By Andrew Frame
How many D/A’s do you do in a day?
No, I’m not referring to the cutie assistant District Attorney on Law & Order. I mean digital to analog conversions. Pre-computer radio folks remember when “DA” meant “distribution amplifier.” But in this day of multiple use acronyms, it also means the devices that slice your soft, fuzzy, warm analog audio into little digital nibbles and bits.
A word first: Before any of you engineering eggheads take me to task with what I say in this article, remember one thing – this isn’t a definitive essay on digital. Although the mechanical discussion may be a little dated, the theory isn’t.
Let’s pull up a chair in your production room, and reminisce. Pre-digital, the analog microphone was connected to the control board. The board was also hooked to an analog open reel tape deck or two, analog tape cart machine, analog turntable, analog telephone patch… do you see the key word here?
When your analog speech leaves your mouth, and makes the route to the analog eight-track open reel deck, it stayed analog all the way down the path. The trouble with analog is that is suffers from degradation and interference. (Ever wonder why balanced audio lines, XLR’s, have three wires? Interference suppression via phase cancellation.) As your analog speech signal goes through the wires and circuits, it degrades. Amplification helps keep the signal boosted to a useable level, but amplifiers generate heat (even tiny amounts in tiny circuits) and heat equals noise – noise that is superimposed on your analog signal. So, now it is louder, but slightly noisier. Then, there are mechanical issues. The signal has to jump the gap between the magnetic heads on the recorder; to the plastic ribbon covered with rust we call recording tape. It has to cross contact points on mechanical switches and pots. Distortion is introduced. When you play the signal back, the noise and distortion increase, as each record/playback cycle adds more. That’s why multi-generation mixdowns sound bad, and why the production guru you learned from taught you to get it to cart with a maximum of clean level, and a minimum of multi-track bouncing. (If you’ve ever dubbed a VCR tape, you see generational losses immediately because the video information is at a much higher frequency than audio, and high frequencies degrade much faster.)
After a time, you heard your first compact disc. No noise floor. None. No surface noise like a vinyl record or cassette – or even open reel. Zip. Dead quiet. And the music came up…
At some point, a short time after the Internet plugged in to your brain, those zeroes and ones crossed into the analog world of your production room. Maybe is was a computer based automation system, or an editor, but you got your hands and ears on that quiet. You merrily bopped along doing all kids of neat non-linear things and impressing the technically disadvantaged among your co-workers, until… we come into the late 1990’s and something called the MPEG file.
When anything is digitized, from a photoelectric signal off a scan platform, or an audio signal from the analog to digital converter in a sound card, you can save it as a file on some sort of storage device.
With the proliferation of computer based recording software, there is a matching proliferation of file formats. Among them, the WAV format. It produced a clean, clear reproduction of the original analog signal. Some audio editors, such as the SAW series, still use the WAV format as their standard. But, WAV files are big. Five and a half megabytes per track per minute. A :60 stereo spot is 11 megabytes—large enough that only the truly patient sent audio across the Internet over a “high-speed” 28.8k dial-up line. Taking it home meant using a spanning program to load it across seven floppy discs. Software and hardware companies worked on compression algorithms to cut the file size down, yet maintain some semblance of sonic quality.
Jumping ahead, we arrive at the MP3 file, today, almost the “standard” for Internet audio transmission. MPEG-3, or Layer 3 does some serious compression – or rather deletion. The process counts on idiosyncrasies in the way we hear, to allow our brains to bypass or even fill in the missing audio information. (Psychoacoustics is a fascinating field, and one that any Production Director might want to consider reading up a little on.)
Your production room is likely a mix of digital and analog systems. A spot comes in to you, a stereo MP3, encoded at a 128-kilohertz bit rate, considered “FM” quality. You save the e-mail attachment, open the file up, and play it back. Doesn’t sound too bad. A little soft in the top and bottom end. Actually, you can pick these things up because you haven’t trashed you ears yet, and you’ve trained yourself to hear these things. You open up a file on your hard disk storage, play the spot from one computer through the control board into the storage computer, and tomorrow you hear it on the air as you drive about town. But it still sounds… not right. The top and bottom ends are “soft.” The station’s air processor is pumping its little multi-band heart out, but…
Face it. You can’t hear what’s not there. Deleted bits are deleted, not compressed—deleted. Hmmm… Time to take out the magnifying glass and do a little detective work. You have a good engineer, so you know the production room and control room audio chains are as clean as they’re going to get.
Let’s follow the trail, at first an analog one back at the studio that produced the spot:
- The talent’s microphone feeds the…
- Control board, which ports out to the…
- Microphone audio processor, which ports back into the…
- Control board.
- Control board it goes back out to the…
- Digital workstation. An analog to digital (A/D) converter (conversion #1) slices the analog signal into digital chunks that are recorded to the hard drive of the workstation. If they have a self-contained unit, they can fully manipulate the audio from within the workstation. All of the mixing, audio and time compression and expansion, bouncing, mixdowns and the like are handled in the digital domain, within the workstation, without being converted to analog. The only thing out of the machine is monitor audio–which has to run through a digital to analog (D/A) converter first, either at the output of the workstation, or at the input of digital speakers.
- Digital audio plays off the hard drive, pipes through a D/A converter (#2) and feeds…
- The control board. Which feeds…
- Another A/D converter (#3) workstation that stores the audio as a WAV file.
- Out of your computer, through a D/A converter in the sound card (#5) into…
- The control board, and out to the input of the hard disk storage, another A/D converter (#6)!
- The D/A (#7) converter sends the audio to the control board, which, still analog, sends it to the air processor. Let’s assume you’re management did a few upgrades, and you have a digital processor.
- Analog audio goes in (A/D #8), digital audio gets processed. The processor sends the signal, whether composite or discrete to the analog STL (D/A #9), which beams it to the transmitter site. Since you have an analog exciter, the signal stays analog through the transmission chain, and out to the listener’s radio – which itself may have some A/D/A conversions before it comes out of the speakers.
Let’s assume they have a digital editor. After the analog microphone signal arrives into the:
The producer does his thing, and the client approves the mix.
A WAV-to-MP3 conversion (#4) and the file is e-mailed to the stations. Four conversions alone, and the spot hasn’t even left the studio. But, wait! There’s more! You’ve got mail!
At this point, the storage system may save it as an MP2, MP3, WAV, or any other of a half dozen formats. Six conversions… so far. The next day, in the control room, the jock fires the spot.
That’s a lot of digitizing. And a digitizer will only work with what you give it. So, if you have a degraded analog signal, you’re going to get a digital reproduction of it. More degradation, more digital reproduction of degradation.
Where is the choke point? I would propose the converters themselves. In the production room I use each day, I have a name brand top quality consumer CD player and a “professional” CD deck. The “professional” deck sounds audibly better—tighter bass, and more of it. Brighter treble, crisp midrange. Comparatively, the consumer deck audio sounds like a piece of gauze were laid over it. You can hear the difference. The converters in the consumer CD player and the “pro” deck are worlds apart. And, the converters in your $15 sound card and your $1500 sound card are worlds apart.
So, lets discuss them for a moment. Usually, you find converters advertised by “bit,” such as a 16-bit or 24-bit converter. The number of bits is a poor method of determining performance. A better measure is the accuracy of the actual bits themselves. Ideally, a 16-bit converter would convert all 16-bits of the sample. (Remember, your signal is chopped up into samples.) However, the error relies on the accuracy of the most significant bit (MSB) of the sample. Inaccuracy can whack the sample’s amplitude in half! So manufacturers thought that converters with high bit rates could overcome this shortcoming along with others through sheer numbers. (It also improves a parameter called quantization performance.) This same error can hit higher rate converters, too. They also have issues with linearity, gain error, slew-rate distortion, and zero-crossing distortion. All of these introduce harmonic distortion and delay, which in turn compromises the audio you ultimately hear. *
And this is just the first conversion. In our example from the studio to the listener, we counted nine—processed by converters from different manufacturers. Get it? If those converters were all built by the same place, with the same specs, you could reasonably expect a balance between what goes in and what goes out. But they’re likely not. On an off chance, they may use the same digital signal processor (DSP) chip in them, but the surrounding electronics will make them all operate differently. So, here we are seventeen hundred words into this article, and you’re asking, “what’s your point?” Does all this really matter to you, in the trenches, running your production M*A*S*H unit?
Yes. As the producer responsible for one-quarter hour of your stations ratings every hour (twelve minutes of spots?) the sound of the production on your station matters. MP3 is here. We will be using it until the better mousetrap is built. Since you are the “ears” of your station, keep your quality control high!
I got curious when I heard a rumour that the BBC had a standard 256k bitrate for MP3, so I dropped a line to them. The rumour was wrong, but the response from Lynda Carter, Delivering Quality Manager of the BBC clearly shows their demand for quality audio. She wrote back to say, “The only real use for MP3 audio at the BBC is to provide browse quality audio for searching audio data bases and auditioning material, where storage or network bandwidth efficiencies are a consideration. It is not used for programme production or storage, but it may be used in extreme situations for work in the field e.g. the delivery of news content from remote locations, where telecom bandwidth is limited. As a basic principle we aim to use linear digital audio systems wherever possible, but where we are constrained by technology or resources, we do compress using MPEG 1 layer 2, with a bit rate of 384kbits/s for Network Radio, and a bit rate of 256kbits/s in Local Radio for audio production and storage. We also broadcast on digital platforms (e.g. DAB & DSAT) using MPEG 1 layer 2 bit rates from 256kbits/s to 128kbits/sec.”
Layer 2 (MP2) is “cleaner” than MP3. If Auntie Beeb is using 384k rates on their network, and 256k rates for local, their quality control commitment is obvious. In the spring of this year, I got in touch with all the group Production Directors in southwest Florida and asked that we all adopt 256k as our “market standard” for MP3. We all did. The Production Barter Bank that I co-manage with Pamal Broadcasting’s group Production Director, Cliff Curtis in Gainesville, Florida and Gold Coast Broadcasting’s group Production Director, Bob Allen in Ventura, California has also standardized on 256k stereo MP3 for produced work, and 128k mono for dry voice tracks. We send and receive production and tracks from over thirty producers and talent all over the world–and don’t have quality problems.
The only problem we do have are the agencies who refuse to send us spots at 256k, and we end up airing their 128k feeds.
There is one common rebuttal to all this though, and it’s true. After all the processing that’s done, in the production room, and on the air, plus the roll-off filters to protect the stereo pilot tone at 19khz, and on and on and on, what difference does it make?
My ears tell me it does.
Keep your standards, and your bit rates, as high as possible to get the job done right.
* Another converter is the low-bit converter. Rather than converting data in parallel in multi-bit converters, low-bit converters do a serial data conversion, converting shorter data lengths at far higher rates. There are some really very clever mathematical actions going on in A/D/A conversions, and they are designed to give you the purest form of what was originally sampled way back in the recording studio. If you have the slightest bit of engineering geek in you, add this to your reading along with the psychoacoustics.
Author’s Note: Every engineer and production person has his or her own take on the MP3 issue, and that’s mine. I would like to thank Lynda Carter (not the actress) of the BBC for her help, as well as a very informative essay on converter technology written by Grant Erickson of the University of Minnesota Department of Electrical Engineering. I was unable to get hold of Grant to secure his permission for using parts of his paper verbatim, so I made my best effort to paraphrase his research accurately without actually using his text so I don’t get my pants sued off.