAC with an async USB input creates latency on purpose.

http://www.merging.com/products/ovation/masscore

 

[dallasjustice]

The JBL's achille's heel is that it is ported - a defect that cannot be corrected even with DSP.

If that's the only defect, I'd be a happy camper. I would never run them full range anyway.

 

[Blizzard]

If that's the only defect, I'd be a happy camper. I would never run them full range anyway.

DJ, what are you currently running for a total system? Did you actually go fully active yet?

 

[Nyal Mellor]

Are you referring to the speakers using the ultra-expensive artisanally built Feastrex drivers? A good friend of mine dreams about building a speaker around these drivers!

Yes exactly.