96kHz or bust

Join Our Community Subscribe to Paul's Posts

While the hornets swarm around the last subject we brought up for a little bit, I thought perhaps we’d turn back to A/D Converters, the topic we started before all this got somewhat sidetracked. When we left off our discussion on the Analog to Digital Converter, I had just finished explaining the function of the Sample And Hold Circuit of older, more traditional A/D Converters. That circuitry used to take a “slice” or “snapshot” of the incoming analog signal, hold its value intact while the A/D Converter converts the sample to a digital number.

Early A/D converters needed the sample held because the method they used to figure out the correct digital number for the sample was time consuming. Many of the original Pro A/D Converters, used to record music, ran at 48kHz and were bandwidth limited to half that sample frequency, or 24kHz, just slightly above the extremes for human hearing. As you may remember, the system had to employ a steep and (mostly) intrusive filter on the input to make sure no higher frequency artifacts were sent to the A/D Converter, or a form of distortion known as Aliasing might occur.

Today’s A/D’s are much more sophisticated and fast, not suffering these types of distortions or limitations at all. For example, the A/D used in our upcoming NPC product has an incoming sample rate of 352.8kHz. To put that in perspective, it means we can have a very gentle low pass filter on the A/D Converter’s input that starts at 80kHz, far outside the band of any usable audio information from any format other than perhaps a live performance (and then “useful” information might be questioned). In fact, we are recommending that NPC users do not exceed 96kHz for playback or recording of bandwidth limited analog source such as vinyl LP’s, tape recorders, FM tuners, etc..

This recommendation no doubt flies in the face of many of us who would immediately go right to the highest available sample rate of 192kHz. I know I did before doing a little research and listening tests. But truth is, 96kHz not only sounds better when playing vinyl LP’s, it makes no sense to go higher. There are several reasons for this: first, 96kHz/24 bits captures any and all info possible on a vinyl LP, tape recorder or just about any analog source you might connect to the A/D. Because the output sample rate on the A/D Converter is unrelated to the input sample rate and rolloff (352.8kHz and 80kHz respectively), there really isn’t any downside to using 96kHz to feed your DAC or computer. One last reason (and this applies to many modern A/D Converters) the decimation digital filters (output digital filter) have guaranteed zero group delay only up to 96kHz. Beyond that most modern A/D Converters have some form of small group delay above a higher frequency. Group delay simply means that some groups of frequencies are slightly out of time with others and in the case of an A/D, it is a consistent gentle change at the upper frequencies. Audibly you might hear a slight emphasis in the upper frequencies, relative to a zero group delay. Recording engineers can (and do) make slight EQ adjustments to compensate when employing higher sample rates with PCM – but for mere mortals like us, we don’t generally have such luxuries available.

I know this is going to raise the hair on some folks necks who are convinced that PCM running at 192kHz is essential for best sound, but I would have to argue that if you handle the incoming musical signals properly (i.e. do not roll them off anywhere near the usable bands) there really isn’t much advantage to it UNLESS you are converting DSD to PCM. Standard rate DSD runs at 176.4kHz and double DSD runs at 352.8kHz. But if you are playing or recording vinyl LP’s, tape recordings, FM tuners etc. there’s no advantage to higher sample rates than good old 96kHz/24 bits. if you’re recording live performance events using the A/D there may be some wisdom to capture using the highest sample rate and bit depth when properly executed, but if I were making live recordings I’d be using DSD and not PCM anyway (and none of these issues apply to DSD).

Perhaps a lot to absorb here, but it would seem to me a very common sense approach to making high resolution copies of your analog sources, saving hard drive space, reducing demands on playback processors, etc. It’s what I am doing and happily so.