Audio for 3D Video

For $600, the Porta Mic Pro is pretty cool. The 5.1 is premixed and encoded to stereo with PLII (didn't find a street price for the PortaMic 5.1 (discrete outputs)). Curious how that sounds when decoded back to discrete channels for post work. A shoot out between the $200 Zoom H2N (5 discrete mics recorded to discrete channels (not compressed or encoded; raw PCM) and higher-priced solutions would be interesting. The H2-Pro 5.1 is $4900 (unmixed discrete output)- that's in the Schoeps range (would need to hear examples of each. Intuitively, I would guess the Schoeps would sound better (higher quality capsules) while the H2-Pro might provide a more interesting spacialization).


Yeah, both the Sony TD10 and Panasonic TM700 do a decent job with 5.1. Might be able to mix in parts of those channels for ambiance if shooting separate audio and recording with a shotgun for dialog.


Found a whole bunch of options for 3D/binaural and 5.1+ and posted results here:
http://cinema5d.com/viewtopic.php?f=32&t=33479


I've been looking for a decent noise reduction plugin (SoundForge 10, Audition CS5.5, Reaper) all have FFT-based filtering, but they all produce audible artifacts (ProTools LE has- nothing). The Voxengo Redunoise plugin works really well and tames the somewhat noisy $99 MXL R144 ribbon (needed for the figure 8 pickup pattern): http://www.voxengo.com/product/redunoise/ . It also appears to be FFT-based, but they are doing extra work to preserve the sound quality without introducing those high-frequency jingles. I picked up a Cloudlifter CL2 to try to help the low output ribbon and dynamic mic, but it so far hasn't helped to reduce noise levels (using an RME Fireface 800 to record). Mogami Gold Quad cables did (surprisingly) improve sound quality (pretty much what was said over on Gearslutz.com).


Starting to think the Audio Technica 4029 stereo shotgun (M+S, effectively provides 3 front channels) along with my Zoom H4N pointed backwards, might provide the highest quality, most flexible solution at that price point (can use just the 4029 if not needing 5.1 (matrixed L+R for a stereo mix, or M+S and mixed in post), if need just mono dialog, can use just the M channel, etc.).



Quote:
Some cameras have 5.1 support built in (such as the Sony TD10), but the audio quality is limited (good, but not pro). After reading about a nice Schoeps surround recording rig ($5500, http://www.schoeps.de/en/products/categories/doublems ) which uses only 3 mics, I have been experimenting with stereo mid+side (M+S). You can get a nice stereo field which you can control in post using this technique. Anything from mono to wide stereo and everything in between is possible, editable with any DAW in post (can also use plugins, but the technique is pretty simple).

Click to expand...

John- It often amazes me that here at AVS so many are quick to evaluate various options solely on the basis of technicals as oppose to the art of the recording. I've always considered the task first as in what sort of environment the recording must be done to select the right mic for the job. The most common stereo choice I have used has been based on the shooting location, sound source location, and the number of discrete audio channels available. Only my broadcast betacam SP dockable recorder has 4 discrete audio channels that I can address. But when was the last time I used that rig? Mid 90's? Since, I have been stuck using video camcorders restricted to 2 addressable channels only, or special internal 5.1 mixed audio track such as the TD10 and SR12.


In nearly all my shooting, I've needed to record the stereo sound source in a location to the front of the camera in the visual stage. This has always dictated an X-Y mic configuration. If I wanted to record the M+S using a three channel recorder, I would require a special recorder capable of 3 channels which is not usually convenient on the run. One of your links provided this:

Quote:
The H2n's onboard 90° X/Y stereo condenser mics are arranged with the right and left mics on the same axis. This design ensures that the mics are always equidistant from the sound source for perfect localization and no phase shifting. The result is brilliant stereo recording with natural depth and accurate imaging.

Click to expand...

While the M+S ( if a 3 channel recorder is used) can improve stereo separation, what sort of shooting environment would benefit from such an arrangement other than special gimmick effects? Instead- I would choose to just go from the stereo XY to the full surround pattern with discrete off visual channels being recorded and presented off visual stage to the rear. (5.1) The use of a front center channel can physically isolate the front center for improved clarity as it avoids frequency phase shifts that results in the mix of left and right when those channels are aimed toward the rear corners.


For the hot shoe, I have an external "5.1" mic that Sony offers and works for the TD10. This mic produces excellent 5.1 separation and accuracy. The advantage of using it over the internal 5.1 system is it is isolated from the camera body and eliminates any handling noise, and in windy environments you can add the fur to cut down on that. The quality? Hey! It's all consumer stuff anyway so for that purpose it does a pretty good job even though it is a 3 mic pickup design. No better or worse ( per my ears) than the internal and can eliminate a couple mentioned problems.


The recording capability of the consumer camcorders should be capable of fairly flat recording IF YOU CAN address the channels for recording to the digital file. This is where the problem is as I see it. You can't! You can get to 2 channels or use the internal for 5.1 or use a consumer hot shoe mic for 5.1; stereo, or mono plugin mics. The final choice for professional mics is the analog stereo connections. So, in my view, once you cross the line to professional 5.1 recording of sound on your projects, you not only need to haul along the pro mic but also use a separate recording device, then sync that up to the camcorder video using the camcorders internal audio channel for sync purposes only.


See some of the accessories I have acquired and tiny enough to haul along on my amateur shoots in the second photo:



Quote:
Originally Posted by bigbarney /forum/post/20827464


I think they're all pretty much FFT based now. The difference however is the quality of the FFT filters out there.... they are not all the same.


A comparison of Adobe Audition and Audacity:

http://www.fadden.com/fft/

Click to expand...

It looks like that person's point is that Audacity has a distortion bug when processing a 256 point FFT (using an Apple II data audio cassette(!)). To process audio with high quality, a 2048 or larger FFT is required. Even with a 2048 or larger FFT, it's much more complex than just the FFT size. The FFT code I used in one of my real-time audio applications also computed phase, which allowed for a much higher quality result than a simple FFT (which quantizes frequencies to FFT bins. When phase is also computed, it's possible to much more accurately compute the frequencies of the partials (where a partial is a sinusoid of a specific frequency. When all the sinusoids are added together, that makes up the full spectral component of sound for that snapshot in time). During the FFT snapshot (called a short-time FFT), all the frequencies are held constant. A simple FFT-based audio filter takes a snapshot, say 10ms or less of audio, converts it from the time domain to the frequency domain (the short-time FFT), modifies the FFT bin values, then performs an inverse FFT to get time domain samples back. Even if no filtering is performed, the audio will be changed due to the frequency quantization when doing the FFT + iFFT. In order to increase the resolution, overlapped windows are used as well as other more complex methods which are more expensive to compute (and apparently hard to implement so that artifacts are not introduced).


For those interested in learning more about FFT's (and the root concepts, the DCT & DFT), here is an excellent resource: http://www.dspdimension.com/tutorials/ (his example FFT-based pitch shifting code is excellent and can be modified for many useful DSP tasks). His top of the line audio processing tools use wavelets instead of FFTs. Since wavelets operate in the time domain, they are not susceptible to the short-time FFT artifacts. While it is possible to work around the SFFT issues, apparently using wavelets allows for a high-quality real-time solution. Conceptually, wavelets are just a series of cascaded low & high pass filters (time domain); coefficients are computed at each level and the filtered samples are passed to the next level for further filtering (building the wavelet coefficients). While wavelets can be used for audio filtering tasks, I haven't found an implementation for general EQ or noise reduction (perhaps the really high end packages used for forensic purposes use wavelets? http://www.cedar-audio.com/ )


One of the challenges of ribbon mics (and low cost ribbons!) as well as recording in an 'untreated' room is broadband noise: mic hiss, mic pre hiss, air conditioning, diffuse traffic, etc. The options are spend more money on quieter gear (and build a treated room) or find a way to fix it in post. Location shots are easier as the ambient sound can be much higher than audio-gear self noise. SoundForge has an OK noise reduction tool, but it produces a lot of artifacts (making it unusable in many cases). Audition has advanced spectral editing, including a healing brush, but it too introduces artifacts (though much less than SoundForge). Audition can remove 100% of broadband noise after taking a noise print. Even after tweaking all the complex settings, I was not able to eliminate artifacts. I recently found Voxengo's Redunoise VST plugin. It's even more complex than Audition in terms of controls, but within a few minutes I was able to remove all broadband noise with no perceptible artifacts- very cool. They market the product as 'high resolution' analog modeled, and that makes sense given that its an FFT-based solution (high resolution is the key to not generating artifacts).



Quote:
Originally Posted by Don Landis /forum/post/20827807


While the M+S ( if a 3 channel recorder is used) can improve stereo separation, what sort of shooting environment would benefit from such an arrangement other than special gimmick effects?

Click to expand...

Hey Don, I've worked with audio for a long, long time (mostly for video game and virtual reality (3D audio); interactive, real-time synthesized applications). I had seen M+S over the years, even have a Sony MS957 M+S stereo mic, but never really looked into it further. When I read how it works, it was one of those- 'sounds too good to be true' moments. Since the MS957 only outputs normal stereo (M+S converted to stereo: the 90/120 degree switch controls how the M+S is processed into stereo), I had no way to test it, so I picked up a $90 ribbon mic (to get the required figure 8 pickup pattern). It works, and works well. The 'stereo zoom' mics of video cameras use M+S to go from wide stereo (zoomed out) to near-mono (zoomed in).


I haven't worked in broadcast, but apparently M+S is very common as it is also mono compatible. The Audio Technica 4029 can output normal stereo, or the raw, unprocessed M+S (just 2 channels). From those two channels you get L+R+C in post. 3 for the price of 2, sounds too good to be true, but it really works. Being able to smoothly control the stereo field in post is amazing.


Check this out: http://www.recording-microphones.co....d%20Side2.html


You don't need plugins- M+S can be done with any DAW (though there are some pretty sophisticated & powerful M+S plugins).


For 5.1, you just need to add another unidirectional mic and do the same mixing with the figure 8 mic to get another 2-3 channels facing the rear (.1 (LFE) is just the low frequency component of all the mics). So for 5.1, you do need 3 recording channels. I have shot separate audio before using the Zoom H4N. It is more work but the audio quality is much better than on-camera (on camera can be greatly improved using something like a JuicedLink: http://www.juicedlink.com/ ).


While this project didn't use any camera audio, the video was from a consumer Canon HF11: http://www.myspace.com/internetexplorer (see the center video, "HD Video"): just another example of using low cost consumer gear in commercial work (I had much higher quality video I shot on a Canon 5DmarkII of the Long Beach Grand Prix; couldn't use it as we couldn't get permission in time (found out afterwards- just need to purchase a photo-pass ticket to get rights)).


When I started my company back in 1997, I put together a $40k ProTools- based sound studio. During the dot com bubble of 2001, I sold the ProTools gear and have found I can get excellent quality from much lower cost tools. Products from Neumann (now part of Sennheiser (for a while now!)), Gefell, MBHO, Sanken, Schoeps, Sound Devices ( http://www.sounddevices.com/ ) are interesting, but until the extra quality warrants the extremely non-linear increase in price, I try to find solutions which provide good enough quality before the cost curve looks like a hockey stick. Now that I can effectively remove broadband noise in post, with lots of EQ options, relatively low cost mics can produce good enough results.

Quote:
Originally Posted by Don Landis /forum/post/20827807


For the hot shoe, I have an external "5.1" mic that Sony offers and works for the TD10.

Click to expand...

Sounds interesting- will have to check that out.

Quote:
Originally Posted by Don Landis /forum/post/20827807


So, in my view, once you cross the line to professional 5.1 recording of sound on your projects, you not only need to haul along the pro mic but also use a separate recording device, then sync that up to the camcorder video using the camcorders internal audio channel for sync purposes only.

Click to expand...

Definitely more work, but worth it if producing something intended for commercial applications (or even just to learn the technique). I've got clapper board software running on the iPad which helps make sync easier (there are also plugins which use audio cross correlation to automatically sync audio tracks (FCP-X, IIRC, has this built in)).

Quote:
No. You need a minimum of 4 discreetly recorded channels for (proper) 5.1


The center channel and LFE can be engineered in post, but the left rear and right rear need to be properly recorded in order to reconstruct left and right directional sound.

Quote:
Mid + side should work in Vegas- see more info here (very simple technique, only tricky part may be figuring out how to do it with your toolchain (when I have a minute I'll look at Vegas)). More info here: http://www.wikirecording.org/Mid-Sid...hone_Technique . Schoeps and others have free VST plugins for Mid + Side.


Here's the Voxengo mid+side free plugin: http://www.voxengo.com/product/msed/


Again, for video work, the Audio Technica is looking very attractive as a stereo shotgun mic (with switchable Mid+Side support)- ~$600.


Barney- check out these bad boys (expensive- $5000+) 3 mic 5.1 rigs: http://www.schoeps.de/en/products/categories/doublems

Schoeps makes some of the best mics in the world (regardless of cost).

Free Double M/S plugin (3 mics in, 5.1 out): http://www.schoeps.de/en/products/dms_plugin/overview


Technical documents (detailed- worth reading!):
http://www.schoeps.de/documents/Witt...o_Surround.pdf
http://www.schoeps.de/documents/SCHO...d-brochure.pdf
http://www.cinesonics.pt/Locmat/Manu...aper_small.pdf


Also check out the Blumlien Pair technique- uses two figure 8's for what is said to be one of the best stereo soundfields possible ("holographic"). Limited uses, but when suitable appears to be amazing.

Quick comment on the figure 8 mic required for M+S: from what I have read, the best results come from a single element picking up audio (It's possible to build a figure 8 mic with two elements). In the low cost, readily available space, that's a ribbon mic. I need to read more, but I suppose it has something to do with effectively being a point source of pickup and perhaps is also related to crosstalk and phase effects (manufacturers of single element mics focus on that as a selling point). It's also possible to do M+S with two cardioids instead of a single figure 8. Instead of panning and setting the phase as with S in M+S, just mix in LS & RS with M to get the desired stereo effect. Not clear if phase and crosstalk issues are significant with that technique, or if it's possible easily wire them together to work as one mic (vs. a single figure 8 mic for S).


Here's a single body double M+S 5-channel unit from Sanken (3 mic capsules), the WMS 5: http://www.sanken-mic.com/upload/pdf/en/wms-5.pdf

The 5 also stands for $K (price).



Quote:
Originally Posted by Don Landis /forum/post/20830335


bigbarney- I think the game plan is they call the figure 8 mic a single mic therefore only 3 channels are required but this system needs to have some DSP before it can be deciphered into stereo.

Click to expand...

The demo is not accurate sound with respect to its directional and depth abilities. Sure...It's adding 'space' to the music... but then that's not really what 5.1 sound is all about.


If I walk around a true 5.1 mic banging a lid, then on replay I should be able to hear an ACCURATE reproduction of that lid-banging and the directional response. At any point in time, just by listening I should be able to pick off where exactly that lid is being banged in relation to me (while standing in the center of the replay room). That can not be done with any less than 4 mics.


The old Dolby Pro logic systems worked on 3 channels... left, right, and rear (center channel engineered) but the problem with that was that accurate DIRECTIONAL reproduction was not possible to the rear. The sound was of course heard from the rear but it wasn't clear WHERE to the rear that sound was originating.... Hence the newer Prologic II systems... with 4 channels was developed.


Now if the interest here lies merely with adding a bit of space and depth to a sound channel... hey... that can be done with just about any mono or stereo track with just about any DAW

Bigbarney-

Quote:
If I walk around a true 5.1 mic banging a lid, then on replay I should be able to hear an ACCURATE reproduction of that lid-banging and the directional response. At any point in time, just by listening I should be able to pick off where exactly that lid is being banged in relation to me (while standing in the center of the replay room). That can not be done with any less than 4 mics.

Click to expand...

Is this your theory or have you actually used and proven that the 3 mic surround systems being sold don't work accurately?

Quote:
The old Dolby Pro logic systems worked on 3 channels... left, right, and rear (center channel engineered) but the problem with that was that accurate DIRECTIONAL reproduction was not possible to the rear.

Click to expand...

I believe you are referring to Dolby Surround. I thought all ProLogic was 4 channels and produced 4 channels of low quality surround that had matrixed front center and center rear. Thus, your opinion that it could not accurately depict surround direction. ProLogicII improved the 4 channels by increasing the mix using steering amplification and amplification to full frequency range and isolation to full 5.1 output. IIX added the 6th and 7th side channels. IIz adds the 8th and 9th front top channels. Earlier this year I upgraded my home theater to IIz.

Quote:
Now if the interest here lies merely with adding a bit of space and depth to a sound channel... hey... that can be done with just about any mono or stereo track with just about any DAW

Click to expand...

Yes, agreed! Just like some people feel that better 3D is that which pushes the extremes of z depth in every shot. I feel this just adds to the gimmick ( not the story) and causes more headaches and fatigue. With stereo sound, (in another thread) it was suggested that spacing omni mics wide apart produced better stereo but I disagreed and felt all this added to an otherwise single location directional stereo mic system was reverb time delays and phase shift distortion. But, here again if that is what others think is better quality stereo, then who am I to disagree.



Quote:
Originally Posted by bigbarney /forum/post/20831297


That can not be done with any less than 4 mics.

Click to expand...

A thought experiment:


Take two stereo mics with a 90 degree pickup pattern and place them in opposite directions. That's four channels of audio, picking up a perfect pattern for real, spatialized surround, right?


Take two Sony MS957 stereo mics & place facing in opposite directions with the angle set to 90 degrees. This setup will capture 4 channels, equally spaced around a circle in 90 degree increments. The MS957 mic uses the M+S design internally and electrically mixes the signals from the figure 8 and cardioid capsules to output 90 or 120 degree stereo.


If we were to take apart the MS957's, we could build a new setup such that both mics shared the figure 8 capsule (the 'S' or 'side' of Mid-Side (M+S)). We still get 4 channels of discrete audio as before, only now we're using the electrical signal of only one figure 8 capsule when mixing with the separate front/rear cardioid capsules. Four discrete channels after decoding from only 3 mics: the double Mid Side technique. Add in the front M channel and compute .1 from all channel low frequency and you have 5.1 (6 channels) from 3 mics.


Mid Side was one of the first stereo techniques used in movies (1933, Blumlein). It's as real and discrete as the MS957 microphone which uses M+S.


The physics of the pickup pattern:


The math is simple:


M = mid (typically cardioid, but can be any pickup pattern mic)

S = side (must be a figure 8 pickup pattern mic)


Converting to normal stereo:


L = M+S

R = M-S


A 'secret weapon' in audio post production is converting a normal stereo signal into M+S, editing the stereo field, editing the center channel directly, etc., then converting M+S back into L+R:


Convert to M+S:


M = (L+R)/2

S = (L-R)/2 or S = L-M or S = M-R


Edit M and/or S directly, then convert back to L+R:


L = M+S

R = M-S


More info:
http://www.uaudio.com/blog/mid-side-mic-recording/
http://emusician.com/mag/emusic_front_center/



Quote:
Originally Posted by Don Landis /forum/post/20831886


Bigbarney-



Is this your theory or have you actually used and proven that the 3 mic surround systems being sold don't work accurately?

Click to expand...

They don't work accurately to the rear. It's the rear directions that get a little 'muddy' with a 3 mic system.

Quote:
I thought all ProLogic was 4 channels and produced 4 channels of low quality surround that had matrixed front center and center rear.

Click to expand...

No. Prologic if I remember correctly was 3 channels.... PrologicII is 4 channels (with engineered center channel)

Quote:
Just like some people feel that better 3D is that which pushes the extremes of z depth in every shot. I feel this just adds to the gimmick ( not the story) and causes more headaches and fatigue.

Click to expand...

It depends on what you're after. If you're merely interested in creating SOME kind of depth to WOW the crowd then you can do it just about any old way... including 2D to 3D conversion. But if you're interested in accurately reproducing the z depth.... ask yourself if this can be done with any serious degree of accuracy with something like a 2D to 3D conversion (the answer is 'no' of course).