DSP speaker/room correction

[dallasjustice]

Sandyk's comments are instructive because they represent an extreme viewpoint regarding DRC. His views happen to be anti-DRC in the extreme and are, therefore, unimportant to this discussion.

 

OTOH, I know of others that hold extreme viewpoints in the other direction with regard to DRC. I know some folks that believe that full frequency target based DRC using one omni-directional microphone at the seated position and computer software can solve all room and speaker non-linearity. They will show you charts, graphs and everything else to prove their point. That point of view is equally extreme and invalid, IMO.

[wgscott]

Yes, but Extremism in defense of MUSIC is no vice.

[dallasjustice]

Yes, but Extremism has been shipwrecked by the laughter of the gods.

 

Yes, but Extremism in defense of MUSIC is no vice.

[One and a half]

For those that have DSP and are happy or sad with it, could they report the typical CPU in % when the DSP software is working? If the CPU use is low, there's no reason why an Atom processor could still be used without causing hiccups.

 

If the CPU is working harder, the noise component (through the 0V and shield of a USB) would increase which eventually leaks out to a DAC. This noise may mask or make an error to the DSP if significant enough.

 

JRiver has provision for DSP correction by using a pointer file, would this be equivalent to say Word opening a macro and making something else work in the background?

 

Also thumbs up for a separate Forum on DRC.

[esldude]

I asked Chris if he would create the DSP forum in a private message earlier today. His reply began with:

 

"What else can I say other than consider it done."

 

So thanks to Chris for honoring our request. And to all of you who replied and voted showing interest.

 

He says it will be a few days and he will try and populate with existing related threads when it is opened.

 

So Hooray, there will be a DSP for speaker and room correction forum.

 

Thanks to all of you and Chris once again.

[targa02]

For those that have DSP and are happy or sad with it, could they report the typical CPU in % when the DSP software is working? If the CPU use is low, there's no reason why an Atom processor could still be used without causing hiccups.

 

If the CPU is working harder, the noise component (through the 0V and shield of a USB) would increase which eventually leaks out to a DAC. This noise may mask or make an error to the DSP if significant enough.

 

JRiver has provision for DSP correction by using a pointer file, would this be equivalent to say Word opening a macro and making something else work in the background?

 

Also thumbs up for a separate Forum on DRC.

 

I run Spatial Computer Room Correction/Pure Vinyl and I am very happy with the results.

 

MacMini:

2.4 Ghz Intel Core 2 duo

8GB 1067 Mhz DDR3

500GB hard drive

 

My typical CPU usage is 25%

[DigiPete]

I run Genelec speakers with DSP build in.

CPU usage is 0% even at 5.1 surround in 24/192 and full room compensation ;-)

[Hipper]

Thanks esldude for pushing this idea.

 

I look forward to it and hope we have interesting and learned discussions so that all our music listening we can benefit.

[Miska]

For those that have DSP and are happy or sad with it, could they report the typical CPU in % when the DSP software is working? If the CPU use is low, there's no reason why an Atom processor could still be used without causing hiccups.

 

It varies a lot depending on material, but from 20 - 75%.

 

If the CPU is working harder, the noise component (through the 0V and shield of a USB) would increase which eventually leaks out to a DAC. This noise may mask or make an error to the DSP if significant enough.

 

This won't happen, because DACs are connected to battery and linear PSU powered ARM machines and processed audio data is transmitted over network from the player machine. CPU load on ARM is well below 10% and all unnecessary hardware like GPU and mass storage interfaces are turned off. So the overall measured noise is around 10 dB less than from bit-perfect playback from PC...

[89reksal]

I'm running Audiolense on Atom based board (CAPSV2) with filters convolved by JRiver with no issues. This is for files from 16/44.1 up to 24/96. My dac has maximum input of 24/96. I'm very happy with the results. I couldn't imagine not using something like Audiolense.

 

I've run this same hardware configuration with Windows 7 Utlimate Pro, Windows 8 pro and now Windows Server 2012. I don't know what my cpu usage is for sure since now I'm running WS2012 in core mode with no gui so I'm unable to check. If I remember correctly from previous times when I noticed, I don't think it ever got much over 10%.

 

I also use the same computer for taking any measurements and for creating the filters.

 

Bill

[VandyMan]

I've been doing computer audio since around 1999. I bought my first USB DAC in 2001 (Stero-Link) and I got into DRC soon after. I've used a Tact 2.0 (now 2.0S) for close to a decade. As far as I'm concerned, DRC is here now and will only be getting better and cheaper. I'd be happy to discuss it in a sub-forum. However, I have no interest in debating if DRC works. For me, it is well beyond proven. Anyone who believes that different lossless file formats sound different from each other is probably not a good candidate for DRC. I have not interest in trying to convince luddites to use DRC. Life is too short.

[Boris75]

I've been doing computer audio since around 1999. I bought my first USB DAC in 2001 (Stero-Link) and I got into DRC soon after. I've used a Tact 2.0 (now 2.0S) for close to a decade. As far as I'm concerned, DRC is here now and will only be getting better and cheaper. I'd be happy to discuss it in a sub-forum. However, I have no interest in debating if DRC works. For me, it is well beyond proven. Anyone who believes that different lossless file formats sound different from each other is probably not a good candidate for DRC. I have not interest in trying to convince luddites to use DRC. Life is too short.

 

Many thanks for your post. You can count me among those who would benefit from you sharing your knowledge and experience of DRC. Indeed, as you surmised in your post, I believe that different lossless file formats sound the same. I already have DRC embedded in my subwoofer (Cabasse Santorin 30), but I use my sub only up to 70Hz, so I guess that still leaves me with a couple of octaves where room correction would make sense.

[VandyMan]

I'm happy to share my experience, but I have a lot to learn too! It is a complex subject.

[One and a half]

This won't happen, because DACs are connected to battery and linear PSU powered ARM machines and processed audio data is transmitted over network from the player machine. CPU load on ARM is well below 10% and all unnecessary hardware like GPU and mass storage interfaces are turned off. So the overall measured noise is around 10 dB less than from bit-perfect playback from PC...

 

Yuh, if you had a battery powered device, what about PC --> USB --> Mains powered DAC?

[Miska]

Yuh, if you had a battery powered device, what about PC --> USB --> Mains powered DAC?

 

That depends on the PC and DAC. Largely on design of the USB interface at both sides and also on PSU design of the PC. Most of the time I've been measuring my PC's, USB noise is not much dependent on CPU load but mostly coming straight from the PSU and also more on GPU and HDD loads. Those Atom machines where PSU is coming as single rail and on-board regulated are most sensitive for noise generation, ordinary multi-rail ATX PSU's are much less susceptible to load correlated noise.

 

My most used setup is:

Core i5 -> Network -> Battery powered ARM -> USB -> Mains powered DAC

Core i5 -> Network -> Linear PSU powered ARM -> USB powered hiFace DAC

 

I also have a setup like:

Core i7 -> Network -> Atom -> Firewire -> Mytek DAC

 

In general, you have to remember that CPU load percentage is only time-averaged figure (typically over period of one second) and doesn't give a good picture of the load patterns. Usually GUI operations generate worst power consumption load patterns because it loads both CPU and GPU simultaneously. Fancier the GUI, more this kind of noise you generate.

 

IMO, having a fast Core i5 or i7 CPU just running operating system's do-nothing-idle-loop is complete waste of money. So I rather use those CPU cycles for doing useful work, especially since the CPU is more capable than typical DSP in audio device.

 

And if you have a DAC or speakers with DSP processor, there you can find constant, typically very close to 100% processor loads and much closer to the sensitive DAC chip than a computer used as source.

[Miska]

In general, DRC makes such a huge difference that I cannot live without it and I definitely want to optimize rest of the system from the point of view that DRC is definitely in place.

 

It is also easy to compare by switching the correction in and out on the fly. Doing it properly requires some time and effort, but it is definitely worth it.

 

My main recommendation for DRC is that don't over do it. Keep it simple and gentle, and if possible at frequencies below 500 Hz (where the biggest room acoustic problems are).

[esldude]

That depends on the PC and DAC. Largely on design of the USB interface at both sides and also on PSU design of the PC. Most of the time I've been measuring my PC's, USB noise is not much dependent on CPU load but mostly coming straight from the PSU and also more on GPU and HDD loads. Those Atom machines where PSU is coming as single rail and on-board regulated are most sensitive for noise generation, ordinary multi-rail ATX PSU's are much less susceptible to load correlated noise.

 

My most used setup is:

Core i5 -> Network -> Battery powered ARM -> USB -> Mains powered DAC

Core i5 -> Network -> Linear PSU powered ARM -> USB powered hiFace DAC

 

I also have a setup like:

Core i7 -> Network -> Atom -> Firewire -> Mytek DAC

 

In general, you have to remember that CPU load percentage is only time-averaged figure (typically over period of one second) and doesn't give a good picture of the load patterns. Usually GUI operations generate worst power consumption load patterns because it loads both CPU and GPU simultaneously. Fancier the GUI, more this kind of noise you generate.

 

IMO, having a fast Core i5 or i7 CPU just running operating system's do-nothing-idle-loop is complete waste of money. So I rather use those CPU cycles for doing useful work, especially since the CPU is more capable than typical DSP in audio device.

 

And if you have a DAC or speakers with DSP processor, there you can find constant, typically very close to 100% processor loads and much closer to the sensitive DAC chip than a computer used as source.

 

Would not all of these concerns be good reason to do the DSP on the music files and store the result as another file? Then use the processed version for playback. Then all you have is a basic music player working with little stress. The processing would have already been done ahead of time.

[89reksal]

Would not all of these concerns be good reason to do the DSP on the music files and store the result as another file? Then use the processed version for playback. Then all you have is a basic music player working with little stress. The processing would have already been done ahead of time.

 

This may be a good approach once you have settled on a filter(s) that you're sure you'll be able to live with over the long haul. Based on my own experience with Audiolense, it can take experimenting with many many (100's) different filters before ending up with one you're satisfied with over a longer period.

 

Once you're exposed to the possibilities of almost infinite tweaking, it can be a long road of experimentation and learning before achieving the true potential of this approach. More often than not, I create a filter, listen and think this is the best my setup has ever sounded. As I get used to what I'm hearing, I find I begin thinking "If this sounds so good, maybe I can get it just a little bit better". And then I try another filter. After about 18 months, I'm at the point where I've used the same filter for over a month. That's the longest I've used the same filter.

 

To me, it's an incredible thing, to have the ability to almost infinitely tweak my system sound at no cost other than the initial expense of the software and measuring setup. As long as I'm willing to spend the time, I can experiment and fine tune to my heart's content.

 

Even taking a correct measurement is unlikely to be done the first time or even first few times, which would result in creation of a new filter each time.

 

If I was doing "the DSP on the music files and store the result as another file", I would be spending a huge amount of time on that, only to have to repeat the process all over again when I created my next "best ever" filter.

 

Bill

[esldude]

lasker98,

 

Sounds like you have done similar to what I have. Spent lots of time tweaking curves. And measuring and learning. I did in time settle down to a narrow range. Doing curves for friends to fit their preference, and observing myself over time I found I might tweak the curve to optimize recently listened to recordings. I guess one could even try and do a curve for each recording if one has no other things to do.

 

Eventually I picked a few of my better curves and lived with it. The equipment I have can store up to 9 correction curves that can be switched between instantly. I use about 5 with slightly different tapers for different quality recordings. Mainly however, I use one for most listening and one special one for low level background music.

 

In the end it is something like picking speakers. You find one you like pretty well with most every recording. It isn't perfect, but you can enjoy it for nearly everything. With room correction it is much closer to perfect, and I do have a bit of instant fine tuning. Once I got to that point, I have been able to just leave it be and simply enjoy music. If I get to thinking the correction isn't important I just hit the bypass button for straight through, and the subjectively large drop in quality tells me the correction is providing plenty of musical benefit.

 

Some times too much adjustment is the road to insanity.

 

I think too for room correction to be more mainstream it needs to be highly automated. Able to pick very good corrective curves right away. Good measurements may be a bit more involved for a lot longer. But if done right you can use it as long as you don't change speaker position or speakers. Mitchco's reporting on Acourate indicates it may be near that already.

[Guidof]

 

I think too for room correction to be more mainstream it needs to be highly automated. Able to pick very good corrective curves right away. Good measurements may be a bit more involved for a lot longer. But if done right you can use it as long as you don't change speaker position or speakers. Mitchco's reporting on Acourate indicates it may be near that already.

 

I tend to agree, based on my experience with the DSPeaker AntiMode Dual Core, which does automatic DRC in the low frequency range -- and does so quite effectively. It also provides the ability to instantly switch to by-pass, thus enabling immediate comparison of corrected and uncorrected sound.

 

Guido F.

[firedog]

One of the reasons that I haven't gone back to using DRC (I used to own a TACT DRC preamp) it that I'm afraid I'll start endlessly tweaking. I just want to get a good RC done and leave it alone; but my compulsive audiophile side won't let leave well enough alone.

[esldude]

One of the reasons that I haven't gone back to using DRC (I used to own a TACT DRC preamp) it that I'm afraid I'll start endlessly tweaking. I just want to get a good RC done and leave it alone; but my compulsive audiophile side won't let leave well enough alone.

 

I feel your pain firedog. On the other hand, it sure is cheaper and more productive than sampling various cable options. Once you get past the upfront cost, it costs nothing to tweak things as long as or as often as you wish. I at one time became vere involved in two or three octave boosts or cuts of .5 db. It could alter the apparent perspective of most recordings quite noticeably. Change the space, the depth etc. etc.

[firedog]

I feel your pain firedog. On the other hand, it sure is cheaper and more productive than sampling various cable options. Once you get past the upfront cost, it costs nothing to tweak things as long as or as often as you wish. I at one time became vere involved in two or three octave boosts or cuts of .5 db. It could alter the apparent perspective of most recordings quite noticeably. Change the space, the depth etc. etc.

 

Yeah, I gave up on comparing cables, too. Just have a few pairs I think are pretty good and I use them as necessary. Just can't drive myself crazy with that stuff. Otherwise, I end up having a hobby of testing audio accessories instead of a hobby of enjoying listening to well produced music.

[VandyMan]

I was able to end my habit of constantly tweaking my Tact's target curves by getting an XTZ measurement system. It allows me to see the real world effect of the curves. It is also really handy for setting the EQ dials on my speakers. Measuring a measurement is one of the main features lacking from the Tact (and most other DRC systems).

[Miska]

Would not all of these concerns be good reason to do the DSP on the music files and store the result as another file? Then use the processed version for playback. Then all you have is a basic music player working with little stress. The processing would have already been done ahead of time.

 

No, because:

1) I would need to have three versions of the files for three different systems

2) Any time I change something I would need to re-process lot of files

3) There are too many combinations of oversampling, delta-sigma modulation and DRC settings to be pre-processed in meaningful way

4) I'm not short of CPU time, whenever CPU is playing back with "little stress" it is running operating system's do-nothing-idle-loop anyway (generating heat while doing nothing useful)

5) In this setup there's no difference in output quality regardless how hard the player computer's CPU is working