You all know, of course, that it is usually mandatory that you start reading at the back of EQ (right here). Well, I will allow you to read other articles first, but not all of the time.

Mac Expo

I went to the big Mac Expo in New York last month. There were some cool things, and some surprises. The new iMac doesn’t have a SCSI connector anywhere to be seen. I hope this is not going to be the trend for all new Mac products. Some vendors decided not to exhibit at the show because Mac no longer supported their SCSI devices.


Is 16-bits Obsolete?

No. That is like asking if dollar bills are obsolete. Fives and tens are better, but there are still plenty of uses for one dollar bills.

20-bit and 24-bit converters have actually helped 16-bit recording. The most important factor in the specifications for an A/D converter is linearity. Linearity is the relative voltage value of each bit in the conversion process. Let me back up a little bit (pun intended).

Converters 101

The analog waveform consists of changes in voltage over time. By now you have all seen audio waveforms in some sort of audio editing software. Lets start at the zero crossing point, the line in the middle that the waveform crosses when going from positive to negative. Right on this line is zero voltage. If you short out the input to your converter and record, you will get a flat line at exactly zero. At the loudest value, up where the waveform clips, the value is usually +10 volts. The clipped waveform at the bottom is –10 volts. With 16-bit information, there are 65,536 steps between the negative clipping point and the positive clipping point. This means that each step is equal to 10 volts/ 32,767 steps, or approximately .0003518509476 volts per step. Each step is represented by a bit. When counting in bits, each bit is exactly twice the size of the bit before it. The smallest bit has a value of 1, the next bit a value of 2, the next bit a value of 4, then 8, then 16, 32, 64, 128, 256, 512, 1034, 2048, 4096, 8192, 16,384, 32,768, and finally 65,536. Here is where the linearity comes in.

When we assign a value of one to the smallest bit, it is very easy to make each bit twice the value of the one before it. But, in our audio converter, the value of the smallest bit is .0003518509476 volts. This means that the value of the second bit must be exactly .0007037018952 volts. This is the basic reason why some 16-bit converters sound different than other 16 bit converters. This is also the reason why some 16-bit converters cost more than other 16 bit converters. Even with a single manufacturer of converters, there are different grades of converters in the product line. The ones that come off the assembly line with perfect linearity receive the highest grade, and get the highest price on the market. So if two companies use the same brand of converter in their product, it doesn’t mean that they will perform exactly the same.

There are other factors that also determine the quality (and cost) of a converter, like shielding, reference voltage stability, clock jitter, and the accompanying analog circuitry, but for this little discussion we will focus on linearity.

When the first 16-bit digital recording machines showed up 20 years ago, there was
no such thing as a 16-bit converter. You had to use a 12-bit converter and 4 bits of an 8-bit converter to get 16-bits. When 16-bit converters started showing up on the market, the top 12 or 14-bits were pretty good, but it was potluck on the linearity of the last couple of bits. Remember that the biggest bit is 10 volts, but it must be 10.000000000 volts. A variation of .01% would give you 10.001 volts. Not too bad, but that kind of deviation in manufacturing would make the smallest bit of the 16 off by more than 300%.

The point of all this? Well, manufacturing techniques have improved over the last 20 years, and we now have 24-bit converters. The tolerances have to be 256 times better than for 16 bit converters. The pay off is that the first 16-bits of a 24-bit converter are going to be 256 times better than 16 bit converters from 20 years ago.

There are a couple of good reasons to record 16-bit. Maybe you haven’t upgraded your 16-bit recorder, but you want to improve the quality of your recordings. Or, maybe you don’t have the 50% more hard disk space for 24-bit files.

Even 16 bit recordings sound better with 20-bit or 24-bit converters because of the increased linearity. Even more improvement can be realized with dithering or noise mapping schemes that are available in both hardware and software versions.

Dithering and Noise Shaping.

Dithering is the process of adding some low level noise to the digital signal to increase the apparent resolution of the smallest bits. Using dithering, it is possible to get acceptable audio quality for some applications with only one or two bits. You know those recordings that you hear at the airport, "This area for loading and unloading of passengers only---?" Those message playback systems are one to four bits at very low sample rates, but they perform their job just fine because of the use of dithering. Adding dither to your 16-bit digital audio will add about 1 1/2 bits worth of additional resolution to your audio. In other words, your 16-bit signal would sound almost as good as an un-dithered 18-bit signal.

There are different methods of adding dither to your signal, and the best results depend on the audio content in your recording. If you can, it is a good idea to listen to your recording with different types of dither to decide which method works the best for you. The most common methods for music are usually triangular dithering or random noise dithering.

Noise shaping is a little more complex, and the methods employed vary among manufacturers. Noise shaping needs DSP power to perform correctly. Sometimes noise shaping can be available right in the converter box that you are using. Because of the DSP processing that is needed, these converters are usually of the more expensive variety.

There are plenty of noise shaping algorithms available, but the two most common noise shaping methods are UV-22 from Apogee, and Super Bit Map from Sony. Sonic Solutions has a Sony Super Bit Map plug-in for their system, and Sony makes a stand alone SBM box. Sony has also included SBM in some of its DAT machines. For Apogee’s UV-22, there is a plug-in for Pro Tools called Master Tools. UV-22 is available in the new Apogee AD-8000 eight channel A/D converter box, as well as their 20 bit AD-1000 converter box. Other manufacturers will also be offering UV-22 plug-ins for their software.

Simply stated, dithering adds noise in the very lowest levels of the signal. Noise shaping takes this noise and mathematically moves it to a portion of the audio spectrum that the human ear is less sensitive to. You pay the extra price because of the complexity of the math involved, but you reap the benefits in the additional apparent resolution in the audio signal. A well noise shaped 16-bit recording will give you about 19-bits worth of resolution. I guess you could say that the noise shaping guys are getting in their two bits worth.

Alternatives.

There is an alternative to all of this. You can buy one of the new 24-bit DAT recorders from Tascam to record your mixes, you can buy more hard disk space and upgrade your Pro Tools to 24-bit, and then you can start worrying about converting to 96kHz or Sony Bit Stream recording. Face it, you can never win when it comes to keeping up with technology.


As a reminder, Roger’s web site has changed to www.rogernichols.com, and his direct email address is roger24bit@aol.com.