As I stated in my opening post, I EQ'd my monitors for flat response (according to the REW analyzer) and it sounded extremely bright. Nowhere near "natural".

I can't imagine anyone mixing/mastering to that curve. I couldn't even listen to it!

(1/24 oct smoothing)

Red = R
Black = L
Green = Summed

This is pretty close to what seems right to me. Id like about 2db more in the 160-225hz range. But other than that, this sounds anything but dull and flat to me.

Its amounts to about +4db in the bass region compared to 10K.

Thanks Jim. That notch at 6K is quite wide which I am sure smoothens the HF a little. No nasty sibilance. Note my very aggressive HF cuts up very high may well be to lose some fairly ugly stuff up there. These ribbons ain't as smooth as folklore suggests…..
Could you expand the vertical scale a lot? I would like to look at your curve more closely…… oooh baby!

DD

Sure. The last time I tested myself about a year ago I could hear 14 KHz but no higher. But rolling off slightly above 15 KHz is not the same as a steady roll-off starting in the midrange as I often see for a "house curve."

--Ethan

As I explained in my Post #10 above, I doubt your speakers were flat after applying the EQ. Without being there to confirm using REW and my precision DPA microphone I can only guess. But I imagine the net response you created is similar to what you set the EQ to. Unless your speakers are far from flat.

--Ethan

14K is not bad at all for a dino Ethan. Same here for one ear, worse for the other, not much action above 12K. And yes, not the same as a house curve, just an interesting aside, particularly in terms of symmetry? Mine are quite different to each other. Should one adjust level and possibly even Eq to deliver the full info equally L and R to the brain as opposed to the ear? Have I ever heard stereo…..;-)

Different slopes in different frequency zones…. very tweaky I have found.
As I said, I don't favour the +3 at 100Hz suggested by Bruel and Kjaer and many many others. But the overall result, my 10K is still 6dB down from 30Hz, similar to B&K. But the two curves don't sound at all alike.
BTW, I would be interested in seeing actual room response if it is to hand.

DD

Flat bass isn't normal IME. Having boost below 80hz - 120hz is far nicer than flat.

Maximum Bass at all frequencies, Jah Bless…..

I like it too Opus. But it steered my mixing Bass light.
I tried it for a good while, and tried to be 'brave', but it really didn't work out.
Similarly I tried all sorts of HF variations in 0.5dB increments.
The current curve delivers uncannily accurate translation for me now.
I expect everyone in their own room will have a unique optimum translation curve. It is also IMO very noteworthy that there is very little tonal difference when I switch between HD650's and the speakers.

EDIT, I am using this technique to connect my CR activity to the outside world. I am using an "Advisor" or even better a "Persuader". Many of us have used all sorts to tools to achieve this same end. e.g. Nearfields, NS bleeding 10s, Horrortones, Ghettoblasters with Line in.
Maybe 'Translation Curve' describes it best.
DD

Do you have modes reinforcing? They will add boost all their own.

Interesting topic. I too am in the midst of honing my control room's response for best translation. DanDan's comments are appreciated, as are others to be sure.

Reading the B&K paper cited by Andre, I find it confusingly written. However, the takeaway I get from it is this: The graph shown in Fig. 5 represents their study's view of the most desirable (listening position) response, measured with pink noise stimulus. Summarizing that curve, it is approximately flat in the first decade, and drops roughly 3dB/decade for the next two decades. If I'm misinterpreting, someone please speak up.

Since we know that pink noise is dropping at 3dB/octave or 10dB/decade, we can add that to the curve to see what we should strive for if we start with a constant amplitude source such as swept sine, rather than constant power source such as pink noise.

My own opinion of optimum house response for best translation is still forming, so I don't have much to say yet. But so far I am inclined to believe that something like this B&K curve may be about right.

Yes! That's a point I have tried to get across a few times here that I don't think is getting through. I believe we are comparing apples and oranges in these frequency response/house curve plots.

When someone tells me that all they needed was a couple dB slope to make their setup sound great I immediately suspect they are using PINK noise as their test tone.

The testing that I did (and is very common in automated FFT-type analyzers) uses a constant amplitude, swept sine wave which is essentially WHITE noise (constant amplitude). If you EQ your system to be "flat" compared to white noise it will be very, VERY bright.

Two different curves with VERY different slopes.

Hi Nonlinear,
sorry I am late to the party...
You have got a lot of good answers already.
More precisely I like to add to easen your confusion

Thats because they (B&K) were using an anlyzer with Relative Band Width e.g. 1/3:rd octave (percentage) which will show a straigt line if excited by pink noise.

If white noise displays flat you have a constant bandwidth analyzer. Every filter has the same BW e.g 10 Hz. Or something similar to that.

At that time: Yes. Constant and percentage BW. This paper was presented 1974.
In the digital world there are probably more types.

No, booth are measuring "amplitude" in different ways. No "power" this time.

Seems like a possible explanation. Way too bright if the result is flat.(Close to White Spectrum)

Hope that helps a little...

Best

Ake

.

Compared to white noise, yes. Then, it depends on how we analyze it.
If you listen to pink noise it sounds flat and quite "music like". But a loud full orchestra is tiring...especially if you are close

The B&K curve was a compromize for the mixture of close miking and distant miking. And nowadays we are even closer

Best
Ake

Opus, yes I do unfortunately have serious low modal action, particular 35Hz, but this perhaps balanced somewhat by nasty BIR nulling. It is a concrete room, and the resulting response despite massive trapping (to be enhanced with VPR or Homatherm soon) is bass light at any reasonable mix position. Put it this way, my S3A's are +4dB at LF and -2dB at HF BEFORE the Dirac Filters are applied.
This is to make the sound with and without broadly similar.

There seems to be a lingering suspicion that the instruments/techniques are being misused or that there is a fundamental flaw undermining us.

None of this is happening. When you look at a frequency response graph, one can assume that is is using similar parameters to every other one presented.
If otherwise, it would be stated.
Many speakers achieve an essentially flat response in an anechoic chamber or outdoors up a pole, usually very close to be sure it is pure speaker we are viewing. The same speaker measured at a reasonable distance in a 1960's living room will show as B&K measured.
So they reckon it is the average actual response experienced in real homes on real but quite high quality Hi Fi.
As such they reckon mix engineers might want to mix 'to' that curve. Or others might want to tune their own Hi Fi to that curve. They made a test LP (Long Playing Vinyl Record kids) Containing third octave bands of pink noise. So if a Hi Fi buff has a third octave graphic and an accurate SLM, she could easily tailor the response.
Basically B&K measured reality and attempted to make it a standard.

White Noise seems to be a great contributor to the confusion. Could we agree to omit it from the discussion? In the context here, it doesn't happen.

There is a lot to be learned from simply playing with the tools in REW.
Note they will all run simultaneously.
So we can Generate Pink. Try different settings on the Spectrum Analyser, including Standard third octave RTA. Run the SPL meter also, see the differences caused by A, C, Fast Slow, Leq.
When calibrated, this tool set is akin to an acoustics laboratory, and not a Fisher Price version. These are real tools and they work.

EDIT, REW can record with the Sweep silenced. e.g. Speakers muted it will display the ambient background noise spectrum. Similarly if Pink Noise is pumped through the speakers and REW is tricked into recording it, we can see the resulting spectrum as REW sees it. Afaik, the software analysing the Sweep (or lack of) will display exactly the same spectrum as the Spectrum Analyser.


The difference in tone between flat and the suggested HF down sloped response is quite dramatic.
It's a big deal, and there are divergent opinions about it.
It took me a long long time to shake the logical assumption that flat has to be the most accurate, the honest representation of what is on the record.
The logical assumption remains correct, but in practice it consistently causes dull mixes.
I started with B&K approximately and over a year or so in 0.5dB increments I backed right off from it and then gradually ended up with the curve presented earlier.
This was measurement, empiricism, and psychoacoustics working together until a stable plateau of certainty was reached.
I commend the same journey to anyone listening professionally. Everyone in their own room will have their own curve. But to get there quickly, the overall trend, 4-6dB slope between LF and HF has to be noticed.

DD

Yes, it DOES happen. When you use an FFT-based (digital) spectrum analyzer you are getting a CONSTANT BANDWIDTH measurement. WHITE noise appears flat on these analyzers.

Some modern, and likely all legacy data (such as the "B&K" curve), were measured using CONSTANT Q (i.e., 1/3 octave) analyzers. PINK noise appears flat on these analyzers.

So the type of analyzer used makes HUGE difference in how the curves look and how a system sounds when these curves are used as references.

I believe the industry, as well as our discussion here, is mixing these two types of measurements - not on purpose - but because it's not well understood. It's not a conspiracy it's just something we have to keep in mind, especially when discussing corrective house curves.

What is the decay times in your CR DD? You may be fighting an uphill battle without strong bass treatments.

@Opus The 35Hz thing is about 0.6 Seconds before the Eq hits it.
I don't know exactly how much it gets shortened by the EQ.
I haven't measured including Dirac except way back, long story.
The room is working brilliantly now, but I fully intend to do some very serious large LF work soon.

@nonlin
Nope, pose take a look at my post 10 again. White Noise viewed on the REW analyser rises exactly the same as on any other analyser.
REW is of course FFT again just like any other analyser.


Again, if the term White Noise were removed from the discussion and considerations here there would be no alleged problem right?

Now, have you seen a single measurement of a speaker, microphone or room, anywhere, anytime, using White Noise?
As I suggested earlier, REW's Sweep Analysis and Spectrum Analyser see things in exactly the same way. This is room ambient noise as seen by both.





DD

That's because they are both using the same FFT analyzer!

Try making that same measurement using a 1/3 octave, analog spectrum analyzer and see what you get.

In the general electrical engineering sense, whenever one speaks of the response of a system, also called transfer function, network response, frequency response, etc., the stimulus is always assumed to be spectrally white, unless explicitly stated otherwise. This applies to audio system components as well as any other electrical system with an input and an output (also known as gozintas and gozoutas).

Now, when we get to acoustic measurements of speaker + room combinations, it seems the default is to use a stimulus source that is spectrally pink, and assume everyone knows this is the unspoken rule. I wasn't so consciously aware of that until this discussion, but I certainly see the reasons.

So then I ask those more in the know: when speakers are measured for on-axis response in anechoic conditions, am I to presume that the input stimulus is pink in this case as well?

@nonlin, the B&K paper describes the use of an analogue third octave process.
Way back, actual physical third octave filter modules were plugged into the SLM.
One at a time. As the paper states, the results are exactly the same.
I feel we have drifted far from what I thought was the original point.
To be frank, I am no longer clear on even that.
Could you restate the original point plse nonlin? I want to get clear that we are on the same page and talking about the same thing, or not.

@jas, Pink Noise has a spectrum quite similar to music.
Thus it is appropriate for testing devices which try to represent music.
White Noise would put a disproportionate stress on the tweeter, necessitating dropping the test level to an impractical extent.



DD

White noise is constant bandwidth, pink is constant Q, simply. That's why white noise is flat on an FFT. For a given octave, there is more energy above it with white noise, where with pink noise there is an equal amount.

LOL - yes, this post has evolved a bit but I think it's good because I learned something here. I hope others have too - or at least got them thinking/researching.

I originally asked why my monitor system that measured "flat" sounded so terrible - extremely bright, little bass.

I can now answer my own question. The reason is because I "corrected" my system to match a white noise curve, not a pink noise curve!

When we discuss audio frequency response plots and corrective house curves we need to be aware of HOW the measurements were taken - with what type of equipment.

A) Analog spectrum analyzers typically measure to a constant Q. Pink noise shows flat on their display and it's what we typically hear as "flat".

B) Digital FFT spectrum analyzers, on the other hand, measure to a constant bandwidth. White noise looks flat on their display. Because of this, audio measurements taken on FFT analyzers have to be "tilted up" by 3dB/octave to emulate what constant Q analyzers would display. SOME digital analyzers that are intended strictly for audio work have this "tilt" built in - but others do not. If this "tilt" is not applied, correcting the monitor system to measure flat on this type of equipment will overcompensate with excessive highs.

These two types of measurement was something I did not understand before this post - but I do now. It makes a HUGE difference in how measurement plots appear and on how corrective house curves are designed from them.

BTW - I am not sure what the REW analyzer displays but I do not believe it is "pink tilted". I have asked JohnM and will advise his reply.

No, it's not adjusted. But it also does not measure with white noise. There is a cumulation of energy in white noise that makes it +3db/oct for our purposes. When a sine is swept, the amplitude is constant, but the cumulation does not occur. It's like this, with white noise, the first octave is 1, the one above it is 1+1, after that 1+1+1+1, an exponential rise in power.

Nonlin, I am very doubtful that you managed to persuade REW to generate filters resulting in a staggeringly rising HF response. Looks to me like a =10dB boost at 10K vs 1K. https://gearspace.com/board/9567439-post19.html
Let's double check what you actually did.
I suggest that you start again. Use REW exactly as intended.
Forget noise entirely. Just measure your speaker response accurately using the constant amplitude 1:1 linear system in REW.
Then apply Eq or simply tweak the speaker HF level, to get a reasonably flat line. Take a good listen. Then adjust your speaker response to say -4dB at HF as jim has. He's a good listener.... Let use know which you prefer.
Then we will be on the same page.
BTW, I am not aware of any device which could be called an analogue spectrum analyser, apart from the historic ones.
In which case, as the B&K paper says they behave in exactly the same expected manner.
I think it is possible to adjust the axes of analysers to display White as flat and Pink as descending level with frequency. On a brief scan I think log/log scales, but I would need to do some homework to ferret that out properly. Or perhaps you will. In any case, apart from learning about obscure possibilities of devices which we don't have at hand, I don't see the point.
REW and all the other commonly used tools work perfectly as designed.
If we simply use them as intended we can achieve which we all understand as a flat response.
My point is that I find that response to be radically too bright to work with. As you can see many share that view.
But I would be interested to see what you think when you compare them.

DD

Just to add to the confusion, err conversation: the human ear/hearing is anything but linear. add to that, there are (probably not too drastic) variances from one person to another.

OK, let's see if I can illustrate my point with a couple of pictures.

On a "classic" audio spectrum analyzer, Pink noise looks like this:


On a digital (FFT) spectrum analyzer (that is not "pink weighted for audio"), Pink noise looks like this:


So, my point is that if you measure your room and EQ your speakers to look "flat" on a digital analyzer you may have greatly boosted the high end. You will then need to apply some drastic house curve (i.e., -20dB low to high) to make it SOUND flat. I believe that is what happened in my case.

This explains to me why some folks require very drastic house curves to make room correction sound "right" while others only need a few dB. You have to be aware of what tool is being used to take the measurements and develop the corrections.

That's a power density graph.
I don't think you will find one of those in REW, FuzzMeasure, ARTA, etc. etc. etc.
Have you ever actually come across any device with this feature?
I cannot figure out a way to do as you claim you did in REW.
There is no 'pinko' hiding in there. 80dB at 100Hz and 80dB at 15Khz read exactly the same on all the graphs, and on the SPL Meter with Z (no) frequency weighting. It is a linear 1:1 system.

Although you haven't described what device was used to eq your speakers, or the extent of the filters (+10dB at 10K?) I am firmly convinced you used REW exactly as you should have. If you try it again, I believe all will become clear.
DD

I'm afraid you are creating a great deal of unnecessary confusion, and mixing up the measurement of transfer function (which is what REW and other audio measurement systems do) and the various ways of displaying the spectral content of a signal, which are not at all the same thing.

REW measures the audio system transfer function, the plot it produces shows how the system changes signals passed through it - for example, if the frequency response is 10 dB lower at 10kHz than it is at 1 kHz that is because the system that was measured reproduces 10 kHz tones at 10 dB lower levels than 1 kHz tones. The transfer function is independent of the stimulus used to measure it, it is a ratio of the measurement to the stimulus. Various types of stimuli are used by different systems, from noise (white, pink or other spectral shapes) to sweeps (linear or logarithmic) with a variety of others also available, but ALL will produce the same frequency response for the same system unless they are broken. As it happens a logarithmic sweep, as used by REW and many other measurement systems as it has a number of useful properties, has constant energy per octave because the rate at which the sweep frequency changes increases exponentially during the sweep - the stimulus looks flat on an RTA display.

That brings me to signal energy displays. The frequency content of a signal (as distinct from the frequency response of a system) can be displayed in various ways, the two most common are spectrum analysers and fractional octave analysers, more commonly called Real Time Analysers or RTAs. Spectrum analysers show the energy in bins of constant frequency width - for example, the energy in each 10Hz span of a signal. On a spectrum analyser white noise, which is defined as having constant energy in any given frequency span, appears as a horizontal line. Pink noise has constant energy per octave - for example, it has the same energy in the 100 Hz span between 100 and 200 Hz as it has in the 1 kHz span between 1kHz and 2 kHz. On a spectrum analyser a pink noise signal shows a level that drops 3 dB per octave. RTAs show the energy in octaves or octave fractions, a 1/3 octave RTA display shows the energy in each 1/3 octave span. On an RTA white noise produces a line that rises at 3 dB per octave, pink noise produce a horizontal line.

Your system sounded bad when you equalised it to produce a flat response at the listening position because you boosted the high frequencies. You should not have done that, speakers that have a flat response in anechoic conditions will have listening position measurements that drop at high frequencies if the measurement includes the contribution of the room, as yours did. Your ears and brain are smart enough to distinguish between the direct sound that arrives from the speaker (which it can tell is flat) and the later sound of the room (which drops at higher frequencies) and be happy with the overall result. When you boosted the high frequencies the direct sound was now unnaturally bright, which your ears were smart enough to tell you.

So, is what your saying is he was basically boosting the direct response in the highs to make up for a room response that was deficient in the highs equating to a overall flat response?

Suppose he equalized the room response via treatment to get a flatter room response to achieve an overall flat response?

I used the correction filters that REW generated - to a flat target curve - which I exported as an impulse file and ran in my monitor chain via convolution processor.

The resultant "corrected sound" - which should have approximated the spectrum of Pink noise - was WAY too bright. Did not sound like a Pink noise curve - it sounded like a White noise curve - and required the application of a target curve with over 20dB roll off from 20Hz to 20KHz to make it sound "right".

It's hard to believe my 17x25 foot room, treated correctly or not, could have that much influence on the sound I'm hearing 5 feet from my monitors. If this exaggerated treble boost was not due to the FFT analysis technique, what caused it?

BTW - My test microphone is a Panasonic WM61 with a dead flat preamp (measured on a o'scope with sine tones).

http://www.panasonic.com/industrial/...61_a_b_dne.pdf