Tuesday, June 12, 2007

The Case for NOT going above 0 dBFS For Digital Playback Systems

I first became aware of this issue back in the early 90's when I was testing a DAC that used a Burr-Brown Digital Filter and D/A Converter. I observed that the 1kHz square wave test from CBS CD1 test disc was being clipped. I knew that the analog low pass filter had plenty of headroom, so I decided to contact Burr-Brown and speak to an applications engineer. When I explained to the applications engineer exactly what I was observing he remarked that this test went over the 0 dBFS limit and was clipped by the digital filter. He also remarked that this was an "illegal state" and nothing should go over 0 dBFS. He also bluntly stated that this was a "stupid test". Twelve years later and that Burr-Brown applications engineer is still right.

In the AES paper titled 0dBFS+ Levels in Digital Mastering by Nielsen and Lund are a number of examples as to what happens when the recorded level goes above 0 dBFS. If the signal is not clipped by the digital filter, and goes above the 0 dBFS level than the signal exhibits gross distortion. A well designed CD player or DAC will have -90 db THD + N or better at 0 dBFS. When you go over that the distortion can easily be degraded by 50 to 60 db. It appears by the numbers and what many of us are hearing is that Linear PCM, just like any other system can be overloaded.

Given the present mentality of the ever obnoxious louder is better crowd, allowing more headroom in digital filters will only cause recorded levels to be pushed even higher until they reach the new limit. This is starting to remind of that one great scene in Spinal Tap, where the guitar player points to his Marshall stack and proudly proclaims that his volume dial was changed so he could turn it up not to just 10 but 11.

There is one last issue that concerns me. The typical voltage coming out of our players at 0 dBFS is typically 2.0 to 2.2 Vrms. Power amp sensitivity can be anywhere from 0.5 Vrms to 1.5Vrms with gains typically from 26 db to 32 db ( Av = 20 to 40). Pre amp gains are typically 6db to 18 db (Av=2 to 6). As we can see from the gains presently used in consumer audio 2.0 Vrms is already high and is pushing at the limit, and with the hyper compression we will find ourselves at the very lower limits of our volume controls with small changes in volume producing very non linear results. If this is pushed higher, than standard systems will have to have adjustable gains to incorporate the higher output from the players. I don't think the rest of the industry is going to have a very positive view on this to put it mildly. For those of us who listen to classical and pop we will now be faced with resetting gains for different program material. For Home Theater users whose speakers are typically 90 db spl the problem is further exacerbated due to the higher efficiency speakers.

Given all of the evidence above I see only greater distortion, and aggravation for consumers and music lovers if recording levels are continually abused and taken to beyond their designated limitations.

http://www.audioholics.com/education/audio-formats-technology/the-case-for-not-going-above-0-dbfs-for-digital-playback-systems