Active Room Treatment (ART) by Dirac

[No_hair_left]

Also last question, am I supposed to use dirac first and then move down the list until I get to art as I just went straight and clicked on art, maybe that is what I am doing wrong, I couldn't even find out what my crossovers were.

 

[smygolf]

My lcr are arendal 1961 monitors so they are all smaller really.

So It would look something like this then?
Increase the sub support a little
Leave the 2 atmos groups all unchecked at their 120hz crossover
Leave the surround channels unchecked at their 80hz crossover
Have l/r at their 80hz crossover support all channels, including atmos? center is a possibility.

The group 6 box for me is the subs I'm assuming I uncheck that completely as I don't want speakers trying to support my subs.

The idea with Art is to use as many of the other speakers to even out the decay.
So LCR does not ad any volume as such.

So let every speaker support support everything with the range you mention.
And use the level Dirac suggest - prob -18dB
Dirac might have 50hz as default, that will prob be a bit low for 1961 Monitor.

But do not atmos give any support or within the group.
Let surround give support but not within the group.

When using ART you can choose it after the sweeps
Dirac
DLBS
ART uses the same sweep but it is different calculations so your good.

*edit*
If you hear that you are pushing LCR to hard change the lvl to -6dB in the sub group.

 

[No_hair_left]

The idea with Art is to use as many of the other speakers to even out the decay.
So LCR does not ad any volume as such.

So let every speaker support support everything with the range you mention.
And use the level Dirac suggest - prob -18dB
Dirac might have 50hz as default, that will prob be a bit low for 1961 Monitor.

But do not atmos give any support or within the group.
Let surround give support but not within the group.

When using ART you can choose it after the sweeps
Dirac
DLBS
ART uses the same sweep but it is different calculations so your good.

*edit*
If you hear that you are pushing LCR to hard change the lvl to -6dB in the sub group.

Right I'll give it another shot, thanks.

I won't be pushing anything too hard though as my listening volume never gets past -30, it's plenty loud and I have neighbours to not annoy.

 

[smygolf]

Right I'll give it another shot, thanks.

I won't be pushing anything too hard though as my listening volume never gets past -30, it's plenty loud and I have neighbours to not annoy.

And do not be afraid to have target that is +10dB in the lower bass

 

[mmares056]

Im completely lost with the grouping, Im assuming I leave the group 6 subs alone and uncheck everything?

In my front l/r (group 1) I have checked the Group 3 box (side surrounds) and the group 6 box(subs) but seeing as the surround are much further back am I supposed to do this?

Or is it really a case of the separate groups support nothing but the subs?

and the atmos just support each other?

I am lost

Also lots of talk about dirac making bass better but in my case with the gravity test the heartbeat in the beginning seems incredibly muted, definitely have more to learn it seems, I have probably watched about 4 hours of videos and still dont understand lol

Perhaps your bass rise target curve is not where it needs to be for your room. I use the same target curve for all my speakers below the transition region, which is the natural bass rise in my room and my +10dB to 24Hz boosted bass sounds great. Bass is impactful, tight, and sounds natural.

 

[Oddball]

Directional Bass is also commented on here at 37:20

Really? I just heard a blurb about "super directional bass" without any concrete pointers. Luckily, we also know ART as well and can help ourselves if needed .

BTW first time I heard about "super directional bass". Now makes my wonder if that is lame and what "hyper directional bass" would be and how to get there?

 

[No_hair_left]

And do not be afraid to have target that is +10dB in the lower bass

Perhaps your bass rise target curve is not where it needs to be for your room. I use the same target curve for all my speakers below the transition region, which is the natural bass rise in my room and my +10dB to 24Hz boosted bass sounds great. Bass is impactful, tight, and sounds natural.

This seems about right for the bass, with xt32 I would always calibrate it so my subs were at -10db then boost it up 10db to get the bass back.

 

[ban25]

Also last question, am I supposed to use dirac first and then move down the list until I get to art as I just went straight and clicked on art, maybe that is what I am doing wrong, I couldn't even find out what my crossovers were.

Just select ART. You only need to select the others if you wish to upload a calibration without ART or without DLBC.

 

[JeffH]

Well compression is just inability of the speaker to produce more output/energy. ART is generally not counting on these circumstances as it will adjust its filters by the sweep that is not at high SPL. Same with distortion - it will not measure distortion, just the FQ response capabilities. So both of these have to be extrapolated into building filters and determining support low levels if the desired SPL is in excess of the sweeps.

I think you nailed it as to the point where you start driving speakers too hard. It is exactly as you note - at least in my experience.

In my exp

Agreed - I wish we had more speaker reviews with compression tests. Solid subs reviews provide that.

I am still worried about distortion at high SPL though, even in the low end. What we have been told is that around/below 20hz distortion does not really matter. Assuming that is right, it still remains a question how much THD above that is audible. I started with 2 subs in my room and they were struggling. SPL was probably almost there, but it was not clean or controlled. 3rd sub would probably do the job, but then decided to go with 4 and do a partial DBA to clean up some decay as well (all before ART). 4 of them are really doing a great job and headroom is evident as they are set between 50-70% output, depending on distance.

Some screenshots from review of SVS PB 4000 to illustrate above. I have 2 SVS PC 4000 which are supposed to be the same in a different form factor. Other two subs are Arendals 1723 2V which are, generally, similar to these.

View attachment 521865View attachment 521866View attachment 521867

EDIT: Link to full review of SVS PB 4000

SVS PB-4000 Ported Subwoofer Review

The SVS PB-4000 is the latest evolution of a powerful subwoofer design that has stood the test of time. Read our in-depth review with measurements to see how SVS improved on their classic design.

www.audioholics.com

In my experience dynamic compression is a bigger issue than distortion, unless the distortion is primarily odd order and over maybe 15-20% or so. The reason is because your room is contributing a lot of distortion, because unless it's a concrete bunker all the walls etc vibrate in non linear ways. Try and do a far field distortion test on a sub playing in a normal room at a reasonable level. Of course mechanical noises and amplifier distortion, especially Class D, are another matter.

 

[JeffH]

In my exp

In my experience dynamic compression is a bigger issue than distortion, unless the distortion is primarily odd order and over maybe 15-20% or so. The reason is because your room is contributing a lot of distortion, because unless it's a concrete bunker all the walls etc vibrate in non linear ways. Try and do a far field distortion test on a sub playing in a normal room at a reasonable level. Of course mechanical noises and amplifier distortion, especially Class D, are another matter.

I'll add that if you think about how much more radiating area the walls have vs even four large woofers you can visualize how little they need to vibrate to produce significant output, and it's never linear.

 

[No_hair_left]

After a fair bit of testing today it seems if I let all my other channels except the centre support my atmos channels it sounds a fair bit better, everyone seems to recommend not doing this, is there a reason why?

 

[Oddball]

After a fair bit of testing today it seems if I let all my other channels except the centre support my atmos channels it sounds a fair bit better, everyone seems to recommend not doing this, is there a reason why?

First, it is difficult to provide general recommendations that are as specific as this one. Atmos channels don't get much love as they carry important but for the most part not so essential role. Most of the soundtrack action happens at bed channel levels.

If this sounds good and measures good, then by all means embrace it. I personally don't really need much support for Atmos as they can support themselves in pairs quite well, with the sub support of course. It really depends what you have as a starting point, where you want to get to, and if you have filters available.

 

[seashell]

As someone that has used ART since it was in beta on the Storm, and believe it works wonders in a variety of setups and spaces... how certain do we know what it is doing?
A similar discussion was on AVS as well but I thought it was interesting enough to see what users think/know here?
Is it using active cancellation, like in headphones?
Is it using other speakers to contribute to the target curve?
Their early documentation seemed to talk quite a lot about active cancellation. It seems many of their later videos in particular don't.
This isn't really a question of if ART works, but do we know how it works and if it is using Active Cancellation or not? Is this possible to measure and has anyone done that which might prove one way or the other if it is using active cancellation?

I do not work for Dirac, and have no insider knowledge of the ART algorithms, but I'm fairly certains ART is a spatial equalizer exploiting transmit diversity.

If you want to understand what ART is doing I suggest you look up "spatial multiplexing" and its use in wireless communications. That will introduce you to the concepts of spatial diversity and MIMO (Multiple Input Multiple Output) communications.

ART works on input audio channels like Left, Right, Center, etc. not on speakers as stated by Dirac in various youtube videos. I believe for each input channel it is a single input, multiple output spatial equalizer exploiting transmit diversity. When you consider that speakers will support multiple channels the system as a whole is MIMO.

If you get a basic understanding of these concepts it will make sense on why in their videos Dirac suggests using any speakers that can play low frequencies to support all the other speakers. The more speakers the greater the spatial diversity and the more likely you will have a path that provides unique information that can be used to fill in nulls and shorten the reverberation time.

This also where the name of the Fsiso parameter comes from (partially explained in a video). SISO = Single Input Single Output which is the traditional form of room correction used by Dirac Live. The signal for a particular channel (like Left) is sent to and played from a single speaker. F_SISO is the frequency ART (MIMO) hands off to Dirac Live (SISO).

I believe this also part of why Dirac is cagey around explicitly stating they are doing active cancellation. It's not that they calculate the whole room layout and say geometrically if we play this frequency at this phase from this speaker it will cancel the reflection near that speaker and thus reduce reverberation. They clearly don't have that information. They don't know the location of any speakers and they explicitly state in youtube videos that there is no need to match the locations on the position diagram to the position of the microphone so they're not even relying on relative position of the measurements. And this is how a spatial equalizer works. It takes a set of signals that propagated through different spatial paths and solves an equation that creates a set of filters that can be applied to each received signal such that the sum of the filtered signals best matches a desired target output (typically the desired target is the original signal as if it was propagated through a single direct path with no reflections). It doesn't matter what the actuals spatial paths were, just that the paths were different and thus provide additional information that can be combined to better recover the desired target output.

The desired target and the limits on support speakers is to me the proprietary special sauce of ART. They state in the videos that they started with thousands of robotically measured positions and worked backwards from there. So I assume they started with a highly spatially oversampled grid of data AND because of the automated positioning they also had the full geometric information. Using that they figured out how using only delay to the given channel's primary speaker (that's what they can measure to link measurements with unknown geometry) what the target across the positions should be. Then they added in limitations like localization, frequency limits, power handling, spectral smoothing (the videos state the support parameter is not about power but about specifity so I think of it like changing smoothing parameters in REW plots) as constraints on the filters when they do the solve for the set of filters that best fits their desired target output. (The use of delay to link positions is also why they have a reference channel for LFE. A narrow bandwidth speaker like a subwoofer will not provide good delay estimates because delay accuracy is inversely proportional to bandwidth so they need a full bandwidth signal to accurately get relative delay between positions.)

Obviously there are differences in the applications of wireless communications and audio room correction, but given spatial multiplexing is common and fully out in the open in the wireless communications world I think it's a great place for people to find articles and videos explaining the concepts.

 

[90sNintendoKid]

I do not work for Dirac, and have no insider knowledge of the ART algorithms, but I'm fairly certains ART is a spatial equalizer exploiting transmit diversity.

If you want to understand what ART is doing I suggest you look up "spatial multiplexing" and its use in wireless communications. That will introduce you to the concepts of spatial diversity and MIMO (Multiple Input Multiple Output) communications.

ART works on input audio channels like Left, Right, Center, etc. not on speakers as stated by Dirac in various youtube videos. I believe for each input channel it is a single input, multiple output spatial equalizer exploiting transmit diversity. When you consider that speakers will support multiple channels the system as a whole is MIMO.

If you get a basic understanding of these concepts it will make sense on why in their videos Dirac suggests using any speakers that can play low frequencies to support all the other speakers. The more speakers the greater the spatial diversity and the more likely you will have a path that provides unique information that can be used to fill in nulls and shorten the reverberation time.

This also where the name of the Fsiso parameter comes from (partially explained in a video). SISO = Single Input Single Output which is the traditional form of room correction used by Dirac Live. The signal for a particular channel (like Left) is sent to and played from a single speaker. F_SISO is the frequency ART (MIMO) hands off to Dirac Live (SISO).

I believe this also part of why Dirac is cagey around explicitly stating they are doing active cancellation. It's not that they calculate the whole room layout and say geometrically if we play this frequency at this phase from this speaker it will cancel the reflection near that speaker and thus reduce reverberation. They clearly don't have that information. They don't know the location of any speakers and they explicitly state in youtube videos that there is no need to match the locations on the position diagram to the position of the microphone so they're not even relying on relative position of the measurements. And this is how a spatial equalizer works. It takes a set of signals that propagated through different spatial paths and solves an equation that creates a set of filters that can be applied to each received signal such that the sum of the filtered signals best matches a desired target output (typically the desired target is the original signal as if it was propagated through a single direct path with no reflections). It doesn't matter what the actuals spatial paths were, just that the paths were different and thus provide additional information that can be combined to better recover the desired target output.

The desired target and the limits on support speakers is to me the proprietary special sauce of ART. They state in the videos that they started with thousands of robotically measured positions and worked backwards from there. So I assume they started with a highly spatially oversampled grid of data AND because of the automated positioning they also had the full geometric information. Using that they figured out how using only delay to the given channel's primary speaker (that's what they can measure to link measurements with unknown geometry) what the target across the positions should be. Then they added in limitations like localization, frequency limits, power handling, spectral smoothing (the videos state the support parameter is not about power but about specifity so I think of it like changing smoothing parameters in REW plots) as constraints on the filters when they do the solve for the set of filters that best fits their desired target output. (The use of delay to link positions is also why they have a reference channel for LFE. A narrow bandwidth speaker like a subwoofer will not provide good delay estimates because delay accuracy is inversely proportional to bandwidth so they need a full bandwidth signal to accurately get relative delay between positions.)

Obviously there are differences in the applications of wireless communications and audio room correction, but given spatial multiplexing is common and fully out in the open in the wireless communications world I think it's a great place for people to find articles and videos explaining the concepts.

To your mentioning of Dirac suggesting speakers that can play low, once I’ve got my system properly setup, I may play about with integrating subwoofers to specific speakers that may need it but Dirac sees one speaker.

I’ve got my eye on the new iloud subwoofer or a Buchardt Sub 10.

 

[Oddball]

I do not work for Dirac, and have no insider knowledge of the ART algorithms, but I'm fairly certains ART is a spatial equalizer exploiting transmit diversity.

If you want to understand what ART is doing I suggest you look up "spatial multiplexing" and its use in wireless communications. That will introduce you to the concepts of spatial diversity and MIMO (Multiple Input Multiple Output) communications.

ART works on input audio channels like Left, Right, Center, etc. not on speakers as stated by Dirac in various youtube videos. I believe for each input channel it is a single input, multiple output spatial equalizer exploiting transmit diversity. When you consider that speakers will support multiple channels the system as a whole is MIMO.

If you get a basic understanding of these concepts it will make sense on why in their videos Dirac suggests using any speakers that can play low frequencies to support all the other speakers. The more speakers the greater the spatial diversity and the more likely you will have a path that provides unique information that can be used to fill in nulls and shorten the reverberation time.

This also where the name of the Fsiso parameter comes from (partially explained in a video). SISO = Single Input Single Output which is the traditional form of room correction used by Dirac Live. The signal for a particular channel (like Left) is sent to and played from a single speaker. F_SISO is the frequency ART (MIMO) hands off to Dirac Live (SISO).

I believe this also part of why Dirac is cagey around explicitly stating they are doing active cancellation. It's not that they calculate the whole room layout and say geometrically if we play this frequency at this phase from this speaker it will cancel the reflection near that speaker and thus reduce reverberation. They clearly don't have that information. They don't know the location of any speakers and they explicitly state in youtube videos that there is no need to match the locations on the position diagram to the position of the microphone so they're not even relying on relative position of the measurements. And this is how a spatial equalizer works. It takes a set of signals that propagated through different spatial paths and solves an equation that creates a set of filters that can be applied to each received signal such that the sum of the filtered signals best matches a desired target output (typically the desired target is the original signal as if it was propagated through a single direct path with no reflections). It doesn't matter what the actuals spatial paths were, just that the paths were different and thus provide additional information that can be combined to better recover the desired target output.

The desired target and the limits on support speakers is to me the proprietary special sauce of ART. They state in the videos that they started with thousands of robotically measured positions and worked backwards from there. So I assume they started with a highly spatially oversampled grid of data AND because of the automated positioning they also had the full geometric information. Using that they figured out how using only delay to the given channel's primary speaker (that's what they can measure to link measurements with unknown geometry) what the target across the positions should be. Then they added in limitations like localization, frequency limits, power handling, spectral smoothing (the videos state the support parameter is not about power but about specifity so I think of it like changing smoothing parameters in REW plots) as constraints on the filters when they do the solve for the set of filters that best fits their desired target output. (The use of delay to link positions is also why they have a reference channel for LFE. A narrow bandwidth speaker like a subwoofer will not provide good delay estimates because delay accuracy is inversely proportional to bandwidth so they need a full bandwidth signal to accurately get relative delay between positions.)

Obviously there are differences in the applications of wireless communications and audio room correction, but given spatial multiplexing is common and fully out in the open in the wireless communications world I think it's a great place for people to find articles and videos explaining the concepts.

For sure some great ideas in there. But perhaps worth simplifying/condensing?

ART obviously knows how to pick frequencies to cancel, as otherwise it would not be able to reduce decay. We did wonder in this thread if it would not make more sense to apply "cancelation" in the DSP mode and just emit the non-offending response (aka net of cancelations), or do a double job of reproducing the frequencies to cancel them by active counter-response from other speakers. Don't remember we had a conclusion on that part. Someone would really have to tear down ART with REW big time to figure that out, which apparently did not happen - yet.

 

[seashell]

For sure some great ideas in there. But perhaps worth simplifying/condensing?

ART obviously knows how to pick frequencies to cancel, as otherwise it would not be able to reduce decay. We did wonder in this thread if it would not make more sense to apply "cancelation" in the DSP mode and just emit the non-offending response (aka net of cancelations), or do a double job of reproducing the frequencies to cancel them by active counter-response from other speakers. Don't remember we had a conclusion on that part. Someone would really have to tear down ART with REW big time to figure that out, which apparently did not happen - yet.

Condensed: Look up how a spatial equalizer works. Imagine the target response is a delta function, which is a single 1. That would mean you would hear only the original signal. (Perfect room and speaker correction.) The equalizer says what filters can be applied to these separate spatial channels such that the filtered sum most closely approximates that delta function. Those are the filters ideal ART creates.

That's a simple as I can make it.

Is that practically feasible? Nope.

Is designing a filter such that the target response is a delta function how ideal single channel room correction works too? Yep.

So where's the gain from ART? The MIMO approach lets you get closer to that ideal in practical scenarios.

If you want to understand how, then again, look up how a spatial equalizer works.

 

[janbth]

Condensed: Look up how a spatial equalizer works. Imagine the target response is a delta function, which is a single 1. That would mean you would hear only the original signal. (Perfect room and speaker correction.) The equalizer says what filters can be applied to these separate spatial channels such that the filtered sum most closely approximates that delta function. Those are the filters ideal ART creates.

That's a simple as I can make it.

Is that practically feasible? Nope.

Is designing a filter such that the target response is a delta function how ideal single channel room correction works too? Yep.

So where's the gain from ART? The MIMO approach lets you get closer to that ideal in practical scenarios.

If you want to understand how, then again, look up how a spatial equalizer works.

Do you have any references or sources for spatial equalizers?

 

[Oddball]

Condensed: Look up how a spatial equalizer works. Imagine the target response is a delta function, which is a single 1. That would mean you would hear only the original signal. (Perfect room and speaker correction.) The equalizer says what filters can be applied to these separate spatial channels such that the filtered sum most closely approximates that delta function. Those are the filters ideal ART creates.

That's a simple as I can make it.

Is that practically feasible? Nope.

Is designing a filter such that the target response is a delta function how ideal single channel room correction works too? Yep.

So where's the gain from ART? The MIMO approach lets you get closer to that ideal in practical scenarios.

If you want to understand how, then again, look up how a spatial equalizer works.

Don't get the point - but also don't get what's the point. Which is not a big deal anyway like as the point seems to be a bit pointless.

 

[JRowland]

Easter Bunny left me a busted RP-1200 sub that I just got recently, I made lemonade. I left the little one in the back and the bigger one where it was on the front right wall as well.

It’s taking full advantage of the left R7, this is the Dirac +5db curve and I’m using +4 which isn’t shown but real similar. Hey Now and Bad Guy I have to limit to -5 not because I heard anything but the woofers are just too active for me. For now it’s fine it’s around 85-90db avg on that NIOSH app depending on the song, I don’t want louder than that all that often.

I just learned you can click on all of the speakers (for me, groups for a lot of you) and lay them on top of each other.

 

[seashell]

Do you have any references or sources for spatial equalizers?

Sadly I thought it would be an easy topic to find lots of illustrative examples, but I was wrong. Lots of pages and videos on the topic, but spot checking several they are all very mathematical and not beginner friendly. A step back is just the concept of "spatial diversity", but even that turns up few illustrative examples. Quite disappointing honestly.

So I'll point you to https://www.cs.princeton.edu/courses/archive/spring18/cos463/lectures/L16-mimo.pdf, specifically slide 17 "Selection Diversity, in Frequency" as one specific example of a spatial equalizer presented as a power vs frequency plot that isn't (too) steeped in math.

Since my first post was deemed too long and complicated there's no way I'm going into any of the math. So I will simply apologize for attempting to answer the question asked of how ART actually works and let everyone return to their discussion of how to use ART.