Why is Weak Bass recommended here?

[-Matt-]

This has the formulas that go into the preference score
Thread 'Speaker Equivalent SINAD Discussion'
https://audiosciencereview.com/forum/index.php?threads/speaker-equivalent-sinad-discussion.10818/

I believe this is how @pierre's with-sub preference score is calculated:

spinorama/src/spinorama/compute_scores.py at master · pierreaubert/spinorama

A library to display and compare spinorama (speakers measurements) graphs. - pierreaubert/spinorama

github.com

View attachment 278921

The lf_x (low frequency extension) for the with-sub score is fixed at 4.31 * log10(14.5) = 5.0 for all speakers.

It seems that for the standard preference score, most of the calculated terms only consider frequencies above 100Hz.

Only the low frequency extension term considers anything lower, and this simply takes the -6dB point into account.

The difference for the "with sub" preference score is that the low frequency extension term is replaced with a fixed value. So nothing at all below 100Hz is considered. (But the low frequency extension term only considers things below 300Hz anyway).

So, my hypothetical tweeter-only speaker (under the "with sub" scheme) would get the standard assumed value for the low frequency extension term. But it might score poorly in the lower octaves of the other terms that make up the standard preference score.

Or would it? I'd need to check how those calculations work out it there is zero signal in the lower octaves.

 

[-Matt-]

How about this approach instead...

Take a target curve for the ideal in-room frequency response.

Apply a standard THX 80Hz crossover to decompose this into an ideal sub's response and an ideal speaker's response.

1) Does that crossover slope (for the ideal speaker response) directly equate to what we should hope to see in the predicted in-room bass response? (Or rather the minimum threshold of bass needed for a given crossover frequency).

2) Or is there some additional transform that must be applied (perhaps to compensate for boundary reinforcement) when we move from a REW target response to a predicted in-room response chart?

If the former, then it should be very easy to superpose slopes with different crossover frequencies over the predicted in room response (showing how low the speaker could be crossed).

If the latter, then it would certainly be interesting to try to do the back-transform, to see the form that the low end of that ideal speaker response would need to take, for a few different room scenarios. Does a bass shelf (as seen in the Kef R3 measurements) actually make sense from this point of view, or are they simply chasing more favourable -6dB specs for marketing reasons?

 

[YSC]

I think there are enough detailed technicanal explanation above, just to step in with some of my basic understanding:

1) The preference score is just a preference system, with more detailed criteria than the common review (x stars out of 5)

2) The resonance and irregular in anechoic response AND directivity is difficult if not impossible to fix IRL, while a lack in bass extension can be fixed rather easily by a subwoofer, especially below ~80hz-100hz where human brain cannot localize the source location, and below like 80hz is commonly dominated by room modes which will need EQ and treatment to fix with aid of proper measurement setup, so that part don't mean a lot in real life, in one room it can be really bass shy but in another it can be very boomy

3) IME Pasives usually have a more gently rolled off bass below 100hz, which put in most realistic homes (highly likely pretty close to corners or front wall at least) would have a more balanced bass response slope, where actives like Genelec and Neumann usually have a extended bass before a sharp roll off, reason is likely that for passive users, the design need to balance out by some tricks in most homes expecting the customer not using active DSP or so to tame the room modes, and for those who heavily treat their rooms, more than likely they will do subwoofer integration to fill in the nulls also, while for actives they would assume more customers have a treated studio, but also taking advantage of the active design, more often leave some "room mode" switches to counter the boomy bass in normal room with excessive boundary gain

and even for say, the Revel F35, those full tower with more rolled off bass, normally they are tuned aiming at higher max SPL than flatter actives, which should be the use case in most conditions, ppl with small rooms generally get bookshelfs, and big living rooms using floor standers and with multiple subs in even bigger rooms

 

[kemmler3D]

to see the form that the low end of that ideal speaker response would need to take, for a few different room scenarios.

So, I think the disconnect here is that room scenarios are not predictable (at least, by a speaker manufacturer or reviewer) in any real way. I think it's generally assumed that a speaker that can produce lower bass is better, because it makes crossing over to a sub with the benefit of EQ or DSP easier, and good bass response generally requires EQ either way.


While the formulas don't totally account for this, it's also generally agreed that flat bass output is better than one with peaks and dips.

But it's somewhat just a matter of whether the bass is there or not.

2) Or is there some additional transform that must be applied (perhaps to compensate for boundary reinforcement) when we move from a REW target response to a predicted in-room response chart?

The relationship between an anechoic target response and a PIR is that the PIR sums up (via some formula) the off-axis measurements as well.

The relationship between a good PIR and a good room response measured with REW - I think they are basically the same thing.

 

[-Matt-]

Here is an overlay for the Kef R3 predicted in-room measurement.

The target curves are made in REW, they have a 1.2dB/octave slope and a target level of 85dB at 300Hz.
The crossover slopes shown all have a slope of 24 dB/octave.
The "Ideal Main + Sub" curve (which goes off the chart) assumes that the sub's low frequency roll-off starts at 20Hz and has a 24dB/octave slope.

 

[Koeitje]

Well, accurate-ish.

Accurate enough.

 

[-Matt-]

...and here it is showing just the version crossed at 80Hz (with the corresponding sub response):

 

[fineMen]

Here is an overlay for the Kef R3 predicted in-room measurement.

Would you mind to let the R3 alone? "That weird shelf in the bass response", referring to your initial posting, is a very good solution for arbitrary room gain, and partly Revel does the same. From the few latest posts one might conclude that good speakers for home use have that shelf, good speakers for studio use extend lower on level. Why is that? Former has room gain, latter has room treatment.
The neglect of bass related distortion in the lower midrange is another topic. There are people who recommend to strech the idea of a sub to real woofers that get up to 150Hz or so. For a reason, me thinks.

 

[-Matt-]

Would you mind to let the R3 alone? "That weird shelf in the bass response", referring to your initial posting, is a very good solution for arbitrary room gain, and partly Revel does the same. From the few latest posts one might conclude that good speakers for home use have that shelf, good speakers for studio use extend lower on level. Why is that? Former has room gain, latter has room treatment.
The neglect of bass related distortion in the lower midrange is another topic. There are people who recommend to strech the idea of a sub to real woofers that get up to 150Hz or so. For a reason, me thinks.

I'm interested in why Kef think it is optimal to have such a shelf - and it makes an interesting backdrop for the overlays. (I'm not bashing the R3 - you seem overly sensitive about it)!

The edge of the shelf at 40Hz doesn't look ideal to me because it looks like it will need over 16dB of attenuation from the EQ system in order to allow it to smoothly blend with a theoretically ideal sub using an 80 Hz crossover. This level of attenuation may be beyond some EQ systems. (Cynically, however, it does allow them to claim a much lower -6dB point in their specs).

 

[fineMen]

I'm interested in why Kef think it is optimal to have such a shelf - and it makes an interesting backdrop for the overlays. (I'm not bashing the R3 - you seem overly sensitive about it)!

The edge of the shelf at 40Hz doesn't look ideal to me because it looks like it will need over 16dB of attenuation from the EQ system in order to allow it to smoothly blend with a theoretically ideal sub with an 80 Hz crossover. This level of attenuation may be beyond some EQ systems.

O/k, I'm not exactly excited about your choice of an example, but I'm not angry likewise.

Let's get on good grounds with that!

I offer you (and the board) to exhibit a plethora of measurements using the KEF R3 w/ and w/o woofers, subwoofers of different kind in-room at my home. You don't have to come over. I'll do it, because I'm actually at it anyway.

What do you want to know? I'm serious

 

[Todd k]

Here is my in room frequency of my salon 2’s. Thanks to member sarumbear for helping me get to this point to see if my room needed any acoustical treatment help. Thanks to his sage advice and help my room is pretty good. This sweep was done with my subs off. Just thought you guys would want to see the revel salon 2 frequency response in a actual users room.

 

[-Matt-]

O/k, I'm not exactly excited about your choice of an example, but I'm not angry likewise.

Let's get on good grounds with that!

I offer you (and the board) to exhibit a plethora of measurements using the KEF R3 w/ and w/o woofers, subwoofers of different kind in-room at my home. You don't have to come over. I'll do it, because I'm actually at it anyway.

What do you want to know? I'm serious

Unfortunately, the things that I want to know - you are going to have trouble finding out with your microphone...

How should we define an ideal target bass response?
How would this ideal bass response look as an overlay upon the predicted in-room measurements?
Should the target curves from REW that I posted be modified in some way to account for room gain?

The easy way out is just to say - there isn't an ideal bass response - it depends on many things.
But in that case I'd like to see some representative calculated examples...
How do they differ from the naive curves that I've generated above using REW?
Obviously this requires assumptions - so what assumptions about room size and crossover frequencies etc are reasonable?


I realise that the term "weak bass" in the title is controversial/triggering - let's just refer to it as the "bass roll-off".

So the original question could be rephrased as: Why do many of the recommended speakers (even large ones) seem to have a relatively high bass roll-off?

Good answers to this could be...

• What appears to be a high-roll off is in fact ideal for crossing over to a sub at 80Hz.
• Additional deeper bass is excessive and tends to reduce speaker clarity (both in the bass and at higher frequencies).
• Those speakers are designed with the assumption that they will benefit from XdB of room gain below YHz.

 

[sigbergaudio]

The "problem" is then, why isn't that evident from the predicted in room measurements where the bass appears to roll-off well before 80Hz?

...and can the review measurements be used by those who don't want to use a sub to identify speakers that offer good bass extension.

Is it a problem though? You have the on-axis measurement, which shows an anechoic -6dB at ~40hz. The official specifications state 38hz as the -6dB point, which matches the measurements well. From this you know that an 80hz crossover will not be a problem.

You also know that without a subwoofer, it is not likely to roll-off much before that point due to room gain. The official specifications state an in-room -6dB point of 30hz, which again looks reasonable.

 

[YSC]

I'm interested in why Kef think it is optimal to have such a shelf - and it makes an interesting backdrop for the overlays. (I'm not bashing the R3 - you seem overly sensitive about it)!

The edge of the shelf at 40Hz doesn't look ideal to me because it looks like it will need over 16dB of attenuation from the EQ system in order to allow it to smoothly blend with a theoretically ideal sub using an 80 Hz crossover. This level of attenuation may be beyond some EQ systems. (Cynically, however, it does allow them to claim a much lower -6dB point in their specs).

as mentioned above, the shelf is likely just a solution to most of the home users who won't do extensive bass treatment (it need very thick bass trap which isn't good looking, so bass room treatment are usually minimal in most customers), the quick roll off after the shelf is just normal ported speaker roll off, which is inevitable for ported design and is what I would say is the real "bass extension" for a speaker.

as for the shelf shape and the degree of roll off, it's just likely individual designer/vendor decision or guess on the "average room", which don't really exist in real life, if you read the in room measure thread, it will wildly vary on the final response anyway, one can'testimate and tackle the nulls in room modes, but guestimation and keep the bass from being too boomy, yet still at reasonable level is what these shelves can/intended to do.

For sub integration, without EQ one wouldn't get it right/perfect anyway, and if you use EQ, either Roon, or even with a PC source, using REW and PEQ could do the trick really well to create a good overall SPL.

It makes perfect sense for me to not taken a lot of account to bass extension in these preference score below 100hz, which will fall below schroeder frequency of most rooms which, vary a lot and can't determine how it will change preference to customer

 

[sigbergaudio]

The edge of the shelf at 40Hz doesn't look ideal to me because it looks like it will need over 16dB of attenuation from the EQ system in order to allow it to smoothly blend with a theoretically ideal sub using an 80 Hz crossover.

Can you elaborate on this? A typical AVR will roll this off at 12dB/octave, so it will be 12dB down at 40hz. How is this more problematic with the R3 as opposed to say a typical floor stander with even more extension?

 

[-Matt-]

Can you elaborate on this? A typical AVR will roll this off at 12dB/octave, so it will be 12dB down at 40hz. How is this more problematic with the R3 as opposed to say a typical floor stander with even more extension?

Because after the crossover is applied, the shelf will still appear as a bump in the downwards slope of the main speaker bass roll-off. If integrated with an ideal sub, this would result in a bump in the overall frequency response unless corrected by EQ. The magnitude of the correction needed is quite large (although it won't be >16dB as previously stated). (After an 80Hz crossover is applied it looks likely that there will also be a bit of a deficit in signal between 80 and 100 Hz).

With a (floorstanding) speaker that maintains the -1.2dB slope down to lower frequencies... After the AVR crossover is applied, the response in the transition region (the bass roll-off of the mains) would follow the theoretical/intended bass roll-off (without any additional bumps) and will therefore only require lower magnitude of EQ corrections. (The bump due to the eventual roll-off of the mains will occur at lower frequency where the main speaker signal is more attenuated and it therefore matters less).

 

[fineMen]

So the original question could be rephrased as: Why do many of the recommended speakers (even large ones) seem to have a relatively high bass roll-off?

They don't. The examples above prove otherwise, except for a more or less pronounced 'shelf', which by the way is an idea that goes back like several decades as to anticipate, you name it, room gain. It mitigates the quite sharp roll-off of ported designs below the tuning frequency.

Maybe your confusion originates in the not so often (well?) explained calculation of 'predicted room response'. They take the assumed reflections and add them all together, but for what I know, not as imaginary numbers aka phase is not accounted for. You might get away with such a simplification in the mids/treble but in bass it's for sure superficial. Regarding the listening experience one has to acknowledge that peaks contribute more than dips.

Now, on top of that you're going to introduce another complication in assuming a subwoofer. Again, the phase relation speaker / reflections @ listening position. With your proposal plus the frequency dependend phase of the sub at another not yet specified placement in relation to the main speakers' position(s!) @ listening position. You see where it comes to?

Anyway, come back to me if you need confirmation by measurement.

 

[sigbergaudio]

Because after the crossover is applied, the shelf will still appear as a bump in the downwards slope of the main speaker bass roll-off. If integrated with an ideal sub, this would result in a bump in the overall frequency response unless corrected by EQ. The magnitude of the correction needed is quite large (although it won't be >16dB as previously stated). (After an 80Hz crossover is applied it looks likely that there will also be a bit of a deficit in signal between 80 and 100 Hz).

With a (floorstanding) speaker that maintains the -1.2dB slope down to lower frequencies... After the AVR crossover is applied, the response in the transition region (the bass roll-off of the mains) would follow the theoretical/intended bass roll-off (without any additional bumps) and will therefore only require lower magnitude of EQ corrections. (The bump due to the eventual roll-off of the mains will occur at lower frequency where the main speaker signal is more attenuated and it therefore matters less).

The variations in response below 100hz in a typical room is extreme, way beyond the deviations you discuss here (referring to "bit of a deficit in signal between 80-100hz").

Here is the R3 in-room response from Erin's review:

Whether you will have a peak or a dip between 80-100hz with the Kef R3 is 100% room dependent, and a smooth subwoofer integration will be impossible without DSP, just as with any other speaker.

"If integrated with an ideal sub, this would result in a bump in the overall frequency response unless corrected by EQ"

There will be no bump to talk about when comparing with all the bumps and dips generated from the room. So if one plans to integrate the R3 (or any other speaker) with a subwoofer without EQ, a shelf like this will be the least of your worries (as evidenced by Erin's in-room response). If you on the other hand DO use EQ, it will not be problematic to compensate for the response bump by simply lowering this frequency range slightly. In reality a shelf like this 12dB down will not add a lot to the total energy. I have no idea where you got 16dB from. I suspect it will be more like 1-3dB at most of additional EQ required in addition to the high pass filter itself.

 

[Steve Dallas]

With DSP that looks pretty nice. Of course, I am already aware of the concept of boundary reinforcement.

The graphs in the OP make it look as though the only speaker able to give full signal right up to an 80Hz sub crossover, is the B&W 805s. (Even though I'm pretty sure the larger Revels should actually be able to give much better bass). (It makes it look like these wouldn't even be suitable as surround satellites)!


The "problem" for me is in trying to use the predicted in-room response data to compare different speakers.

How, in the current measurement scheme, is a speaker with superior bass extension to be recognised? The B&W 805S was criticised for being too bassy! How is this quantified within the preference score? Is it given sufficient weighting relative to the smoothness of higher frequencies?

If the theoretical boundary reinforcement can be calculated easily then it should be possible to do something like this...

View attachment 278837

Where the different colour curves to show the minimum bass response needed (in different sized rooms, small, medium, large m3) in order to produce an ideal 80Hz crossover to a sub.

Obviously there would be assumptions, like standardised distances between the speaker and boundary for each of the room size scenarios.

This would then show visually, that whilst Kef R3 might be fine in a smaller room, it might struggle in a larger one (for example).

The dips you see in my uncorrected measurement are SBIR centered at ~120Hz, and a bad room null centered around 64Hz. All speakers I have tried in this room look very similar in those two FR regions. The way the R3s slant down in the bass has little to no effect once room gain is factored in. With other speakers, I just have to cut more when I apply EQ / DSP.

Here is the Philharmonic BMR, which does not slope down in the bass, in the same position:

It is worth noting that everyone who measures speakers has trouble getting accurate measurements below 100Hz. Sterophile, for example, as a curious hump in every speaker JA measures. The NRC (or maybe Axciom, who used NRC's chamber until they cloned it) is on record saying their chamber is accurate above 150Hz. The only ways to accurately measure bass extension are the ground plane method and placing the speaker VERY far away from any reflecting surface. As a result, pretty much all bass measurements should be read with a bit of skepticism along with the knowledge that the room changes things more than a little bit.

 

[-Matt-]

It is worth noting that everyone who measures speakers has trouble getting accurate measurements below 100Hz. Sterophile, for example, as a curious hump in every speaker JA measures. The NRC (or maybe Axciom, who used NRC's chamber until they cloned it) is on record saying their chamber is accurate above 150Hz. The only ways to accurately measure bass extension are the ground plane method and placing the speaker VERY far away from any reflecting surface. As a result, pretty much all bass measurements should be read with a bit of skepticism along with the knowledge that the room changes things more than a little bit.

Indeed, and Amir even says the following (emphasis mine):

The room dominates the response in bass region, not the speaker. You can't make any judgements about that region from measurements. The PIR gives clues as to tonal balance, not exact frequency response in a real room. It represents a room with no room modes.

Many of the comments keep referring to the specific measured response for particular speakers in particular rooms. Of course I know these are filled with many hideous looking bumps that are nevertheless, not that audible (I've made measurements - I was horrified the first time)!

What I was hoping for was a discussion, and perhaps general consensus, on (approximately) what speaker designers should be aiming for at the bass roll-off of the main speakers?

I.e. A hypothetical target - a dashed line that averages out all those room interaction bumps.
Perhaps a few different curves for different room size scenarios.
Different target curves based on whether subs will be present or absent.
But it looks as though maybe the naive crossover curves I posted from REW may be as close as we get.
(Does it even make sense that those crossover curves have a 24dB/octave slope or should instead they be 12dB/octave)?


Some are arguing that the ideal bass response is:
* For a domestic property (with smaller rooms) - the predicted in-room response shown by the Kef R3
* For a treated studio - the predicted in-room response shown by a Genelec 8361

If those statements are true then I'd like to understand further why that might be the case.
How have Kef and Genelec arrived at those optimal responses?
What assumptions have they made about the room in each case?