Phase Cancellation, Stereo, and "Room Simulation"

[SMcD]

Yet another thread in what's turning out to be a SMcD Series.

 

Long story short: I permanently lost the use of my left ear in late 2020, which forced me to learn the hard way about all of the problems that arise when summing stereo sounds into mono. Of course, all of it - mixes changing on recorded music, DP patches sounding lifeless, etc. - is worst through headphone(s), when L + R channels are summed electronically at the source.

 

However, I've found that it's never a problem for me with external stereo speakers. Even when they're fairly close together (like in my Traynor K4), the fullness of stereo sounds lands in my good ear without anything being noticeably lost or cancelled. As I understand it, this is because of complex physical interactions happening in the matter between the speakers and my ear. The sound waves bounce in unpredictable ways off speaker cases, walls, my head, etc. - so the final sound that arrives at my ear includes both channels, but with enough "room randomness" to smooth out any potential for egregious phase cancellation.

 

In other words, letting sound waves combine "in the air" produces better results than combining them electronically. With that in mind, I ordered a custom fit "single-sided stereo" IEM from Ultimate Ears. The L + R channels each have separate armatures, all crammed into one earpiece. The hope was that this would be enough physical separation for some "room randomness" to be introduced in my ear canal: an internal, scaled-down version of the phenomenon I experience with my K4.

 

Unfortunately, this did not turn out to be the case. Out-of-phase signals aren't totally cancelled as they would be with electronic summation, but they end up being so faint that the difference from mono sound is ultimately negligible. Sensaphonics has a similar product that promises compromise-free stereo, but I fear that the physics just aren't on my side. No matter which product I go with, it's a tiny physical separation between channels and a relatively small amount of matter between the source and my eardrum. Maybe there's just no way to introduce enough "randomness" to the sound waves with that setup.

 

But...the fact remains that I can hear a reasonable stereo image externally, in a room, with my K4. And it has me wondering if the "room" can be simulated.

 

I'm talking about some intermediate step(s) in a signal path that would introduce a significant degree of phase randomization, maybe a tiny bit of reverberation...whatever's going on between my K4, my walls, and me. You'd put one side of a stereo DP input through that (I think it'd suffice to do only one and leave the other side pure), then route the outputs as normal. What comes out of my single-sided stereo IEM would be both DP channels, but with enough simulated room randomness to keep the full image.

 

Does such a product exist? Could it be approximated with some combination of effects? What would be involved in the kind of "room simulation" I'm describing here?

[SMcD]

I wonder if it would suffice to apply some subtle, fast, small-room reverb to one of the DP channels (or separately to each of them), and maybe a tiny amount of phase shifting on one of the channels.

 

On the other hand, if it was that simple, wouldn't everyone do it?

 

EDIT: the more I think about this, the more baffled I am that the onboard reverb in a DP doesn't usually prevent phase cancellation. When applied subtly to a stereo patch, shouldn't that act as the kind of "room simulation" I'm talking about here? What's missing compared to a real room?

[mynameisdanno]

I like this idea.  Unfortunately, to your last question, artificial reverb (whether mechanical or digital) usually puts out a "100% wet" signal, with the assumption that you're mixing in just a taste of it with the original dry signal still mixed in fairly prominently (and therefore the "core" of the sound is unaltered and still subject to phase cancellation problems).  

What you'd need is a "room simulation" that models a pair of virtual stereo speakers being captured by a virtual mono mic some distance away, probably in a slightly asymmetrical layout, and in a regular room, not one explicitly designed to be highly reverberant.  UAD's Ocean Way Studios plugin might be suitable.  There are probably others.

[SMcD]

Thanks!

 

Ideally I'd love a hardware solution that's easy to use when gigging. Bonus if I can rig headphones into it for listening to music as well.

Unfortunately, a Google search for "room simulation pedal" or "room modelling pedal" mostly gives amp sims. As you can imagine, neutrality is a must given that we're dealing with DP's.

[ElmerJFudd]

What about a sample accurate delay between the left and right speaker.  You can raise the delay on just the left speaker by increments of say 200 samples until you get the effect that feels right in your listening situation. I suspect somewhere around 1400 samples would arrive at something that simulates the width you’re looking for.  
 

This can be done with a plugin on your master output.  

[SMcD]

Not a bad idea, but I do wonder if a small delay on one channel would introduce phase cancellations at some frequencies while mitigating others. Is the "room effect" as simple as a delay between channels, or would other factors be necessary to simulate the walls and such?

[ElmerJFudd]

9 minutes ago, SMcD said:

Not a bad idea, but I do wonder if a small delay on one channel would introduce phase cancellations at some frequencies while mitigating others. Is the "room effect" as simple as a delay between channels, or would other factors be necessary to simulate the walls and such?

There’s no way to tell until you try it.   If you do find build up you can try eq/filter on the delayed channel and try to tweak to suit as well.

 

A lot it hearing is psychological. So when working with one ear I’m not surprised that you’re still able to sense distance and proximity.  So a delay and whatever reflections are generated in patch reverb effect - on the K4 may give you something you like.  Also try reducing the volume of the delayed channel to suit.   

[John Tweed]

31 minutes ago, ElmerJFudd said:

What about a sample accurate delay between the left and right speaker.  You can raise the delay on just the left speaker by increments of say 200 samples until you get the effect that feels right in your listening situation. I suspect somewhere around 1400 samples would arrive at something that simulates the width you’re looking for.  
 

This can be done with a plugin on your master output.  

I have experimented with something similar in order to be able to use stereo (piano) samples in mono. Image you are at your slab piano with speakers, so your ears are equidistant from the speakers. Using a stereo piano sound play middle C - sounds nice. Now you move to the right hand end of the piano, head down, so the left speaker is directly aligned with the right, and you play the note again - still sounds good. Now the speakers are no longer equidistant, but they are coming from the same direction. Now you take the right channel directly into a mono system, apply a short delay (approx. 3ms to roughly equal the distance between the speakers) to the left signal, and then merge them. Now you are hearing a mono signal that sounds as good as the “mono” you simulated by positioning yourself with the speakers in alignment.

 

I did manage to set up my CP4 in this way - it worked, at least it gave rise to some interesting sounds as the delay was changed. It did not end up being something that I used as the CP4 has plenty of mono pianos already. I have been meaning to go back and experiment a bit more with it, in readiness for the next piano stereo to mono thread.

[SMcD]

Thanks for the ideas and input, folks!

 

UE claims that they can improve the stereo imaging from the IEM by adjusting the fit, so first order of business is to ship it back and let them try. I wonder if it'd suffice to change the angle of one of the holes a bit? Say, have the right channel bounce off the inside of my ear canal at a different angle so it doesn't interact with the left channel as directly. We'll have to wait and see.

 

Failing that, I'll start playing with delay and reverb. Of course, the "delay one channel by 3ms" approach bypasses the need for stereo playback entirely (i.e. it would also work when electronically summing to mono).

 

[One does wonder why they'd charge full price for a "single-sided stereo monitor" if the effect is doomed by physics to be virtually identical to mono. I can't think of any situation where you wouldn't be able to sum at the source, so what's the point of offering separated channels in one unit if there's no real difference?]

 

20 hours ago, John Tweed said:

I have experimented with something similar in order to be able to use stereo (piano) samples in mono. Image you are at your slab piano with speakers, so your ears are equidistant from the speakers. Using a stereo piano sound play middle C - sounds nice. Now you move to the right hand end of the piano, head down, so the left speaker is directly aligned with the right, and you play the note again - still sounds good. Now the speakers are no longer equidistant, but they are coming from the same direction. Now you take the right channel directly into a mono system, apply a short delay (approx. 3ms to roughly equal the distance between the speakers) to the left signal, and then merge them. Now you are hearing a mono signal that sounds as good as the “mono” you simulated by positioning yourself with the speakers in alignment.

 

I did manage to set up my CP4 in this way - it worked, at least it gave rise to some interesting sounds as the delay was changed. It did not end up being something that I used as the CP4 has plenty of mono pianos already. I have been meaning to go back and experiment a bit more with it, in readiness for the next piano stereo to mono thread.

 

In your experience, was the effect more or less uniform across the range of the piano? In other words: it works well for middle C; does it work as well for the C three octaves down? Three octaves up?

[jazzpiano88]

It’s a good idea, and I think it could be done through a plug in like a speaker or amp sim (of course one would need to develop it).  

I toyed with a similar idea a few months back over on SSS about a mastering plug in to insure all the instruments could be heard in the shopping mall by modeling the speakers and various building configurations. 

 

in your case I think it would be a lot easier so add some random phase to each channel; probably as a function of time and frequency.  Good project if you can do some prototyping in matlab or python.