Redefine your budget room EQ 'flat' target curve to Harman's pro curve

Apr 19, 2015
51
0
0
#1
Many people have indicated that consumer-grade Audyssey room EQ is worthless to them because the target curves are wrong. They leave it disabled, in many cases losing the Dynamic EQ loudness compensation control and Dynamic Volume loudness compression control at the same time. Presumably other budget auto room EQ could have similar issues.

Until this moment the only alternate options to the typical 'flat' room EQ of budget receivers have involved additional investment:

$$ Pro grade receiver with built-in pro room EQ (e.g. Audyssey MultEQ-XT32 Pro) with adjustable target curve
  • increases cost of the receiver
  • allegedly has a clumsy interface with limited flexibility

$$ Dirac Live Front-end DSP (e.g. Dirac Live nanoAVR-HD)
  • has limited inputs and must understand all your audio streams
  • more functionality == much more expensive product

$$$ Dirac Live back-end DSP (e.g. Dirac DDRC-88A)
  • requires more expensive receiver with line out
  • requires external amplifiers
  • introduces an additional A/D/A conversion to the signal path

Budget-conscious audiophiles might also have a fourth option with little to no additional investment required.

$ Use your sound card to tweak the frequency response of your measurement microphone and fool your 'flat' budget EQ into a new default curve.

For example, the Harman curve seems to have about 3dB per decade of lowpass-ish treble rolloff between 20Hz and 20KHz. Invert that to 3dB per decade of highpass-ish bass rolloff in the microphone's signal path, and normalize to unity gain at 600Hz. When the room EQ is done with its auto calibration, the resulting room EQ curve will look very much like Harman's.

Alternatively, the JBL system in @amirm linked article uses about 2.5dB per decade slope. You can pick whatever slope or shape sounds good in your room.

There might be some unforeseen complications such as:

  • tweak-induced skew on the -3dB algorithmic cutoff endpoints beyond which room EQ will not correct the response
  • saturation of the gain coefficients in the FIR EQ chain

It might be necessary to go with a less steep curve and live with less final treble attenuation in the resulting speaker power response, but it should still work in principle as long as such limitations of the built-in algorithm are avoided or minimized by choosing an appropriate tweak. Several people have already indicated on forum posts elsewhere that the Harman curve has more treble attenuation than they like.

Let me know what you think about this tweak. At this time it is only conceptual but I expect some enterprising person might come up with enhancements to the concept and maybe even help get it working. My experience with room measurement and sound card EQ applications is nearly identically zero or I would jump in and finalize the solution right now. I need your assistance figuring out the implementation if you want to avoid the wait while I get up to speed.

The only thing that is required to do this tweak is a single analog channel of your computer sound card (preferably one with quality microphone input or at least very good S/N ratio such as 24 bits and sensitive line in) and some sort of equalizer algorithm to run on it, such as parametric or graphic EQ VST plugin.

This is a one-time tweak with fixed results so iteration may be required to fine-tune it.

Here are the issue and the existing tweaks that I am aware of for Audyssey. (I only have used Audyssey but I assume competing budget algorithms have similar issue).

My current receiver (TX-NR929) allows me to use Audyssey Dynamic EQ and Audyssey Dynamic Volume even with Audyssey room EQ disabled, unlike some lesser models I own. Without these algorithms, the sound would suffer here in an apartment where I need to be conservative with volume levels, particularly at night.

Audyssey room EQ default target curves failed the preference DBT at Harman International and was deemed by their expert listeners to be inferior to no room EQ at all. Many audiophiles avoid budget-grade Audyssey because it only has the default curves. In many cases this leaves budget systems with only tone controls and maybe manual graphic EQ that is difficult to tune without laborious room measurements.

Here is why the default curves are deemed wrong.

When a speaker is placed in a normal, somewhat reflective room, the on-axis frequency/phase response and the off-axis frequency/phase response are blended into a single combined power signal at the microphone that is taking the room measurement. Since off-axis frequency response typically differs from on-axis (usually treble rolloff and maybe some contours as well), the composite in-room power response is treble-attenuated.

The treble rolloff in the power response of the speaker is intentional on the part of the speaker designer, not accidental. Audyssey 'flat' or 'music' curve picks an apparently flat target power response curve that is apparently unaware of this intentional treble attenuation. It equalizes the power response flat and that over-boosts the treble.

Only the on-axis response is supposed to be flat and that generally requires an anechoic chamber to measure with a simple microphone. Flat EQ is only appropriate for on-axis speaker response compensation during the speaker design cycle and the result is generally measured in an anechoic chamber where the off-axis response is completely removed by removing as much reflection as possible with aggressive absorption. Absolutely no one listens to any programming in anechoic environment because such rooms are very expensive testing and measuring chambers that also make unpleasant listening rooms.

Audyssey 'reference' or 'movie' curve is also wrong, modeled after the theater X curve treble attenuation or some such that begins at approximately 2KHz, with intentional 'BBC midrange dip' that is not usually part of the power response of a modern quality speaker taking the place of the first few KHz of attenuation and the rest of it implemented as two discrete rolloffs at higher frequencies.

One very simple tweak to the default curve is to use the Audyssey 'flat' or 'music' curve and put approximately +4dB of bass boost plus -4dB of treble cut on it with tone controls. The result is better but still wrong.

Another less simple tweak is to use the Audyssey 'reference' or 'movie' curve and skew the reference level by boosting the total gain of the signal path +5db<-->+12dB with the input attenuator (Intellivolume on Onkyo). Then when Dynamic EQ is enabled, the bass will be somewhat boosted and the treble will be somewhat flat. This is also better but still wrong.

The final tweak is to fix the Dynamic EQ function so its own peculiarities are not so bothersome. Like THX Loudness Plus, Dynamic EQ boosts surround levels at lower volumes but does so far more aggressively. This can be annoying because it causes an 'inside the head' headphone-like surround sound stage from the overly loud surrounds and drowns out the dialog as well as the rest of everything in the channels of the front sound stage. I suppose this gain boost is applied so new owners of such receiver will not complain that they cannot hear the generally quiet surround channels when they turn the master volume down.

The fix is relatively simple though. Just crank down the surround speaker trims by about 0.5dB<->1.0dB per every 10dB below reference on your master volume whenever Dynamic EQ is engaged.

Instead of settling for better but still wrong curves, now you have another option that might let you get closer to the better curve that Harman developed. Combined with the Dynamic EQ tweak it could prove a highly effective way of rehabilitating Audyssey.

Between these tweaks, even owners of budget receivers with Audyssey or other simple room EQ should be able to approximate the customized target curves that the pros are using. For those receivers that demand Audyssey room EQ be enabled in order to use the rest of Audyssey functions, you can also preserve your access to Dynamic EQ and Dynamic Volume functions too without compromising the sound of your system with the default curves that so many learn to hate.

Now for the final piece de resistance. We can also do something similar to what Onkyo's new AccuEQ is doing, something that addresses the transition region.

Instead of picking a curve based on Harman study, why not pick a curve based upon the in-room response of your best-sounding speakers (probably your front l/r)?

Just measure the native in-room response of your front L/R with an application such as REW and then invert the curve with the zero crossing (unity gain) point referenced to 600Hz. Use that as your target curve and derive the corrective microphone filter from that native speaker response. Then you get the native power curve directly off your speakers instead of guessing what it is with something like Harman's preferred curve. Now your whole system is EQd to match your best-sounding speakers.:cool:

If you want, you can even smooth the room modes out of the bass too. Maybe the measurement file can be parsed with a script that loads it into a spreadsheet and applies a least-squares sort of smoothing fit to the bass below the transition frequency of the room (probably transitions between 200 and 300 Hz for the average-sized living room or home theater). Or you could just eyeball the bass with a straight-ish line manually in the spreadsheet if you want or maybe even splice Harman's target curve onto the bass of that curve before running the auto EQ with the resultant compensation filter on the microphone.

With the right VST plugin, you might even be able to manually switch between stored default curves and apply different ones on each channel as the calibration proceeds around the room from speaker to speaker. You can keep the native sound of every speaker in your system if you want to, and just smooth out the response below the transition region.:cool:

Let's complete the concept so we can give it a try.:D

Inform us of which sound card application or plug-in will allow us to implement such EQ curve(s) on a real-time signal.

Most people probably use Windows and this tweak is probably somewhat simple to do with that OS but my system uses Linux so I will not have so much access to the native sound card applications and built-in hardware DSP, else I would just use the Patchmix DSP application that shipped with my E-mu 1212m so many years ago. Any Linux expertise is especially welcome.

Thanks everyone and I look forward to your thoughts.
 
Apr 3, 2010
16,022
0
0
Seattle, WA
#2
Hi Cheryl. This is a clever approach. To summarize, you are letting Audyssey perform the Auto EQ as it normally does, then overlay the target correction using upstream transformation out of the PC. I assume in this context the PC is the source. The playback software needs to have built-in EQ or ability to accept plug-ins such as the VST you mention.

I am a little unclear what you are using REW for. Target curve can be programmed using ear/preference alone. Just dial in what Harman and others use as the starting point, and experiment up and down with your selection of content. Direct inverse correction using REW can be tricky in that it may not always sound good even though mechanically would seem "perfect."
 

LL21

Active Member
Dec 26, 2010
10,550
2
38
#3
Interesting...i could go and find the research, but perhaps someone could explain in non-techie language...

1. Is this the Harmon Target Curve?

Harmon Curve.jpg

2. What exactly does 'target' mean? What is supposed to actually to measure this way? In-room/at seating position, anechoic full-range speaker response?

3. If stronger bass is indicated in the curve for 'in-room ideal target response'...is this because the human ear is less sensitive to bass and therefore it needs to be targeted at several db higher to sound 'balanced' to a human ear?

Sorry for the ignorance, but i dont know and would like to ask the experts here. Thanks...I find it interesting.
 
Apr 3, 2010
16,022
0
0
Seattle, WA
#4
That's correct. A target response is the desired frequency response of the system. It is called "target" because at the detail level, it is never achieved. But if you step back enough to see the forest from the tree, then you would see the kind of target graphs shown. In technical terms, you want to significantly filter the measured frequency response so that you don't see these kinds of variations:



If you ignore the ups and downs, you can see an overall trend. That is the response of the system and one that target curve attempts to modify. If you are using a program like REW, you would want to use 1/3 octave filtering.
 
Apr 19, 2015
51
0
0
#5
1. Is this the Harmon Target Curve?
The Subjective and Objective Evaluation of Room Correction Products

Slide 24: Top (red) trace is actual measured "Average Magnitude Response at Primary Listening Seat" of the Harman room EQ solution. You can infer the target curve from the slope and smoothness of this result. They used lots of DSP to get the line that straight.

2. What exactly does 'target' mean? What is supposed to actually to measure this way? In-room/at seating position, anechoic full-range speaker response?
Every negative feedback system has a reference input signal to bias its operation toward a desired operating point. Auto EQ uses a contoured frequency response curve as the desired operating point of the room EQ and the convergence algorithm attempts to shape the system response to match that curve at the measurement microphone location(s).

3. If stronger bass is indicated in the curve for 'in-room ideal target response'...is this because the human ear is less sensitive to bass and therefore it needs to be targeted at several db higher to sound 'balanced' to a human ear?
No, it is not loudness correction. It is attempting to create a perceptually neutral frequency response using a speaker in a room. The non-neutral reference contour aligns the feedback better to the actual non-flat spectrum of the total sound power emitted by the speaker (sum of all direct and reflected sound).

Sorry for the ignorance, but i dont know and would like to ask the experts here. Thanks...I find it interesting.
:)
 
Apr 19, 2015
51
0
0
#6
Hi Cheryl. This is a clever approach. To summarize, you are letting Audyssey perform the Auto EQ as it normally does, then overlay the target correction using upstream transformation out of the PC. I assume in this context the PC is the source. The playback software needs to have built-in EQ or ability to accept plug-ins such as the VST you mention.

I am a little unclear what you are using REW for. Target curve can be programmed using ear/preference alone. Just dial in what Harman and others use as the starting point, and experiment up and down with your selection of content. Direct inverse correction using REW can be tricky in that it may not always sound good even though mechanically would seem "perfect."
Amir, you get it on the concept but the language is still contorted and you missed the whole point of using REW. Sorry, I was in a rush and probably confused the issue with too many things going on at once in the first post.

I think the following is a better explanation, simpler and maybe even more practical:

Plug the measurement microphone into an analog 31 band graphic EQ and run monotonic 3dB/decade treble boost with 600Hz band at unity gain. Run the EQd signal into the receiver microphone jack.

Run calibration routine. The opposite curve results on the output of the receiver's room EQ function. Any magnitude and phase anamalies of the analog EQ will also be mirrored on the final room EQ so it might pay to use a really good EQ for this.

I was thinking it over this afternoon and realized that any delay in the microphone feedback path would skew the distance settings in the receiver. DSP may have too much latency to do this without saturating the receiver distance settings unless maybe the EQ is run directly in the sound card on quality processor. Analog might have a decent chance of doing it though.

Whether using an analog or digital filter, the calibration will need to be normalized back to its typical non-tweaked distances or the lip sync may be off. This means a manual intervention is required after the calibration completes to restore the non-tweaked cal's distance settings.

I actually have access to a 31 band EQ at my neighbor's and we are thinking about giving it a try. I will let you know.

The part about REW was a suggestion to disregard Harman's target EQ contour and just model the tweak off the native l/r main speaker power response in the actual room. Assumption is room and speaker are already optimized with absorption and placement etc. First-order (linear) least squares fit above Shroeder should do it, with an extrapolation for the subwoofer response. Then the actual slope of the applied EQ will actually match the overall average sound power contour of the speaker in the room rather than shooting for an arbitrary target.

Never know, it might actually work better if the subwoofer acts as an extension of the speaker's own native response. Maybe improves matching or something.

OK so what plugins might do the trick and can we pre-emptively determine if the additional microphone delay will saturate the distance setting?

My distance settings range from 0.15m<-->9.0m (0.5'<-->30'). That translates to about 26.2ms max delay.

I already have distance setting ranging from 4' to 8.5'. That means my system has about 19.1ms of delay remaining for the EQ.

Now, where can I get a handle on the actual realistic delays that analog or DSP 31 band EQ will have? Maybe we can figure out up front what equipment and software might be capable of this tweak.
 
Apr 3, 2010
16,022
0
0
Seattle, WA
#7
The complexity of what you are describing is losing me halfway in the story :). I am unclear what you are trying to do and how. Can you give a one sentence statement of what problem you are aiming to solve?
 
Apr 19, 2015
51
0
0
#8
ULTRAGRAPH PRO FBQ6200

The S/N is rated at 94dB/+4dBu so it will probably need preamp and output attenuation to work with microphone. Not sure how to handle that but he has a 32ch mixing board too so I am sure we can come up with something.

Regarding the use of the speaker's inherent sound power in the room as the target rather than Harman's curve or tuning by ear, the intent is to preserve the overall spectral balance of the speaker. You paid for it, you might as well get the benefit of it.:)

Most of the complaints about the 'wrong' EQ curve I have seen on forums note that if it changes the speaker's perceived response (rather than just smoothing it) it is doing something wrong. These are the claims of the 2 channel purists anyway.

The assumption I made on Harman's own study is that they picked their own target curve based on the sound of the B&W, possibly by default with listening trials prior to the DBT if not by measurement and design.

If you compare Harman's EQd result to the un-EQd speaker on slide 24 (one listening position) you see that averaging the speaker's native response with a linear fit above the transition ~300Hz gives a rough approximation of Harman's EQd response.

I just think it might be more appropriate to let the speaker itself tell you where its own happy operating point is than overlaying a preference on it. Not sure about that, just supposing.
 
Apr 3, 2010
16,022
0
0
Seattle, WA
#10
The assumption I made on Harman's own study is that they picked their own target curve based on the sound of the B&W, possibly by default with listening trials prior to the DBT if not by measurement and design.
Harman EQ system is part of the JBL Synthesis. They do not support its use with any other speaker so for sure it is not based on EQing B&W. The B&W was used in that report to avoid people accusing them of their EQ only working with their own speakers. They showed that the same target curve worked as well on the B&W as it does on their own speakers.

The history of this type of curve goes way back.

Here is a good thread we had five years ago (!) documenting B&K and Tact. B&K goes back decades. The TacT which I personally used dates back to year 2000. Harman follows accepted practices here. New guys on the block didn't. http://www.whatsbestforum.com/showthread.php?723-Target-Curves



I just think it might be more appropriate to let the speaker itself tell you where its own happy operating point is than overlaying a preference on it. Not sure about that, just supposing.
That is not a sound assumption, pun intended :). I used my TacT on my Paradigm speakers for example and was quite surprised to see a) the default not being flat and b) flat sounding so bright.
 
Apr 3, 2010
16,022
0
0
Seattle, WA
#11
I should make a side remark that once you correct the response of the speaker in the room with EQ, the target curve produces far more consistent response. Without it, on one track you think you have too much bass, on another not enough. These are resonances getting hit or not with the music being played. Once you smooth the response, these variations go away and your target curve becomes much more stable and resilient across a lot of content.
 
Apr 3, 2010
16,022
0
0
Seattle, WA
#12
Plug the measurement microphone into an analog 31 band graphic EQ and run monotonic 3dB/decade treble boost with 600Hz band at unity gain. Run the EQd signal into the receiver microphone jack.

Run calibration routine. The opposite curve results on the output of the receiver's room EQ function. Any magnitude and phase anamalies of the analog EQ will also be mirrored on the final room EQ so it might pay to use a really good EQ for this.
Your analog EQ is going to have a phase response that is all over the map. So there is no way to draw any conclusion from using it, and then attempting to correct phase.

Fortunately, phase response is not important. We simply are not sensitive to it despite all the talk online and marketing collateral pushing it.

From David Clarks old but good report in AES Paper, MEASURING AUDIBLE EFFECTS OF TIME DELAYS IN LISTENING ROOMS


"circuit consisting of cascaded first-order all-pass sections was constructed
(Fig. 11). Two channels, each use 15 cascaded opamp sections to achieve a delay
of nearly 8700 degrees
over the audio bend with taps at intermediate delays.
Approximately 7 ms of delay et low frequencies is also produced. The circuit
centers about 630 Hz, the geometric center of the audio range. Although this
circuit does not represent all possible time delay and phase shift
characteristics, it is felt that the amounts and rates of change off these
characteristics are worse than what ts encountered in high quality equipment.

No one to date has been able to detect the presence of this circuit in
hundreds of sensitive double-blind tests on speech or music program material
with peripheral equipment including time coherent speakers and headphones.

A change in sound can be detected using a narrow pulse as a test signal by a
alight TEE-OOM sound. Even this clue disappears however below a mere 1000 or
so degrees. The inescapable conclusion is that frequency response is what we
hear
when phase shlft, time dispersion and Frequency response are all occuring
simultaneously."


At higher frequencies than he tested there can be some audible effects but with respect to where you want to apply strong correction, i.e. bass frequencies, it just isn't a problem: From more authoritative work from Professor Vanderkooy's paper, On the Audibility of Midrange Phase Distortion in Audio Systems:

"Phase distortions accompany many links in the audio chain. The most frequent manifestation of these phase nonlinearities occur near the low- and high-frequency cutoffs of the system, where significant differential time delays occur. At these extreme frequencies, however, the bulk of the evidence indicates that quite sizable phase distortions are inaudible."

[...]

On normal musical material heard via loudspeakers in an average listening room, we have not thus far detected the effect of midrange phase distortions of up to two cascaded all-pass networks of Q <= 2* sqart(2)... but it is clear that the effect, if audible, is extremely subtle.


---

"All-pass" is two filters cascaded back to back which cancel each other out but leave us with a programmable delay. It is the circuit behind the "phase" dial on a subwoofer for example which is a simple delay.

My strong advice to anyone looking at speaker and room performance is to strictly stay on frequency response. Get rid of phase response out of your vocabulary. Focus on getting a smooth, sloping down frequency response and you are golden. Frequency response is easy to understand to boot.
 

LL21

Active Member
Dec 26, 2010
10,550
2
38
#13
Thanks, Gents. The B&K Chart is interesting, and i do recall seeing it before. So in simple man's terms if i were to run a test signal from 20hz to 20khz thru my system, B&K is saying this is what the graph should look like if i put a microphone and recorded it at my listening position?

B&K Target Curve.jpg
 

LL21

Active Member
Dec 26, 2010
10,550
2
38
#14
Yes absolutely ,it won't look anything like that, but it would be perfect if it did!
Using DRC you can achieve an almost perfect curve, with very little +/- variation .
I find a flat response does makefor a far more enjoyable listen, but it may take a little time to get used to.
Keith
Thanks...but just so i get this right for future reference, this curve is not flat...it is 3db higher in bass, 0 in mids (relatively speaking) and then 3db lower in treble.

Are they suggesting NOT to make it ruler flat but to speaker to have a stronger bass response in db terms and roll the treble? sorry...no techie here, so just asking based on my simple man's take on the chart.
 

dallasjustice

Member Sponsor
Apr 12, 2011
2,090
0
0
Dallas, Texas
#15
I've owned several different target based DSP softwares. I always use the downward tilted curve others described supra.

I've found that at the extremes, it's always best to modify the target to match the speaker's natural response at seated position. Acourate does this best, IME. For example, most speakers beam at very HF. Moreover, most all rooms will disproportionately absorb HF above 10khz with pretty much all speakers. So, the overall response at seated position drops off fast above 10khz. It's best not to fight this behavior with the target since the problem is really not fixable with EQ. Even the typical tilted curve is usually too flat at these frequencies. Acourate allows for a small hook at the very top end. This works well.

I've also played with different high pass filters at the extreme low end. This is much trickier due to the frequency. Ive tried very steep high pass filters to filter out unwanted infrasonics. This seems to be a good idea. However, steep LF filters will cause preringing if they are FIR and post ringing phase shift if IIR. IMO, it's best to make a more gentle LF roll off.
 
Oct 15, 2012
100
0
0
#16


Instead of picking a curve based on Harman study, why not pick a curve based upon the in-room response of your best-sounding speakers (probably your front l/r)?
I described a manual method for determining the best curve for your room nearly ten years ago here. However, most people these days seem to be inclined to use automated systems like Audyssey. I imagine most people would be happy with Audyssey’s flat curve, along with a suitable boost of the subwoofer level and perhaps some outboard EQing for the same.


Plug the measurement microphone into an analog 31 band graphic EQ and run monotonic 3dB/decade treble boost with 600Hz band at unity gain. Run the EQd signal into the receiver microphone jack.
The measurement mic requires phantom power and a mic pre amp. You can’t plug it directly into an equalizer.


I actually have access to a 31 band EQ at my neighbor's and we are thinking about giving it a try. I will let you know.
ULTRAGRAPH PRO FBQ6200
Trust me, you don’t want to connect a cheap equalizer to your system. If you want to use an equalizer, get a high-quality model like a vintage Yamaha YDP2006.

Regards,
Wayne A. Pflughaupt
 
Apr 3, 2010
16,022
0
0
Seattle, WA
#17
Those are phenomenal articles Wayne and must read for anyone interested in this topic. They are just as valuable today as they were then for our audience here which does not use AVRs or Audyssey.

We agree on the recommendation to simply boost the sub output as a cheap/easy/free approximation to a sloping down curve for people who do use Audyssey.
 
Apr 3, 2010
16,022
0
0
Seattle, WA
#19
My pleasure. It is well deserved. I recall landing on those and your other excellent articles on directions of measurement mics and such.

Would you consider copying those here and I make them sticky? Or do you prefer that people to go HTF and read them?
 

JackD201

[WBF Founding Member]
Apr 21, 2010
10,994
8
38
Manila, Philippines
#20
Long time no see Wayne, nice to see you posting here again.
 

About us

  • Founded in 2010 What's Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals we enjoy learning about new things and meeting new people and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing