Experimenting with opposite-ear cancellation (like SDA or Sonic Holography) in software

[duncan]

I wasn't sure where to post this, as it's a no-soldering-necessary (and no-pants-necessary) enterprise. Admins feel free to move it elsewhere if appropriate.

Having never actually heard any Polk SDA speakers, nor any of Carver's sonic-holography products, I have sometimes been curious about how - and, frankly, whether - the effect worked. As I understand it, the idea is to make each ear more exclusively hear its corresponding speaker, reducing confusion of auditory cues. Normally the left ear hears something from the left speaker, then the concurrent signal from the right speaker slightly delayed (because that speaker is further from the ear) and slightly attenuated (because part of your face is in the way). That second sonic event is undesirable for an imaging standpoint. The cancellation effect works by introduing an echo that is out-of-phase, channel-reversed, delayed and slightly attenuated. Also, possibly with an EQ curve applied, or a low pass or bandpass filter, because we're mostly interested in the midrange.

When an unwanted cross-signal from the R channel is about to hit the L ear, the out-of-phase echo should just be arriving from the L speaker to nullify it. And vice versa. Et voila, improved imaging?

Here's a two-line example that attempts to add opposite-ear cancellation to an existing FLAC file. It uses the freely available "SOX" utility.

sox -v -0.75 infile.flac tempfile.wav bandpass 500 4o pad 0.0002 swap
sox -m infile.flac tempfile.wav outfile.flac gain 2

Explanation of parameters, in the first line:
"-v -0.75" - this is both a volume and phase specification. 0.75 would be a slight attenuation, as it's a multiplier where 1 would have no effect and 0 would be silent; -0.75 negates the signal, i.e., takes it out of phase.
"bandpass 500 4o" selects four octaves centered at 500hz, so approx. 125-2000hz.
"pad 0.0002" introduces a delay in seconds; 0.0002 is 0.2 milliseconds.
"swap" refers to channels, left <-> right.

In the second line, where the main and cancellation waveforms are combined, "-m" means mix, and "gain 2" is a tweak that I thought it seemed to need to restore about the original volume.

There's a lot here that can be adjusted. Unfortunately it can't be done in real time like turning knobs while listening: you have to run a new conversion and then play the resulting file. I find that it takes almost as long to convert a file as it does to play it, at least on the pipsqueak Rasbperry Pi that acts as my home media server. Probably it would run a lot faster on a Real Computer.

I'm just getting started and won't comment on listening results just yet. But I'll be curious if others would try it too, see if they can confirm whether this is on the right track, and if so, what parameter tweaks might hone in on the best effect possible.

Note: the files being worked on don't strictly have to be flac format. They can be in any stereo format that your version of SOX was compiled with support for. In my case it also works for mp3 and probably ogg, but I prefer flac where possible. If your SOX install doesn't speak flac, it should always work for wav.

[late edit for a math mistake]

 

[SDA SRS 1.2]

The SDA effect is definitely real as long as the speakers are placed properly and obviously must have the interconnect cable between the speakers. Certain well recorded program material helps to enhance the effect as well. Stereo Review tested them in the late 80’s and noticed the effect, but was mainly impressed with the overall sound quality on all types of music. Placement away from any side interference is probably the most important criteria as the owners manual has a very detailed set up procedure “to obtain maximum enjoyment of the SDA effect.”

The sonic holography feature on my Carver C-16 Preamp is interesting, but I’ve found it to alter the sound a little too much for my tastes. The C-16 is a great preamp even without using the sonic holography features.

 

[duncan]

The sonic holography feature on my Carver C-16 Preamp is interesting, but I’ve found it to alter the sound a little too much for my tastes.

While using SDA speakers (just noticing in your signature that you have some), I guess I'd expect "doubling up" the effect using the preamp would sound strange, because the delayed signal generated by the preamp would then be repeated with further delay by the outer SDA drivers... for a second set of ears maybe farther back on the head?



Or are you talking about using the Carver feature with different speakers? But what do I know... again this is all new to me, just messing around with a software implementation and playing into conventional speakers.

 

[sKiZo]

I can attest to the Carver holography having a major influence on imaging here, but ... it also makes for an extremely small "sweet spot". Plan to spend a lot of time positioning speakers and tuning the room for maximum effect.

Oh ... and once you get it right, might want to get one of these to lock in said sweet spot when listening ...

PS ... I think you'll find that a software approach may not be as easy as one might think. Yes, the tech might now be considered primitive, but the outcome is an effect that would be hard to recreate without knowing exactly what frequencies are being tweaked, as well as when and how much, and what exactly triggers it. I remember a few folk playing with a software version of dbx expansion and abandoning the project as it slowly drove them bonkers.

 

[SDA SRS 1.2]

Or are you talking about using the Carver feature with different speakers? But what do I know... again this is all new to me, just messing around with a software implementation and playing into conventional speakers.

The C-16 was initially connected to non-SDA speakers and the sonic holography effect was like sKIZo said, an extremely small sweet spot. Even then, it reminded me of too much “reverb” and just wasn’t something I would want to listen to for an extended period. The SDA effect is much more “spacious” and less intrusive.

 

[duncan]

Thanks for the clarification. Just curious, did Carver provide any adjustments on the effect? Say, in the length of the delay for the cancellation signal, or maybe its relative volume? Were there knobs involved, or was it a single switch?

To elaborate on the first post, I'm hoping somebody will try the supplied SOX calls and compare notes on how the effect sounds to them. I do think I'm hearing something promising, but it's not convenient to A/B so I can't be sure it's not just, you know, hearing what I expected to hear. Initial impression is that sources between the speakers are perceived with better focus, such that when I go back to listen to the original file it sounds slightly smeared by comparison. What I'm not getting, at least not yet, is the expanded stage that's been described for the commercial implementations. Also there isn't a sweet spot per se -- although it's clearly best when listened to from the center, it doesn't fall apart as you move away from there, just gets gradually less effective. Doesn't really stop working until I get outside the "box" - like, sitting to the right of the right speaker.

Your observation that Polk's speaker implementation works better than Carver's signal-processing implementation makes sense to me. It would seem to correct more for difference in listening position at least on the Z axis: as long as you were facing more or less forward, the distance between the ears always remains about the same as between the drivers, regardless of how wide or narrow the base of the listening triangle is. Whereas in a signal processing approach, the delay is determined by time instead of distance, and ... (wanders off muttering, looking for a trigonometry textbook)

 

[sKiZo]

Thanks for the clarification. Just curious, did Carver provide any adjustments on the effect? Say, in the length of the delay for the cancellation signal, or maybe its relative volume? Were there knobs involved, or was it a single switch?

Bit of both. Either on or off, but there were several versions available. The C9 processor was considered the best by most, and had no adjustments (not even a bypass switch) but I find I prefer the H9AV. That was just at the head end of the AV revolution and works better with surround sound - important to me as I have a quad plus center system. For me, the effect is subtle, but sorely missed if I bypass it.

The H9AV does have an extra adjustable feature for height, but - not my cuppa tea. That gives you something of a reverb effect that I'm sure would enhance center stage vocals in a movie track, but which IMHO detracts from the music.

 

[SDA SRS 1.2]

Thanks for the clarification. Just curious, did Carver provide any adjustments on the effect? Say, in the length of the delay for the cancellation signal, or maybe its relative volume? Were there knobs involved, or was it a single switch?

To elaborate on the first post, I'm hoping somebody will try the supplied SOX calls and compare notes on how the effect sounds to them. I do think I'm hearing something promising, but it's not convenient to A/B so I can't be sure it's not just, you know, hearing what I expected to hear. Initial impression is that sources between the speakers are perceived with better focus, such that when I go back to listen to the original file it sounds slightly smeared by comparison. What I'm not getting, at least not yet, is the expanded stage that's been described for the commercial implementations. Also there isn't a sweet spot per se -- although it's clearly best when listened to from the center, it doesn't fall apart as you move away from there, just gets gradually less effective. Doesn't really stop working until I get outside the "box" - like, sitting to the right of the right speaker.

Your observation that Polk's speaker implementation works better than Carver's signal-processing implementation makes sense to me. It would seem to correct more for difference in listening position at least on the Z axis: as long as you were facing more or less forward, the distance between the ears always remains about the same as between the drivers, regardless of how wide or narrow the base of the listening triangle is. Whereas in a signal processing approach, the delay is determined by time instead of distance, and ... (wanders off muttering, looking for a trigonometry textbook)

The sonic holography activation was by push button, but it’s effect could be manipulated by the BLEND and ABSOLUTE PHASE adjustments.

 

[a_retent]

My Carver C4000 pre only allowed for settings of normal or theatrical, meaning some or over the top flavor if you will. I typically set mine to normal and yes, one had to find a limited "sweetspot" to realise the effect.
In the latter 80's Lexicon produced the cp-1 environmental processor that had a selectable panarama effect that addressed the sound cancellation effect you refer to but did it in the digital domain vs. analogue approach like Carver's Sonic Holography or Polk's passive SDA/SRS approach. I have one on 3 of my (4) 2ch systems.

The unit allows you to adjust the listener position and speaker phase angle in a "normal, wide and binaural" settings with individual memory for each. The technology works if you know what your looking/listening for and if you don't then you ain't gonna hear it.
Something else to look at would be the Hughes AK100 (SRS) unit which also utilizes Head Related Transfer Function (HRTF). I wish but cannot help you from the software aspect on this subject but have been a believer in the HRTF effect since the 80's and understand what it is and what I hear...it works for me and after I explain it to other audio enthusiasts that hear it on my system, they too get it, especially when I turn it off.
Best of luck, I look forward to a software version of this concept with even more acoustical adjustments and control.

 

[rothwellaudio]

Normally the left ear hears something from the left speaker, then the concurrent signal from the right speaker slightly delayed (because that speaker is further from the ear) and slightly attenuated (because part of your face is in the way). That second sonic event is undesirable for an imaging standpoint.

Your initial assumption is both right and wrong in my opinion. Yes, sounds from a source on your left will reach your left ear first, then reach your right ear slightly delayed and slightly attenuated (and frequency balance altered), but that is how human hearing works to locate the origin of sounds. Without those delayed signals we wouldn't know where sounds were coming from and "purist" stereo recordings are made with two microphones (or a stereo microphone) and designed to be reproduced by two loudspeakers with both ears able to hear both speakers. If you listen to the same recording on headphones, the left ear only receives the left channel and the right ear only the right channel, which achieves what your software is attempting to achieve. Does that improve the sense of stereo? Most people think not.
However, binaural recordings (using a dummy head microphone) are designed to be heard over headphones and do give a very convincing stereo effect, but they don't sound so good over loudspeakers. So, is left/right isolation a good thing or not? That depends on how the recording has been made.
Of course, in reality 99% of today's recording aren't made with any kind of stereo microphone techniques at all. Instead isolated tracks are recorded in mono and panned left or right during mixing to create a sort-of stereo effect, and stereo reverb is added to enhance the effect, but they aren't "purist stereo", so playback equipment designed specifically for purist stereo seems like a waste of time to me.

 

[Hyperion]

so playback equipment designed specifically for purist stereo seems like a waste of time to me.

 

[rothwellaudio]

Let me re-phrase that - wouldn't want you to have a heart attack
What I meant to say was that maybe only 1% of music is recorded in purist stereo, or maybe none of the recordings you have are purist stereo if you listen to pop/rock etc. Ok, if you're a classical music buff the percentage will be higher.
Anyway, modifying your system to optimise that 1%, possibly to the detriment of the other 99%, seems like a futile pursuit to me.

 

[Hyperion]

Thank you! <phew> - heart attack averted.

 

[rothwellaudio]

My earlier post may have sounded like I was being dismissive of purist stereo and the OP's attempt to improve it, but that actually isn't the case - I do like the sense of space that good recordings give you. I've even experimented with it myself and made stereo microphones and even started to make a dummy head stereo microphone (though that's yet another unfinished project), but there's never been any consensus as to what's the best approach. There are various techniques, such as X/Y, A/B, ORTF, dummy head, Jecklin Disc, M/S etc, but none seems to have emerged as a winner. The recording industry on the whole doesn't seem to care about purist stereo and nearly all pop/rock recordings are made without any regard to purist stereo at all. Their recordings are made with modern techniques and mixed to give a pleasant, if artificial, sense of space. And if it sounds good, isn't that good enough? Trying to optimise the playback system on the assumption that all recordings are in pure stereo when in fact very few are is what casts doubt on the wisdom of attempting it. Of course, that's just my opinion.

 

[duncan]

Your initial assumption is both right and wrong in my opinion....

That's not my own assumption per se, rather repeating a point from Matthew Polk's article in Audio Magazine, June 1984.

Anyhow -- not necessarily to interrupt discussion about the whys and the whethers, but I'll keep putting out the plea: would someone be willing to try out the code from the first post (possibly on a variety of recordings), listen to the result on conventional equipment (i.e., not through any kind of additional spatial processing, and not into SDA speakers), and comment back so I can get a little bit of confirmation (or not) of the effect so far?