[sdiy] Ray tracing hardware for audio simulation

[Guy McCusker]

On Mon, Aug 1, 2022 at 2:41 PM Matthew Skala via Synth-diy
<synth-diy at synth-diy.org> wrote:
>
> On Mon, 1 Aug 2022, Mike Bryant wrote:
> > 1mS delay is too much for live voice processing so they need to work on
>
> Move the speaker a foot closer to the listener

I was intrigued by the discussion around the significance of such
small latency and started reading about human's ability to perceive
such timing differences. Some internet searches led me to the Haas
effect, aka precedence effect, as described e.g. here
https://en.wikipedia.org/wiki/Precedence_effect

I found it quite interesting to play around with adding a few ms of
delay to one channel of an audio track and listening to the effect.
For me, anything up to 40ms, and a bit more in fact, manifested itself
as a shift in the perceived location of some of the sounds. High
frequencies in particular seemed to shift left as I made the right
channel play later. I did not perceive the two tracks as being offset
from one another at all. Beyond 40ms delay the two tracks did begin to
separate and sound like a very short echo. If you have not experienced
this before it's quite an interesting use of a few minutes! I am
starting to think that this is also the reason that manufacturers of
hifi speakers recommend placing the speakers a couple of feet away
from walls for a "wider soundstage" -- reflections will cause this
kind of effect and spread out the perceived location of the sound;
presumably at the cost of a weird frequency response thanks to
cancellations. If someone here knows more about this stuff I'd be glad
for any good references to read up on it.

On the topic of avoiding latency, the above strongly suggests that 1ms
here or there is not a problem. But that does not mean you should not
try to avoid introducing latency if you can. These systems will be
chained together, latencies will add up, and uncontrolled latency
can't be a good thing, especially if you don't know which paths have
what latency. So it seems smart for engineers to keep it down as much
as they can, and it is obvious that manufacturers of low-latency
products will trumpet that as a benefit, and others had better catch
up.

So I guess I agree with both Mike and Matthew, but Matthew's message
made my day, as well as prompting me to look into this, which I have
enjoyed. Thanks!

Guy.