So this almost never happens...

[Bennett Prescott]

Re: So this almost never happens...

Stuart,

I am not sure who is more the fool, the engineer who only listens or the engineer who only measures. I do not know any audio person who knows how to run a measurement system that does not run it in sympathy with their ears. One must play the strengths of each against the other... my ears tell me things I cannot measure, some subjective (should I add 3dB more HF boost?) and some objective (even though the MF measures flat, I need to kill 2.2kHz for this to sound balanced). The measurement system tells me things my ears cannot (there is a 30 degree phase lag between the LF and MF passband).

 

[Timo Beckman]

Re: So this almost never happens...

I've challenged you to design crossover settings by ear, which is one thing. Let's see someone design a flat-phase tuning with all-pass filters by ear alone.

Impossible . The human ear is not sensitive enough to phase to hear the difference . I gave a small measurement coarse a few day's ago . At the end of the seminar i did a experiment . 1 of the students stood on axis of the demo system and i changed the phase response of the entire system by putting in a 2nd order all-pass @ 1 khz via the main input filter section . No 1 could hear the difference between the AP2 being on or off .

Also there's no change in the frequency domain only in the time domain of the system under test . Meaning you have to measure it otherwise do not try it.....

 

[Ivan Beaver]

Re: So this almost never happens...

Impossible . The human ear is not sensitive enough to phase to hear the difference . I gave a small measurement coarse a few day's ago . At the end of the seminar i did a experiment . 1 of the students stood on axis of the demo system and i changed the phase response of the entire system by putting in a 2nd order all-pass @ 1 khz via the main input filter section . No 1 could hear the difference between the AP2 being on or off .

Also there's no change in the frequency domain only in the time domain of the system under test . Meaning you have to measure it otherwise do not try it.....

It is not so much the exact phase-but rather how the phase response "plays" with the other parts.

The flatter the phase response is-the more likely the loudspeaker is going to sound like a single device-rather than multiple devices stuck together in a cabinet.

Sometimes there is a good bit of difference-sometimes there is not so big a difference.

 

[John Roberts]

Re: So this almost never happens...

Human audition is not a literal 1:1 mapping of complete sound characteristics into our meat computer, but a progression of extracting simpler macro impressions from the complex sound. Phase doesn't matter until it does. Mostly as Ivan says in interaction between and compared to similar sources (in crossovers), etc.

A common mistake in design engineering is to extrapolate too much from such experiments. Linear and accurate is always better while most playback is full of compromises to both while focussing on parameters known to be more important to our meat computers. The interesting part is that the meat computer can learn to hear more subtle stuff within it's bandpass and dynamic range, so this is not a rock solid, stable target.

Note: this is not meant to support all the claims from people that they hear stuff they probably don't. IMO 95% of what people hear is simple frequency response or spectral energy content of signals. When this is not controlled for, frequency response differences can be interpreted as all kinds of random phenomenon.

So much of audio technology is about tricking the meat computer, and I have seen huge differences between individuals ability to hear subtle stuff, mostly connected with their training. A musician can focus in on some aspects of sound that they use for negative feedback while playing, a capability that normal listeners never developed. This doesn't necessarily make musicians more critical listeners, just more sensitive to some specific cues based on practice and experience. Human perception is only tracking a fraction of the input it receives. These musicians focus on a slightly different fraction of that whole. Since getting the whole perfect is pretty much impossible we mainly need to try to get the right fractions correct. .....or not.

====

Phase relationship between similar sources can create comb filtering which alters the spectral energy content... or frequency response. This not so simple frequency response can not be easily corrected with simple EQ

JR

 

[Timo Beckman]

Re: So this almost never happens...

It is not so much the exact phase-but rather how the phase response "plays" with the other parts.

The flatter the phase response is-the more likely the loudspeaker is going to sound like a single device-rather than multiple devices stuck together in a cabinet.

That's why i go to the trouble of correcting the phase response of a loudspeaker if given a well build one and proper processing . Also having loudspeakers with "flat phase" makes combining different models with similar behavior a lot easier .

This is a screenshot of the system I used during the experiment .
The red trace is the same system as the other 1 only with a 2nd order all-pass implemented . So there is a change in the time domain but no real change in the frequency domain .
To have 2 systems play together 1 with the response of the purple trace and 1 with the response of the red trace would give me a rather large cancelation around the 1kHz area provided they match in level .