Recording Formats: Definition of Original

[Axiom05]

I've been confused about what is an original recording format for some recent recordings that are being sold on sites such as NativeDSD. For example, the recent Brahms 4th Symphony from Reference Recordings (Honeck/Pittsburgh Symphony) was recorded in DSD256 and post processed in DXD. The recording is available as a download in the following formats for the 2 ch stereo version: DSD512/256/128/64, 24-bit DXD and flac at 192KHz & 96KHz. So what is the "original" format? Is it DSD256 or DXD or does it not really make a difference? It seems strange to me to go from DSD256 to DXD and back to DSD256, can this still be called native DSD? It seems that what is an original recording format is not so simple anymore.

[yamamoto2002]

DXD is 352.8kHz 24-bit PCM.

 

Technical considerations

 

• IMO editing DSD256 signal with DXD is okay, but ultrasonic range signal above 170kHz will be lost and it is impossible to recover full DSD256 info from converted DXD data. Therefore DXD is the master data format in this case.
• I think it is okay to call DXD edited DSD data to native DSD64/native DSD128 because DXD contains almost all the info to fill DSD64/DSD128 information bucket. But very high-end ultrasonic range of DSD256 info has been lost when converted from DSD256 to DXD, and final DSD256 deliverable contains silence on the very high frequency range and it is waste of signal capacity of DSD256.

 

But, practically there is no meaningful musical info recorded above 100kHz because even microphones specifically designed for high-resolution recordings (such as Sanken CUX-100K) has output frequency range limited to 100kHz. So if there is some info above 170kHz existed in "original" DSD256 data, most probably it is noise created by analog circuit of head amp, ADC or by digital processing, and it is not by vibration of the air of recording booth, scoring stage or concert hall.

 

[John Dyson]

15 hours ago, yamamoto2002 said:

DXD is 352.8kHz 24-bit PCM.

 

Technical considerations

 

• IMO editing DSD256 signal with DXD is okay, but ultrasonic range signal above 170kHz will be lost and it is impossible to recover full DSD256 info from converted DXD data. Therefore DXD is the master data format in this case.
• I think it is okay to call DXD edited DSD data to native DSD64/native DSD128 because DXD contains almost all the info to fill DSD64/DSD128 information bucket. But very high-end ultrasonic range of DSD256 info has been lost when converted from DSD256 to DXD, and final DSD256 deliverable contains silence on the very high frequency range and it is waste of signal capacity of DSD256.

 

But, practically there is no meaningful musical info recorded above 100kHz because even microphones specifically designed for high-resolution recordings (such as Sanken CUX-100K) has output frequency range limited to 100kHz. So if there is some info above 170kHz existed in "original" DSD256 data, most probably it is noise created by analog circuit of head amp, ADC or by digital processing, and it is not by vibration of the air of recording booth, scoring stage or concert hall.

 

On normal consumer material, there is a rolloff that starts slowly below 30kHz and is pretty much gone at around 35kHz.   This doesn't include boutique stuff or material that has been specially copied from an original *MIX* tape master.   Some distribution masters might also have wider response.

 

Also, a lot of hidef 'normal consumer material' has noise injected so that it seems like there is more info above 25kHz. (This is from an *industry* individual!!).   Also, the little blips of mostly noise that one sees on spectrographs comes from the noise reduction system 'opening up'.   Pure digital material, of course, doesn't do that.   However, Ill bet you that almost all material produced before the late 1990s will also produce the false high frequency measurement.

 

True boutique material or derived directly from mix tape is something to be treasured.

 

John