This all came up because I wanted to figure out how ART and similar reduce decay times and see if I could do it manually. Originally DIRAC said they were going to use "cancellation" but that has been walked back. AI said it could make a filter that targeted decay similar to how art and other MIMO systems work and it was different that just a amplitude filter. AI claims the the decay filters it's script generates are inherently stable. Now I am more confused than ever but still curious about how ART works even if I can't do it manually, I would still like to understand it.Ran the simulations in Python. The HTML version of the Jupyter notebook viewable with an browser is in the ZIP archive. I will add my explanations later when I have more time.
Yes. As expected, the unstable filter blew up. Also, in this case it is simply not possible to implement pole/zero cancellation. It is because the correction filter processes the electrical signal, and the room effects are induced by the room in the acoustic domain. The idea is to pre-adjust the electrical signal to compensate for the room effects. The correct filter acts alone, and if it is unstable, there is nothing to stop it from blowing up. The room effects the correction filter was to correct are way downstream. The filter and room can't "work cooperatively" with each other (i.e. for the room zero to cancel the unstable filter pole) by having the room inside the feedback control loop, which is how typical control systems are constructed to correct for errors.
Regarding whether or not room modes are minimum phase (MP), they aren't. They sort of behave like MP, but Dr Earl Geddes, who has a PhD in acoustics, said this:
Multiple Small Subs - Geddes Approach
John I had not thought about the electrical filter situation, but agree that the same thing could occur. And different decay rates of the other modes is a good point. In school we did a calculation of a coupled room where the coupled room was very low damping and the main room was highly...www.diyaudio.com
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Dirac Live ART vs. Custom DAW-based MIMO Calibration Objective and Subjective Comparison
- Thread starter Rmjesty
- Start date
3ll3d00d
Addicted to Fun and Learning
- Joined
- Aug 31, 2019
- Messages
- 502
- Likes
- 363
IME regular peq aimed absolutely correctly does the job , ie even being few tenths of a Hz off makes a difference
3ll3d00d
Addicted to Fun and Learning
- Joined
- Aug 31, 2019
- Messages
- 502
- Likes
- 363
I thought the measurements posted elsewhere on this site show use of delayed energy to cancel out? I got the impression it's therefore a fancier, more elaborate and automated, bagend etrap sort of approach
I’m not sure I get this. I would think you can factorize the transfer function into a factor describing any given room mode and «the rest» to a very good approximation. The factor of the mode should be minimal phase, shouldn’t it?Regarding whether or not room modes are minimum phase (MP), they aren't. They sort of behave like MP, but Dr Earl Geddes, who has a PhD in acoustics, said this:
Multiple Small Subs - Geddes Approach
John I had not thought about the electrical filter situation, but agree that the same thing could occur. And different decay rates of the other modes is a good point. In school we did a calculation of a coupled room where the coupled room was very low damping and the main room was highly...
I started a thread a little while ago looking at the theoretical possibility of using minimum phase PEQ to correct for SBIR. SBIR is a multi-path process which means it is not minimum phase (MP). I was surprised that PEQ's could do a decent job (theoretically) to flatten some of the SBIR dips. So SBIR largely behaves like MP, but not right at the dip. (See Keith_W's simulations on post #5)
However, for my example case, the wisdom of applying a huge amount of boost to correct for deep null that are very localized in the room is another matter.
Correcting Room Interference Dips with Minimum Phase PEQ
There are discussions in the miniDSP Tide16 thread (started about here) regarding whether or not to correct a room response dip, and I can't help myself and want to add my 2¢ :p Since the topic is not really related to the miniDSP product, I think a new thread would help me stay on the good side...
www.audiosciencereview.com
Yes, SBIR appears to be a non-minimal phase phenomenon that can the handled by EQ to some degree if you do it «right». Floyd Tools has a section about it in Sound Reproduction. In the 4th Edition it’s in Sec. 13.3.I started a thread a little while ago looking at the theoretical possibility of using minimum phase PEQ to correct for SBIR. SBIR is a multi-path process which means it is not minimum phase (MP). I was surprised that PEQ's could do a decent job (theoretically) to flatten some of the SBIR dips. So SBIR largely behaves like MP, but not right at the dip. (See Keith_W's simulations on post #5)
However, for my example case, the wisdom of applying a huge amount of boost to correct for deep null that are very localized in the room is another matter.
Correcting Room Interference Dips with Minimum Phase PEQ
There are discussions in the miniDSP Tide16 thread (started about here) regarding whether or not to correct a room response dip, and I can't help myself and want to add my 2¢ :p Since the topic is not really related to the miniDSP product, I think a new thread would help me stay on the good side...
Regarding room modes and minimal phase, I was of the impression we were discussing peaks? These are resonances, and should be minimal phase and invertible, I believe. But of course they still ring even if they are notched down, so it’s an approximation.
So here is what I found out taking measurements at LP (one position only) with one sub. I only targeted the mode at 29 Hz. See spectographs below.
1. The modal filters AI came up with are stable. I had no problems with multiple sweeps with 2 versions.
2. I did not see much if any difference in the Modal filters and a plain old PEQ as far as decay goes.
3. It seems to me modal filters are more about targeting a decay time instead of an SPL level. Maybe useful as I like the looks of the "aggressive" spectrograph overall even though the SPL is lower, I have not tried listening yet.
4. Seems like ART is doing more than just cutting amplitude in order to cut decay as mentioned. Still hope to emulate it manually.
Raw Sub
-12 dB 29 Hz Q10 PEQ
Standard AI Modal Filter
Aggressive AI Modal Filter
SPL Raw=purple, PEQ=Violet, AI Modal Standard= Orange, AI Modal Aggressive= Green
1. The modal filters AI came up with are stable. I had no problems with multiple sweeps with 2 versions.
2. I did not see much if any difference in the Modal filters and a plain old PEQ as far as decay goes.
3. It seems to me modal filters are more about targeting a decay time instead of an SPL level. Maybe useful as I like the looks of the "aggressive" spectrograph overall even though the SPL is lower, I have not tried listening yet.
4. Seems like ART is doing more than just cutting amplitude in order to cut decay as mentioned. Still hope to emulate it manually.
Raw Sub
-12 dB 29 Hz Q10 PEQ
Standard AI Modal Filter
Aggressive AI Modal Filter
SPL Raw=purple, PEQ=Violet, AI Modal Standard= Orange, AI Modal Aggressive= Green
Attachments
My AI friend said that the most accurate measurement for decay in REW was to read the decay peaks off the spectrogram, can't remember why but the reason sounded good.... doesn't mean it is correct. The only thing that changed between measurements was the filter. See below for waterfalls:
OP
Rmjesty
Member
- Joined
- Jul 28, 2025
- Messages
- 53
- Likes
- 110
- Thread Starter
- #90
If possible, the difference in decay time should be compared when the frequency response is processed to be as flat as possible, just as I posted earlier. This way, no additional normalization processing is required, and at the same time, a better signal-to-noise ratio can be achieved
Keith_W
Master Contributor
Can you please clarify what you mean by "EQ done right"? Because my simulation that NTK linked to required a 50dB boost and a very narrow Q filter (Q=50). Note: 50dB boost.
3ll3d00d
Addicted to Fun and Learning
- Joined
- Aug 31, 2019
- Messages
- 502
- Likes
- 363
It's true in general but modal peak decay is really obvious on a waterfall and, perhaps more importantly, you can overlay them in rew. In your example, your need to fix the y axis range (it's way too wide) and also adjust the level to match at the frequency range of interest (as @Rmjesty says, preferable is equalise the response to match)
Keith_W
Master Contributor
May I suggest that a superior way to view the decay (and the waterfall) is to use REW's RT60 Decay?
Rather than squint at the waterfall and wondering if the decay is faster or slower, position your pointer in the lower graph and look at the decay in the upper graph. You can clearly see the decay slope that REW used to extrapolate to obtain the T60 (green), and you can also see the noise floor. You can also see how bumpy the decay is. Each bump is a reflection arriving at the microphone. You can really go to town with this display.
Yes, basically you need to make a spatially averaged measurement, either with a MMM technique or an average of sufficiently many point measurements. The idea is to measure how the speaker couples to the boundary, and disregard any effects of the listener. If you need to boost with a lot of dBs and/or with narrow filters, then the set-up is not suitable to this kind of EQ, so better find some other solution. Floyd Toole probably explains it better than me, so I recommmend seeking out Sound Reproduction, as mentioned above.
Keith_W
Master Contributor
EQ alone of that channel will never fill a non-minphase dip. This is because EQ is itself cancelled. If you want to fill that dip, you have to employ another speaker.
I do have Toole's book, but mine is the 3rd edition. I took a look at Chapter 13.3 as you suggested, but it appears that Toole has rejigged the chapters between 3rd and 4th editions. I really should go and buy it
SBIR is a cancellation, yes, but it's not complete. Floyd Toole describes is as "radiation resistance" in this context.EQ alone of that channel will never fill a non-minphase dip. This is because EQ is itself cancelled. If you want to fill that dip, you have to employ another speaker.
I do have Toole's book, but mine is the 3rd edition. I took a look at Chapter 13.3 as you suggested, but it appears that Toole has rejigged the chapters between 3rd and 4th editions. I really should go and buy it
The chapter numbering is different between 3rd and 4th. I still have a copy of my 3rd edition, so I will have a look when I get home later.
- Joined
- Mar 17, 2026
- Messages
- 67
- Likes
- 54
Even if it were minphase, beyond a few dBs is still a bad idea since that drives up distortion.
Keith_W
Master Contributor
Thank you, I just read it. All that Toole says about the Allison effect is that it can be "hidden" by other room effects, so the only way to reveal SBIR is to take an averaged measurement from multiple locations. He says it can be mitigated by better speaker positioning and the only role of DSP is to remove the peaks. I didn't see him say that the dips can be removed by DSP. Well, they can - but you need other speakers to help.
One of the most damaging misconceptions about DSP quality is evaluating it primarily through target-tracking performance. This notion has unfortunately been heavily promoted as a marketing tactic by mediocre automated DSP tools for decades, since it's easy for users to verify and easy to "optimize" with filters. Consequently, many people who believe they are making a technical assessment of DSP end up falling into this trap.
In reality, nothing harms good sound more than aggressively boosting dips merely to force a flat target curve. It's often wiser to accept certain response dips as unavoidable limitations of the room or speaker placement (frequently they aren't even audible) and focus instead on what can genuinely be improved above that baseline.
As @umamiaudio and @Keith_W notes, even careful boosting of minimum phase dips can reduce maximum SPL, while any broader boosting degrades overall sound quality. Yet people still treat "how well dips are filled under the target" as the primary metric for judging a calibration, which is deeply misleading.
An even worse habit is extending target tracking to higher frequencies above the room's transition zone. This has also been falsely marketed for decades for the very same reasons. It's easy to demonstrate and easy to sell and has become deeply ingrained in consumer thinking.
There is some merit in using crossover design to avoid problematic dip regions, though it remains an imperfect solution. Similarly, systems employing additional support speakers with MIMO filtering have significantly improved the situation. However, traditional SISO approaches remain constrained by these misconceptions and will likely continue to suffer from them since it's easier to believe than to understand and the majority will automatically opt out of the latter.
In reality, nothing harms good sound more than aggressively boosting dips merely to force a flat target curve. It's often wiser to accept certain response dips as unavoidable limitations of the room or speaker placement (frequently they aren't even audible) and focus instead on what can genuinely be improved above that baseline.
As @umamiaudio and @Keith_W notes, even careful boosting of minimum phase dips can reduce maximum SPL, while any broader boosting degrades overall sound quality. Yet people still treat "how well dips are filled under the target" as the primary metric for judging a calibration, which is deeply misleading.
An even worse habit is extending target tracking to higher frequencies above the room's transition zone. This has also been falsely marketed for decades for the very same reasons. It's easy to demonstrate and easy to sell and has become deeply ingrained in consumer thinking.
There is some merit in using crossover design to avoid problematic dip regions, though it remains an imperfect solution. Similarly, systems employing additional support speakers with MIMO filtering have significantly improved the situation. However, traditional SISO approaches remain constrained by these misconceptions and will likely continue to suffer from them since it's easier to believe than to understand and the majority will automatically opt out of the latter.
Similar threads
