posted 09 July 2001 15:18            

Stuff like the Trueverb plug-in (Waves) always, naturally, makes samples sound as if placed in a room, but Iīd like something resembling someone standing in a field about a hundred meters away and shouting my name. etc.

posted 09 July 2001 16:19            

Does the atmosphere disperse the frequencies? (I don't think so... but it does seem like the bass lessens with distance...)

- DM

posted 09 July 2001 18:29            

that's a pretty good idea! In fact if you ever get a recording of an outdoor impulse we could tease it apart and construct reverb units that would behave like this too! That would save on the post-processing and allow realtime, assuming you have enough memory for the delay lines...

[Ach... sorry, I see that you had in mind a real-time approach using the impulse as a filter. Hmmm... But it seems that would require very large block convolutions -- and that's why I assumed about post-processing...]

[...actually, this gets to another idea, request, whatever... Using the FFT of the log of an FFT gives you something called the Cepstrum. Some people claim this can be used to analyse/remove reverb.

I had a boss years ago who always poo-pooed the ideas involved, but I keep seeing it mentioned in print. So maybe there really is something to it. On the surface the math holds together just fine. In question is what the results actually sound like...]

- DM

posted 09 July 2001 18:53            

Removing the reverb in the recording shouldn't need to be done very exactly, because the voice will be at low volume anyway. So if you can get rid of the dominant reverb, then add back in just a faint amount of secondary reflection for the ground-bounce then you might be close.

Also, ground bounce is going to cause very slight comb filtering because of the differential delay in the multipath. How much? and what frequencies?

For large distances, the path-length for purposes of attenuation will be nearly the same. A six foot person (2 m) at 100 m has a ground-bounce pathlength of 100.08 m. Hardly any difference as far as attenuation goes. So I guess the comb filtering will be quite strong!

What comb frequencies? The pathlength difference from primary ray to groundbounce ray is about 0.08 m. So the comb will cut those frequencies that have odd multiple half-wavelengths equal to this distance. So that would be wavelengths of 0.16 m, 0.053 m, 0.032 m, etc. At 340 m/sec speed of sound, this means frequencies of 2125 Hz, 6415 Hz, etc. So it looks like the cancellations don't happen till fairly high up in frequency.

Hence, I must have been incorrect about the fading bass frequencies. In fact, those should propagate without much trouble at all.

At 44.1 KHz sample rate, the comb filtering could be done with a delay line with nearly unity negative feedback and length of about 21 samples. [I always get that wrong! A half-wavelength of 10 samples with positive feedback! -- or did I do it wrong yet again?]

[PS. I wouldn't make the comb filtering too strong... the grass and weeds are likely to cause major scattering of the sound, hence weakening the groundbounce signal. A bowling alley would have stronger comb filtering, if you could remove the box effects of the room]

[argh... that's 21 samples of delay between the direct and groudbounce waves, period. No feedback needed. The mixing of the two will cause the appropriate comb filtering on its own... nicht wahr?]

- DM

posted 09 July 2001 19:27            

Carla's voice doesn't sound distant no matter how much I reduce the overall volume... Why? Because she isn't shouting! A SHOUT has entirely different formants than a soft speaking voice. Just guessing, I would estimate that they are higher up, and hence my impression that the bass suffers...

The delay line had almost no noticeable effect no matter how strong or weak I made the mix of the secondary sound. So I would bet you that 90+% of the effect is in the proper use of psychoacoustics!

posted 09 July 2001 21:53            

quote:

---

Originally posted by sm:
i guess you could also employ impulse response filters.
you would need two soundfiles: you shouting and a recording of a natural soundfield impulse respons. e.g. shoot a gun in the grand canyon and cut the attack. voila.

then your shouting would be the exciter of the filter which is the analysed fild-recording.
but i guess itīs not plug&play
cheers

---

This technique has been widely used for some time. You have to convolve the signal with the impulse response of the space you want to put it in.

Until recently this was not realtime a effect. (e.g. - I used to do it with SoundHack.)
With faster DSPs, the technique is feasible with little latency (which most reverb has anyway, so there is a way to conceal some processing time).

posted 10 July 2001 07:11            

Its also true that the timbre that a voice makes when shouting is entierly different to the spoken word. I've been trying to transform a voice to add effort by manipulating the spectrum with an algorythem, but the closest I've got so far is the sound of someone straining or being strangled.

Also a wide open space tends to give a very dry sound, dryer than most booths are capable of producing.
Because we are humans, its not just the shouting voice that tells us the distance but the whole enviroment (the ambiant noise that goes along with it, and the relative level).

Here at the Tape Gallery, we simple put a mic and the voiceover on the roof . You can never get the VO to shout with such agusto when he's standing in a comfortable booth in any case.
Cutting the lower frequencys and adding an extra ambiance track can help.

posted 10 July 2001 08:41            

A) The Vatican
B) St. Pauls Cathedral
C) The White House (Try to *miss* him please! )

Would you have time Carla?
I'll record it

posted 10 July 2001 09:28            

quote:

---

Originally posted by photonal:

Would you have time Carla?
I'll record it

---

Sure! But perhaps you'd better make certain it's loaded with blanks.