I don't agree to the latter. In symphonic halls I'm always experiencing a huge amount of spaciousness. Without visual clues it's nearly impossible to localize individual instruments.
That said I don't think natural auditory scenes are the benchmark for stereo. Stereo is capable of pinpoint images. If the engineer doesn't want pinpoint images it's easy to make them more diffuse in production. Nevertheless pinpoint imaging is one quality criterion for stereo. If a setup can't do sharp images it can't faithfully reproduce anything else that's on the recording.
Here's what I'm more interested in though, what's the perceptual mechanism that lead to better imaging when digital room correction is applied? This is something a lot of people report. My current thinking is that it's a <1.5kHz phenomenon. It seems that ambiguous localization cues are reduced by reducing in-room amplitude response deviations between the left and right speaker. Interchannel amplitude differences lead to interaural phase differences with 2-speaker stereo. This effect might dominate recorded interchannel time delay cues.
Markus
The acoustic center of the drivers in native state effectively moves in and out along axis of driver travel dependent upon frequency.
Effective apparent source width and depth produced by speaker becomes fuzzy blob. This may be viewed as temporal aberration contributing to both the non flat frequency response and phase response seen in measurements.
With two speakers as fuzzy blobs, resulting image is full of phantom sources appearing as fuzzy blobs.
Speaker drivers are not minimum phase.
Correction by PEQ filters all aligned to same starting point are helpful, but do not address the issue nearly as well as DSP solution based on directly calculated inverse transfer functions for the individual drivers as mounted in a given speaker system.
Manually tweaking PEQ filters is possible, but a very time consuming proposition. Each PEQ needs its own time offset.
The acoustic center of the drivers in native state effectively moves in and out along axis of driver travel dependent upon frequency.
Effective apparent source width and depth produced by speaker becomes fuzzy blob. This may be viewed as temporal aberration contributing to both the non flat frequency response and phase response seen in measurements.
With two speakers as fuzzy blobs, resulting image is full of phantom sources appearing as fuzzy blobs.
Speaker drivers are not minimum phase.
Correction by PEQ filters all aligned to same starting point are helpful, but do not address the issue nearly as well as DSP solution based on directly calculated inverse transfer functions for the individual drivers as mounted in a given speaker system.
Manually tweaking PEQ filters is possible, but a very time consuming proposition. Each PEQ needs its own time offset.
A single speaker provides a different acoustic input to our hearing than two speakers. So I don't think it's safe to make an inference from the single "fuzzy blob" case on the two "fuzzy blobs" case.
Just look at unsmoothed phase response of any driver at resolution of a few degrees over a narrow frequency range.
Group delay is convinient view of same problem:
If driver is minimum phase line would be smooth. In above the raw response deviations reflect the effective relative start time for any individual frequency. In above a 3ms window is used with no additional smoothing effects, thus measurement is direct sound of tweeter as mounted in speaker.
The group delay v wavelength is the acoustic center wobble.
Correction by PEQ would require at least a filter for each peak, and likely would need tweaks to start times. Each filter would need a fairly high Q, adding to difficulty of getting result nearly as good as FIR filter used in corrected view.
Group delay is convinient view of same problem:
If driver is minimum phase line would be smooth. In above the raw response deviations reflect the effective relative start time for any individual frequency. In above a 3ms window is used with no additional smoothing effects, thus measurement is direct sound of tweeter as mounted in speaker.
The group delay v wavelength is the acoustic center wobble.
Correction by PEQ would require at least a filter for each peak, and likely would need tweaks to start times. Each filter would need a fairly high Q, adding to difficulty of getting result nearly as good as FIR filter used in corrected view.
Mitch's comments about piano and violin are spot on. Also sound of drumstick making contact with skin, rim, or metal become clearer.
That's a speaker after EQ has been applied, right? Could you post that IR as WAV please?
Last edited:
^
Thanks but this is after EQ so it's not clear what causes non-MP behavior. Can you post an IR without EQ?
Thanks but this is after EQ so it's not clear what causes non-MP behavior. Can you post an IR without EQ?
Understood. Many posts sound more like advertisements however.
And it may also be that reporting "better imaging" is just hype. A lot of people don't even know how to recognize it. If you look at Olive's blind results of room EQ they are anything but "ideal" and/or ubiquitous improvements. But people "report" wonderful things about them.
Advertisement for what?
Currently it's all just anecdotal evidence but a couple of people which ears I trust - inluding me 😉 - report the same thing.
Not sure what test results you're referring to. I've never read an Olive paper that would warrant your conclusion.
Here is link to stereo file of raw tweeter and woofer impulse responses:
https://www.dropbox.com/s/sytyifp1tn32dg5/tweeter and woofer impulses no EQ.wav
Yes, and many people report miniture PA speakers as being high fidelity.
Most of the reports supporting this view point also are more like advertisements too.
Thanks. When looking at the excess group delay, it's virtually flat. What's the mic distance?
Well I don't trust anyone's ears as you know.
I don't think that Sean's work was ever a paper, but I believe that he reported it on his blog. I am really surprised that you didn't see this because I found it quite enlightening. I think that he took a lot of heat for it since it was not very complimentary of the technology.
You are confusing group delay and excess group delay. Flat excess group delay shows a system is minimum phase. A system can be minimum phase and still have lots of normal group delay.
When trying to determine acoustic centre and whether it moves with frequency it's excess group delay (excess phase group delay) not group delay you want to be looking at.
Last edited:
Link? I read his blog and I've read all the Harman papers and I can't remember him saying what you claim.
The majority of "conventional" drivers are almost perfectly minimum phase.
One exception is a whizzer cone full range driver - most definitely not minimum phase in the treble, and I've posted phase measurements of two such drivers before.
The reason being is that in reality its not one driver but two different decoupled cones with a mechanical (approx 2nd order) crossover between the two cones. Therefore the phase response looks very much like a 2nd order crossover with about 180 degrees of excess phase rotation through the transition region. (Centred at about 6Khz on the 8" drivers I measured)
Makes getting good phase tracking when crossing over to a tweeter "interesting" 😛
Last edited:
- Status
- Not open for further replies.