Reply posted by Steve Bradbury on June 14, 2002

Wayne,

Historically it was considered that the relative phase of the harmonic components has no audible effect, a premise widely accepted for some time, but which is now known to be incorrect. Later this idea was modified (sorry I don’t know by whom). The new theory stated that in general the relative phase between two signal components should be irrelevant if more than a certain critical bandwidth separates the two components. This is because the ear behaves as if it contains a bank of filters, each filter passing frequencies within the critical band. Above 1 kHz the critical bandwidth is close to 1/6 octave. Below 100 Hz it is approximately 30 Hz. When greater than the critical bandwidth phase should not matter then because there is no element [of the cochlea] that sees both components.

However this revised argument is also considered to be incorrect in the face of experimental evidence. There is, however, considerable claim and counter claim that a lot of these tests are contrived and can’t be compared with real life situations. In an AES paper a Dr Floyd Toole stated, “within very generous tolerances, humans are insensitive to phase shifts. Under carefully contrived circumstances, special signals auditioned in anechoic conditions, or through headphones, people have heard slight differences. However, even these limited results have failed to provide clear evidence of a 'preference' for a lack of phase shift. When auditioned in real rooms, these differences disappear..." This quote was lifted from the internet several months ago so I haven’t seen the rest of the paper. Maybe others here have and can provide additional comment.

There was an interesting paper by Nelson and Bilger –“ Octave masking in normal-hearing listeners”. Masking is when a loud tone makes an otherwise audible weaker tone completely inaudible. In the paper they show that the masking level of a tone for its 2nd harmonic depends on the relative phase of the tones. This difference, which can be as large as 30 dB, is apparently a result of a second harmonic produced by nonlinearities in the ear itself, adding constructively or destructively with the externally produced 2nd harmonic. The effect varies quite a bit among individuals. So if you are sensitive to this effect, it could change the perception of harmonic distortion produced by a sound system for better or worse. It could also alter the perception of harmonics naturally produced by musical instruments. This is yet another valid argument that the fidelity of sound reproduction can be audibly degraded by a realistic phase change. The paper was written back in the early seventies and I don’t have a copy; the above is all I can remember.

The problem in your case, with a graphic equaliser, is that you are, by the very nature of equalisers, altering the amplitude of the signal because you want to alter the sound. Even in experiments where the only the phase is being altered it is difficult to prevent subtle amplitude variations that could contribute to the differences people hear.

What does all this sound like? Try doing a search on the internet there are audio files out there that demonstrate the effect. As stated above these may or may not resemble any real world situation so make sure you know how they were done.

Hope this helps,

Steve


Reply posted by Wayne A. Pflughaupt on June 14, 2002

Okay, consider this:

About 10 years ago, in the early 90s, I read an article by Stephen St. Croix in Mix magazine where he related his experiences with a true digital equalizer (i.e., one with digital-domain processing, not just digital manipulation of analog filters). This was a device costing (at the time) tens of thousands of dollars. St. Croix expressed no uncertain amazement that he could boost a frequency as much as 6-8dB and hear hardly any difference in the speakers system. Based on this startling experience, he observed that what you hear when adjusting conventional analog equalizers is primarily phase shift.

I’ve noticed my TDM EQ (which, by the way, I acquired because it was highly recommended by members of this forum) seems to exhibit “similar” qualities - much less in the way of audible change when adjusting filters, compared to most 1/3 octave EQ’s I’ve used in the past. I often find myself checking to see if I have accidentally switched it from ±12dB to ±6dB range.

So based on the things you guys have said, this equalizer has very good phase characteristics – or filters with a tight Q -- ?

Which isn’t necessarily a good thing if your goal for equalizing is to “improve” the way things sound.

So it looks like perhaps we’re chasing our tails.

Regards,
Wayne A. Pflughaupt


Reply posted by John Roberts on June 15, 2002

I believe Mr. Danley has identified the more important relationship... phase and amplitude are intimately related in all minimum phase filter designs. This is useful because phase and amplitude are linked in many naturally occurring frequency response errors, and corrective equalization that restores flat frequency response will also restore phase response.

Without meaning any disrespect to Mr. St Croix (his Marshall Time Modulator was popular in the day), amplitude is far more audible than phase. Digital filters without phase shift are interesting, but without the phase shift wouldn't be right for most corrective EQ. (Relax most digital filters do have phase shift and act like analog filters, just with more accuracy and repeatability).

JR

PS: TD also noted some non-minimum phase naturally occurring FR problems, so perhaps the phase shift related to typical EQ might not be beneficial in those cases... but IMO trying to EQ away non-minimum phase errors is like trying to fix a 3 dimensional problem with a 2 dimensional solution... such problems are better eliminated at the source than EQ'd away, if that's an option.

PPS: I don't know that I've ever heard a "sound" associated with pure phase shift... The common pedal/studio effect "Phasor" is created by an all-pass (phase shift) circuit, but what you hear is the amplitude peaks and notches when the phase shifted signal interferes with the dry signal.

I do not deny that phase shift could be audible. I suspect it may cause subtle interaction with how harmonics constructively and destructively interfere but is probably down there with absolute polarity. Yes, it's audible in theory but far from significant in practice with typical PA.


Reply posted by Curtis H. List (Too Tall) on June 15, 2002

“About 10 years ago, in the early 90s, I read an article by Stephen St. Croix”

I remember this and your description could be interpreted several ways. (I wish I could remember the device. I’ve tried more then once.) The device was an early FIR digital device (Finite Impulse Response) . They cost so much because they took a ton of processing power. The typical IIR (Infinite Impulse Response) takes less, but has some of the same flaws as analog filters including that they both “ring” when you use a high “Q” filter.

Your statement above- “he could boost a frequency as much as 6-8dB and hear hardly any difference in the speakers system” could be interpreted as he could not hear any change at all. That was not the case. An example of what he wanted was to take out a narrow very high peak centered around 8kHz in a recording. An analog filter or an IIR digital filter set to a high “Q” value would “ring” and cause more trouble then it cured. He was happy that the FIR digital filter could erase the 8kHz narrow peak with no ill effects. (As I recall some of the filters examples he cited were even more extreme in cut.)

“Based on this startling experience, he observed that what you hear when adjusting conventional analog equalizers is primarily phase shift.

“I’ve noticed my TDM EQ (which, by the way, I acquired because it was highly recommended by members of this forum) seems to exhibit “similar” qualities - much less in the way of audible change when adjusting filters, compared to most 1/3 octave EQ’s I’ve used in the past. I often find myself checking to see if I have accidentally switched it from ±12dB to ±6dB range.”

To me this MAY be a sign of a poor piece of gear…or it could be you are not using it properly (Trying to do something that can not be accomplished by that method). The cheap EQs I have used can exhibit similar problems to what you describe. I reach for a filter slider to get rid of a peak at 2kHz and either I have to pull the filter to the bottom or it sounds like I have mistakenly moved the overall gain instead. With a decent EQ, used where it can change what you want to effect, it usually takes a very small movement of the slider and the change is of positive benefit.

“So based on the things you guys have said, this equalizer has very good phase characteristics – or filters with a tight Q -- ?”

I know them more by performance then a knowledge of their unique phase properties. For performance the best thing available is a “full parametric” where you can precisely match the filter to the problem. A third octave is easier to use and understand, but will not do the job near as well unless you are veeeeeeery lucky. The biggest problem with a parametric is you typically need many filters and they are expensive.

The above is a large thread in itself and I strongly urge you to do an archive search for “parametric.” To come back to your question on “what does bad phase sound like”, in relation to cheap 1/3 octave EQs the main problem is for want of a technical term “sloppy” filters.

I have heard a couple of my friends point to the fact that the filter labeled “2.5kHz” on their 1/3 octave EQ is actually somewhere below 2kHz as a symptom of a “poor” EQ. While that is confusing I believe the part that kills you is that the filters themselves are poorly shaped and interact in a less then advantageous manner. When you pull two adjacent filters the response is a mess (not a smooth drop at frequency in the center between where you hoped for) and the sound is worse then the peak you tried to excavate.

For instance if your peak is 2.2kHz (between the 2kHz and 2.5kHz slider) it might be digging narrow holes either side of the peak and leave a narrow peak in the middle. This is just a made-up example. My point is for the poorly designed filters they don’t play well together.

Too Tall