A few newb clarifications

[Blade Borge]

So I'm reading "Sound for Digital Video" right now, good stuff.

I'm still on the first few chapters and it talks about 48khz usually runs on 2 channels. I'm still confused about what they mean by channel.

I mean, my camera that runs on 48khz has 2 xlr channels.

They say that consumer dv cameras usually run on 32khz, which runs on 4 channels, but I don't see 4 xlr channels on a consumer camera.

Balanced microphones are balanced because they block out magnetic hum from other wires?

32khz means you hear less than 48khz right? Is it by a significant amount? I mean, I'm assuming if you recorded a high pitched singer, you must record onto 48khz.

All the questions for now, thanks so much in advance!!!

 

[Barry_Green]

Channels means individual audio tracks. DV cameras usually run 48KHz on two channels, but they have the option to run 32KHz on 4 channels. Very few DV cameras have the ability to actually record with 4 inputs (the HVX and the XL2 being the only two cameras I can think of that had the ability). Usually the extra 2 channels were used for dubbing audio after the fact; you'd record your live audio on tracks 1 and 2, and then you could perform an audio dub onto channels 3 & 4.

32KHz means there's less fidelity to the audio than 48KHz. For dialogue, you'll probably not notice much of a difference, but for a wide dynamic range environment like an orchestra, 48KHz is much better.

 

[Rippie]

A channel and a track should be considered 2 separate things.

A track being a recording term. IE: I have a 4 track recorder that lets me record 4
separate items separately.

A channel being more of a "live" term. IE: I have a mixer with 24 channels, letting me
"mix" 24 sources down to 2 channels.


Combination of both.. IE: I own an 8 channel mixer that I use to mix down 8 sources
to 2 channels (Left/right) I then record them to my 2 track recorder.



Hmmm trying to explain what a channel/track is without using the word "track or channel" in the description is kinda like explaining what the color blue is without using the word blue

 

[Alex H.]

Hmmm trying to explain what a channel/track is without using the word "track or channel" in the description is kinda like explaining what the color blue is without using the word blue

How 'bout this?

An audio channel can be thought of as a pipeline, be it an input or an output. It is merely a pathway for the sound signal to travel. Most mixers have more inputs than outputs, you you may have a mixer with three channels in and two channels out. (Mixers, as the name implies, are designed to consolidate input channels with a proper balance between the sources).

An audio track is a recording path. It is, in tape terms, the width of tape on which any single source, or mixed channel of sources, may be recorded.

Let's also look at sample rate, though. Because we're talking about digital audio, it's all 1s and 0s. And instead of a constant, magnetic stream (analog tape), we're recording data that is made up of those 1s and 0s (bits). And the device is not recording a constant, unbroken stream, but rather a series of data recordings that are taken so many times per second. You can plot them on a graph and they become kind of like a connect-the-dots picture, but in a waveform.

When the device plays the recording back, it looks at each point on the graph, and then tries to fill in the blanks. Each dot is a sample, and the more samples per second the less guessing is left up to playback. So at 48kHz (1 sample per second = 1Hz), the device is plotting a point on the graph 48,000 times per second.

To fit four tracks of recording in the same amount of space, the device has to record less data for each. So dropping down to only 32,000 samples per second makes room to add two more record tracks. This is further assisted by changing the bit rate.

Bit rate (aka word length) indicates how many 1s and 0s are contained in any given sample. So if each sample contains 16 characters (which make up a word), the audio is 16-bit. The fewer the bits in a word, the less fidelity and clarity in the recording. And in dumbing down the signal for the extra two tracks, DV cameras drop to 12-bit along with the 32kHz sampling.

As Barry said, there's less fidelity with 32kHz... because there's less actual data recorded and more that has to be filled in upon playback. Furthermore, the highest sound frequency that a digital device can record is equal to half the sample rate. So at 48kHz, the device can record frequencies as high as 24kHz (human hearing goes to 20kHz). But at 32kHz, the highest frequency attainable in the recording is only 16kHz. Again, as Barry said, this won't be noticeable for voice, but will be very much audible for something like an orchestra.

And to echo others, most often in 32kHz/12-bit mode, the camera only makes tracks 3 and 4 available for overdubs after the footage is shot. This was really created for soccer moms and dads to be able to edit in-camera by adding music and voice-over to recordings of little Jimmy's Pee-Wee-League games. When you have full post-production capability, there's no need to overdub in-camera.

Stick with 48kHz/16-bit.

 

[Ted Spencer]

Hmmm trying to explain what a channel/track is without using the word "track or channel" in the description is kinda like explaining what the color blue is without using the word blue

I'll give it a shot:

Track = a separate, discreet recording or memory space in an analog or digital audio capture device. Depending on the device it could be one (mono), two (stereo) or more/many (multitrack).

Channel = a physical (or software representation of an) audio signal path on a mixer, amp, preamp, effects unit, or other audio processing device. As above, it can be one or two channels (typical stand-alone preamp, reverb, etc.) or more (small mixer, large mixing console, etc.).

An HVX200 might be described as a 4-track recorder that only allows two channels of externally connected input sources (tracks 3 and 4 are restricted to the built-in mics only)

Software DAWs like Pro Tools are computer-based stereo or multitrack audio recorders, all versions of which have substantial multichannel mixing capability in software. Higher end versions can have much larger numbers of recordable tracks, vastly greater numbers of hardware input/output channels, and somewhat less 'vastly greater' numbers of software mixing channels.

Whew... : )

 

[Blade Borge]

wow you guys are really making things clear, much to digest. Thank you!

 

[Rippie]

you guys put way too much effort into that

 

[David G. Smith]

How 'bout this?

An audio channel can be thought of as a pipeline, be it an input or an output. It is merely a pathway for the sound signal to travel. Most mixers have more inputs than outputs, you you may have a mixer with three channels in and two channels out. (Mixers, as the name implies, are designed to consolidate input channels with a proper balance between the sources).

An audio track is a recording path. It is, in tape terms, the width of tape on which any single source, or mixed channel of sources, may be recorded.

Let's also look at sample rate, though. Because we're talking about digital audio, it's all 1s and 0s. And instead of a constant, magnetic stream (analog tape), we're recording data that is made up of those 1s and 0s (bits). And the device is not recording a constant, unbroken stream, but rather a series of data recordings that are taken so many times per second. You can plot them on a graph and they become kind of like a connect-the-dots picture, but in a waveform.

When the device plays the recording back, it looks at each point on the graph, and then tries to fill in the blanks. Each dot is a sample, and the more samples per second the less guessing is left up to playback. So at 48kHz (1 sample per second = 1Hz), the device is plotting a point on the graph 48,000 times per second.

To fit four tracks of recording in the same amount of space, the device has to record less data for each. So dropping down to only 32,000 samples per second makes room to add two more record tracks. This is further assisted by changing the bit rate.

Bit rate (aka word length) indicates how many 1s and 0s are contained in any given sample. So if each sample contains 16 characters (which make up a word), the audio is 16-bit. The fewer the bits in a word, the less fidelity and clarity in the recording. And in dumbing down the signal for the extra two tracks, DV cameras drop to 12-bit along with the 32kHz sampling.

As Barry said, there's less fidelity with 32kHz... because there's less actual data recorded and more that has to be filled in upon playback. Furthermore, the highest sound frequency that a digital device can record is equal to half the sample rate. So at 48kHz, the device can record frequencies as high as 24kHz (human hearing goes to 20kHz). But at 32kHz, the highest frequency attainable in the recording is only 16kHz. Again, as Barry said, this won't be noticeable for voice, but will be very much audible for something like an orchestra.

And to echo others, most often in 32kHz/12-bit mode, the camera only makes tracks 3 and 4 available for overdubs after the footage is shot. This was really created for soccer moms and dads to be able to edit in-camera by adding music and voice-over to recordings of little Jimmy's Pee-Wee-League games. When you have full post-production capability, there's no need to overdub in-camera.

Stick with 48kHz/16-bit.

Excellent answer. It took a while for me to wrap my melon around digital audio, but as for the sample rate/bit depth, remember that this is the equivalent to the resolution of an image. So for standard definition (NTSC) video , the resolution is 720X480 (lines or picture elements horizontally by lines or picture elements vertically) and high definition video is either 1280X720 (lines or picture elements horizonally by lines or picture elements vertically) or 1920X1080 (lines or picture elements horizonally by lines or picture elements vertically).

So for audio "Resolution", 32kHz/12bit (32,000, 12 bit samples a second) is a lower "Resolution" than 48kHz/16bit (48,000 16 bit samples a second). There are readily available inexpensive digital audio records that record audio at up to 96kHz/24 bit and some mid priced digital audio recorders that now support up to 192kHz/24 bits. 48kHz/16bit is really all you need now for film and video audio, while the higher resolution digital audio is designed more for music production.

 

[Noiz2]

I'm still on the first few chapters and it talks about 48khz usually runs on 2 channels. I'm still confused about what they mean by channel.

The other usage of "channels" in digital audio is the number of "tracks" as it were that are wrapped into a single file. So for instance if your four track recorded is recording four inputs and you end up with a single file, that would generally be referred to as a "four channel" sound file.

They say that consumer dv cameras usually run on 32khz, which runs on 4 channels, but I don't see 4 xlr channels on a consumer camera.

At would be a better word than "on" since it is the sample rate your talking about. 32K sounds fine for most things BUT it is a non standard rate. Things recorded at 32k on one device may not play on another device. 48K is the DV standard and is what you should use. Other rates come in if your recording sound on a separate recorder and there are some reasons to use them but recording to camera stick with 48K.

Balanced microphones are balanced because they block out magnetic hum from other wires?

Kinda sorta. The balancing allows the receiving device to cancel out many common hums. The balancing doesn't actually block anything. The shield (on both balanced and unbalanced) cables blocks hums and other boise.

32khz means you hear less than 48khz right? Is it by a significant amount? I mean, I'm assuming if you recorded a high pitched singer, you must record onto 48khz.

No, the human voice doesn't go that high. Human hearing in adult males drops off at about 12K (women usually have better high end hearing but the ear doesn't hear pretty much anything over 20K) so even 32K would record pretty much everything most people can hear. You want 48K because you want to be compatible and 32K can make your life a living hell. You will also see 44.1K popping up in your reading. It is the standard sample rate for CD's. But again you don't want to use it if you are recording directly to camera. Stick to 48K.

 

[Ted Spencer]

Human hearing in adult males drops off at about 12K (women usually have better high end hearing

This gender distinction is inaccurate as far as I know. Human hearing is generally described as 20Hz-20KHz in non gender-specific terms. Most younger people with undamaged hearing approximate that (I tested myself to over 18K when I was a teen - we won't talk about now...), and I know many male adults who I have personally verified at around 20K (some even higher).

...but the ear doesn't hear pretty much anything over 20K) so even 32K would record pretty much everything most people can hear.

I bet you didn't really mean to put it that way, because it's a bit misleading without further clarification - the 20K and 32K numbers here are not directly related.

So, while I know you know about this (Noiz2), nonetheless, for the benefit of others:

In digital audio sampling technology, the "K" suffix indicates the sample rate, or "Nyquist Frequency" which is double the effective audio frequency bandwidth of the system in question (and then some, due to decimation filtering of frequencies above it). In 1928, Harry Nyquist proved the theory that all digital audio relies on today: any frequency can be reproduced accurately with only two voltage samples recorded per cycle (Hz), plus roughly 10% extra for filtering (thus the 20-20K bandwidth is theoretically covered by the CD sample rate of 44.1KHz: twice 20KHz + 10%. Never mind the .1 part, or whatever reason it's there for...). When stored digitally as numbers, these voltages can then be read back at the same rate they were captured and reconstructed as an audio stream.

So a 32K sample rate allows an effective audio bandwdith up to a little above 14Khz (32K - 10% = 28.8K, /2 = 14.4K). This will sound pretty good, but will fall well short of what many if not most young, healthy humans can hear.

I should also add (again, for folks other than Noiz2, who undoubtedly already knows it), that all of these upper frequency ranges are strictly the domain of overtones - not actual instrument notes. Even the highest pitched instruments don't produce notes (aka "fundamental frequencies") above 5K or so. But the high overtones are what gives music it's sparkle and airiness, and certain non-pitched instruments like cymbals are rich in loud, high overtones.

Interestingly, and quite relevant to filmmaking, the human voice, particularly relatively low volume male speech at close range (3-6" away), contains a wider frequency range overall, including very present high and low overtones, than almost any musical instrument. Professional voice actors (like the late, great, Don LaFontaine, the all-time king of movie trailer VOs) often capitalize heavily on this fact...

 

[SMuir]

In digital audio sampling technology, the "K" suffix indicates the sample rate, or "Nyquist Frequency" which is double the effective audio frequency bandwidth of the system in question (and then some, due to decimation filtering of frequencies above it). In 1928, Harry Nyquist proved the theory that all digital audio relies on today: any frequency can be reproduced accurately with only two voltage samples recorded per cycle (Hz), plus roughly 10% extra for filtering (thus the 20-20K bandwidth is theoretically covered by the CD sample rate of 44.1KHz: twice 20KHz + 10%. Never mind the .1 part, or whatever reason it's there for...). When stored digitally as numbers, these voltages can then be read back at the same rate they were captured and reconstructed as an audio stream.

The use of "K" to denote the Nyquist Frequency is a new one to me. I had always understood that particular "k" (generally given as lower-case) to be a PREfix denoting the metric "kilo" (one thousand), followed by "Hz" which denotes "Hertz" (cycles or actions per second). 48 kHz, therefore, refers to the 48000 digital samples taken per second. Similarly, one kilometer is made up of a thousand meters. While I have heard frequencies described as "48 k" verbally, I had always assumed that the "Hz" was implied.

 

[Ted Spencer]

The use of "K" to denote the Nyquist Frequency is a new one to me. I had always understood that particular "k" (generally given as lower-case) to be a PREfix denoting the metric "kilo" (one thousand), followed by "Hz" which denotes "Hertz" (cycles or actions per second). 48 kHz, therefore, refers to the 48000 digital samples taken per second. Similarly, one kilometer is made up of a thousand meters. While I have heard frequencies described as "48 k" verbally, I had always assumed that the "Hz" was implied.

It is an implied KHz number, 48KHz, that being the typical sampling rate in the camcorder world, just as you said. But in common pro-audio parlance "48K" or "44 1" usually indicates that (sample rate), as opposed to other "K" or KHz numbers, generally used to indicate frequency response (like "20-20K") or a specific frequency itself (like "a 1K tone").

I was attempting to clarify the difference between frequency response and sampling rate in particular, but may have inadvertently caused some confusion in the process. There is no formal K versus k protocol that I know of regarding this, and I mentioned the Nyquist Frequency as background information to explain why a 32K sample rate only records up to a 14KHz frequency response bandwidth.

It seems you already understand what I was trying to convey anyway. Sorry for the excess information...

Too many K's on this bus, huh? : )

 

[Steve House]

...
Too many K's on this bus, huh? : )

Exit left to Funway....

 

[SMuir]

There is no formal K versus k protocol that I know of regarding this...

In the metric system, the "k" in "kilo" is always lower case.

 

[beatspinazz]

the trade off is between storage, er the amount of MB used vs Quality.

record more channels, then u will need more MBs.

sample rate directly affects amount of storage needed. so does bit depth.

trade offs: low sample rates results in less fidelity. low bit depth results in quantization distortion and of course reduced dynamic range capability. less channels equals to less flexibility.

jus match and mix to achieve the best combo.