[Chester]

For the past 3-4 years I have been listening to all of my music, tv, movies etc. with a special EQ 'house curve' I created derived from the ISO 226:2003 equal loudness contours. My filters are designed upon the offsets of the 80 Phons graph shown below. If you are familiar with the "loudness" button on old receivers, this is that but from 20 to kHz and accurate to ~+-1 db (may be a bit less accurate below 100 hz as the curve is quite steep but overall I'd say +-3 db for the entire spectrum).

To implement this sort of setup you will need some experience with digital eq filters and must be mindful (and have good understanding) of the headroom you have to work with. The results do sound quite amazing though Here is the graph from a PDF I used (along with some of my measured points and a more accurate scale on the bottom), along with a Transfer-Function EQ graph from WinISD Pro.
Full-size

Full-size

Comments/questions are welcome I will try to whip up some sample audio files to give a comparison on your systems

 

[Chester]

Here is a demo of the Ear EQ, this also has a highpass filter at 70 hz, 12 db/oct; typically that works well with headphones, usually use ~40 hz for average sized room, hopefully to use the 'cabin gain' effect in conjunction with the 12 db/oct cut to get some more headroom... Ideally 70 hz would be where cabin gain kicks in inside the room you are listening to the file though either way the difference in clarity should be apparent.

You should turn off any house curves you have applied if possible when listening to this however any EQ applied to make the speakers more 'flat' would be beneficial, try it with no eq though too if you want.

http://rapidshare.com/files/310699194/Hotel_California_-_Ear_EQ.mp3.html (good for 10 downloads)
(it's 11 mb, I can make a smaller/shorter demo too if that would be helpful)

http://rapidshare.com/files/310702775/Hotel_California_-_Original.mp3.html
This is the original Hotel California for comparrision purposes

EDIT:The mp3's appear to be down... Does anyone know a good place to upload files to for free?
EDIT2: I put them on rapidshare, they are good for 10 downloads each apparently... if anyone has any suggestions for a better 'dump' site please let me know: otherwise once the 10 downloads have been used let me know and ill repost

 

[Chester]

Now that the links work, please give it a listen I am anxious to hear some opinions

 

[JohnM]

If this is in connection with headphones you might find this post interesting: http://www.head-case.org/forums/hea...32-stax-srm-monitor-history-appreciation.html. I had a set of the Stax Lambda Signatures many years ago and borrowed an ED-1 equaliser to try with them, it worked remarkably well but was rather expensive, so I measured the frequency response and returned it but never got around to implementing the EQ. The ED-1 was specifically matched to the Signatures, but it was extraordinary how much more natural music sounded with it in circuit.

 

[Chester]

It works for both headphones and for a 'typical' stereo system (so this is my 'house curve' I use for all of my systems), this is more compensating for the non-linearities of how we hear however I do want to figure out how to EQ my earbuds with REW too I need to make an artificial ear or something...

EDIT: I will definitely check out that guys post, lots to read though

 

[terry j]

hi chester. luv to see people giving different things a try and reporting back, well done.

have not tried it yet, so all of this is very much in the theoretical/what about world ok?

What I am curious about is, what is your natural in room response?? These thoughts are based on quite a few assumptions, any of which could be wrong. I presume a graph of this nature would be done on headphones, and further assume that FR would be a lot flatter and accurate than any 'normal' in room response...ie so does it transfer across because of that?

Where I am going, have you tried ending up with this sort of response in room??? Obviously I have no idea if it would be better or worse, but if it is true that headphones are more accurate, then the response you are aiming for with your eq would be (effectively) that response at the ear. and would be highly unlikely that you'd get that response 'at the ear' in room.

So, I wonder if you eq'd to get that response at the LP how it would compare??

Anyway, thanks for that, yee haa more stuff to try and experiment with. Will let you know what I think when I finally get around to it.

 

[Chester]

Hey; yes, I do 'shoot for this as an in room response' however I do it 'indirectly', I apply whatever EQ is necessary to get as flat a response as possible; then I also 'throw in' the 7 (or 8 since I usually need 2 filters for the 20 hz value ) parametric eq's to modify the sound from being what I consider (relatively) 'flat in a power sense' (flat db output for a constant input signal) to being 'flat in a perception sense', so if you hear a frequency sweep, the perceived volume (to your ear) should stay the same across the frequency band. There are factors affecting the accuracy

1) how 'flat' did you get the initial system (is the equalized response +- 10, 6, 3, 1 db?) without applying the "EarEQ" points?
2) is the equalized response flat *at the ear* or is it flat where the measurement microphone was?
3) is cabin gain taking over (the bass gets quite 'boomy' if you do not compensate for where the pressure mode takes over in the room...)
4)can the system in question even PLAY the lowest frequencies at an appreciable level (remember you need A LOT of headroom, even with the cabin gain (typically a highpass 12 db/oct filter @ 30-50 hz depending on the room) *so technically I do not have 38.9 db of gain @ 20 hz*, the highpass does need to be *before* the bass gain for headroom, but that would be the typical way of running things anyways...
5)are you going to be listening to the music at 80 phons (on average)? remember, the graph at the top is basically showing the equal intensity to equal loudness conversion; I chose 80 phons to be my reference since it is a pretty average listening volume. I have actually considered doing some sort of secondary eq chain which would be determined by a compressor (and base the compressors gate on the time it takes for the ear to respond to changes in loudness) so I could then figure out how much to add based upon what a 0 db signal is on the card however I just have not had the time...

Anyways; in my experience it does work great on speakers too along with headphones (remember, the sample I uploaded has a 12 db/oct filter @ 70 hz), like I mentioned before, this is basically trying to "EQ for the non-linearities of your hearing" once you get the non-linearities of the stereo (reproduction system) ironed out... Even if you do not EQ your speakers at all it appears to have a pretty big difference (like the "loudness button" on many old receivers).

As a less technical description: my Mom's aunt came to our house and was quite impressed by my stereo setup (it has improved since she saw it even but relative to what the 'average person' has ... if you are on the forum you probably understand ...) Anyways, when she came into the room, she told me something to the effect of "oh I do not like *loud* music, so please don't turn things up loud; I have very sensitive ears"... I said "ok, but this is going to be much different than any stereo you have heard before because of the way I run it"... Long story short, we ended up turning things as loud as they would go and she thought it sounded great (at quiet and loud volumes). One thing she mentioned to me was that "it didn't sound harsh as you turned it up louder". So for lack of a better way to verbalize how something sounds; the EarEQ I made with the 7 parametrics listed above makes the sound "less harsh"...

EDIT: remember: if you are listening on headphones, there is no binaural processing; I typically use the B2SP (http://bs2b.sourceforge.net/) plugin in foobar2000 in conjunction with a VST plugin running the parametric filters I use; along with a resampler to convert what is usually 44,100 samples/sec to 48,000 samples/sec so if you usually have some sort of processing enabled for your headphones, run it the exact same way as you always would. The bass (lower than 70 hz) may be a bit off on your system since I wanted to have the average signal be of decent volume relative to the original file and 70 hz is usually pretty good for headphones: remember though, this is a 20-20kHz eq, so there will be a big difference even if the bass is a bit off.

Addition: I mentioned (above) about converting the flat room to something that is flat to the ear so that a frequency sweep should sound flat; when I listen to music with the eq I do notice that bass lines, piano notes, and instruments do sound much more even (one note is equally as loud as another note, with bass (really deep 'rap' bass) for example, the bass line does not 'fade away' as the sound gets deeper (as I typically notice with rap music on an un-EarEQ'd system))
there is a song by Bent Fabric called "Relax Boy" that has a harp being 'strummed' (I think that is the correct term) starting at a high frequency and ending at a low one; with everything being eq'd it sounds much more even and you can "hear everything well" where normally there would be some parts that are harder to hear... It's sort of a different song but it may help you to hear what I am talking about... Or just download both files I posted (22 mb total) and listen to the original first and the processed one second; the processed one will require a bit more volume since the entire signal is a bit quieter due to headroom issues

 

[Chester]

I should add, I am *pretty sure* that the responses used in the ISO 226:2003 were measured on axis to the ear so there would be a bit of deviation from the differing angle of an on axis response and where the sound is actually coming from; in practice it still works quite well though. check out my posts here: http://www.hometheatershack.com/forums/general-discussion/23089-ceiling-rear-speaker-placement.html (made yesterday coincidently) which are on the topic of "converting" the frequency response at one angle to another with impulses measured by the MIT media lab... I am about 40% there to getting some actual eq values to test

 

[Chester]

Has anyone tried it yet? lol

 

[Synthsayer]

I'm happy to see this thread. It seems that so many of us get too wound up in the goal of 'flat' response from zero Hz to infinity. Myself included.
I hook up a test mic and place it in one spot and measure. Then I do it again. And again...
I first herad about the equal oudness curve when I was started building serious? speakers using Dynaudio drivers in the mid 80's. I am in Kentucky and have been a big fan of the Thiel brothers for a long, long time.
Along the way somebody mentioned the Fetcher-Munson curve of equal loudness and also about how our hearing changes over time.
To get the sound that I find most pleasing, I used my laptop and Room EQWizard's Tone Generator.
I set the signal generator to 'Follow Cursor' and to generate a sine wave. Using the mouse, I changed the frequency moving through the frequencies where my 5 surround speakers crossed over into the subwoofer. This let me hear exactly where peak frequencies were. It turned out they were exactly where the Mark Taw Room Mode Calulator predicted they would be.
A great side effect of this method was that it helped me get my subwoofer pahse adjusted 'correctly'.
In the process of manually sweeping the tone from low to high frequencies I found a significant null around 65Hz. I slowly adjusted the phase control of the sub amp until the null was no longer present and the response 'sounded' even to my ears. I didn't even bothering to hook up the test mic setup again.

 

[Chester]

Thanks for the response Synthsayer! Your methods appear to be a bit more 'subjective' than the way I do things however the end result should be something similar I'd hope

The reason I use the ISO:226:2003 curves [rather than 'eqing each system based upon perception' as I would probably call your system] is because I can then apply the results (the work of figuring out the EQ values) to every system I listen to. In my writeup, I did focus primarily on the idea of EQ'ing for the ear, however most all of the systems I listen to music on are also EQ'ed to be as flat as possible. So the 7 parametric eq's I listed are more like my 'house curve' than the actual EQ that is being applied to the signal. I do program in the 7 filters in precisely, and then also apply the separate system equalization: I do not use them as a house curve, however I have considered doing so as it may save one or two filters... I believe that there may be more variation between stereo channels to do things that way though... Some of the filters (especially the one at 12.3kHz) are a little 'steep' to be used for a house curve IMO.

EDIT: in your other post (http://www.hometheatershack.com/for...-test-mic-needed-you-have-pair-your-ears.html) you also mention using your ears to set the phase delay on your sub (and then not needing the mic again...) do you do any EQ to the rest of the audible spectrum?... Not that I would consider settings phase as EQ per-say...

 

[Synthsayer]

I only use the lowest +2 Treble adjustment on the AVR154 Harman Kardon currently used.
I also use the parametric EQ on the O-Audio sub amp to smooth out a peak around 43Hz.
Recently, I did the unthinkable and hooked up an old Behringer Ultrafex II enhancer between my cable converter box audio outputs and the AVR154. I only use a slight amount of enhancement; setting at "1".
This helps to make up for the poor sound quality of my cable TV provider in rural Kentucky, and my less than perfect hearing from playing in bands for many years.

 

[Chester]

Ahh gotcha; yeah, the EQ I typically do is very 'precise' relative to using the 'treble adjustment'... If that is what you need to be satisfied with your sound though it is probably a lot easier to setup than all the EQ I do

 

[JerryLove]

I'm confused.

If a violin gives out a specific sound, and a speaker is flat enough to exactly reproduce that sound, how will unflattening the speaker "based on how we hear" make it sound "better"?

Or did I miss something important?

 

[Synthsayer]

JerryLove: You will notice the curves at the beginning of the post represent how our ears 'perceive' or notice given frequencies across the range of human hearing. This is based on the assumption that range goes from 20Hz to 20kHz.

Knowing the concept of octaves may help understand what Chester is getting at here.

This is something I have asked people in the DIY and professional speaker design field about for years.

When you look at specs for a speaker an example would be 70Hz to 35kHz. I know what you may be thinking already. Isn't 35kHz well beyond the range of human hearing? Yes, but I do have both Scanspeak and Fostex tweeters that measure up to 35kHz-50kHz.

So, this is supposed to mean the speaker reproduces ALL frequencies between the low and high cutoff equally. Wrong :-(

First of all you are likely familiar with f3, the frequency where a speaker driver or system's low end has rolled off by a factor of -3db. With a low end cutoff of 70Hz, the bass is already less loud before reaching 70Hz. It most likely would start rolling off around 90-110Hz depending on whether the system is sealed or ported and the Qts goal of the designer.

By measuring a speaker system with a test mic and signal generator using a program like Room EQ Wizard you visually see there are little peaks and dips along the way.

Okay, nothing new with that. This just restates what Chester has been getting at.

What our ears hear is not flat. Re: Chesters original post. That is why different speaker manufacturers and designers use different goals of Qts. Without oversimplifying things, a speaker system with a Qts of ,707 is supposed to represent the most even roll off. But, for years there have been systems designed with higher and lower values of Qts that basically have a stronger,( .9-1.2Qts ) low end peak than the .707. Or, a lower value of Qts with a steeper bass roll off of .5-.6Qts.

Higher Qts means more bass punch at the lower limit and lower Qts means less bass punch and smoother roll off. Generalizing, JBL and Cerwin Vega would be higher Qts speakers with what some people refer to as a boomy sound, and Thiel speakers would represent the .707Qts, while most British and European systems have lower Qts and a less punchy sound.

This addresses only the low frequency roll off characteristics. Chester can explain better than me why EQ a 'loudness' and tone controls have been historically used to get speakers to sound the way we like.

After almost 30 years of building speakers and amps, selling some of the most highly regarded gear available, and doing studio recordings I have found that hardly anybody agrees about what sounds 'right'.

I used to listen to a lot of reggae music, but I did work recording a heavy metal band at one time. My friends that liked reggae had totally different systems than the headbangers. Go figure.

After all of this long explanation, the point is that even though a speaker system MAY reproduce a perfectly flat range of frequencies, you, Chester, and me, are going to hear different things from them.

Another topic for later is room modes and room resonant frequencies. That is another can of worms.

Great thread, Chester.

By the way, my subjective measuring method seems to work out like this: I run the tone sweep and adjust the system as previously described. But, when I first play music, a DVD, or listen to TV sound, the subwoofer will be WAY to loud. :-(

I understand that different radio stations and TV networks use what you have described as different 'house curves'. If you go from one city to another, in my case from Lexington, to Louisville, Kentucky you will notice there are differences in EQ and enhancement or compession from one city to the other.

This is really anoying when switching from one cable channel to another, the CBS channel in my area has much more bass than ESPN, or ABC. But, ESPN2 and Fox Sports channel sill have different sounds, yet again. It seems like I get one station set where I like and switch channels only to have a completely different sound.

I won't start on volume levels between cable TV channels. We all know what a nightmare that has been for years.

I didn't expect this to become such a long posting.

I hope I haven't contradicted anything in your postings Chester. It wasn't my intent to do so. Let me know if I have. In any event I defer to your expertise.

 

[JerryLove]

Thank you for the in-depth response, but I'm afraid I'm still confused.

Perhaps it will help with an example? Let's take this speaker http://www.salksound.com/ht1-freq.html.

Now there's a +-1.5db from 35Hz to well, the chart isn't clearly labeled but let's say 16k.

So if I record an insturment making noise between 50Hz and 12kHz, and play it off this speaker, and am sitting on-aixs, the sound that hits my ear should be, withing 1.5db, identical to the sound that I would hear from the insturment itself were I sitting where the mike was when it was recording (though perhpaps at a different overall volme based on my amplification).

Is any of that wrong?

And if that's right: why would I modify the sound... unless I don't like the sound of the actual insurment?

 

[Chester]

Hey, thank you Synthsayer for your lengthy explanation; I agree with the majority of what you have to say the differences in the way we both express ourselves probably accounts for what I wasn't quite able to follow.

In regards to JerryLove's question; the 'reasoning behind this' is not necessarily to play back the sound as if it were in the room, it is trying to achieve the goal of making the sound you hear from 20 hz to 20khz equally perceived by you; so say the violin you alluded to was playing a very high frequency note, your (human's) ears may not have perticularly great sensitivity at the higher strings frequency (or maybe just the harmonics of a lower note) relative to another note. The EQ curve I made is basically trying to make things sound as if we (as humans) can hear every note equally well (assuming the reproduction system is relatively flat and we are hearing an on axis response of the speaker). Try giving it a listen and let me know what you think... I was actually considering putting the two files on youtube or something so that people could more easily play them though I have never used youtube as an uploader, if someone were willing to do that I would appreciate it

JerryLove and Synthsayer, have you listened to the file I put up (the rapidshare file of Hotel California) the link to the EQ'd one still works for me (I believe it needs to be re-posted after 10 downloads)

 

[JerryLove]

In regards to JerryLove's question; the 'reasoning behind this' is not necessarily to play back the sound as if it were in the room, it is trying to achieve the goal of making the sound you hear from 20 hz to 20khz equally perceived by you; so say the violin you alluded to was playing a very high frequency note, your (human's) ears may not have perticularly great sensitivity at the higher strings frequency (or maybe just the harmonics of a lower note) relative to another note. The EQ curve I made is basically trying to make things sound as if we (as humans) can hear every note equally well (assuming the reproduction system is relatively flat and we are hearing an on axis response of the speaker).

But isn't that already in the performance? If I'm listening to an orchestra, and the composer/conductor wants me to experience a bell at a given volume, they play it at the volume that will get that effect.

Perhaps it just boils down to different preferences. I'll admit I'm pretty focused on getting as accurate a reproduction as possible; and that will color my preferences accordingly.

Now if I can just get the engineers to stop "fixing" natural to accomidate iPods...

** edit **

I did DL your mp3 and play it on my computer rig (admittedly, one already running some equalization). The voices were very muddy and difficult to make out, and the drums kicked hard but imprecisely. Perhaps it has a different effect on your rig. For reference: I have the song ripped off CD and did compare them.

 

[Chester]

It is definitely a 'preference thing'...

If your rig is equalized to be as flat as possible then that would be "how I intended it to sound"... granted I use REW to equalize the full range, not just the bass end (IDK how you setup your stuff)... You should also be aware that for signal headroom purposes, I did roll off the bass at 70 hz (so things weren't so quiet relative to 'normal' volume); if you have some headphones with what most people would consider 'good bass' or a pair of earbuds (the in-ear type, not the kind that come with an ipod) you may want to try listening via that route.

If you don't like it that's cool with me too you may be listening at a low volume too though, the curve is based on the 80 phons graph also; your ears could be non-average also; remember, the ISO226:2003 chart is based on what should be average for humans, your ears could be shaped a bit differently than what was average when they did their measuring. Not that anything is 'wrong with you' per-say but the data I used may not work well for you as an individual

From what you mentioned about trying to get as accurate a reproduction as possible, you could also be used to a 'different' listening environment than me; I don't know how much EQ you do or what your system is like, but from what I have heard, if someone has a pair of 'studio monitors' they (the speakers) work great to master music on, however I have been told that most people should not just 'go buy a pair of monitors' for normal listening since they tend to sound sort of 'dead' (I hate the fact that I am having to use subjective 'audiophiley' terms here to get across my point I like to think of myself as a very objective person...)

EDIT: agreed about the engineers 'fixing' things to be normal for ipods; if there were two releases (one for ipods, one uncompressed (greater dynamic range, not the music format)) I wouldn't have a problem; with the advent of MP3's as a distribution system I do not see why that would be so hard to do...

EDIT2: when you said your system has some EQ already, is it eq that you 'did by ear' or did you use a microphone? If you did it by ear, you basically (probably) were doing the "Ear EQ thing" inadvertently; I eq my system in 2 'steps', first setting things to be flat to the microphone, then I apply the ear eq to the signal in addition

 

[JerryLove]

There are three lisening envyronments in my house. The bedroom is a HT and was SPL balanced then adjusted with some "reference DVDs" to make it the most appeling for me for movies.

The computer, though it would theoretically be the easiest to EQ, is deliberately off to adjust for the fact that I'm considerably off-axis (I'm working to replace the computer desk with one that will let me set the near-field monitors at the appropriate positions).

The family room is the most balanced, and most powerful. It is the one that I should play your MP3 on to, assuming your base equipment approches flat, experience what you intended. Sadly, it's in mid deconstruction as I have aquired new speakers and will be entirely redoing the room for them.

Where I listened, it's entirely possible that your changes repeated changes I made and so caused problem. If I get a chance to listen on a known-flat rig I'll post my impressions from them.

And don't get me wrong about my opinion on "flat vs not-flat". I full agree that most people will prefer some less realistic but "better sounding" arrangement. That's why there's a "loudness" preset on even the most basic of AVRs, that's Bose's trick to make that first impression "wow, these sound better" (conversely "midnight mode"). There's nothing at all wrong with that: but I don't think it will be the same settings for everyone.