An intriguing sound design question....

BeatNik

DJohn Mustard Project
Messages
5,723
Reaction score
246
Location
London
How would you simulate a sound constantantly travelling past you e.g:

ears parallel and facing a traintrack and a long freight train drives past you for a minute...

Now let's say i had a sound collage which i wanted to travel past from left to right as though it were a train... how'd ya do it?

i.e. the entire sound doesn't move... there are always elements at either end of the stereo field.. except these elements are constantly panning (maybe even dopplering) past you...

Reason I ask is I found a studio engineered sample of a train doing exactly this... and i'd like to apply it to something else....


hmm...

tricky un...
(i may not be thinking straight though lol :Wink3: - undoubtably there's some serious stupid flaw in this question which I fail to understand yet)
 
matt said:

:ilol:

*bastard*

:Wink3: just joking....

this is vexxing... i'm sure i have the answer up here in my noggin somewhere *knocks on head* but my brain is functioning super slowly today lol (must be the bunged up cold)
 
No, you're right, I am a bastard :Wink3:

In stereo all I can think of is doppler and delay, probably with some very mild filtering to emulate front/behind - sound from behind is never direct as it's partially absorbed by ear flesh and will be slightly less bright, but that's mega hard work to get right imho, better look for some sort of dynamic spatialiser.

If it's just panning, you'll have to "pulse" constantly from one channel to another to emulate the "long train passing by" effect since that's what you're really hearing, individual pulses of that characteristic 2-step noise, fading one pulse in as another fades out... doppler is either a one-off or a circular thing.

I hope I know what you're on about :Smile3:

Edit: that's not to say you can't use two linear (one-off) dopplers to reenforce the panning.
 
yeah exactly what I'm on about...

been sitting here thinking abit more...

You'd have to emulate azimuth... and pan different events, as well as simulating the pulsing...

thinking more i'd say you could key different events to doppler past in whichever direction, but not simulataneously - so you always have activity at either end of the stereo field..

meh... more work than I'd planned on :Wink3:
 
Just a thought, a stereo granular synth might be what you're looking for, but will need to support per-granule panning, dunno if anyone ever thought of that? I mean instead of just triggering a mini AEG, have that mini-sample pan...

A hardcore solution would be to sample something with those characteristics (eg train) and convolve your target sound with that, but that's pure trial and error until you get the right impulse response. Dull and very CPU intensive but kinda ultimate solution for any DSP problem...
 
matt said:
Just a thought, a stereo granular synth might be what you're looking for, but will need to support per-granule panning, dunno if anyone ever thought of that? I mean instead of just triggering a mini AEG, have that mini-sample pan...

A hardcore solution would be to sample something with those characteristics (eg train) and convolve your target sound with that, but that's pure trial and error until you get the right impulse response. Dull and very CPU intensive but kinda ultimate solution for any DSP problem...


Per granule panning would take up a ferocious amount of CPU also... unless it were in some way possible to do offline... great idea though! Abit too brain confusing to think about visually though "all those little granules floating past" :Wink3:

Convolving is a suuuper idea actually, only need to find some way in which i could use the impulse response of the train sample and then apply that offline to the sound i want to process... the filtering, pulsing, azimuth and panning should be replicated exactly to be applied to the target... only thing which boggles my mind is having instances panning at the same time as others remain... that should in theory be 'caught' when convolving though... :S

Any ideas of something I can use for this?

Interesting thread :Smile3:
 
actually what am i on... could i use the impulse response of a sample!??!?

would that be possible... or would it have to be an own recording... ?

aaargh... i'm confusing myself to an hilarious extent!:iconfused
 
Cool Edit/Audition has convolution. Impulse response is really just another sample to convolve the target sample with (eg church echo of a click as reverb, like that sampling reverb unit by Yamaha) so anything goes really... Convolution is kinda base method in DSP, everything else spins off it (at least in theory - IIR filters are back-to-front approach afaik but that's just maths and that). You can do anything with it as long as you've got the right impulse response... sounds easy but in my experience can be very hit and miss. Cool for using your gf's favourite/sexiest vowel as reverb :Wink3:
 
I'm way out of my depth hear, but I know that Leslie speakers (or their software versions) use the doppler effect, essentialy the speaker is constantly rotating, and you could maybe set it up to send like something was constantly passing by with some stereo delay. Could that help at all? Or am I way off track (bad pun ha ha)
 
To get the actual impulse response though of a sound... would you not need (a la that sampling reverb unit) a dry click which is then compared to the click with reverb from the suroundings?
I don't have a dry version of a train going past, so how'd I get the impulse response of it?
(unless the sound of the train going past by itself can convolve my target sample)...

hmm...

Will have a go though using the train sample in Cool Edit though anyway :Smile3:

Cheers though! :iyes:
 
It sounds to me very similar in concept to a Risset tone and I would suggest approaching it in much the same way, first by creating the sound of one wagon passing from left to right (which should be fairly easy) and then by using as many instances of this sound (separated in time by an appropriate amount which would depend on the speed of each wagon) as needed, depending on how long you wanted the train to be. I would imagine using a sampler containing the one-wagon sound (set to 'play to end', or one-shot) mapped across the whole keyboard, with keyboard follow turned off; I'd then trigger a new key every few seconds, starting at the bottom of the keyboard. This would enable the creation of a train 127 wagons long very easily.
 
BTW if you have Waves IR1 then you can convolve anything with anything. It's not just limited to reverbs. I think convolution is a red-herring here though, except maybe to add appropriate ambience to the final train sound.
 
Colin, that's exactly the goal here... And convolution *is* an anything-with-anything process by definition, no convolver is limited to reverbs only :p Sampling reverbs are simply the most common use as it's relatively easy to understand what's going on.

It's definitely not for the faint-hearted, but it *is* the only way to impose arbitrary characteristics over any source. The moment you start breaking down the sound as a whole into small bits you'll be limiting yourself to whatever approach you've chosen, whether it's clever granular synth or manual trigger & pan MIDI massage. What if you want to do this to your favourite William Blake quote? :p

A quick overview for BeatNik:

Impulse is an infinitely short and infinitely loud signal. In digital (ie discrete & finite) systems this is just single 0dB sample, ie a click followed by silence. It contains all harmonics (like saw wave) but with flat spectrum. Running this through any linear system yields the impulse response, ie tells you how the system responds to a pure impulse.

So for reverb purposes, the impulse response is the echoing of say church hall after the impulse (in practice, can be a short burst of white noise).

In this case though, you want to create a properly looping file that describes your desired effect, ie the train rushing past. As an aside, convolution is just a mathematical conversion of one system to another, there's never any comparison (see http://www.dspguide.com if you dare :Wink3: ).

When you apply that to the target, you'll get something akin to vocoding - but using the entire possible spectrum, not just a few filters... This should explain why convolution is so hard to use it yet essentially unlimited in potential. In this case you want the spatial distrubution characteristics, not the gain/freq part of the deal, otherwise it'll just sound like some overly refined vocoder.

So you need to remove information out of the train sample, leaving something like white noise with the stereo information preserved. How? Run it through Cool Edit's Noise Reduction but keep the noise instead :Smile3: The trick is in retaining just enough information to keep the spatiality, so that's the most important point and where most work will be. Make sure it loops seamlessly and that's your impulse response :Smile3:

Edit: not dissing any methods here, just pointing out the obvious - single triggers will always be samey while convolving affects entire sound, and a billion other effects too :Smile3:
 
I am trying to get my head round Wave arts Panorama www.wavearts.com/Panorama5.html
I used it in a sound effects thing I did for halloween simulating the movement of a galloping horse running round woods. Speakers where positioned amogst the trees and people who where near to them commented how great the movement of the galloping was. This was using the X-Y screan + doppler effect.
 
Wow!
Cheers so much everyone!

Big big thanks Matt, very very interesting stuff to read there *things click in head*... convolution is definitely something to experiment with.
Thanks Colin (as per :Wink3: ) thought about the risset tone actually, the illusion of an endlessly rising tone... that kind of fading can definitely be applied to samples! So will have a play with that...

Cheeeers!! :Grin:

Wavearts panorama looks really really interesting too!... been looking for something like that (not just Ambisone)... so will defo check it out! Cheers!
Just looking at the manual - it has different HRTF settings... from the KEMAR (which i was actually reading about 2 days ago) and from another persons head, which work differently for loudspeakers and headphones!
Interesting :P
Really really want this prog!

:ibiggrin:
 
BeatNik said:
Wow!
Just looking at the manual - it has different HRTF settings... from the KEMAR (which i was actually reading about 2 days ago) and from another persons head, which work differently for loudspeakers and headphones!
:ibiggrin:

On headphone setting i think its something to do with the crosstalk bieng canceled or the other way round. Far too technical for my little brain.
 
Back
Top