'Flat' is not correct for a stereo system ?

Status
Not open for further replies.
I feel as if Earl is asking me "Have you stopped beating your wife?" - for which there is NO safe reply.

No Earl (or maybe, yes Earl) contrary to the implication of your reply, I am not a monumental idiot, as you imply, suggesting a floor tapper be used as a music simulation.

I will admit to not understanding your post or why your were showing that link, hence I was trying to make sure that everyone understood what these machines are used for. I did not intend to be rude, I just didn't understand.
 
That sounds very clever and interesting. I vaguely remember reading about that, somewhere. Got any links or clues where to look?

Sorry, not off hand. I believe that one could use any MLS based measurement system since this noise signal just gets lost in the background. The key is to do a lot of averaging to compensate for the low SNR (noise in this case meaning error, not the MLS noise signal).
 
seem like this thread.. completely missed the original intent...
why loudspeakers should not be flat...
And turned into a how to make speakers in a room fit a specific curve..I must say that if your'e in for good musical experiences and live like performance I have yet to hear any room adapted system deliver it....

Tried to state some misconceptions that makes it an impossible dream....the main being the assumption that all SPL at a certain frequency sound the same. next is that you utterly confuse the psychoacoustic brain when you try to tell it that it's in a different place when you play music....

It's too complex and we can't interpret what's going on..and we get tired by doing so...listening fatigue is not far away...

Speakers of different topologies must have different curves to sound right...Eg electrostatics and panel types can be more flat than dynamic multiways...and within the multiways types with ribbon tweeters can be more flat than speakers with soft dome tweeter...so frequency response is also dependent on speaker and driver technology...
by testing drivers it's quite evident that when membranes start to have breakups they may measure flat, but they sound louder..this is then in turn compensated in the crossover and it's also one hell of a good reason why the filters must be tuned by listening to music and not just by measuring...

Good and involving speakers does not have a flat response, but to fit a universal target curve to it is not possible, because its speaker and speaker driver dependent..
 
Last edited:
Praxis can do that. Even without the noise and just music as the source.

Well I remember something like that being done in HOLMImpulse, but I've never tried it. Use music as the source, instead of sweeps or noise.

I was thinking - or hoping - that in the case Earl mentioned that the noise could be mixed into a "live" audio stream then extracted and measured. Sort of a DiffMaker on the fly.
 
Aren't you just talking about power response?

Have we come full-circle? I don't think so, because my poor understanding of these things is:

1. using a power or other DI-influenced curve, the physical loudness of a tone is conditioned by the duration of that tone, in a given room, and

2. that and other known and unknown factors further influence the perception of loudness (and it is perception we are trying to predict, not physical loudness after all).

Therefore, the stimulus used to produce the power or other traces needs to resemble the statistical characteristics of the kind of music (and maybe other parameters) representative of sources played in a given room*.

And if the stimulus does so, we have the possibility of a non-tweaked mic set-up.

*or be referenced to it
 
Last edited:
I get confused when engineers talk about "power response", even though I believe I know what they mean, because it's a fairly general term. I'd bet I'm not alone on that. I think what they are abbreviating is total energy from the speakers reaching the listener, with the effects of both dispersion and room acoustics added in. That's a highly variable thing. I think it's important, but I'd rather do gated burst measurements on the speaker drivers, and non-gated tone burst measurements on the room acoustics, to get a graph of RT60 vs. frequency, and guess at the psycho-acoustic effects of ringing variations over frequency, and perhaps fine tune any tonal calibrations based on that. I'm very convinced that frequency selective ringing can cause us to think a certain frequency is louder than a calibrated mic and pink noise would suggest, but ringing also has a start-up time that needs to be considered.
 
As long as there is no "standardized speaker", all that
discussion about "flatness" is difficult and sort
of discussion on "pseudo flatness".

As long as there is no "standard room", we have the
same problem.

Real speakers usually have

- frequency dependent DI,
- frequency dependent power resonse
- frequency dependent "on axis" SPL
- frequency AND angle dependent group delay,
with serious discontinuities at almost all angles


Real rooms have frequency dependent reverberation
over frequency. IMO RT60 is pretty useless to describe
a living room for "Hifi speaker listening".

Plotting early decay time over frequency using lots of
averaging over possible listening positions and using
an adequate speaker for that measurement tells
much more.

Has anyone ever achieved a good result by EQ'ing
ungated inroom response to flat or "nearly flat" ?

I seriously doubt it ... the result obtained from doing
so is mostly worse than doing nothing.

Unfortunately there is always a particular
speaker in a particular room to be investigated.

So commonly valid recipes may be hard to find.

If on the other hand a speaker would have

- flat on axis response

- approximately constant directivity, All planes,
and 360 degrees taken into account)

- reasonably flat group delay from say
500Hz to 10Khz, independently from the angle of
measurement ...

- early reflections (floor!) handled in a proper way ...

You can place that speaker into most living rooms,
without any EQ or room treatment and it will sound
"natural" or "realistic" or "authentic"
(whatever term you prefer) with recordings being
"unobtrusive" and mixed with "genre adequate"
loudness characeristics.

If EDT shows serious imbalance (in a consistent manner
verified by avering over different excitation/measurement
positions) e.g. a lower midrange rise - there might be
two options:

- improve that room's EDT balance over frequency
by diffusers/absorbers

- apply mild EQ

This is in order of preference ...

---
I am not talking about the "room mode game"
below Schroeder frequency here, which is another
story.

Here a balanced mode excitation has to be found
regarding speaker positions and some preferred
listening positions.

And last but not least: Forget about "Digital Room
Correction", it is a drug in the best case, but no real food.

Invest the time and money saved in solving the problems
at its roots, whenever possible.

DRC maybe nice sometimes when sitting here, but 40cm behind ...

Cheers
 
Last edited:
" MiiB
Denmark.gif
diyAudio Member

Join Date: May 2005
Location: Denmark


seem like this thread.. completely missed the original intent...
why loudspeakers should not be flat..."

okay, it is not too hard to put a short end to it.
-> impossible to make it flat anyways
-> even if it was, it would be just junk.

hmm.. 72 pages wasted?
not at all.
if this was the reply in the first place, then someone would ask wy it is impossible, and if it was not, why would it sound just as bad as having roast rat as dinner.

actualy i would say most probably -at least to my current knowledge- this thread contains every answer, including a lot of intresting theorys.

At the end of the day, this thread/artice has unimaginable information value for those who care to read it.
Thank you for every single one who contributed to it 😀
 
I don't pretend to know what this thread is really about, but I know what I see in it and what I want to get out of it.

For me it's simple. I'd like to set up a microphone somewhere in the room, run a few measurements and be done. If there is some signal/curve/method that will allow me to do that and get a subjectively balanced result - I'm all in! (Probably not alone in that).

But that seems unlikely to happen. Some of us seem to be sure why, others are not so sure. It does look like an easy question. If it were easy, there would be no need for this thread. It does seem that most agree that a flat on-axis response in a typical listening room is not going to make us happy. The question remains - why? We've gone some way in understanding the reasons.
 
Therefore, the stimulus used to produce the power or other traces needs to resemble the statistical characteristics of the kind of music (and maybe other parameters) representative of sources played in a given room*.

And if the stimulus does so, we have the possibility of a non-tweaked mic set-up.

*or be referenced to it

Ben,
This only matters if the system has appreciable non-linearities or is severely limited in headroom. This way, the stimulus excites these non-lins in a manner that is statistically equivalent to "typical" program material.

In telecom audio, we used artificial voice with the same long term statistics (frequency content) and peak/ave stats as typical speech:
http://eu.sabotage.org/www/ITU/P/P0050e.pdf

This was only useful in that the channel was non linear (A and u-law compression).

For distortion measures using multi-tones (PhiTone), it may also be useful to define the tones to have a spectrum mimicking the IEC 268-1, 1985 power density spec. I've attachment the measurement of a pre-amp I designed and built for my PC, where the stimulus was 12-tone and representative of the long term spectral density in the IEC spec.

But for linear systems, it doesn't matter what you use as a stimulus as long as you aaccurately capture the speaker o/p with the mic.

Dave
 

Attachments

  • 845 Phi12r IEC.gif
    845 Phi12r IEC.gif
    25.9 KB · Views: 121
Sorry, not off hand. I believe that one could use any MLS based measurement system since this noise signal just gets lost in the background. The key is to do a lot of averaging to compensate for the low SNR (noise in this case meaning error, not the MLS noise signal).

The only real value of this technique is if the system is non-linear or you're trying to measure it in-situ under end use.

If the goal is just to measure with an arbitrary stimulus, just use any signal you want and a dual channel analyzer like the old days. The old HP analyzers were a gift for this, as long as you eq'ed teh group delay on both ports to be equivalent.

The value of using test signals which represent actual source material is in measuring non-linear systems. In the good old days I developed an audio test system that used music as stimulus and then buried tone markers and narrow band noise in, with the amplitude modulated by the bulk waveform. By tracking these markers, you could see all the time varying behaviours, compression etc first hand. Very powerful for monitoring time dependent transfer functions and non lins but of dubious value for measuring linear components.

It worked so well, I reverse engineered the DSP algo in a Polycom in a week. They threatened to sue us thinking I'd hacked their code. 🙂

Dave Dal Farra
 
...
Phase content maybe even more important than frequency content..
The point is that frequency areas with phase shifts stands out...simply because of our sensitive awareness...
...

Although that point of view does not seem to be very
popular in here, i would at least agree in group delay
having influence on perception and tonal balance.

There seem to be frequency ranges which are more
critical regarding discontinuities in group delay.

Optimizing for flat group delay in the 1..4 Khz region
with high priority and with somewhat lower priority
from 500Hz...10Khz seems advantageous.

Unfortunately this PDF is in german language (only):

http://forum2.magnetofon.de/bildupload/goosphase.pdf

There is a brief table form summary on audibility
thresholds of group delay found so far by different
researchers.

The "Institute for Broadcast Technology" , Munich
(above Paper) recommends even narrower tolerances
in frequency dependent group delay for monitor
loudspeakers, than could be derived from earlier
findings mentioned e.g. from Blauert/Laws 1978.


Blauert/Laws ( as cited in the above paper):

500Hz............3.0ms

900Hz-3Khz...0.5ms

4Khz..............0.4ms

7Khz..............1.1ms


The current recommendation from IRT Munich is to have
the group delay for monitor loudspeakers
within +/- 0.2ms "from mid to high frequencies"
for each speaker and apply pair matching additionally ...

---
I ran over this while optimizing my own flat panel
bending wave loudspeakers, which have considerably
flat group delay above 500 Hz.

But the rear radiation caused some irregularities in
certain underdamped rooms, which was hard to capture
by (front/rear) frequency response comparison.

Front and rear radiation is not perfectly symmetric, as
the motor covers some area of the panel on the rear,
like in conventional dynamic drivers too.

Optimizing for flat group delay - even for the rear
radiation - yielded the desired result: Tonal balance
was more independent from listening position.

The (rear) frequency response could be judged better
or worse after applying the "countermeasures" ...

Group delay (only) showed an improvement
(by considerable flattening) , which could be correlated
with the audible result.

I cannot say whether that IRT Munich recommendations
are too strict, but currently i'd like to support the standpoint
of group delay being important also from own developmental
experience.

Nearly nothing in audio can be seen in isolation ...

Especially when entering psychoacoustically critical
frequency ranges of hearing, as the "presence region" is
surely one of those.


Kind Regards
 
Last edited:
I get confused when engineers talk about "power response", even though I believe I know what they mean, because it's a fairly general term. I'd bet I'm not alone on that. I think what they are abbreviating is total energy from the speakers reaching the listener, with the effects of both dispersion and room acoustics added in.---
.

No, it is not a general term. It is a technical term, and one that is frequently mis-used.
You are correct, that it is the total amount of energy and not just the on-axis component. However, it does not normally include room effects. It is a property of the loudspeaker.
 
The only real value of this technique is if the system is non-linear or you're trying to measure it in-situ under end use.

Thats not actually true, the system can be linear or not it does not matter (you'll only measure the linear part in either case), but there is a reason to want to known how the system is working with the audience in place. You don't want to make them listen to sine sweeps, so you just do it while they listen to the concert and they never know.
 
I don't pretend to know what this thread is really about, but I know what I see in it and what I want to get out of it.

For me it's simple. I'd like to set up a microphone somewhere in the room, run a few measurements and be done. If there is some signal/curve/method that will allow me to do that and get a subjectively balanced result - I'm all in! (Probably not alone in that).

But that seems unlikely to happen. Some of us seem to be sure why, others are not so sure. It does look like an easy question. If it were easy, there would be no need for this thread. It does seem that most agree that a flat on-axis response in a typical listening room is not going to make us happy. The question remains - why? We've gone some way in understanding the reasons.

You might be better w/o measuring in a sense. I mean your going off a subjective metric anyway. No one can build a metric based on a temporary thing. The laid out objective metrics that correlate to preference over the long haul have been done and put to print. There is no mystery anymore. They're even posted in this thread at your request. I don't think most people buying or building speaker would agree that flat is bad. It's more like the goal. I bet a number of people would buy just of they saw a flat graph accompanied by some techno babble. 😱

Dan
 
Status
Not open for further replies.