• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Dirac/Atmos upfiring target + MultEQ-X question

In my system the best target for my Atmos-enabled upfiring speakers is:

  • Same full-range curve as my base layer speakers

    Votes: 0 0.0%
  • Curve replicating Fig 3 of the patent

    Votes: 0 0.0%
  • Other

    Votes: 0 0.0%

  • Total voters
    2

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
3,757
Likes
5,806
The current version of Dirac doesn’t properly support upfiring Atmos speakers in that it doesn’t distinguish between the initial impulse of the speaker versus the reflection which captures the ceiling height. Whether this makes a big or small difference probably depends on your ceiling heights, etc.

That said, the other wildcard is the target curve. Since a lot of Dirac owners run full bandwidth correction (with its pros and cons), correcting the upfiring Atmos speakers to flat can remove the Dolby science. Audyssey does apply the Atmos crossover at 7/12 kHz in its correction but it also adds the midrange dip (at least with the mobile app).

Anyway, the original Dolby patent: WO2014107714A1.pdf (storage.googleapis.com) gives us a few different strategies in creating a curve.
  • body of the patent talking about 5 dB peak at 7 kHz and 7 dB notch at 12 kHz
  • Figure 3. This shows "HRTF responses averaged across a large set of subjects" where the peaks and notches are greater in magnitude. This shows the variability from person to person and how we are just getting an average
  • Figure 15. What Dolby calls a digital target with smaller amplitudes at 7 and 12 kHz. This is more similar to Audyssey’s curve though Audyssey has its midrange compensation and doesn't seem to match anything from the patent exactly and seems to be more a mix.
I went ahead and did a rough version of Fig 3 and 15.

Instructions
1) You will need to rename the .targetcurve.txt to .targetcurve
2) You may need to play with the settings for the frequency response below 100 Hz and curtains below that level.

I haven't had enough time to experiment to see if either of these work better than using a standard target without HRTF. Using Top Gun: Maverick as my reference and using the scene when they are launching Darkstar, I do think the modified HRTF curves help.

I am curious if MultEQ-X has a detailed target curve for Atmos upfiring speakers. I think MultEQ App was the only time I have seen auto EQ software specifically include the HRTF data.
 

Attachments

  • UpfiringAtmos_Fig15.targetcurve.txt
    332 bytes · Views: 108
  • UpfiringAtmos_Fig3.targetcurve.txt
    415 bytes · Views: 72
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
3,757
Likes
5,806
nice try troll
You might want to do some background research on who you’re calling a troll before calling anyone a troll… But in case you really think my post was a troll… it means that I have assumed too much and haven’t given enough background explanation.

So here goes.

1) The argument for upfiring Atmos elevation speakers are in homes with plaster walls not drywall, apartments or homes where you may be renting temporarily and not want to deal with patching up any work you’ve done (remember there are rentals like this in the world), or for aesthetic reasons in rooms that are multipurpose. We all want a 32-ch Dolby Atmos cinema at home, but budget/space/reality all play a role

2) ASR is about science not only to identify the best measuring hardware but also how to best utilize the gear we have.

3) The argument against Dolby HRTF is that it’s not the same for anyone. The figures about half way down show how two individual ears will hear height differently.

4) The argument for Dolby HRTF is that Dolby put a ton of research into it and HRTFs as a concept are well documented in decades of research

5) Most of the articles talking about the Dolby target are limited in detail when reviewing the actual patent.

Audioholics talks about Fig 14 which is the actual speaker that Dolby designed using an analog crossover. They show that their specific driver and specific crossover gave them curve 1422 in measurement where they wanted to hit 1424 as their goal when designing the crossover.

What isn’t clear is if the goal was specific to the speaker they designed or all Atmos elevation speakers. The 12 kHz dip goes to -4 dB in Figure 14B.
1693110213665.jpeg



Archimago referenced Figure 9. This is describing how a two way crossover can be used to generate the Atmos HRTF. 908 is the target that is supposed to be added and you can see the 12 kHz dip goes to -9 dB.

1693110405604.png


Audyssey used this as a target. You see Audyssey’s mid range compensation, but note the 12 kHz dip seems a tiny bit deeper than Figure 14B and it never comes back up to +0. So it’s like the room target PLUS the HRTF.

1693110372391.png


So which is it?

We can look at the patent. Paragraph 91 mentions a 7 dB drop
IMG_8973.jpeg


Fig 3 shows you the raw data which is closer to -8 dB at 12 kHz
1693110866587.jpeg


And Fig 15 is a better version of Fig 14 because it’s a simple digital filter, with 1522 being their equation and 1524 being their target.
1693110961677.jpeg



Here’s the thing about patents. You actually never want to give all the secrets away. Most patents will focus on “ranges”rather than being too narrow in response. You also focus on patenting embodiments that have been reduced to practice.

What you care about is the claim set and it gets super broad.

10…
“a virtual height filter applying a frequency response curve to an audio signal generated by a renderer and transmitted to the upward-firing driver, wherein the virtual height filter at least partially removes directional cues from the speaker location and at least partially inserts the directional cues from the reflected speaker location.”

So we see plenty of data on this to convince the patent examiner. But the interesting parts are:

15. The system of claim 10 further comprising a detection component configured to detect the presence of the virtual height filter in the listening environment.

16. The system of claim 10 further comprising a bypass switch to bypass the virtual height filter during a calibration process that prepares audio playback equipment to transmit the sound waves to the listening environment.

17. The system of claim 10 further comprising a room correction component performing a pre-emphasis filtering operation on the sound waves transmitted to the listening environment to compensate for the virtual height filtering applied to the signal transmitted to the upward- firing driver.

18. The system of claim 17 wherein the speaker has a default virtual height filter and wherein the room correction component uses the speaker location to modify the default virtual height filter curve based on a frequency response curve optimized for the speaker location.

19. The system of claim 10 further comprising a room correction component generating a target response of the listening environment by use of a probe signal and adding a default virtual height filter response to a target response of the listening environment.

20. The system of claim 10 further comprising an array of audio speakers including respective upward-firing drivers for distribution around a listening environment, and wherein each respective upward firing driver is oriented at a unique inclination angle relative to the ground plane.

21. The system of claim 10 wherein the virtual height filter implements an algorithm using a scaling factor to compensate for height cues present in sound waves transmitted directly through the listening environment in favor of the height cues present in the sound reflected off the upper surface of the listening environment.

22. The system of claim 21 wherein the virtual height filter represents a unique frequency response curve, and wherein one or more characteristics of the frequency response curve are changed based on the value of the inclination angle.

——/
This is where they get really vague, but it seems to me, that there are a lot of unknowns.

The fact that Dolby didn’t mention any specific numbers in the claim set means they want to claim infinite options. For example, using Fig 3 they could have said a boost of 2-6 dB in the range between 5-9 kHz and covered the whole range, but instead they went all in.

Since the bolded claims are there, they can articulate to an examiner that they haven’t restricted their claim because it can be anything and certainly they are the first to come up with this idea.

1) to my knowledge, I cannot do a REW sweep of an Atmos channel since HDMI PCM cannot address the height channels directly. I need to encode in an Atmos container. There is no free Atmos encoder.

2) In Audyssey setups, assigning Dolby Atmos elevation speakers and giving a ceiling height helps with distance calibration. For the Monolith HTP-1, I see no such ability to define the ceiling height. I can work with delays.

3) Dirac just thinks about front height or ceiling, not reflections that benefit from the HRTF and doesn’t account for the delay which might affect positional queues if you have a really crazy height difference between the front upfiring and rear upfiring ceiling.

In a dream implementation, when assigning Atmos elevation, the AVR or AVP detects that there is truly an Atmos speaker with the physical crossover tweak. It instead sends an inverted digital signal that undoes the Atmos HRTF so that the room correction can do full EQ to the room curve. Then in real-life, there is no digital inversion and so the full correct EQ is then manipulated by the analog crossover to add the HRTF. But we don’t know if this is occurring in any setup since we cannot send Atmos sweeps in REW.

4) if we know that angles and in room calibration works, in theory, with a curve editor, we might want to tweak up and down 7 kHz or 12 kHz to enhance the ceiling effect more.

5) The best image of what Audyssey chooses as a curve is what I posted, which comes from the mobile app. I’d love to see if MultEQ-X has a factory target that you can zoom in to see the vertical scale with more precision in its interface.
 

Flak

Senior Member
Industry Insider
Joined
Nov 18, 2018
Messages
382
Likes
587
Good questions...

there is a special target curve modifier for Atmos upfiring speakers when using Dirac Live,
it's up to the user to "tell" the AVR about the speaker type:


MicrosoftTeams-image.png


Also, on Denon (and other AVRs/Processors) you can reduce the delay post Dirac in order to account for the bounce of Atmos upfiring speakers
 
Last edited:
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
3,757
Likes
5,806
Good questions...

there is a special target curve modifier for Atmos upfiring speakers when using Dirac Live,
it's up to the user to "tell" the AVR about the speaker type:
1) The modified target is automatic on Audyssey and other EQ tools like Yamaha YPAO RSC which you can see in the various displays. However, are you saying that on Denon (and Marantz), Dirac software applies a flat correction and then the Denon/Marantz will after-the-fact apply the HRTF since the AVR knows its upfiring and Dirac does not?

2) I made my own target curve modifier from reading the Dolby patent which works for me (posted above) -- is that doubling up the correction? Is there a better target curve modifier?
 

Flak

Senior Member
Industry Insider
Joined
Nov 18, 2018
Messages
382
Likes
587
The measurement will not "see" the applied dolby atmos patented upfiring coloration filter, and therefore the Dirac Live filter will not attempt to "correct" for it.
In other words... it will remain as intended
 
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
3,757
Likes
5,806
The measurement will not "see" the applied dolby atmos patented upfiring coloration filter, and therefore the Dirac Live filter will not attempt to "correct" for it.
In other words... it will remain as intended

So,

Atmos processors that identify Dolby upfiring speakers
a) will disable the added HRTF
b) send a flat signal to the speakers, which Dirac will correct to the user's chosen curve
c) re-add the added HRTF post-Dirac
d) but will not correct the delay to account for the ceiling bounce for timing

Is this true for all setups? In that case, I am adding double HRTFs which may exaggerate the effect?
 

Kakerlake

New Member
Joined
Nov 12, 2023
Messages
1
Likes
1
Interesting question. I think the most scientific way to answer your questions is a REW measurement.

However, I put your fig 3 and fig 15 target curves to slot 2 and 3 of my Maranz Cinema 50. For sound comparison I used the Helicopter and Rain Storm Atmos demo tracks.

The Helicopter demo is particularly well suited to reveal the weaknesses of upfiring speakers due to the highly directional sound. For me, the target curves with HRTF lifted the helicopter sound significantly when comparing to the standard curve I use for the base layer. The difference between the fig 3 and fig 15 curve was little. I slightly preferred the fig 3 curve which may be a psychological effect because I expected that the stronger correction helps more.
I also asked my two sons (14 and 16 years old) for a (blind) test. Both preferred the curve without HRTF :(. This was really surprising for me because for me the improvement due to HRTF was clear and not subtle at all. But also with this perceived improvement I clearly have to state that the capability of upfiring speakers to produce directional sound is very limited (at least with my setup in my room)

The Rain Storm demo is much more compatible to upfiring speakers. I have impressive results with and without HRTF. I slightly preferred HRTF.

I will keep the fig 15 curve. Thanks a lot for sharing.
 
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
3,757
Likes
5,806
Interesting question. I think the most scientific way to answer your questions is a REW measurement.

I don’t know how to send a sweep to the height layer. Someone with the Dolby encoder software would have to make a few sweeps for everyone to use.

I also asked my two sons (14 and 16 years old) for a (blind) test. Both preferred the curve without HRTF :(. This was really surprising for me because for me the improvement due to HRTF was clear and not subtle at all.
The interesting part of the patent AND consistent with the science is that everyone’s ear is a little different.

Some people take that to mean that the HRTF is useless and that you should just using in-ceiling speakers, but others should see it and think “I wonder if I can tune to my own ears”.

But also with this perceived improvement I clearly have to state that the capability of upfiring speakers to produce directional sound is very limited (at least with my setup in my room)

The Rain Storm demo is much more compatible to upfiring speakers. I have impressive results with and without HRTF. I slightly preferred HRTF.

I will keep the fig 15 curve. Thanks a lot for sharing.

Glad at least one person found it helpful!

It would be interesting to see how an Audyssey default setting sounds to you compared to Dirac-default specifically for those demo tracks. Audyssey puts a D&M HRTF which doesn’t match perfectly either.
 

geodon005

New Member
Joined
Nov 28, 2023
Messages
3
Likes
0
I am quite new to Dirac, and would like to try the fig.15 curve for my upfiring Atmos speakers. Do I apply the curve ONLY for those speakers? Thanks. . .
 

mishmushkela

New Member
Joined
Dec 8, 2023
Messages
4
Likes
0
Hey,
Im new to Dirac and just read this article. I applied fig.15 and now I'm asking myself if I need to adjust the time alignment in my denon for my atmos enabled speakers.

They measurment values for my standing speakers is 2.4ms and surround 5.9ms.
Front dolby is at 1.8ms and surround dolby at 5.4

They are about 1.35m away from the ceiling
 
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
3,757
Likes
5,806
Hey,
Im new to Dirac and just read this article. I applied fig.15 and now I'm asking myself if I need to adjust the time alignment in my denon for my atmos enabled speakers.

They measurment values for my standing speakers is 2.4ms and surround 5.9ms.
Front dolby is at 1.8ms and surround dolby at 5.4

They are about 1.35m away from the ceiling
There is not clarity if Atmos-certified vs “Upfiring”, non Atmos(TM) certified speakers has differences or if the difference is due to how each person hears HRTF.

You should look at the impulse response and can look for the peak from the on-axis and then the second peak from the ceiling bounce. You can then adjust the timing so that the ceiling bounce is how you time the alignment.

Dolby talks about the reality being that you hear a bit of on axis and bounce together. It’s also not clear if you should correct the delay fully so that the timing aligns with the ceiling source or if it should be somewhere in between the two.
 

D8hS

New Member
Joined
Nov 18, 2023
Messages
4
Likes
3
I'm a little confused. I have a 5.2.4 system and Denon x4800h with 4 "atmos-enabled" upfiring speakers. If I decide to get Dirac somewhere down the road what exactly will I be missing if Dirac doesn't differentiate between height and ceiling vs upfiring?
 

mishmushkela

New Member
Joined
Dec 8, 2023
Messages
4
Likes
0
Just to clarify...
This is the Impulse response for my front atmos.
Blue beeing the corrected one.
The second bump is at 10.7ms and the corrected second bump at 20.7ms...
And what do I need to do in the Denon menu?

I own the Dali Alteco
 

Attachments

  • 17020534842058138391045498399981.jpg
    17020534842058138391045498399981.jpg
    248.3 KB · Views: 42

mishmushkela

New Member
Joined
Dec 8, 2023
Messages
4
Likes
0
Maybe this is a better picture
 

Attachments

  • 17020540486021454854025152663881.jpg
    17020540486021454854025152663881.jpg
    242.3 KB · Views: 33
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
3,757
Likes
5,806
1702056457725.png


So, the dashed line is what Dirac is aligning you to. We don’t have exact numbers, but it looks like 4.0s to me.

But it is focused on aligning the initial on-axis. The ceiling bounce is louder. We see that around 7.7 or 7.8 seconds.

The difference is 3.7 to 3.8 seconds give or take.

“Front dolby is at 1.8ms and surround dolby at 5.4”

So if the front Dolby is what this graph shows, you just set delay to be 0 ms. This is where Dirac doesn’t do a good job since you needed more correction. If the surround Dolby is what you are showing, in theory, you want to change it from 5.4 seconds to 1.6 or 1.7s. That way the alignment is based upon the sound that is coming from the ceiling. Right now, the ceiling bounce happens after the Atmos content passes the desired point.
 

mishmushkela

New Member
Joined
Dec 8, 2023
Messages
4
Likes
0
Thanks for the help.
I put those values in and it felt better. Will do some additional testing. Maybe in time there will be an update to Dirac.
 

geodon005

New Member
Joined
Nov 28, 2023
Messages
3
Likes
0
GXAlan:
if you wouldn't mind looking at my graphs. . . .
 

Attachments

  • Screenshot 2023-12-09 at 4.31.45 PM.png
    Screenshot 2023-12-09 at 4.31.45 PM.png
    844.4 KB · Views: 35
  • Screenshot 2023-12-09 at 4.31.58 PM.png
    Screenshot 2023-12-09 at 4.31.58 PM.png
    841.2 KB · Views: 33
OP
G

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
3,757
Likes
5,806
GXAlan:
if you wouldn't mind looking at my graphs. . . .
That’s fine. You can boost the bass if you want. My curve is flat but you can match the speaker’s native bass boost if you like.

See the top right in blue “Set Target”? Click the impulse response option. That lets you know the delay from on axis to ceiling bounce.
 

geodon005

New Member
Joined
Nov 28, 2023
Messages
3
Likes
0
That’s fine. You can boost the bass if you want. My curve is flat but you can match the speaker’s native bass boost if you like.

See the top right in blue “Set Target”? Click the impulse response option. That lets you know the delay from on axis to ceiling bounce.
Dumb question: boost the bass curve, or in the AVR itself? Or do you mean that my graph already shows a bass boost?
 

D8hS

New Member
Joined
Nov 18, 2023
Messages
4
Likes
3
@GXAlan Where are we able to adjust the delays? It seems like all of the distances, levels, and crossovers are grayed out in my 4800 and I'm not seeing any delay values in Dirac itself.

Edit: Bah, it's a limitation of using Bass Control. The distances are not greyed out if I simply use DLRC
 
Last edited:
Top Bottom