Shrinking audio files

A few weeks ago, I recounted how listener Gary Freeland suggested a program on file sizes. "With the proliferation of digital audio and digital video files," Gary wrote, "perhaps you could address the subject of compressing these files so they can be sent via email or IM messages." A few weeks ago, I talked about some of the applications you might consider for digital pictures. This week, it's sound files.

Nearly any audio editor (from the free Audacity to the far-from-free Sony Sound Forge can shrink an audio file. In working with sound, you have 3 primary variables that control the size of the resulting file in addition to methods used by the file format to store the data. These are the sampling rate, the bit depth, and the number of channels.

Sampling rate: The sampling rate is expressed as a frequency in Hertz. It describes the number of samples per second used to store a sound. The rates vary from 2 KHz to 192KHz. Higher rates mean better fidelity, but they also mean the file size will be larger.

Bit depth: This is the number of bits used to represent a single sample. For example, 8- and 16-bit are common sample sizes, but the bit depth can range (depending on the application) from 8 to 64 bits. A high bit depth will display less noise and will provide better dynamic range, but it will create a larger file.

Number of channels: The number of channels is the one that's easiest for people to understand. Applications may allow the user to record on dozens of tracks, but the resulting output file will typically be monaural (1 channel) or stereo (2 channels). Whatever the file size is for a mono recording, converting the file to stereo effectively doubles the file size.

If your goal is the best possible fidelity for music, the file size will be large. If your goal is the smallest possible file, you'll have to give up some quality. The trick is to find settings that provide acceptable quality at the smallest possible size. The first decision is an easy one: The output file should be mono unless the stereo effect is important. The TechByter Worldwide podcast is recorded and rendered in mono.

I made 2 brief recordings that illustrate the differences made by the sampling rate and the bit depth. The first was recorded with a 8KHz sampling rate and an 8-bit depth, the second was recorded with a 192KHz sampling rate and 32-bit depth.

• Recording sample 1 (AudioTest-008khz-08bit: 100 KB original recording)
• Recording sample 2 (AudioTest-192khz-16bit: 8,000 KB original recording)

The sound card I use doesn't support 2KHz at the low end or 64-bit depth on the high end. I rendered both files in MP3 format using a 128Kbps bit rate and a 48KHz sampling frequency. You may be wondering what I mean by that, so let's look at the export process.

Exporting audio

Exporting audio generally requires you to specify a bit rate and a sampling rate. The first is expressed in kilobits per second (Kbps) and the latter in KiloHertz (KHz).

The bit rate is the speed of the data stream—the average number of bits used in each second of audio. The most common bit rate for MP3 audio is 128Kbps, which is generally considered CD quality, while 64Kbps is considered FM radio quality. Some people say they can hear the difference between 128Kbps and 320Kbps. I can't. Higher bit rates mean better audio, but streaming audio (even on a high-speed connection) won't work well at high bit rates.

The sampling rate for output is analogous to the bit rate used in recording. Higher numbers mean more samples per second and better audio quality, but larger file size; lower numbers produce smaller files, but reduce the quality of the audio.

Putting it all together

Because you can never get higher quality output than exists in the recorded file, it's wise to record at the highest reasonable quality. That doesn't mean recording everything at 320Kbps with a 64-bit depth; it does mean that if you believe you'll need "CD quality" at some point, the recording parameters should be at least 128Kbps at 16-bit depth. For TechByter Worldwide, I use 44.1Kbps at 16 bits and reduce the rendered file to 32Kbps at 22.05KHz. Because of tests conducted for this program, I may further reduce that rate to create smaller files that will stream without choking on dial-up connections.

Each of the sample files came from the same source file, which was recorded with a 192KHz sampling rate and a 32-bit depth. I then rendered the file several times. The file size is shown in parentheses.

• AudioOutput032-22-podcast.mp3 (245,133)
  This file uses the parameters that are standard for the TechByter podcast, 32Kbps at 22.05KHz. The file size is about 250KB per minute, which is small enough to be handled by any high-speed connection. It's acceptable for faster dial-up connections, too.
• AudioOutput008-11.mp3 (61,388)
  8Kbps with an 11.025KHz sampling rate is rough by anyone's standards. I would use a file such as this only if the size of the file was the primary consideration.
• AudioOutput020-11.mp3 (153,465)
  This file is about 37% smaller than the standard podcast file. I output it at 20Kbps at an 11.025KHz sampling rate. Although the quality is somewhat rougher than what I use now, I believe that the quality is adequate.
• AudioOutput128-48.mp3 (979,968)
  128Kbps at 48KHz is overkill for the spoken word. Yes, it sounds a lot better than any of the other files, but it's also nearly 4 times the size of the standard podcast file and more than 6 times the size of the file that would probably be adequate.
• AudioOutput128-48-Normalized.mp3 (979,968)
  I included this file simply to show the effect of normalizing the file. This process finds the highest peak in a file and then amplifies the entire file to make that peak's volume 100 percent. There are several settings for the normalization process, but the most important one is deciding whether to use the peak level or the root mean square (RMS) of a file. RMS is a measurement of the intensity of the sound over time and the RMS power of a sound corresponds to the loudness perceived by a listener when measured over small intervals. RMS sounds better to me, but I didn't like the default setting in the application I use. It suggested normalizing to -10dB, but the resulting file sounded distorted. I tried -12dB and eventually settled on -15dB.*
• AudioOutputVBR001.mp3 (547,680)
  Variable bit rate output can conserve space for music files, but the result is needless large files if the audio consists of the spoken word. VBR has a quality setting. This file is the lowest quality VBR.
• AudioOutputVBR050.mp3 (712,992)
  Here is the middle quality VBR.
• AudioOutputVBR100.mp3 (994,944)
  And here is the highest quality VBR. Did you hear much difference? Neither did I.

  * What's a dB? It's a vexing unit of measurement that represents a logarithmic ratio between two numbers. I thank Sony Electronics for this example: "When comparing the numbers 14 and 7, you could say 14 is two times greater than 7; or you could say 14 is 6 dB greater than 7. Where did we pull that 6 dB from? Engineers use the equation dB = 20 x log (V1/V2) when comparing two instantaneous values. Decibels are commonly used when dealing with sound because the ear perceives loudness in a logarithmic scale."

There's a lot more to this topic

It seems like a long time ago now, but I mentioned formats. As with image file formats, some audio formats are lossy and some are not. MP3 is a lossy format. It throws away some data in the process of saving the file, so if you're editing an audio file, your work file should be in the program's native format (Sony Sound Forge uses frg files, the generic Windows lossless audio format is wav and Apple's standard file is aiff.)

When it's time to output the file, your choices include mp3 and mp4 (which Apple calls aac), ogg vorbis, Real Media's rm format, Windows Media Player wma, and others. Because nearly every computer has an application already installed to play mp3 files and because podcasting seems to have standardized on that format, mp3 is what I use. You will find older audio on this site in Real's rm format

Black folks aren't black and White folks aren't white

This isn't about technology; mainly it's just some rambling thoughts and a bit of travel advice based on recent trips to New York City. When I mentioned to a guy from Cincinnati who was attending the same Direct Marketing Association conference I was that I was staying at a bed and breakfast in Harlem, his response was, "That wouldn't have been my first choice." When I mentioned that the place I was staying had cats, he wondered if they were there to give people something to kick at night. We didn't talk a lot after that.

Some people don't care much for cats, but I've lived with cats most of my life. I like them and I enjoy having one (or more) lean on my feet, arms, shoulders, or head at night. Staying at the Harlem Flop House means that I might have one or more cats in bed with me (if I leave the door open at night so they can come and go). And although Harlem is no longer exclusively Black, it's also not exactly White. Folks of my complexion are a distinct minority on the streets of Harlem and on the trains once they're north of 96th Street.

Being White and from the Midwest, I've been in the majority most of my life. Staying in Harlem gives me an opportunity to see things from a different perspective. Going "home" in the evening, I'm usually the only honkey on the D train after 59th Street or, if I'm on the 2 or 3 train, the only honkey on the train after 72nd Street. I can walk from 125th Street to 135th Street on Malcom X Blvd (the extension of Lennox Ave. north of Central Park) and never see a White face. Some people glance at me with a "what are you doing here?" expression. A few stare. Most ignore me.

Whenever I'm in Harlem, I sometimes think back to a Nikon program that I attended in the early 1970s. It was a class on color photography and one of the instructors was from deep in the South—Alabama, I believe. I thought I wouldn't like him because he was from Alabama and most certainly (I thought) must have been a racist. In talking about color, he discussed the problems involved in photographing a light-colored White person and a dark-colored Black person standing side by side. "They're the same color," he said in that Alabama accent, "and you can prove that with a color densitometer. The difference is density. "

His point was that the pigment is the same and, because color depends on pigment, the color was also the same. It was a moment of epiphany for me. The man from Alabama may have been a racist, but he certainly didn't sound like it. At the time, I didn't sound like a racist, but I probably was. But, as usual, I'm digressing even in my digression from technology.

Why Harlem?

Harlem is far quieter than Midtown. I used to stay in the Times Square area because it was convenient, but then I discovered the land north of Central Park North. It's residential. It's quiet. It's easy to get to. And you get far more space. In the past 10 years or so, I've stayed at 3 Harlem locations. Why 3? Each is small, with just 2 or 3 rooms. As people like me stay and then spread the word, getting a room is harder and harder.

For example, in January I wanted to book a room at the Harlem Flop House, but found that it was unavailable. I booked for February, but still needed a January location. That's when I found Harlem 144 and stayed there for several days. In late February, I asked about booking Bed and Breakfast Mont Morris in May and found that I would need to change my travel plans by a few days to fit the schedule.

So, consider this TechByter Worldwide's Travel Service, with a disclaimer or two: I'm not a travel writer, although I did work for the Ohio Bureau of Travel and Tourism (part of the Department of Economic and Community Development) for a few years. TechByter Worldwide is a technology program, not a travel program. That being said, I'd like to share with you some information about 3 Harlem locations you should consider for your next trip to New York City.

Bed and Breakfast Mont Morris

Two suites are available, one with a kitchenette. You'll stay a few blocks from the 116th Street 2/3 station, so getting to Midtown will take only a few minutes. Breakfast is at the bakery (delicious!) on the corner. Visit the website.

The Harlem Flop House

Rooms here are large enough, but you'll share a bathroom with guests in the other room on the floor. You'll also share the staff cats, who are most attentive to your needs. The closest subway station is on the A/B/C/D lines and the travel time to Midtown on the D is just a few minutes. No breakfast is included. Visit the website.

Harlem 144

You'll share a bathroom and a kitchen with the other guests on the same floor and you'll find breakfast in the kitchen area in the morning. The fresh fruit cup is a pleasant surprise. The location is convenient to the 2/3 line at 116th Street. No website yet, but you can get more information.

What is a racist, anyway?

I've mentioned Katie Melua on a previous program. She's Russian by birth and Irish because her family moved to Ireland. For a young lady in her 20s, she has a remarkable voice. In "Spider's Web" on her 2005 CD called "Piece by Piece", she sang these words that she wrote: "If a Black man is racist, is it okay when it's the White man's racism that made him that way? Because the bully's the victim they say. By some sense they're all the same. Because the line between wrong and right is the width of a thread from a spider's web. The piano keys are black and white, but they sound like a million colours in your mind." (She sings it better than I write it.)

Wow.

Nerdly News

Spam Control

I used GoodbyeSpam and have written about it. Recently I received a question about the application and I had to recommend using something other than GoodbyeSpam.

How do you buy GoodbyeSpam? It asks what type of mail box and has a drop-down menu listing some, but not Compuserve. I use CompuServe 7.0 and get 300-400 spams a day.

When I try to register for the 30-day trial it rejects any selection I make from the "mailbox type" drop down window. It will not accept the application. I have e-mailed their support 4 times in the last 10 days and never had a answer.

It was CompuServe support that suggested using Goodbye Spam.

GoodbyeSpam is a server-based application. You set it up to check your CS e-mail account and then you set up your e-mail account to check GoodbyeSpam. Although I like GoodbyeSpam, the service began to have reliability problems and I eventually stopped using it. Support, as you have seen, tends to be slow.

I'm now using a similar service called SpamArrest (www.spamarrest.com) and have been pleased by its reliability. The SpamArrest documentation seems to be a bit better than what GoodbyeSpam provided, too.

You can continue to send mail via CS or you can use SpamArrest's server to send mail. I have my portable computer set up to do that because sending e-mail via WiFi connections can be tricky.

Microsoft rivals continue to attack through the European front

Microsoft has 4 weeks to respond to a threat from the European Commission to fine the company millions of euros because the company demands unreasonable royalties from rivals who request information from Microsoft about how to make their software function with Windows. The commission has already fined Microsoft $655.3 million in 2004 and $370.4 million last year for failure to comply. Microsoft has appealed both fines.

The commission charges that Microsoft has not cooperated, but Microsoft says it has been trying to cooperate. Microsoft’s general counsel, Brad Smith, said the company had been fair and reasonable in setting prices.

The commission says that the new fines could be as much as $3.9 million per day and some of the fines could be retroactive to December 2005. The European Committee for Interoperable Systems (which includes IBM, Oracle and Sun Microsystems) says it is pleased by the commission’s position.

This is the only ad you'll ever see on this site. It's for my website host, BlueHost in Orem, Utah. Over the past several years, they have proven to be honest, reliable, and progressive. If you need to host a website, please click the banner below to see what BlueHost has to offer.

TechByter Worldwide receives a small advertising payment for each new client that signs up with BlueHost but I would make the same recommendation even if the affiliate program didn't exist. (If you don't see a banner ad above and you would like to know more, this link takes you to BlueHost.)