Active Room Treatment (ART) by Dirac

[KenMasters]

No, I'm not saying excactly that. If the goal is to time-align different subs with different shapes for their impulse responses, then aligning the peaks is not a reliable method. It may or may not work. Instead focus on the frequency response, and ensure that it is smooth.

Are you not perhaps missing the context in which my post was made i.e. the post Dirac result? I've articulated my thoughts as well as I can in my interim posts.

 

[JRowland]

I'm pretty sure everyone using ART has noticed this or perhaps not everyone has noticed it because not everyone measures their system post ART calibration. I always measure post calibration as RCS don't always get it right.

In my experience ART does an excellent job with phase and time alignment. This particular example is for 12 point measurement calibration for a 6ft 3 seater sofa. ART's speaker time alignment is impeccable, all speakers are within 0.53 inches of each other and phase alignment is flat and my system sounds great.

I'm an enthusiast and not expert but I think ART makes my subs start super early, 15ms early(Circled in red) to accomplish its goal of optimizing based on a flat phase. It is my understanding that for subs the start of the impulse response which in my case is 15ms should be aligned at t=0 as this is when energy from the subs starts to arrive at our microphone(Ears). But since Dirac locks us out of delay adjustments post calibrations there is nothing we can do about this early start. I wish Dirac would reconsider and allow for post EQ time alignment at least so that we can correct for this.

View attachment 529642

View attachment 529643

For my own curiosity can you post your Dirac impulse response screenshot showing uncorrected and corrected overlays like so (same checked boxes)? I know the sub wavy lines won't show but where Dirac times it will via the dotted vertical line(s).

 

[mmares056]

For my own curiosity can you post your Dirac impulse response screenshot showing uncorrected and corrected overlays like so (same checked boxes)? I know the sub wavy lines won't show but where Dirac times it will via the dotted vertical line(s).

View attachment 529810

Here it is. Curious to see if you spot any inconsistencies in the Dirac Impulse screen and compare to my REW Impulse Response screenshots.

 

[JRowland]

Here it is. Curious to see if you spot any inconsistencies in the Dirac Impulse screen and compare to my REW Impulse screenshots a few posts back.
View attachment 529839

Thank you, I'm trying to process as this is out of my wheelhouse by a big margin. Interesting first thing to me is your subs fire 2ms early like the fargate guy on AVS (Ren something on here IIRC). Like I said to him when I had that Klipsch wireless (WA-2 maybe) thing on one of my subs I had a 4ms early firing sub impulse.

I removed the wireless thing and added a cord but at the same time replaced my front left sub that broke with a different brand so more than one change between measurements. The new measurement shows everything lining up and I believe I can feel and hear a difference but it is purely subjective as I have no evidence. Zoomed in now I'm .2ms off vs 4ms so what I think I hear/feel subjectively should align with what I am seeing on my Dirac charts.

Maybe more data will prompt someone with a high level understanding of timing ('cause that certainly ain't me) to jump in?

 

[JRowland]

@mmares056 one more request if you would be so kind, can you post a screen shot of just the LFE group impulse response so I can see the dotted line and where Dirac puts it.

Like this

This was from the same overlay with the 4ms early firing subs, I was wrong about the 3rd sub being added at that time. It was just the front wired sub and the rear wireless while I was waiting for the new one to show up.

 

[KenMasters]

@JRowland, if useful in your investigations:

Again with the roughly 5ms discrepancy, a feature of your above graph as well (speaker alignment centred between sub peak and dashed line).

 

[JRowland]

@JRowland, if useful in your investigations:

View attachment 529845

Again with the roughly 5ms discrepancy, a feature of your above graph as well (speaker alignment centred between sub peak and dashed line).

I see 4 ms on the bottom, you can use your cursor and check the cursor coordinates box to make sure. How did you get the wavy lines from the sub to show on the bottom, I can't seem to make them overlay?

 

[KenMasters]

I see 4 ms on the bottom, you can use your cursor and check the cursor coordinates box to make sure.

Looks to me to be about 10ms between impulse peak and dashed line. Dirac doesn't say it's 5ms on the dot, just that the delay is about 5ms relative to the other channels. I'd say all three of our graphs are in that ballpark.

How did you get the wavy lines from the sub to show on the bottom, I can't seem to make them overlay?

Photoshop

 

[JRowland]

Looks to me to be about 10ms between impulse peak and dashed line. Dirac doesn't say it's 5ms on the dot, just that the delay is about 5ms relative to the other channels. I'd say all three of our graphs are in that ballpark.


Photoshop

Photoshop makes sense, I was wondering if I needed to try Dark Mode or something

Ignoring the peaks and any lines other than the dashed (dotted) vertical ones that signify the start of the signal to Dirac, the first dashed line is the LFE group and the second is all your other groups. If you place your cursor on the dashed line it will give you the ms and then place it on the other one, I think I'm seeing the difference is 4ms. @mmares056 shows 2ms, and my current one shows the lines right on top of each other (zoomed shows the .2ms difference)

My current (see the 2 dashed lines on the bottom chart side by side compared to yours compared to mmares)

Zoomed same image (scroll wheel on the mouse) and you see the scale at the bottom is way zoomed as well (sub dashed line sits on the 99.4ms mark, it's baby blue and hard to see)

So for this piece which I am assuming is what Dirac is using for total system timing mine, mmares, and yours do not look similar because the subs are firing early for him and even earlier for you. My subs are .2ms later, but my last measurement with different gear they were 4ms early just like yours. After taking that wireless kit off and doing a new set of measurements my dotted lines align. I was assuming the wireless kit was causing my issue but recently more folks (you included) have shown the same discrepancy and with wired subs so I'm rethinking my previous conclusion and wondering what else could have improved the timing if it wasn't that?.

 

[mmares056]

@mmares056 one more request if you would be so kind, can you post a screen shot of just the LFE group impulse response so I can see the dotted line and where Dirac puts it.

Like this

View attachment 529842

This was from the same overlay with the 4ms early firing subs, I was wrong about the 3rd sub being added at that time. It was just the front wired sub and the rear wireless while I was waiting for the new one to show up.

Here it is the LFE group impulse response.

 

[KenMasters]

Ignoring the peaks and any lines other than the dashed (dotted) vertical ones that signify the start of the signal to Dirac, the first dashed line is the LFE group and the second is all your other groups. If you place your cursor on the dashed line it will give you the ms and then place it on the other one, I think I'm seeing the difference is 4ms. @mmares056 shows 2ms, and my current one shows the lines right on top of each other (zoomed shows the .2ms difference)

My current (see the 2 dashed lines on the bottom chart side by side compared to yours compared to mmares)

Zoomed same image (scroll wheel on the mouse) and you see the scale at the bottom is way zoomed as well (sub dashed line sits on the 99.4ms mark, it's baby blue and hard to see)

So for this piece which I am assuming is what Dirac is using for total system timing mine, mmares, and yours do not look similar because the subs are firing early for him and even earlier for you. My subs are .2ms later, but my last measurement with different gear they were 4ms early just like yours. After taking that wireless kit off and doing a new set of measurements my dotted lines align. I was assuming the wireless kit was causing my issue but recently more folks (you included) have shown the same discrepancy and with wired subs so I'm rethinking my previous conclusion and wondering what else could have improved the timing if it wasn't that?.

What I'm seeing is that all three of our graphs exhibit roughly the same behaviour. The dashed line for LFE sits somewhere in the region of 10ms apart from the sub impulse peak, and the speakers find themselves somewhere roughly in the middle. I've illustrated the phenomenon using your graph - it's the same across all three of our graphs:

The impulse peak of the sub/s finds itself occurring roughly 5ms later that the speakers - I believe this peak is intended to align with the rest of the speakers rather than the dashed line, given Dirac's wording around the issue:

• The 120 Hz low-pass filter introduced in version 3.13.2 applied to the LFE channel introduces a delay of about 5 milliseconds compared to the other channels.

I think the dashed line is where Dirac identifies the signal rise and from that point the signal is modified. I believe it's then the intended behaviour for the impulse peak to be aligned with the rest of the speakers, but the algorithm seems to currently place the speakers between those two points.

I might be way off base, I'm not an expert, but Dirac's wording of the issue leads me to believe that to be the case.

 

[JRowland]

Here it is the LFE group impulse response.

View attachment 529854

Using my tiny phone for one screen and comparing your earlier REW pics you circled a spot and said 15ms, the line Dirac is using as the start is just to the left of the bottom of the 1st trough as seen on the above image. It would be just to the left of your other speakers if you could overlay the above image and your REW image (early by about 2ms).

Help me understand the correlation to the spot you circled on your REW image to the dashed line Dirac says is the green start flag so to speak. From the Dirac shot above that dashed line isn't where your circle would be either correct (your circle would be at the top of the first hump I'm thinking)?

 

[JRowland]

What I'm seeing is that all three of our graphs exhibit roughly the same behaviour. The dashed line for LFE sits somewhere in the region of 10ms apart from the sub impulse peak, and the speakers find themselves somewhere roughly in the middle. I've illustrated the phenomenon using your graph - it's the same across all three of our graphs:

View attachment 529856

The impulse peak of the sub/s finds itself occurring roughly 5ms later that the speakers - I believe this peak is intended to align with the rest of the speakers rather than the dashed line, given Dirac's wording around the issue:

• The 120 Hz low-pass filter introduced in version 3.13.2 applied to the LFE channel introduces a delay of about 5 milliseconds compared to the other channels.

I think the dashed line is where Dirac identifies the signal rise and from that point the signal is modified. I believe it's then the intended behaviour for the impulse peak to be aligned with the rest of the speakers, but the algorithm seems to currently place the speakers between those two points.

I might be way off base, I'm not an expert, but Dirac's wording of the issue leads me to believe that to be the case.

I did see what you meant earlier by in the middle (those pretty drawn lines above ^^^ reassured me that's what you were seeing too) I just haven't even gone there yet in my head I'm still stuck on should the dashed lines on the predicted corrected be on top of each other for the best result and if so how does one make that happen?

 

[janbth]

Are you not perhaps missing the context in which my post was made i.e. the post Dirac result?

That’s perfectly possible of course . I guess my point is that impulse responses are unintuitive and hard. There’s more to sub integration than simply aligning peaks. The algorithm that Dirac ART uses to set relative time delays is not known, but it it probably involves a number of other ingredients as well, like phase, frequency response, etc.

 

[mmares056]

Using my tiny phone for one screen and comparing your earlier REW pics you circled a spot and said 15ms, the line Dirac is using as the start is just to the left of the bottom of the 1st trough as seen on the above image. It would be just to the left of your other speakers if you could overlay the above image and your REW image (early by about 2ms).

Help me understand the correlation to the spot you circled on your REW image to the dashed line Dirac says is the green start flag so to speak. From the Dirac shot above that dashed line isn't where your circle would be either correct (your circle would be at the top of the first hump I'm thinking)?

I was basically pointing out on the REW verified impulse response where the energy from the Subs starts (Circled in red) to arrive at the mic or ears and that this starting energy should be aligned at t=0. My guess is that ART tries to optimize based on a flat phase and apparently to be able to do that it makes the subs super early so it appears based on actual measurements.

Now as for the Dirac corrected/predicted impulse response graph shows (dotted line) the bass rise signal starts 2ms early, before the impulse peak of the speakers and on my REW verified impulse response the bass rise signal starts around 15ms early. If the predicted response is not accurate we could try to correct this problem if Dirac allowed for post EQ time alignment tweaks.

 

[JRowland]

I was basically pointing out on the REW verified impulse response where the energy from the Subs starts (Circled in red) to arrive at the mic or ears and that this starting energy should be aligned at t=0. My guess is that ART tries to optimize based on a flat phase and apparently to be able to do that it makes the subs super early so it appears based on actual measurements.

Now as for the Dirac corrected/predicted impulse response graph shows (dotted line) the bass rise signal starts 2ms early, before the impulse peak of the speakers and on my REW verified impulse response the bass rise signal starts around 15ms early. If the predicted response is not accurate we could try to correct this problem if Dirac allowed for post EQ time alignment tweaks.

I understand your point about not having access to the distance settings to correct things. The way my mind works it is moot so what can be done then with what is available? It’s subjective but things are better now that my impulses start together. It was good before, it’s better now. Totally subjective but it makes sense to me that the better “impact” I get from percussion now is a result of better timing. So if/then in my mind what did I do? New measurement so all of the variables involved with that, new subwoofer in a similar location with similar internal delay based off uncorrected impulse responses, I moved my tube trap over to cover the corner of my recessed door frame more, and I made a wireless sub wired all at the same time. I liked Ken’s dark mode so here again is my current set up

Just subs, orange trace is rear and closest physically, blue is sub 1 LFE.

So Dirac chooses a little later than the actual start again based on your circle explanation from earlier correct? I notice it hits directly in the trough this time instead of to the left a bit just writing “out loud”.

Total timing

Zoomed

I went back to some pics from when I only had 2 subs and my initial ART set up. I didn’t know about overlays but I see my subs were 2-3ms early then with no wireless involved.

The other thing I’m noticing and don’t understand is the numbers on the scale. Initially my first Dirac was low numbers and now they’re higher, why?

 

[jhaider]

Why doesn't Dirac ART include loudness compensation / Fletcher Munson out of the box or as an option ?

Dirac is not calibrated to absolute SPL. It’s not a closed system, unlike Audyssey—Audyssey knows the mic gain and (within reasonable tolerance) sensitivity, so it can calibrate to near enough absolute SPL.

 

[KenMasters]

The other thing I’m noticing and don’t understand is the numbers on the scale. Initially my first Dirac was low numbers and now they’re higher, why?

Dirac changed the way they display the latency figure. Previously the display only referenced the speakers, with the first impulse marked as zero. Now it takes the entire chain into account, so zero is the moment the signal is sent to the device.

I like the change, useful for setting lip-sync if you know the latency of your display.

 

[JRowland]

Dirac changed the way they display the latency figure. Previously the display only referenced the speakers, with the first impulse marked as zero. Now it takes the entire chain into account, so zero is the moment the signal is sent to the device.

I like the change, useful for setting lip-sync if you know the latency of your display.

That’s nice to know Ken, thanks! I love it when this thread delivers.

 

[geox]

That is an interesting question. I do think that Dirac SHOULD consider various options in this area as part of the package (i.e. not left to implementation of specific brand). At a minimum, it should offer dynamic compression options as that can be useful for many people with late night sessions, or with material that has dynamic range that is not to one's liking. They should not only cater to reference - this is consumer grade product so should provide some preference options.

I was recently engaged in discussion on a different forum how Interstellar audio track (ATV, 5.1) is posing a great challenge to many HT systems. Dialogue levels are overall very low, and for lots of people whispering and low talk levels are ineligible, while the peaks are very significant. So the only option is to constantly adjust the volume - manually, if there are no dynamic compression tools. This is something that TVs and soundbars can do very effectively. Can't really think of many movies like these, but there are certainly some movies where dynamic range is beyond what some people would want to experience at home.

Sony has had dynamic range compression - drc as they call it in quite a few of their audio products.
Frankly , it seems elementary for the top sound processing algorithm in the market not to support loudness compensation and drc imo.

HTP-1 is no longer made only option is crazy expensive storm audio atm