The quantification of stereo imaging

[bluefuzz]

Many of you still have a "test-switch" built into your amp, it´s the mono-switch.

A few years ago I realized I had been listening to music for several months with the mono switch on. I honestly never noticed, although I do remember thinking at the time that some tracks didn't quite sound like they usually did. But I never put 2 and 2 together ... ;-)

 

[audiophile]

If you have ever performed any acoustic measurements you will know that there are massive variations created by just moving the mic a few mm.

Here I posted some graphs, when I was trying to measure the effect of tube rolling and realized it's not possible to get anything meaningful due to background noise affecting the results.

 

[March Audio]

Here I posted some graphs, when I was trying to measure the effect of tube rolling and realized it's not possible to get anything meaningful due to background noise affecting the results.

Yes. Whilst this might be an interesting subject for dshreter to explore, its not a realistic prospect to get any meaningful data in a domestic environment.

 

[Blumlein 88]

A few years ago I realized I had been listening to music for several months with the mono switch on. I honestly never noticed, although I do remember thinking at the time that some tracks didn't quite sound like they usually did. But I never put 2 and 2 together ... ;-)

You lose all of your audiophile credentials. You must not pass Go, and do not collect $200. You can only use the included with your phone earbuds for the next 10 years. Sorry pal.

 

[Davelemi]

Whatever Dirac is doing it certainly seems to be a fairly reliable method of creating a realistic and believable soundstage in my room with wildy different speakers.

100% percent agree with you on how Dirac effects soundstage at least in my near field set up.

 

[dshreter]

I set out to produce a set of proof of concept measurements with the methodology i suggested before. My hypothesis was that you would see a meaningfully higher SPL from the microphone when placed on the side closest to the speaker while playing a tone from a typical left speaker angle. When repeating the measurements from dead center, the difference between left and right microphone placements would disappear.

I played a 3000Hz tone through the left speaker, and measured the SPL on the left and right side of a dummy head (a rubber ball). I repeated this measurement two times, alternating sides for each pass, and compared the recorded SPLs from 5s-10s so the measurement was settled in. I then did the same thing with a speaker centered with the ball. These were not expected to get highly repeatable measurements because the microphone was simply placed by hand on a cushion, and I wanted to see results before fussing around with a test rig.

As you can see in the results, left ear is distinctly louder than right ear on the 60 degree measurements. When placed at 90 degrees, the measurements are closely bunched. I consider this validation of the hypothesis that the amplitude contribution to localization could be easily measured.

To improve the design, this would be much better if performed with simultaneous dual channel measurements. As well, a consistent test rig would be needed for the repeatability necessary to compare results from one pass to the next. Lastly, a more realistic dummy head would make for a better representation of the head transfer function of the measurements. Like suggested earlier, something the miniDSP EARS should be able to perform this adequately if it can be used in a dual channel mode.

Finally, I would be very interested to perform a similar test with phase comparison - alas, this will definitely require dual channel measurement. I'll think about whether or not I'm motivated to go out and get the equipment necessary to take this further.

 

[dshreter]

Yes. Whilst this might be an interesting subject for dshreter to explore, its not a realistic prospect to get any meaningful data in a domestic environment.

Actually I went ahead and started to make measurements. For a 3kHz test tone I found that there was ~8dB difference from side to side on a crudely constructed dummy head and measurement approach. That's at a level that's very easy to detect, and with a more precise rig there should be no problem performing a medium SNR series of measurements.

That's not to say I've answered how we could measure "imaging" performance of a system, but I have determined we can easily measure one of the three variables known to contribute to localization. That seems like a great starting point.

Further tests will require a better testing apparatus, and could include a number of studies:

• SPL differential vs frequency
• Phase differential vs frequency
• Comparison with and without acoustic absorption panels on side / front / rear walls

I hope to discover reliable measures of in-room stereo separation and focus within a system. Given they don't exist today, that's only going to happen by trying new things and seeing what I find. Helpful advice and coaching will be greatly appreciated because I'm a novice! I've learned a lot on this site already. Shooting down of ideas out of hand will be gladly accepted as a challenge.

 

[bluefuzz]

You lose all of your audiophile credentials.

LOL! I know, I know. Not that I had any audiophile credentials to begin with. I freely admit to not being able to hear the difference between Spotify 320 bps and Tidal lossless, let alone Tidal MQA. Nor the difference between redbook 16/44.1 and 24/192 'hires'. I maybe lacking in imagination ...

And I don't miss vinyl LPs one bit, except for the nice big covers. Definitely a sorry excuse of an audiophile ... ;-)

 

[Hipper]

dshreter, it would be worth you reading pages 159-166 of Floyd Toole's 'Sound Reproduction. He discusses the phantom image and how interaural crosstalk (the left ear hears sound from both the left and right speakers) can lead to some comb filtering which reduces the decibel level at around 2kHz by perhaps 6dB. This is because the sound coming from the right speaker to the left ear has to travel a bit further, round the head, then that from the left speaker etc.. The frequency of 2kHz has a wavelength of around 7", 3kHz around 5".

Is it possible that the 8dB loss of the 3kHz signal is caused by the 'HRTF' of the rubber ball?

Toole goes on to say that some of this dB shortfall can be made up by side wall reflections.

 

[dshreter]

dshreter, it would be worth you reading pages 159-166 of Floyd Toole's 'Sound Reproduction. He discusses the phantom image and how interaural crosstalk (the left ear hears sound from both the left and right speakers) can lead to some comb filtering which reduces the decibel level at around 2kHz by perhaps 6dB. This is because the sound coming from the right speaker to the left ear has to travel a bit further, round the head, then that from the left speaker etc.. The frequency of 2kHz has a wavelength of around 7", 3kHz around 5".

Is it possible that the 8dB loss of the 3kHz signal is caused by the 'HRTF' of the rubber ball?

Toole goes on to say that some of this dB shortfall can be made up by side wall reflections.

Great insight. With this crude setup, there could be multiple effects like you’re suggesting. Part of it is the opposite ear doesn’t have line of sight to the speaker, it’s blocked by the “head”, so it’s unsurprising it would have a lower level. I also expect the opposite ear to the speaker hears more reflected sound as a proportion of total compared to the near ear, which will have a greater proportion of direct energy.

 

[eliash]

The LEDR test gives you some idea if you speakers are setup well. Here it is online. Was on early Stereophile test CD's and Chesky demo CD's.
https://www.audiocheck.net/audiotests_ledr.php

Also the recently discussed JA interview, he mentions playing pink noise on both channels and seeing if you get a very thin pinpoint location in the middle with no sense it is coming from the sides.

...managed to extract the LEDR files now from the website (with some browser tricks) and play it on my foobar-based CAS.
"left, right, up, over and lateral" work more or less on speakers, "behind" won´t do...similar impression with Stax L700 (in-head localisation, also w cross-feed), "up" not very high...my personal HRTF???

...the real killer for my listening room though was the HD MATT test, predominant "muddy" or non-accented sound in the listening position, improves only really close to the speakers...as expected...


Anyway, a great website, congrats to Stephane!


One late remark:
The above MATT result seems to be a quite normal acoustic behaviour in a room with measured reverberation time T60 = ca. 0.36s and a calculated critical distance of rH = 0.75m.

 

[TankTop]

If you want to experience great imaging play Call of Duty Modern Warfare. I hear footsteps behind me, helicopters fly over my head and and bullets whiz by me.

Strangely I get the best 3D sound when my speakers are slightly different distances and they are towed in at different angles. The addition of a subwoofer I feel increases the soundstage dramatically.

 

[dasdoing]

horizontal placement of the sound is all about treating early reflections.
about vertical I have an intresting effect here in my early reflection treated room and my rew corrected TV(!):
as soon as I engage the rew filter the sound goes up into the middle of the screen. when I remove the filter the sound moves to the bottom.
in fact the speakers are below the screen on my TV

 

[onion]

1. In real life, a point source that emanates sound transmits a sound wave that reaches the left and right ear at different levels and different times depending on its position relative to the ears. Those differences enable the brain to localise the position of the point source.

2. When listening to stereo speakers playing back that point source, two separate point sources reach the left and right ear. This is not a perfect replication of 'real life'; spatial information is lost and the brain is unable to localise as well as from the original point source.

3. So anything that replicates the interaural time difference and interaural level difference the same as the original point source should reproduce the stereo image perfectly. There should be a way of quantifying the effect of speakers on this replication using binaural recording.

The only complication to this is that sound engineers usually mix sound using a stereo setup that suffers the same problem as described in 2. So if the information that was lost from going from 1. to 2. is restored, choices made by sound engineers when mixing multiple point source recordings could result in unpredictable stereo imaging effects.

 

[eliash]

1. In real life, a point source that emanates sound transmits a sound wave that reaches the left and right ear at different levels and different times depending on its position relative to the ears. Those differences enable the brain to localise the position of the point source.

2. When listening to stereo speakers playing back that point source, two separate point sources reach the left and right ear. This is not a perfect replication of 'real life'; spatial information is lost and the brain is unable to localise as well as from the original point source.

3. So anything that replicates the interaural time difference and interaural level difference the same as the original point source should reproduce the stereo image perfectly. There should be a way of quantifying the effect of speakers on this replication using binaural recording.

The only complication to this is that sound engineers usually mix sound using a stereo setup that suffers the same problem as described in 2. So if the information that was lost from going from 1. to 2. is restored, choices made by sound engineers when mixing multiple point source recordings could result in unpredictable stereo imaging effects.

Additionally to the direct sound, as mentioned in 1 + 2, also wall, ceiling and floor reflections alter the aural impression of a center point image, when played through a stereo set-up. From my personal perspective these "un-natural" reflections are the disturbing and annyoing ones, especially in a rather asymmetrical listening space. Therefore a proper room treatment to control these reflections will enable an optimum recording room impression, as it was thought to be by the engineer.
A recent impressive example for me was listening to ECM´s "Playing The Room" by Avishai Cohen & Yonathan Avishai, recorded in a large, but empty auditorium. Looking at the liner photos, I felt somehow "home" in that recording environment when listening.
I think this was also what Ethan Winer had in mind, discussing room treatment somewhere before here...