Active Room Treatment (ART) by Dirac

[JRowland]

Interesting, I wonder what is the Dirac view of this.

I know that support levels are not related to pre-EQ of subwoofers, but my understanding is they dictate how much Dirac is going to use a subwoofer (or speaker).
I'm guessing support levels are more important than whether a subwoofer (or speaker) is pre-EQ'd or not.

Joss Walker represents Dirac, he doesn’t use support levels much and prefers F Support Low. I’ve messed with them a bunch, I use -18 all around now. It’s fun to play with ART, everything sounds good just varying shades of color.

 

[JRowland]

Interesting, I wonder what is the Dirac view of this.

I wish they had a presence on one of these audio forums, lots of folks would love to know instead of the mental masturbation we’re forced to deal with.

 

[RenPa]

Same full range response, Var smoothing, but with sub EQ on and sub EQ off.

See the post edited if it is what you asked for. I don't have full range 20Hz - 20KHz SPL eq in the subs and without Dirac . I could xo front left channel and measure the bass managed SPL without and with EQ in the subs if you wish.

 

[Talisman]

Fact or just assumption? Why would this occur? I can't think of any reason why pre-EQ resulting in a flatter response would be harmful. If anything, it means that Dirac doesn't have to do as much "work". Dirac measurements are not of the sub(s), they are of the sound in the room.

Because you're thinking about old calibration systems, which only focused on flattening the frequency response. Art works differently. Pre-equalizing the subwoofers means limiting the output and, if necessary, correction capabilities of that subwoofer. If there are frequencies that need cutting, Art will handle them perfectly, but it can use the full power of each speaker for cancellation if necessary.
Trying to "help" Art is a no-brainer; it makes no sense on such an advanced system.

 

[Satanica]

Because you're thinking about old calibration systems, which only focused on flattening the frequency response. Art works differently. Pre-equalizing the subwoofers means limiting the output and, if necessary, correction capabilities of that subwoofer. If there are frequencies that need cutting, Art will handle them perfectly, but it can use the full power of each speaker for cancellation if necessary.
Trying to "help" Art is a no-brainer; it makes no sense on such an advanced system.

No I'm not thinking about older calibration techniques but I get your point and can understand what you're saying could be put to use by ART.
In the effort of getting better results, does anyone know if there is any official word from Dirac to not pre-EQ speakers and subs before calibration where ART is the target?

 

[mmares056]

Joss Walker represents Dirac, he doesn’t use support levels much and prefers F Support Low. I’ve messed with them a bunch, I use -18 all around now. It’s fun to play with ART, everything sounds good just varying shades of color.

This is my experience. I've experimented with different support levels and after many hours of critical listening and measuring using -18 across the board sounds the best in my room with my equipment.

 

[markus]

Sorry to bring this up again, but I can be a bit slow on the uptake, and my post https://www.audiosciencereview.com/...oom-treatment-art-by-dirac.40743/post-2591811 might be proof of that. Please go easy on me for this one.
So, to sum up, let’s focus on the signal itself rather than the moment when we apply the EQ.
When I apply EQ to the subs, I do it BEFORE Dirac analyzes them, but the internal filters of the subs still are applied after the signal has been sent and processed in the receiver by Dirac, which is therefore counterproductive. It’s as if we were applying the sub filters twice before they produce the sound, isn’t it?

Would you be so kind as to confirm that I have (finally!) understood (I hope!)?

Translated with DeepL.com (free version)

Pre-equalization isn’t inherently counterproductive. There are valid reasons to apply EQ before running a DL calibration on top of it.

For example, if a subwoofer has a strong resonance that DL struggles to correct effectively (slot-loaded subs are a good example) pre-EQ can help smooth things out beforehand. Likewise, if the response contains significant irregularities, pre-equalization can bring the sub closer to the level of the loudest speakers, which may improve overall system headroom.

What you should avoid, however, is changing any EQ settings on your subs after an ART/BC calibration has been completed.

 

[janbth]

See the post edited if it is what you asked for. I don't have full range 20Hz - 20KHz SPL eq in the subs and without Dirac . I could xo front left channel and measure the bass managed SPL without and with EQ in the subs if you wish.

OK, that is close enough. It seems to me that the low frequency behaviour in you room is already quite good. Not very offensive peaks or valleys. This could possibly explain the small difference in your results.

 

[janbth]

Fact or just assumption? Why would this occur? I can't think of any reason why pre-EQ resulting in a flatter response would be harmful. If anything, it means that Dirac doesn't have to do as much "work". Dirac measurements are not of the sub(s), they are of the sound in the room.

Assumption - the ART algorithm is proprietary.

One example that can illustrate what I mean is this: A room with two subs, one on the left wall, one on the right. If played separately, they will excite the lowest lying mode (in the width direction) and produce a peak. If played together in mono, in phase, they will beat down the mode. (This is a "mode manipulation technique" that is described in Sound Reproduction by Floyd Toole.) If you separately EQ the subs flat first, then, when you play them in mono, there will be a dip in the response at the resonant frequency. You could then EQ the subs back up again of course, but that would defeat the purpose of the first EQ. The concusion I draw from this is: Be careful with "smart" pre-EQing if your room correction system - Dirac ART or whatever - can handle it.

 

[RenPa]

Be careful with "smart" pre-EQing if your room correction system - Dirac ART or whatever - can handle it.

totally agrre. Art is a lot smarter than me! A wild animal since I get it and I need to found new bearing to evaluate how the most little change in a filter Art setting might sounds in my room.

 

[JRowland]

In the effort of getting better results, does anyone know if there is any official word from Dirac to not pre-EQ speakers and subs before calibration where ART is the target?

After a few hours of my own time and an hour or so of fighting with AI "hallucinations" all I can find is this snippet at the 15:25 time stamp of the Mad Audio vid.

Pros and cons is the jist of it, no meat.

 

[dlaloum]

I thought processors already exist that have high or nearly no cross term limits (HTP-1, Storm, and possibly Tide16)? The only reason I can think of for Loudness to be ignored is because the Dirac feature set has to be similar for ALL processors (least common denominator). Sometimes its hard being an OEM and making sure the product is consistent across all platforms.

The smoking gun will be when Dirac ART strictly for PCs comes out. If that doesn't include Loudness as a feature, I'd imagine it's never coming. If we're counting from the StormAudio days, ART has been around for 3+ years!

Dirac does RoomEQ... Loudness is perceptual Loudness EQ - the two don't actually overlap - I don't think Dirac wants to get involved in the latter!!

Audyssey include it as a bonus in their package, Dolby have some options in their decoders software suite that do some perceptual loudness processing, so does THX and DTS...

In my worldview, it should be provided by the processor manufacturer - so that a single universal mechanism applies regardless of which decoder you use or which RoomEQ - the Perceptual Loudness mechanism should be a completely seperate "layer" - just like Bass/Treble controls or even parametric equalisation (seperate from Room EQ!!) should be a completely seperate layer.

 

[dlaloum]

I wonder if and when they will introduce AI to improve their software and make it easier to use

I strongly prefer solutions that are NOT dependent on outside software, and therefore subject to various market vagaries, which could lead to your processor becoming non-functional when the cloud software is retired!

 

[Nivag123]

Hello,

i've been using DIRAC ART with a Storm audio ISP since Jan 2026. I came from an Anthem AVM90 with ARC Genesis.

DIRAC ART has both impressed me with the bass decay and clarity, and at the same time frustrated me with the weird and unpredictable bass impact performance with HT and multichannel concerts content.

In stereo, the bass decay, texture and transient clarity is very very good. However, this strength does not translate to HT and multichannel concert content. Some HT content have very good bass transients and impact, similar to stereo, yet other HT content the bass sounds smeared, sluggish and vague. This was absolutely confounding to me.

After a lot of digging and troubleshooting, measurements, i think i have finally figure out the explanation. The hint was the 'known issues' in the DIRAC ART release notes.

This has been mentioned many times in this thread, but so far i don't think anyone really understand what it means and how it impacts the performance of DIRAC ART with actual bass content.

View attachment 532790

First, let me share measurements of what this 5ms LFE delay means. It means that LFE channel content, played through the subs, is literally delayed 5ms, as stated. This is in contrast with the channels, where with bass management, the subs will also play redirected bass from channels.

The diagram show the acoustic timing reference step response of the LCR channels, and the LFE channel. I have labeled the peak of the subwoofer (playing redirected bass ) step response from LCR channel as '1', and the subwoofer (playing LFE content) step response peak from LFE channel as '2'. One can immediately see the ~ 5ms delay that is stated in the DIRAC release notes.

View attachment 532791


In the next diagram, it is the same acoustic timing reference measurement of the LCR and LFE. However, i have muted the LCR channel, leaving only the Subs active and playing the bass management redirected bass from LCR. These still peak at '1', and the LFE with the delay peaks at '2'.

What does this mean, and how does it explain why some HT content bass impact sounds clean and tight, while other HT bass sounds smeared and vague, less precise and less impact?

View attachment 532792

I think it has to do with the relative amount of bass information in the (LCR) channels, vs the LFE channel. This is what i mean, and i try to break it down to a few scenarios.

1) No significant bass in LCR channels, bass mainly mixed in LFE channel. In this scenario, DIRAC ART will perform marvelously, bringing all the clean decay, precise texture, transient impact that ART excels in. This is because there is not redirected bass from LCR, playing in the LFE, so there is only one bass transient peak, at '2' during playback.

2) Significant bass content in LCR channels, complimenting bass mixed in LFE channel. For eg, in 13 hrs First Wave scene, the gunshot impacts. There is significant bass transient peak from gunshots in LCR channel, in addition to LFE channel bass transient peak.

In this scenario, the 5ms LFE delay bug wreaks havoc in the bass playback, for both HT and multichannel concerts, especially gunshots and kick drums.

One common example is Roy Orbison's Blue Bayou from Black and White Night. The song starts with a very nice transient acoustic bass + kick drum impact, that is very noticeable smeared and less precise on ART.

Another common track is Green Day's Boulevard of Broken Dreams concert track, another bass kick drum monster test track. With DIRAC ART, the bass kick drum is a pale shadow of what it could be, what it should be.

The reason this happens, is the bass impulse peak content from LCR, being redirected and played by subs, is peaking at '1', and then at the same time, the same complimentary bass impulse peak from LFE channel, played by the same subs, is peaking at '2', a good 5ms later. The result is the transient impact peak is smeared across 5ms.
Another way to think about this, with multiple subs. What happens when you intentionally delay half the subs 5ms from the other half? There will likely also be associated cancellations between the first and 2nd peaks.

3) Significant bass content in LCR channels, little / no bass content in LFE. This is common for stereo 2.0 tracks, or multichannel 5.0 or 7.0 concert tracks, often classical concerts are mixed / mastered these way. With these content, ART again shines with bass reproduction. The subs (redirected bass) will peak at '1', there is no LFE channel bass to peak at '2' to cause smearing or cancellations.

I hope this post raises awareness and understanding of this issue and DIRAC can implement the fix very soon in the next release.

I truly believe ART brings the next level of bass performance in room, and already enjoy the incredible results with stereo content. I hope to enjoy similar with all my HT and concert content as well.

Thanks for reading.

Gavin

Hi Everyone,

i have some additional data from REW Phase align tool to give another view of what i'm describing.

First plot is Phase Align Tool, summing full range L+R. Full range because of bass management, LFE subs used to make the L & R full range. This is scenario 1 with stereo 2.0 content, no issues. Both channels are well aligned, any bass present in both channels (mono bass) sums perfectly.

Next, is the scenario of L + LFE sum. Imagine there is a 50 cal sniper rifle going off in the LH channel. And the content is mastered with significant bass on the LH channel, which is bass managed and redirected to LFE subs. And at the same time, there bass content of the 50 cal impact is also on the LFE channel. What happens when both are played simultaneously, how do they sum?

Again, using REW Phase Alignment tool, summing L ch + LFE ch. Instead of the expected Green summation with everything in phase, we get the Red cancellation, due to the 5ms shift on LFE. This is also visible in the phase plot below.

Next, i will add a 5ms shift to the LFE channel, same L ch + LFE channel summation. I have highlighted the added 5ms shift in green on the alignment tool controls. We get very nice summation again, when L and LFE plays bass at the same time.

Hope this helps illustrate what i mean, and the impact of the 5ms LFE delay issue in REW.

Effectively, it is like running a multiple sub LFE setup, with a known 5ms shift between the subs, instead of time aligning them.

Nobody would accept multi-sub with 5ms delay misalignment. But that is what we are getting with ART now with the 5ms LFE delay.

i recognise that this may not be immediately audible in all setup, but it is very audible in my setup. As mentioned in a previous post, 5ms is half wavelength at 100hz, and we see that strong cancellation centered at 100hz due to this issue.

i hope this also explains why the issue only happens with 0.1 LFE content, and does not happen in stereo 2.0 or even multichannel 5.0/7.0 content.

 

[KenMasters]

I hope this also explains why the issue only happens with 0.1 LFE content, and does not happen in stereo 2.0 or even multichannel 5.0/7.0 content.

You say that, but it does in mine. I’ve passed on my data to Dirac.

 

[Nivag123]

You say that, but it does in mine. I’ve passed on my data to Dirac.

Can you elaborate? I do have some other observations about how ART is phase aligning Subs to Mains channels - it is somewhat odd, and there are some phase cancellation issues too. It gets worse when you try to 'crossover' the mains lower, but where i have them now about 80hz the issues are visible in measurements, but not really that audible in listening. When I tried to cross low at 60hz, the problem becomes worse, the phase mismatch and misalignment thru the crossover region between mains and subs. It is not a true traditional LR2 type bass management crossover. There is much more overlap, and FIR phase correction is applied to linearise / match the phase between the mains and sub.

But in the larger scheme of things, the 5ms LFE delay is a much much bigger problem. So i'm focusing on that first, one problem at a time...

 

[KenMasters]

Can you elaborate?

You can look at my posts here, here and here (in that order).

 

[RenPa]

Pre-equalization isn’t inherently counterproductive. There are valid reasons to apply EQ before running a DL calibration on top of it.

For example, if a subwoofer has a strong resonance that DL struggles to correct effectively (slot-loaded subs are a good example) pre-EQ can help smooth things out beforehand. Likewise, if the response contains significant irregularities, pre-equalization can bring the sub closer to the level of the loudest speakers, which may improve overall system headroom.

What you should avoid, however, is changing any EQ settings on your subs after an ART/BC calibration has been completed.

This could possibly explain the small difference in your results.

I think you will find this interesting. I said the pre-equed subs in SVS didn't sound as usual non eq sub in Art and analysis in REW didn't show evidences why.
I'd lile comments about this.
Here are the filters of the subs group inArt with Dr Toole Trained lestenr TC (red is sub 2 blue is sub 1 LFE):

Here all the same but with Subs pre-Equed in SVS:

 

[JRowland]

I think you will find this interesting. I said the pre-equed subs in SVS didn't sound as usual non eq sub in Art and analysis in REW didn't show evidences why.
I'd lile comments about this.
Here are the filters of the subs group inArt with Dr Toole Trained lestenr TC (red is sub 2 blue is sub 1 LFE):
View attachment 534550
View attachment 534553

Here all the same but with Subs pre-Equed in SVS:
View attachment 534551

View attachment 534554

Similar but different looking to me. Lower low end with the pre eq, barely but lower boosts above 0 but bigger cuts on the non lfe with pre eq. LFE looks a little better boost/cut wise with pre eq to me but I’d use my ears as it appears 6 of one 1/2 dozen of the other to me. Corrected spread same same as well?

 

[RenPa]

Corrected spread same same as well?

Not pre-EQued in SVS LFE

Pre-EQued in SVS LFE