Search Gear
 

Bit by Bit

May 1, 2000
share

As more and more sound cards support digital audio transfers, it's important to understand what various audio interfaces may (or may not) do to your carefully crafted sounds. The primary objective of a digital transfer is to create a bit-accurate clone of a digital recording (as when you copy a DAT tape to your computer or vice versa). However, a number of devices with digital interfaces can only perform this task under certain circumstances, and some can't do it at all.

If you find this situation a bit confusing, you're not alone. I was surprised by the amount of misinformation I came across while researching this column. But don't fret-it's not that complicated if you're familiar with the different digital audio formats and their characteristics.

Last month, we examined the four most common types of digital audio formats and described the features and capabilities that each one provides (see "Square One: Spare Interchange" in the April 2000 issue of EM). This month we'll look at some real-life situations and describe how you can avoid common problems when making digital transfers.

A FOUR-PACK OF FORMATSLet's briefly recap the primary formats for exchanging digital audio data. Digital interfaces currently come in two types: stereo and multichannel. Each has two major formats. The 2-channel formats, AES/EBU and S/PDIF, use different cables and connectors, although both use the same method to encode data. In addition to using traditional coaxial audio cables, S/PDIF can also transmit over fiber-optic cable. The two multichannel formats, ADAT Optical (Lightpipe) and TDIF, can each transfer eight channels at once.

At the dawn of the digital age, the bit-accurate piracy of copyrighted works loomed large, leading to the requirement of a form of copy protection in digital interfaces that target consumers. The Serial Copy Management System (SCMS, fittingly pronounced like scums) defined two hidden flags in the S/PDIF data stream that prevented consumer machines from making digital copies of digital copies. AES/EBU, the professional interface, carried no such burden. Today, consumer DAT decks are essentially extinct, and SCMS is all but forgotten, so most S/PDIF devices ignore it completely.

Although issues surrounding SCMS have faded into the background, other problems remain when different types of digital audio devices attempt to communicate. Therefore, the first rule of digital transfer is knowing which format your gear supports. Check your connectors: each format uses a different type of connector, with some exceptions-ADAT Optical and S/PDIF optical use identical fiber-optic connectors-so you need to know which optical format you're dealing with.

The proliferation of format converters makes it easy to connect different formats. For example, Tascam makes a device called the IF-TAD that converts from TDIF to ADAT Optical, and you can also use MOTU's 2408 or Soundscape's SS8IO-2 as stand-alone TDIF-to-Lightpipe converters. For stereo transfers, Midiman's CO3 transmits your signal between any combination of AES/EBU, coaxial S/PDIF, and optical S/PDIF devices.

GUMMING UP THE WORKSIt all sounds so orderly-if A plugs into B, you're set to clone, right? That's what I expected when I tried to use my Opcode Sonicport (S/PDIF-to-USB connection device) to transfer some tracks from my notebook to my desktop computer. I connected the coaxial S/PDIF output of the Sonicport to the coaxial S/PDIF input of my desktop sound card, punched Play on one and Record on the other, and watched helplessly as everything arrived 6 dB quieter than it started out.

We've all chuckled at the phrase "It's not a bug, it's a feature!"-but this time it's true. The decibel attenuation of my tracks makes sense when you consider that from a consumer perspective, USB audio is intended more for connecting speakers than for recording. Therefore, the USB audio fader goes into the Windows Mixer applet with a default setting of 6 dB below maximum to avoid blowing out USB speakers. All I had to do was raise that fader to its maximum setting to make my Sonicport send in the clones.

Mixers, outboard A/D/A converters, effects, and other devices with digital interfaces are handy when you want to have control over input and output gain. If your goal is to produce bit-accurate copies, you need to override the gain controls or set them to the proper level. Often this simply means setting a software fader to unity (0 dB). Check your device's manual or consult the manufacturer for the appropriate setting.

Guillemot has simplified this process for its Maxi Studio ISIS card. The mixer applet features a simple check box labeled Backup. The Backup option disables the S/PDIF output faders to ensure that your mix doesn't suffer any unintended gain adjustments. Here's proof that intelligent design doesn't need to cost an arm and a leg.

WHO'S IN CHARGE HERE?It's a fact of digital life that no two clocks agree completely. This can cause glitches when you're transferring digital audio. If the receiving device's clock runs faster than that of the sending device, it will eventually expect to receive a sample that you haven't sent yet. If it runs slower, it will need to drop a sample to keep up. These discontinuities in the data stream appear, respectively, as flat spots or jumps in the waveform and often cause audible clicks (see Fig. 1).

You can avoid this problem in one of two ways. The most common solution is to designate one device as the master clock, so that the slave device receives timing information from the master. This leaves no room for disagreement between the two devices and ensures accurate transfer of the waveform. Most sound cards refer to this master-slave relationship in terms of internal-external clock or sync. Set the master device to use its internal clock and the slave device to receive timing information from an external source. If your interface doesn't offer an external option, it probably can't be a slave to another device and therefore won't do bit-accurate transfers.

The high-end solution to synchronization is to slave all devices to a superaccurate external clock. If your gear has word-clock or Superclock inputs, that's what they're for. Of course, this requires that everything else in your studio be capable of operating in slave mode.

E-mu has an interesting approach to this master-slave relationship. Its Audio Production Studio (APS) operates only on an internal clock, which is fixed at 48 kHz. Any audio coming through the APS device's S/PDIF port undergoes a sample-rate conversion to 48 kHz and another sample-rate conversion if you're saving it as anything other than a 48 kHz file. (To avoid piling up conversions, E-mu advises users to work exclusively at 48 kHz and then convert the final mix to 44.1 kHz before burning a project to CD.) Because this conversion affects the sound less than an analog conversion would, the APS manages to make high-quality digital copies but not bit-for-bit clones.

A BIT OF CAUTIONNaturally, it's important to make sure that your source and destination devices are set to the same word length-and that they can handle it. Try sending a 24-bit signal to a device expecting only 16 bits, and you could get an ugly surprise. Also, never assume that all digital interfaces support the same resolutions. All four formats can pass 24 bits, but not all manufacturers implement the formats that way, even in their newer, high-profile products.

The ISIS, for example, only supports 16-bit transfers through its S/PDIF interface. For many purposes, this is not a limitation, but if you're hoping to transfer higher-resolution audio, you need to buy a card that supports the format's full bandwidth.

Since current ADATs are 20-bit devices, some ADAT Optical interfaces don't support 24-bit transfers. One particularly high-profile device that passes only 20 bits is Digidesign's ADAT Bridge I/O for TDM systems.

If you have an older S/PDIF-equipped DAT deck with SCMS, you'll want a digital interface that knows how to deal with the extra bits. TerraTec's audio cards, such as the EWS88-MT, give you the option of keeping or ignoring the copy protection bits. Of course, this isn't an issue with an AES/EBU interface or with either multichannel format.

CHAMPING AT THE BITTo paraphrase Murphy's Law, things can and will go wrong. For example, if your DAT or MDM encounters errors on the tape, it will try to correct them. Other than maintaining your decks and taking proper care of your tapes, you can't do much about this.

Once the data has left the source device, bad cables or interference can corrupt it in transit. Installing quality cables, maintaining them properly, and using the shortest cable runs possible limit opportunities for this sort of error (see the sidebar "Cloning 101: Ensuring Bit-Accurate Transfers" for some additional tips). You can easily test the reliability of your digital transfers by comparing files in an editing program. Transfer a file from your computer to DAT (or to another computer with a matching digital interface), then transfer it back. Some audio editors offer a function that compares two files and reports any differences.

You can do this manually by inverting one file and pasting it over the other (see Fig. 2). Phase cancellation causes two identical files to leave a straight line on 0 dB. Anything left over is an error introduced in the transfer process. If you see a miniature version of the original file or its inverse, you have a level mismatch. Periodic bumps in an otherwise straight line could indicate a clock problem, such as an improperly set master-slave relationship. Nonperiodic remnants may result from bad connections or line interference.

If you're shopping for a digital transfer device, consider a few basic questions. How important is bit-accurate transfer to you? Does the device support the sample rates and bit depths that you require? Is it capable of slaving to an external clock? Some of these questions may require researching the manufacturer's Web site and possibly even e-mailing or calling tech support with specific queries.

FINAL TIDBITSDigital transfer may not be the best way to move your audio around, at least between computers. You could burn an audio CD, then use a ripper program to retrieve the track, or you could save the file as a data file to a CD-R or a Zip or Jaz disk. Ethernet connections are becoming more popular in personal studios because they facilitate the sharing of large audio files. I've even used my notebook's infrared port to share audio files with others.

Nonetheless, if you have digital tape in the mix, you'll want reliable digital transfers. Evaluate your needs, run some tests, compare the results, and adjust accordingly. Once you've worked it all out, write down the procedure so that you won't forget a step when the heat is on. When cloning time comes, you shouldn't need a bit of luck.

To transfer bit-for-bit clones consistently between digital devices, keep the following checklist handy:

1. Check your connections; make sure that they're snug and secure.

2. Use the shortest possible run of the highest-quality cable.

3. Make sure that you have both devices set to the same bit depth and sample rate.

4. Set the source device to its internal clock (master) and the destination device to external clock (slave).

5. Disable input and output faders if possible; if not, set them to unity.

6. Test your results in an audio editor to confirm that accurate transfers are occurring.

Show Comments

These are my comments.

Reader Poll

Do you use music streaming services?


See results without voting »