Paul McGowan: In my post Climbing the ladder I pointed out that something fundamental changed between the introduction of digital audio and today with respect to DAC architecture. DACs of old (we’re only talking 25 years or so) used a fixed architecture to handle the PCM signal. That architecture relied on a string of 16 precision resistors to set the voltage steps for audio.
As the pressure to have greater dynamic range and signal to noise performance pushed the industry to 20 bits, then 24 and even 32, it quickly became evident that use of the traditional DAC architecture would no longer do. We needed something different and we got it. The Pulse Density model (known under a number of different names) Delta Sigma, 1-bit, multi-bit, DSD, SACD, is an entirely different model than the classic ladder DAC. It gives us the higher bit depth and increased dynamic range we’re looking for. But that’s not the whole story.
In the late 1990′s there were two divergent paths being taken by DAC chip companies to achieve these better specifications: Delta-Sigma and Sign-Magnitude. In the end, Delta-Sigma won but not before a long torturous path was travelled. That path still evokes emotional responses even today.
As best I can tell, here’s what happened. Once engineers figured out that building a straight ladder DAC was a dead end street, they started looking for alternative methods to achieving higher bit depths and faster sampling rates. The higher bit depths problematic because of the precision needed in components, the higher sample rates not easy because most of the schemes required processing time for each sample.
One group stuck with the traditional approach to PCM and came up with a derivation of the classic ladder DAC called Sign-Magnitude. Sign-Magnitude sounds pretty technical but it’s simple. If you remember our discussion on how to count in binary we are always going from 0 to a higher number. So a legacy PCM Ladder DAC does the same thing – it is always either at 0 or something higher – but this doesn’t accurately represent how sound works in the real world. Sound centers around a zero point and is subtracted and added to from this zero point (picture a loudspeaker pushing out and pulling back). This means we could more accurately represent lower level signals by having negative numbers and positive numbers. Therein lies the Sign-Magnitude.
We sacrifice one of our precious bits to denote a positive or negative number and our DAC output sits at the mid point, and subtracts and adds numbers (eventually voltages). A typical ladder DAC can only add. This technique, coupled with improved laser trimming of parts, resulted in a breakthrough part from Burr Brown (later purchased by Texas Instruments) released in 1998, the famous PCM1704 23 bit R2R DAC. Quickly adopted by Wadia and Theta and years later still used by MSB, this part is said to have far better sonics than anything else ever produced and it is a ladder style (R2R).
Limited to 96kHz and 23 bits (some refer to this as a 24 bit DAC because the sign bit still gives the extra 6dB, but technically it’s a 23 bit DAC) this was an amazing technological feat by a group of pretty savvy engineers at BB. The part is still available but there are big warnings and cautions on the TI website about it becoming obsolete soon – and not to use it in a new design.
Many Audiophiles and even a few manufacturers still cling to the superiority of this technology as the be all to end all. Ahh, we Audiophiles sure like to cling onto our cherished beliefs. I am as guilty as the next.
But as a designer, I would suggest that while there are some good advantages to this architecture, it’s scope and days are limited and basically doomed.
For my money, I like DSD and Pulse Density. There lies the future of high performance purist audio and tomorrow, let’s take a look.