• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). 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!

The Pre/Pro to buy under $5000?

  • Thread starter Deleted member 46062
  • Start date

dlaloum

Addicted to Fun and Learning
Joined
Oct 4, 2021
Messages
866
Likes
564
I feel your pain, but I view the situation differently. I believe, to a point, that the best sounding speaker after correction is the best sounding speaker before correction. Other factors are the speaker’s driver layout and the port design; or lack thereof. I have never heard a single uncorrected stereo system that could compete with a properly tuned, corrected system. Right off the bat, unless the speaker uses a horn or waveguide that recesses the tweeter into the cabinet, the tweeter and midrange will not have anywhere near perfect time alignment no matter how well engineered the crossover is. Besides impulse response correction or a DSP with delay; a plinth or stand that angles the speaker are the only other honorable mentions to address this issue.

The best systems I have ever experienced all shared a common element: externally powered active speakers; zero passive crossovers and no internal amps. The benefit is being able to DSP each driver individually. REW let’s you measure a driver and set the PEQ target to compensate for the crossover filter slope. That way, you can focus your PEQ on the area near the crossover, probably the speakers weakest area for even response, helping to blend that driver with the other drivers. Once you integrate each of the active drivers with gain, x-overs, delay and PEQ, a lot of DSPs will let you PEQ or Dirac on top of that, but some won’t.

The best case scenario of an uncorrected system is to have the size and design of the speaker perfectly match the environment. The closest I have ever gotten was with Focal Chorus 705v. The inverted tweeter dome, small 5” woofer and front port design worked excellently in a medium-large size room that was 12x24x10…but people want the bass from their Kanta 3s to sound just as smooth and even in the same size room.

While not completely popular, my opinion is the goal should be flat frequency response from 20-20k. The lone exception to this rule is big bass, but I never boost bass. Big bass should only come from big amps and big woofers.

Dirac has the unique ability of essentially unlimited filters because of its mixed-phase design. Dirac is not limited to IIR and PEQ filters like Audyssey. People talk about which target curve is the best. I do not run a target curve in Dirac. I delete all of the “control points” and let it do it’s thing. The result is even response from 10hz-24k, as long as you have the speakers to do it. A lot of “full range” speakers are really only useful from 40hz to about 18k and room correction can not improve that. The most important thing I have found is that the results heavily depend on the calibration microphone and the accuracy of its cal file. UMIK-1 was a monumental step up from the mic that came with the M33.

I use the EQ function in REW to interpret the graphs. If REW can automatically determine the target and generate PEQ for 20-20k without any warnings about the target being too high or low, that’s a pretty flat speaker to begin with. I’m not an engineer, so I can’t really interpret the graphs visually, I also don’t have the background knowledge of what to look for in the graphs. By tweaking parameters in REW EQ, I can easily gather data I can understand. If I run REW EQ at +/- 1db and it only generates 1-3 PEQ plots for the entire 20-20k spectrum, I know I’ve got it pretty dialed in.
The issue is, we are not talking about the "pure" speaker response, but the "in room" response

Dirac does do some impulse processing - but it's major impact (along with audyssey) is in tuning the frequency response ie: the Target Curve.

And the Target curve includes room absorption, reflection, etc...

Optimal "best sounding" room response target curves, from flat F/R speakers are in most cases tilted - ie: higher in the bass and lower in the highs - setting the room target curve to flat (as you propose) is likely to result in a very very bright sounding room.

Whether or not you like it that way is a different issue - but it is certainly a very unusual setup - and not reflective of what most people would consider optimal. (The Harman studies identified an optimal curve preferred by 60%+ of the population leaving 30%+ with different preferences!)

Personally, I find most Harman curves too bass heavy.... I am still experimenting with my Target curves, but I tend to prefer the Dirac default curve bass to the Harman boosted ones (the default dirac still has the tilt, the bass is higher than the highs, but it is a straight line tilt, rather than a further boosting beyond 100Hz of +4 to +10db (so circa +2db bass maybe...)
 
OP
D

Deleted member 46062

Guest
The issue is, we are not talking about the "pure" speaker response, but the "in room" response

Dirac does do some impulse processing - but it's major impact (along with audyssey) is in tuning the frequency response ie: the Target Curve.

And the Target curve includes room absorption, reflection, etc...

Optimal "best sounding" room response target curves, from flat F/R speakers are in most cases tilted - ie: higher in the bass and lower in the highs - setting the room target curve to flat (as you propose) is likely to result in a very very bright sounding room.

Whether or not you like it that way is a different issue - but it is certainly a very unusual setup - and not reflective of what most people would consider optimal. (The Harman studies identified an optimal curve preferred by 60%+ of the population leaving 30%+ with different preferences!)

I think it’s important to articulate the idea of a target curve is to attain repeatable results with different rooms and speakers. Dirac is automatically calculating the filters required for a flat response anyways. The target curve merely lays over that calculation and adjusts it so the filters achieve that specified, custom target. The default target curve yields different calculations with different speakers, signal chains, microphones. The technology is better than human capability but still imperfect:

Different speakers sound different with the same target curves. This concept loops back to the idea that speaker measurements matter and room correction is a must.

I definitely wouldn’t call any of my stereo systems bright. I am very sensitive to harsh tweeters and my hearing range is roughly 25-25k. That said, I definitely require power and detail on the top end of the spectrum for my ears to be satisfied. I imagine a lot of people on this forum have hearing that extends past 20k. Logic would dictate that Hi-Fi enthusiasts have Hi-Fi ears. The idea of High fidelity is better reproduction of the audio spectrum. The larger the spectrum being reproduced, the more realistic the reproduction, especially for people with extended hearing ranges.

If I use the Harman or Nad target curves, the bass sounds bloated and the whole sound in general is nasally and midrange heavy. My 603S2’s can handle all the bass boost you can throw it with Dirac.

When I boil it down, the theory comes down to the job of the recording engineer. If they did their job correctly, a stereo with perfectly flat response is going to yield the ideal sound reproduction.
 

dlaloum

Addicted to Fun and Learning
Joined
Oct 4, 2021
Messages
866
Likes
564
I think it’s important to articulate the idea of a target curve is to attain repeatable results with different rooms and speakers. Dirac is automatically calculating the filters required for a flat response anyways. The target curve merely lays over that calculation and adjusts it so the filters achieve that specified, custom target. The default target curve yields different calculations with different speakers, signal chains, microphones. The technology is better than human capability but still imperfect:

Different speakers sound different with the same target curves. This concept loops back to the idea that speaker measurements matter and room correction is a must.

I definitely wouldn’t call any of my stereo systems bright. I am very sensitive to harsh tweeters and my hearing range is roughly 25-25k. That said, I definitely require power and detail on the top end of the spectrum for my ears to be satisfied. I imagine a lot of people on this forum have hearing that extends past 20k. Logic would dictate that Hi-Fi enthusiasts have Hi-Fi ears. The idea of High fidelity is better reproduction of the audio spectrum. The larger the spectrum being reproduced, the more realistic the reproduction, especially for people with extended hearing ranges.

If I use the Harman or Nad target curves, the bass sounds bloated and the whole sound in general is nasally and midrange heavy. My 603S2’s can handle all the bass boost you can throw it with Dirac.

When I boil it down, the theory comes down to the job of the recording engineer. If they did their job correctly, a stereo with perfectly flat response is going to yield the ideal sound reproduction.
Having experimented with the NAD & Harman curves - I agree with you! - Even though I started with a mere +4db bass boost, I have gone back to the Dirac Default bass boost.

The Harman curve sounded good on a number of movies, but on music they sounded completely wrong (bass bloated is a good description)

The assumption of Engineer/Mastering and perfectly flat response is fundamentally flawed, as frequently part of the mastering engineers work is to make the recording sound good for the "target system" - which could be a boom box, or mid fi home system.... and there are a bunch of assumptions there, which will differ from mastering engineer to mastering engineer, and from studio to studio and target customer to target customer.... - certain frequency ranges will have been boosted, or restrained, based on these assumptions....
Which basically means there is no baseline - at least with the movie standards we have reference levels, and some definitions around things like THX and Imax.... but with Music - there is no way of knowing the original intent - so you might as well tweak it to whatever sounds best to you!

The target curves are an overall "voicing" - from past experience, differences of 0.5db can cause perceptual variations in how individual instruments sound, while keeping the overall system sounding as if it is the exact same "loudness" - and given the levels of variation in speakers and rooms, this type of voicing (Dirac) remains a relatively crude overall tonality instrument. we are nowhere near being able to control the sound in the average domestic space down to below 0.5db accuracy!!!
 
OP
D

Deleted member 46062

Guest
The assumption of Engineer/Mastering and perfectly flat response is fundamentally flawed, as frequently part of the mastering engineers work is to make the recording sound good for the "target system" - which could be a boom box, or mid fi home system.... and there are a bunch of assumptions there, which will differ from mastering engineer to mastering engineer, and from studio to studio and target customer to target customer.... - certain frequency ranges will have been boosted, or restrained, based on these assumptions....
Which basically means there is no baseline - at least with the movie standards we have reference levels, and some definitions around things like THX and Imax.... but with Music - there is no way of knowing the original intent - so you might as well tweak it to whatever sounds best to you!
Ah, but allow me to spin this the other way around. If the recording process already involves EQ, compression, crossover filters, ect to tweak the sound of the recording, why should we tweak it more? By having a stereo with even response, you’re essentially providing the recording engineer with a level playing field. If we develop our own individual target curve to perfectly satisfy our ears, what signal is that curve based on? If the signal is always changing, so is the effect of the target curve, yielding a non repeatable result. I don’t know about you, but I regularly listen to more than one album.

However, I do use the same albums and songs over and over to test audio systems. Steely Dan Aja comes to mind. I would go on a limb to say it’s the most popular album for solely for its being recorded very well.

No matter how well a system is tuned, poorly recorded music will still sound poor. If anything, better systems expose more of the mistakes than average ones.

Until we have a technology that can develop target curves with a scientific method that measures the frequency responses of our individual ears, a goal of perfectly flat response is the only way to go.

The target curves are an overall "voicing" - from past experience, differences of 0.5db can cause perceptual variations in how individual instruments sound, while keeping the overall system sounding as if it is the exact same "loudness" - and given the levels of variation in speakers and rooms, this type of voicing (Dirac) remains a relatively crude overall tonality instrument. we are nowhere near being able to control the sound in the average domestic space down to below 0.5db accuracy!!!

Audio as we know it is very crude. I would be thrilled to see what a stereo system of 2122 would feature.
 
Last edited by a moderator:

peng

Major Contributor
Forum Donor
Joined
May 12, 2019
Messages
2,798
Likes
2,305
Until we have a technology that can develop target curves with a scientific method that measures the frequency responses of our individual ears, a goal of perfectly flat response is the only way to go.

My issue with that Harman study and the so called Harman curve is that we don't seem to know at what spl the listeners listened to. It is important because of human's sensitivity to low frequencies vary much more than to the higher frequencies.

Audyssey tried to use their DEQ curve to compensate for that, and based on my numerous REW FR plots I would say they have done a very decent job, but its accuracy would depend on other factors and would be reasonably accurate if the user leave everything unchanged after an auto calibration run. Audyssey's DEQ tilt the curves at various SPL, but my understanding is that rely on the volume setting for the "dynamic" part. I actually plotted my own curves to check if it did what it claimed to do, and found that they did look very close to the ones published by Audyssey on their website (when they first introduce the feature, can't find them now).

I have yet to see any published information on that Harman study in which they identified the most preferred tilt, aka "Harman curve".
In my experience, when using Audyssey, DEQ is a must because I typically listen to 60 to 75 dB spl (average), volume setting in my case would be about -23 to -15 most of the time and if I crank it up to -12 and -10 I would turn DEQ off. I guess that's because I find movie mix generally have exaggerated bass (to my liking..). DEQ would have been better if instead of an offset that can be set in steps of 2 instead of 5, but it is still much better than nothing.

I would love to hear more from @Floyd Toole or @Sean Olive on about that tilted curve they talked about such as did they vary the slope to different spl the listeners were presented with, or just one and they stick with the slope and in that case what spl (average) was fixed at. Is it possible that the Harman curve may actually be sort of, or effectively a partial loudness compensation/equalization, because the participants were listening to significant quieter than say 85 dB average?
 
Last edited:

Floyd Toole

Active Member
Audio Luminary
Technical Expert
Industry Insider
Forum Donor
Joined
Mar 12, 2018
Messages
285
Likes
2,500
My issue with that Harman study and the so called Harman curve is that we don't seem to know at what spl the listeners listened to. It is important because of human's sensitivity to low frequencies vary much more than to the higher frequencies.

Audyssey tried to use their DEQ curve to compensate for that, and based on my numerous REW FR plots I would say they have done a very decent job, but its accuracy would depend on other factors and would be reasonably accurate if the user leave everything unchanged after an auto calibration run. Audyssey's DEQ tilt the curves at various SPL, but my understanding is that rely on the volume setting for the "dynamic" part. I actually plotted my own curves to check if it did what it claimed to do, and found that they did look very close to the ones published by Audyssey on their website (when they first introduce the feature, can't find them now).

I have yet to see any published information on that Harman study in which they identified the most preferred tilt, aka "Harman curve".
In my experience, when using Audyssey, DEQ is a must because I typically listen to 60 to 75 dB spl (average), volume setting in my case would be about -23 to -15 most of the time and if I crank it up to -12 and -10 I would turn DEQ off. I guess that's because I find movie mix generally have exaggerated bass (to my liking..). DEQ would have been better if instead of an offset that can be set in steps of 2 instead of 5, but it is still much better than nothing.

I would love to hear more from @Floyd Toole or @Sean Olive on about that tilted curve they talked about such as did they vary the slope to different spl the listeners were presented with, or just one and they stick with the slope and in that case what spl (average) was fixed at. Is it possible that the Harman curve may actually be sort of, or effectively a partial loudness compensation/equalization, because the participants were listening to significant quieter than say 85 dB average?
I just stumbled into this discussion. I addressed the essence of this in Jan 2020: https://www.audiosciencereview.com/...ut-room-curve-targets-room-eq-and-more.10950/

Above about 500 Hz room curve "targets" are of little use in describing the sound from a loudspeaker except in terms of overall spectral balance. To detect (and hopefully correct) timbre-distorting resonances one needs a set of anechoic curves. Below that frequency we encounter three issues of note: room resonances, boundary effects, and the equal-loudness curves. The first two are linear in the sense that sound level is irrelevant, and steady-state room measurements are useful.

The sensitivity of ears to perceived bass level as a function of overall volume level is always with us, even at live performances. Obviously ANY subjectively-determined "room curve" will at low frequencies be determined by the playback level in the room, as well as the bass balance built into the program material, because it had to be monitored at a sound level determined by the mixers and musicians - part of the "circle of confusion". There are no standards for sound level in the music industry so recordings vary in how much bass is needed for gratifying playback at any volume setting. This clearly is something that cannot be decided once and for all in a sound system. We need access to adjustable bass if we are fussy. Fortunately most people regard excessive bass as a forgivable sin.

The Toole and Olive listening sessions were conducted at a comfortable "foreground" listening level - no single number is adequate to describe it. It was a level agreed upon by numerous listeners as being what they would use if listening seriously. That said, I know a few people who listen at (to me) uncomfortable sound levels, and a few others who clearly aren't interested in the details that are lost below hearing threshold. Maybe they have hyperacusis - lots of people do. Look it up.

None of this has anything important to do with the choice of loudspeaker for quality listening - the absence of timbre-distorting resonances and the uniformity of the direct sound is paramount in that context.
 

peng

Major Contributor
Forum Donor
Joined
May 12, 2019
Messages
2,798
Likes
2,305
The sensitivity of ears to perceived bass level as a function of overall volume level is always with us, even at live performances. Obviously ANY subjectively-determined "room curve" will at low frequencies be determined by the playback level in the room, as well as the bass balance built into the program material, because it had to be monitored at a sound level determined by the mixers and musicians - part of the "circle of confusion". There are no standards for sound level in the music industry so recordings vary in how much bass is needed for gratifying playback at any volume setting. This clearly is something that cannot be decided once and for all in a sound system. We need access to adjustable bass if we are fussy. Fortunately most people regard excessive bass as a forgivable sin.

Many thanks Dr. Toole for chiming in and included more details then I expected.:) Also thank you very much for the link to the other ASR thread. The highlighted part I quoted above was the very point I tried to make, that in that particular Harman comparison of preference by the listeners to several room correction (RC) software (one being the widely used Audyssey at the time), the listening level would have been a factor.

It seems reasonable to say that the ones (Room correction software) that did not have the tilted bass response that increased the bass response level gradually from say below 120 Hz aka the "Harman curve" in the audioholic (or bassoholic) world, was one of the reasons why people did not prefer the RC's that had the flat bass target curves, and favored the ones that have a tilted target curve because most people may not only regard excessive bass as a forgivable sin, but actually enjoy committing the sin.

Edit: For better clarity, I would add that since Dr. Toole has now confirmed that "The Toole and Olive listening sessions were conducted at a comfortable "foreground" listening level", that means if Audyssey was one of the several RC software compared, it would likely have suffered even more of a disadvantage in the bass response because of its "flat" default target curve, "more", because the effects of the equal-loudness curves factor would have become even more significant at just comfortable foreground level.

If that's the case, then certain RC software of today's might have addressed, or partially addressed that disadvantage by including the so called dynamic EQ feature that would put such a tilt that increases the low bass gradually towards 20 Hz, (adjustable to some extent) based on the listening level. That's just my thought, not sure if that makes any sense, just seem logical to me.
 
Last edited:
Top Bottom