Help understanding group delay

[Chris A]

Goodness...

Just look at the data.

One might ask how I thought about the zero-all-pass crossover idea (without having to invoke FIR filtering all around, and the latency problems that come with that approach using PCs to do the FIR filtering). Then I bought a Danley SH-50 and did some tests and looked very closely at their (quite involved) passive crossover--after measuring the transfer function (which was difficult to believe how flat it was--in a passive network). That's when the light bulb went on.

By the way, I've got a 2018 Danley SH-50 in pristine condition in storage right now--looking for a new home... I bet it will be the best sounding center loudspeaker that you've ever heard.

Chris

 

[Arindal]

This effect suddenly appeared when I flattened the phase response of the entire setup, (i.e., well after flattening the amplitude response of the 5.1 array). This was as big of a surprise as anything that I had encountered before in my home hi-fi adventures.

Have only once dialed in a setup with Klipsch Jubilees, dealing with a similar issue of ´making them sound transparent instead of crisp with classical music´, so I have no doubt about your report and how you achieved just that.

My question was more referring to: how you can be sure, that it was solely the group delay linearization achieving what you have experienced in listening tests? From how I understand it, differences in amplitude at this or that point could always occur as a result of phase manipulation and replacing x-over filters, so I would not rule out the theory that it was solving these problems laying the base for improved clarity and transparency. Maybe I am biased here, as I usually deal with environments not showing any signs of harshness, and if ever, they could be traced down to typical issues such as edge diffraction, narrow-banded cancellation, discrete reflections, imbalanced indirect sound or alike, which are clearly amplitude-related or non-linear issues.

Can confirm what you were writing about importance of constant directivity and suppression of early reflections.

 

[audiofooled]

This is why I present a very "conventional" view whilst privately holding different beliefs.

I love the way you say it.

I wouldn't worry about group delay unless you're after bass performance and such room control where you actually want to be able to test for yourself what's audible or not.

Things such as transient response at MLP, transient response non related to MLP, bass performance outside of the room, outside of the house, stereo bass, AE (Auditory Envelopment), very loud and tactile bass without overloading the room, or disturbing the neighborhood, etc.

IME, things that involve strategy of how you get energy out of the room, provided that your system has enough headroom available and you actually want attenuation. Measurement at single position would hardly tell anything useful if this is the case.

Time alignment across frequency, with a hint of a speed of sound, milliseconds, and feet:

 

[3ll3d00d]

I would answer your question directly, but it seems that you might use this moment to continue your learning path in this subject area (passive IIR, active IIR crossover filters and the difference between those filters and PEQs/shelves used in DSP equalization).

I skimmed the thread, the no of words used for what appears to be described makes it not worth reading in detail so onus is surely on you to point to specific points or summarise. As far as I could see the thread seems to be "rely on the acoustic response of the drivers in their enclosure with a bit of additional filtering to suppress any stopband anomalies" and then attributing major improvements to the phase response changes that result from not running with a crossover but relying on acoustic filtering providing by the cabinet/horn/etc arrangement.

 

[Chris A]

Fair enough...

(Then for those that have read the thread...), look at this post for the procedure used to avoid all-pass phase growth induced by IIR-type crossover filters--via use of PEQs and shelf filters only:

https://community.klipsch.com/topic...hase-loudspeakers/page/4/#findComment-2388972

The difference that's apparent is in step 2 of that procedure, i.e., "Set the HF or LF channel delay to get perfect impulse response in the time domain--as seen in the spectrogram view."

Starting at step 3 in the following thread ("3. Take REW sweep with all [crossing] drivers and chosen crossover filters"):

Using REW to Determine Time Delays Between Drivers

Note that nowhere are you trying to use a microphone to determine absolute arrival times separately from each set of crossing drivers, then trying to guess relative time delays to use. Rather, you're using relative arrival times by playing both sets of crossed drivers at the same time to see the resulting time delays/phases.

You're going to have to do this step anyway at some point to verify that any induced polar driver lobes are oriented for flat amplitude response on or off-axis, And you avoid the possibility of absolute timing issues. More on that here: https://community.klipsch.com/topic...e-delays-between-drivers/#findComment-2372726

Chris

 

[Chris A]

My question was more referring to: how you can be sure, that it was solely the group delay linearization achieving what you have experienced in listening tests? From how I understand it, differences in amplitude at this or that point could always occur as a result of phase manipulation and replacing x-over filters, so I would not rule out the theory that it was solving these problems laying the base for improved clarity and transparency.

To be honest, I wasn't concerned with separating the effects of flattening the rate of change of phase with frequency (i.e., group delay) with only phase vs. frequency curve steepness (without large group delay changes). The human hearing system hears both--perhaps in different ways. I do know what I heard. Others have also attested to this effect. That's what I felt was important: an overall improvement in sound quality.

Maybe I am biased here, as I usually deal with environments not showing any signs of harshness, and if ever, they could be traced down to typical issues such as edge diffraction, narrow-banded cancellation, discrete reflections, imbalanced indirect sound or alike, which are clearly amplitude-related or non-linear issues.

If you're using mainly direct radiating loudspeakers, I've noticed that there is a "dulling effect" (for lack of a better phase) that occurs relative to using fully horn loaded drivers in significantly higher efficiency applications. That's been my direct observation after experiencing the effect myself, and reading and listening to others talk about it. You may not hear the effect due solely to this.

What I've heard from fully horn loaded loudspeakers is that you really do need to correct/flatten their phase response (relative to much lower efficiency direct radiating loudspeakers) in order to avoid that subjective harshness. I've also observed that virtually no one is currently doing that.

By way of example, the JBL M2 is using a hybrid horn-direct radiating woofer setup for instance, and therefore the harshness effect thereby greatly diluted by that particular detail--since the fully horn-loaded loudspeakers I use and experience that effect were corrected for phase at frequencies below ~400 Hz when I started to hear this effect.

Just an observation: I've found that many people want to forget or even suppress the effects of directivity control and higher efficiency below 400-500 Hz. I've instead discovered that there is much to learn and experience in this area.

Another interesting effect is the difference in fully horn-loaded loudspeakers playing at very low SPL and hearing details and overall tonal balance not heard using much lower efficiency direct radiating loudspeakers--which seem to exhibit a definite change in clarity and tonal balance at some SPL threshold. I'm not sure that the exact mechanism of "why" has been fully investigated. But if this effect is subjectively true (via listening tests like Toole and Olive used), then I'm sure there are identifiable reasons why this might be so.

Chris

 

[audiofooled]

Just an observation: I've found that many people want to forget or even suppress the effects of directivity control and higher efficiency below 400-500 Hz. I've instead discovered that there is much to learn and experience in this area.

I agree. But it may be important to note that small room acoustic environment may require some additional means in order to explore/experience the potential psychoacoustic effects, which there are many. Sadly not all that well documented.

 

[Arindal]

I do know what I heard. Others have also attested to this effect. That's what I felt was important: an overall improvement in sound quality.

Absolutely agree, and no issue with that. I am just reserved when it comes to assigning the improved transparency/reduced harshness to the group delay or phase linearization, as my personal experience is telling me that in most cases it was amplitude-related or nonlinear flaws causing such.

If you're using mainly direct radiating loudspeakers, I've noticed that there is a "dulling effect" (for lack of a better phase) that occurs relative to using fully horn loaded drivers in significantly higher efficiency applications.

Not sure we are talking about the same thing, but intuitively I would associate a ´dulling effect´ with differences in directivity, reverb tonality, dissipation and wavefront form (spherical vs. planar vs. cylindrical), or with room-related issues (e.g. imbalanced absorption). Used some speakers in the past which are infamous for phaseshift and delay issues (such as MEG) alongside others which are de facto linear phase (like KSD and Kii), or even ones with switchable DSP (such as K+H, predecessor to Neumann), and I really could not figure out any dependance of tonality vs. phase.

In contrary, looking at directivity index measurements prior to tuning speakers in-room, I could pretty reliably predict which speakers will have not dulling issue at all - those with constant directivity or slightly decreasing D.I.

By way of example, the JBL M2 is using a hybrid horn-direct radiating woofer setup for instance, and therefore the harshness effect thereby greatly diluted by that particular detail--since the fully horn-loaded loudspeakers I use and experience that effect were corrected for phase at frequencies below ~400 Hz when I started to hear this effect.

Absolutely do not doubt your experience, but that would be one of the examples I would not even start fighting harshness or horn colorations/dullness as it appeared to be futile.

Another interesting effect is the difference in fully horn-loaded loudspeakers playing at very low SPL and hearing details and overall tonal balance not heard using much lower efficiency direct radiating loudspeakers--which seem to exhibit a definite change in clarity and tonal balance at some SPL threshold.

I know which effect you mean, but have always associated it with a more advantageous ratio of direct sound vs. indirect sound, not with sensitivity and power. Reason being, I have worked a lot with line sources and cardioids, which in my understanding offer an even better subjective detail resolution at lower SPL than horns, but are notoriously inefficient.

 

[levimax]

Just an observation: I've found that many people want to forget or even suppress the effects of directivity control and higher efficiency below 400-500 Hz. I've instead discovered that there is much to learn and experience in this area.

I agree with this. To me it seems that there is a lack of "settled science" on LF audibility issues and since reproduction of LF is inherently difficult and expensive compromises are made and then the excuse of "not audible" is used even if there is no science backing it up. People are looking for electronics with SINAD of 100+ but yet tolerate SINAD of <10 and group delay of over 100 ms for LF? In addition to group delay I think the other areas where LF audibility thresholds are unsettled is distortion, directivity, and localization as well as related topics like "stereo vs mono" bass and AE.

 

[john61ct]

Yes just from my reading I agree these are all very thorny and interesting topics.

So far group delay (within limits) is in itself of relatively low importance overall within the larger LF topic domain

 

[Arindal]

To me it seems that there is a lack of "settled science" on LF audibility issues and since reproduction of LF is inherently difficult and expensive compromises are made and then the excuse of "not audible" is used even if there is no science backing it up.

Wholeheartedly agree. There have been some controlled experiments taking place when K+H launched their first DSP-controlled loudspeaker, as you could switch it from minimum-phase behavior at lower frequencies to true linear phase. Having taken part in one, I would say the difference was clearly audible, not as dramatic as expected in terms of ´bass timing´, but audible.

What is a problem for controlled experiments, in my understanding, is the interaction between group delay distortion and decay. If you get the first wavefront right from technical perspective, little is said about how long the lower frequencies will decay, which I personally tend to think of as bigger problem than GD, but they certainly add up.

 

[thcdru2k]

I’m trying my hand at reducing excess group delay and keeping it as smooth as possible through the crossover region. Sanity check welcome.
Setup: miniDSP Flex HTx, Canton Karat 300, dual HSU VTF 15H (Mk1 + Mk2). This is an in room REW sweep at MLP, Excess Group Delay, no smoothing. Sweep is both front channels and both subs.
Filters: mains HPF 60 Hz LR24. Subs LPF 80 Hz LR24 with 16 Hz BW12 HPF. I’m keeping the subs at 80 because my center crosses at 80 and surrounds at 90, so 80 is the common bass point across channels.

 

[Flaesh]

I’m trying

The scale of the graph is daunting; even a leisurely pedestrian would move 4-5 feet in a second. Can you upload .mdat?

 

[thcdru2k]

The scale of the graph is daunting; even a leisurely pedestrian would move 4-5 feet in a second. Can you upload .mdat?

 

[ZolaIII]

Despite my efforts understand GD I'm still baffled. So, rather than waste people's time here I invoked ChatGBT to get jump startged. That actually made it worse, so maybe we can start over. Here are a few nOOb questions.

• What scans do I take with REW? I can use REW just fine, but it's not clear what to measure. Individual subs (I have 2)? Individual mains? Everything everywhere all at once?
• Then how to interpret? ChatGBT suggested that just a little sub movement would do a lot. I'm sceptical.
• Subs are ported. Seems like some plug the ports. Is this really a thing, and how sealed is sealed? Really airtight or just block with foam or something?
• What else to be aware of?

To be clear, I don't even know if I have a problem. I used REW for EQing the systemn, XOing subs with mains and setting delays. Honestly it sounds pretty damn good, but the idea of leaving some performance on the table grinds me. Thanks for any and all comments, suggestions, links, tutorials etc etc. Cheers,

As this went very far off (typically for ASR) let's get it back. Your question whosent about room decay times or refraction ratio of course they do play role especially regarding final results you receive and are classified to the room type and size. What you aim at is group delay of subwoofer driver and how to keep it into good to very good range (7~10 ms). First of all it happens that delay increases when he is doing what's very hard for him to do to its FS or under it with port. Excursion becomes unlinear and he can't stop on time to pick next impuls alone as it should. Cure is don't let him do it and cut it above to where he says no. That's of course where phase/impedance meet each other at the bottom.
For example take a look at this Kickers driver measurements to closed buffle (they all are to closed buffle and will differ for open bufle ones to their design [how and where port is tuned]).

KICKER Solo-Baric L7R 45L7R154 - 15x15″ Subwoofer 2x4Ω

Technical parameters and comparison of the KICKER 45L7R154 15x15″ Subwoofer 2x4Ω · Resonance frequency 22Hz · Power handling 1800W · Sensitivity 87dB · Xmax 20.25mm

loudspeakerdatabase.com

In this case impedance/phase crossing is at 30 Hz and when you cut it there you end with under 7 ms GD. You cut it with self low filter as pass filter will already be taken by crossover. Principle is same for both filters (decay under or beyond the frequency limit). We in principle to counter time delay increase amplitude of signal but when driver says no (and you can hear it not by delay but THD) we can't increase it anymore at least not to sound good. So you do oposite thing and let him do better (lower THD or higher SPL) what he can.
I hope that makes it clear.
It's important to have good and to the index room decay times, above 18 is good enough, 25 is rather very good and 30 and above is great along with refraction index especially regarding back to front ones (to ISO 3382 and to the room type). It's not a big problem in small to medium sized rooms it becomes problem in large closed ones (cathedral, movie theatre and so on). You won't be able to do much regarding natural in (your) room decay times (you can improve it to certain amount by accustic treatment) but sure as hell you don't want it worse than it need to be by not having impulses align and or pre ringing. I always say to even do crossover orders to it not speakers design as that's what you gonna get anyhow.