Skip to main content

AD Converter for extreme slow motion stretching. please help.


I am continuing a project, call it art for lack of anywhere else to put it, that involves colliding hollow steel and copper boxes and recording the collisions. I film the visuals with old 16mm high speed film cameras, filming at up to 5,000 frames per second - super slow motion. The audio I have been recording using high sampling rates, taking vibration off the skin of the metal boxes using piezo sheet contact mikes (vibration sensors actually). I am running into problems slowing down the audio - I am slowing it down up to 100 times (6.5 fold compounded doublings of length) and the sample length is very gritty and stuttery even when I record at 192 kHz.
I tried slowing the audio down through analog methods using a UHER report (4-speed), but after such extreme stretching, the comparative difference between the sounds is barely discernable (it seems the differences are halved every time I stretch it twice!)
My end result is slow motion audio-visual of the collisions between the boxes, showing the ripples of impact, which I would very much like to hear.
I found this perfect 1000kHz mono AD converter for recording bats and insects, but it is very pricey - http://www.avisoft…"]Avisoft-UltraSoundGate 116Hm[/]="http://www.avisoft…"]Avisoft-UltraSoundGate 116Hm[/]

Does anyone know of any possible avenues of research? AD converters with higher than 192kHz sampling that are not ridiculously expensive? Or DIY options? Or getting someone to make me something up? Or other left-of-field possibilities?

In thanks,


TheJackAttack Sat, 01/14/2012 - 08:35

What you are doing with the video inre slowing it down by running high speed film at slower rates does not have a direct correlation in the digital audio realm. Recording with a reel to reel machine this could used in the same way IF the initial recording were fast enough and I doubt it is on your machine. That said, there are plugins that can adjust the time/pitch relationship in a logarithmic fashion. This doesn't have anything at all to do with the AD converter. Most DAW programs have this feature and there are various VST plugins that will do this with various functionality.

albertpaca Sat, 01/14/2012 - 16:02

Thanks IIRs - yes I have tried Pauls Extreme Sounds Stretcher. It makes amazing sounds, but they are quite "smeared", even with smaller sized windows. I would like to keep the starting punch of the collisions.
TheJackAttack, I am not that knowledgeable in audio, especially digital, but I thought that if I had a higher sample rate, smaller snippets, that I could stretch ou the time base longer without getting glitchy artifacts...??? THis is right, yes? I have used the time and pitch stretchers and I thought that if I stretched the sound 50 times, then I would be relying on a high sampling rate to give me something smooth(ish)???

I am really quite new to all this, but I have been using the time stretch feature in Soundtrack Pro, but it gives me glitchy artifacts even at 192kHz, I have used Pauls Extreme Sound Stretch, but it gives me smeared smoothed attack and lack of definition, and I have used reel-to-reel recording, recording at 15ips, playing back at 1.875, then going back through (stretching 64 times effectively), but the difference between the sounds is diminished in a direct relationship to the stretching amount (which makes sense - every frequency is halved every time the speed is halved) - everything sounds the same.
So, what I would really like, is to have a way of recording at a super high sampling rate (like the 1000kHz sampler linked to above) and then stretch the time base and still have a reasonably artifact-free sound, at the original frequency range.

I hope this makes sense - I was wondering if anyone has any possible places for me to look at or explore, hardware possibilities, or a different way of thinking about it all?
Sorry this is so convoluted....

TheJackAttack Sat, 01/14/2012 - 16:30

The flaw in your reasoning is this: a pcm wave form is not a continuous form. Recording at 44.1k for the purposes of stretching is no different than 176.4k. The computer extrapolates the sine wave in between and generates the FFT image displayed in your DAW with the sample markers in place. If you are introducing artifacts then it is a problem with the plugin or hardware box. Alternatively, artifacts appear if the computer is not streamlined for audio or have sufficient resources to begin with.

As to getting mushy slowing down an analog reel could it not!??!?!?!?!?! The attack is no longer a millisecond or fraction there of and is now spread out-especially if you are slowing 100% or more. If you want an articulated no BS sharp beginning to a sound you will need to overdub one in.

IIRs Sat, 01/14/2012 - 16:37

TheJackAttack, post: 382756 wrote: If you want an articulated no BS sharp beginning to a sound you will need to overdub one in.

Here's a possible approach: start by slowing your audio down as far as you can without it sounding "smeared" to you. Then slow down another copy enough to line up with the video. You can then chop up the first file, and manually line up the transients where you expect to hear them.

albertpaca Sat, 01/14/2012 - 16:38

Thanks for the reply TheJackAttack. Please do not get me wrong - I do not think I know what I am doing. Far from it....
I have shown this project four times now, twice at solo shows. The main emphasis was always the slow motion 16mm film, but now I really want to complement it with audio that matches. I have been buggering around and over-dubbing to my heart's content, and of all the things I have tried, the only one that kind of works is slowing down high sample rate recordings. I had always thought that the digital editor would edit in digital, thus the need for higher sampling rates for greater accuracy. I did not realise it extrapolated a sine wave before editing - this is obviously a flaw in my own knowledge....
I guess I just need to read up on the basics of digital recording.

IF YOU OR ANYONE ELSE HAS ANY IDEAS ON RECORDING AND EDITING SLOW MOTION SOUND, though, I would be greatly appreciative of any tips or pointers....

In thanks,

albertpaca Sun, 01/15/2012 - 00:34

thankyou all for your replies - I am looking at melodyne to play around and see if it can help at the editing stage, and i will do a lot of tests with the recording at various rates, getting a feel for the nature of the slowdown.
i realise now after much reading about the maths involved, that what i am asking is quite impossible, as, by nature, an extremely slowed sound will sound "unnatural", whether slowed by tape, sampling rate, or fourier maths.... but i just want to know what it is actually like in those small moments, really, before i make up something that seems more believable....

dvdhawk Sun, 01/15/2012 - 18:28

Stretching and sample rates aside... am I wrong in thinking that even under perfect conditions the resulting sound could be subsonic anyway?

For every halving of speed, you lower the sound an octave and you would get below 20Hz pretty quickly.

The film is shot at 5000 fps and played back at an undisclosed speed - but 5000 fps is 166.66 x faster than normal frame rates.

These arbitrary numbers to illustrate my point.

For example:
A 5kHz sound at 5000 fps would result in 1Hz per frame of playback speed. In other words played back at standard 30fps that 5kHz sound would be 30Hz.

Start with a lower fundamental, or slower playback speed, and you're well below what most speakers could plainly reproduce without just rattling and farting.

Am I missing something? (other than knowing the playback speed)

Anyway, maybe a pitch shifter would yield better audible results than stretching the time beyond practical limits.

albertpaca Mon, 01/16/2012 - 15:13

dvdhawk, no you are not missing anything. the playback speed is 25fps. i do not know enough about audio and was basically wondering if there was a way to stretch audio musically that was not either analog tape slowing, or sample rate stretching.... you are not missing anything - i do not know enough about this stuff - i was basically fishing for possibilities. all the pitch shifting i have tried has ended up sounding ridiculously phase-shifted. i was probably asking for the impossible. i was really hoping that someone who had worked on slow-motion sound may share his or her views on possible approaches to the problem. i do not want to be told it is impossible or that i am asking too much - what i want to be told is possible alternate ways of approaching the problem, or tips from someone who has actually tackled the problem....

just as further info, the last work i did i used old milliken cameras, shooting at 500fps, slowed to 25fps playback - 20 times slow down, which equates to about 4.4 octave shift. the newer work will use a NAC camera, capable of shooting 10,000fps, but i want to use it at 2,500fps - equating to about a 6.7 octave shift. yes, it is a lot. the sounds from my copper and steel boxes colliding are mostly within the 500Hz - 8kHz area, so at worst, a 6.7 octave shift, this equates to about 5Hz - 80Hz.
as posted earlier, i think that simply stretching it out like this at a super high sampling rate would at least give me an impression of what the sound really is like slowed that much - the nature of the transients, fine details that are missed at high speed - the reverberation of the steel , and, most importantly, an audio equivalent of the visual images that i am making. and then perhaps i can learn from that and craft some sounds that use that information and portray it in a more musically interesting way, editing and overdubbing as everyone is saying and as i have done in the past also....

still, i would of course love to hear from someone who has actually tried to play with slow-motion sound, and has some ideas of how to tackle, overcome, disregard, or give up on these problems....

TheJackAttack Mon, 01/16/2012 - 17:44

You are going to do a combination of things. If you have the reel to play with then start there. Slow it down until it has the rough sound you want regardless of whether it is slow enough. Now you go into the DAW. Dub in the intial transient sequence you desire. Now you are going to perform a combination of digital stretching and looping/overdubbing cut up segments of the original. You insert these as needed to lengthen the time span of the audio byte you need as well as covering up splices etc. You won't use fast crossfades in this case.

The masters of this technique are Foley engineers. I would look for any writings on this subject with "foley" in the search engine.

dvdhawk Mon, 01/16/2012 - 18:33

This is a fascinating challenge to be sure Sean. I hope you don't think I'm trying to dump on your prospects. I shoot video myself, so I'm intrigued.

I wish I had a better answer for you, but I believe when we see super-slo-mo video with audio, the sound matches the footage in length (duration) and has actually been shifted up in pitch to put it within an audible range when necessary. It may sound unnatural, because it is unnatural and likely to exceed the sub-sonic capability of your sound monitoring system. Or if you have seen a clip where the sound is excellent I suspect the "Hollywood magic" that Jack refers to, came from a Sound Designer, or Foley stage that may or may not have used any part of the actual sound as it's original source.

5Hz - 80Hz spans 4 octaves (two of which are below the accepted normal range of hearing for the vast majority of human beings). So, I'm merely suggesting you focus your attention on the parts within the range of human hearing.

Very few manufacturers make equipment [speakers / headphones / amplifiers] capable of anything below 20Hz. The low frequencies hog a disproportionate amount of the power and would really muddy up the amplifier's ability to accurately control the movement of the voice-coil and cone of the subwoofer. A lot of amps would have a filter to prevent anything sub-sonic from getting to the outputs, because it gets dangerously close to those DC signals which will destroy a speaker in a matter of seconds.

However, having said that, [[url=http://[/URL]="http://www.fullcomp…"]BagEnd ELF[/]="http://www.fullcomp…"]BagEnd ELF[/] builds speakers that claim to reproduce cleanly down to 8Hz. They're the only company I can think of that aims that low. Speaking strictly for myself, I may feel 8Hz, but I couldn't discern a note from it. You may need something servo-driven to get super low.

The BBC and the folks at iMax do some of the most spectacular super slow motion film I've ever seen. Have you attempted to contact anyone associated with either of those fine organizations? - It might be worth a shot.

albertpaca Mon, 01/16/2012 - 22:47

dvdhawk, thanks for the reply.
maybe i have been a bit misleading, but this is an art project (whether i like the term or not, unfortunately) and it has no clear outcome - it is led by experiment and exploration. my expectations are always changing depending on what is going on with the work and what tech i can get my hands on and what i can get my head around....
the last show i did was in a small sandstone gallery space with rear projection, and had a pair of tempest 15" drivers mounted in large vented boxes tuned to 15Hz (they of course went lower, but unloaded dangerously quite quickly....) the amplifier was a behringer ep4000, with the subsonic filter switched off. the final edit of the sound had a quick roll-off filter at 14Hz. so the sound was very bass heavy and held a lot of subsonics (which you could quite easily feel, as impact, though not hear), but i also edited a lot of higher frequency grabs from the original recording on top, so it wasn't all bass and had some sort of semblance to the original pitch, at least in places.
i felt it was a messy way of doing the sound.
my question on this forum comes from a desire to be "truer" to the original sound - in a way more like a science experiment than a foley film artist - because what interests me is the actual movement of the boxes as they collide, in slow motion - and ultimately what i really want to visualise, and sonify, is the vibration and eventual collapse of the boxes as they collide into each other.
i realise am asking too much (from myself, my tech, and the laws of physics), but i really think that if my expectations were any lower, i would have given up long ago. just for this one project, i film with and maintain 50 year old 16mm high speed film cameras (i have four millikens, an hitachi, an NAC and recently offloaded a hycam), work a digital high speed weinberger i bough cheap, record the sounds with condenser mikes and piezo sheet film placed on the boxes (through an editol 192kHz ADC and through an old reel to reel), work out the lighting (6 1000w par cans at close range wired off three separate phases to equalise the 50Hz flicker that is easily visible on the film), develop the b&w film (or send off the colour neg to be developed), get it telecined, edit the video, and here, where i am having real problems, edit the audio, the master it all to HD, set up a space with a projector, build my own subs (just about to make two 15" driver horn subs that do actually go low).... and of course i have made a machine which collides the small (30cm long) boxes in mid-air, using pneumatic pistons held with electromagnet to seal them for firing. crazy. you get the picture - i am not going to give up.
at the very least, what i want to explore is a way of recording the sound that lets me understand what is happening in those small fragments of time. the video lets me see a separate frame 2500 times per second - i want to be able to get a similar sort of understanding of the vibrations, through audio, as well.
i understand that the resultant audio may be useless, and may have to be re-engineered to sound "natural" or even just make sense. but lets call this research for now - i want to get as detailed as possible with the sound - to be able to slow it right down and hear what is going on millisecond by millisecond.
sorry - that is a bit of a rant - but you get the picture - this is not really a rational undertaking! i think what i need to do is find ways to explore what the sound can tell me, and that may have to be visualised (spectrometer) or through the use of servo-driven subs as you suggest (though the horns i am about to make should be good for even more tactile subsonics....)

i think several people now have mentioned that i contact people from the film industry, and this is a wise idea. also audio research may be another avenue.

why do we do what we do? who knows, but we have to do it i think....

Boswell Tue, 01/17/2012 - 02:28

The way I would do this is to use a two-stage process. Slow down the audio (both pitch and time) until the sound is what you want, ignoring any timing to do with the video. This may initially only be a factor of 5 or so. Then stretch the time of the audio without altering the pitch to fit the length of your shot.

You may need several tries with different initial division rates as the acoustic perception will be different when time-stretched. Any professional DAW has the time stretching without changing pitch facility. It's done by replication in the frequency domain.