![]() ![]() |
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
In Reply to: Re: P-3A doesn't convert to 24bit. posted by steve b on January 08, 2001 at 14:26:44:
I think you are still missing the point: .there is a difference in how the P-3A (as well as most other upsampling DACs) and the P-1A approach creating a 24-bit signal from 16-bit input.While, as you indicate, the Crystal sample rate converter chip in the P-3A "maintains 24 bit precision in its calculations," it does this by adding dither to the signal, which is different from creating a 24-bit signal by interpolation using DSP, which can be done by the P-1A but not by the P-3A. .Please note, I'm not saying better, just different, as I've not heard this yet. .(I note, however, that Dan Wright and most of the others posting here who have listened to both feel that the P-1A's interpolated signal is preferable; .some users, as indicated earlier in this thread, are not so sure.)
In the recent review of the P-1A/P-3A in SoundStage!, Doug Blackburn quotes one of the P-Tech founders as follows, which is one of the most concise explanations I've seen of this distinction:
"Peter Madnick clarifies the operation of the P-3A and other so-called 24/96 DACs and 24/96 outboard processors this way: .The P-3A does not interpolate 16-20-bit data to 24-bit data -- nor does any 24/96 DAC (e.g., the Bel Canto DAC1) or add-on 24/96 external device (e.g., Assemblage D2D-1). .All of these products, so far, will output 24-bit data via dithering of the 16-bit input data. .This results in data that has roughly 17 bits of equivalent resolution within a 24-bit datastream. .The P-3A and other 24/96 products do upsample, for example, from 44.1kHz to 96kHz if you select that mode. .People tend to assume that 24/96 products that advertise upsampling and/or interpolation actually convert 16/44.1 data to full 24/96 resolution, which is not accurate. .To actually get 24-bit data in a 24-bit datastream, you must use digital signal processing (DSP), which is included in the P-1A and is referred to as "Resolution Enhancement" in the P-1A's user manual."
Hope this helps explain the distinction between the two processes, adding 8 bits of dither versus interpolation of 24-bit data by software running on DSP chips.
cooney
Follow Ups:
Thanks for the notes cooney :)Usually when discussing sample rate conversion 'interpolation' refers to the extra samples created during upsampling. Precision is used to describe the 'resolution' of the samples after processing. How the value of the samples are calculated is the difference between the 2 PT units, and that is understood.
steve
Hi cooney,No, it only adds to the confusion because it's not quite accurate. The CS8420 does do interpolation, as all up/oversampling chips do - in fact interpolation is implicit in up/oversampling. I'd be amazed if the CS8420 didn't pass 24 bit data (without dither), as it takes more effort and expense to dither the signal rather than pass the expanded wordlength to the output. It just wouldn't make sense.
Here's what PT's own literature says on the subject:
"The CS8420 device actually has a 24-Bit data path to its
outputs, with some dithering options to reduce the
word length. We are not using the CS8420 device’s
dithering option, but instead use our own proprietary
“adaptive dithering algorithm” in the DSP. For a
44.1K sample CD source, the CS8420 device is
passing 24 bits, which essentially track the 16-Bit
source signal into its new output sample positions."So, in the P-1A the CS8420 is passing 24 bits, but they are not using it's dithering options.
What the P-1A does that no other DACs do is what PT call resolution enhancement. This isn't interpolation in the sense that the term is usually used in, it's a process of estimating the value of the original data using psycho-achoustic techniques. It may work very well, but can we call any value that is the result of "estimation" (PTs own term) True 24 bit data?
Note: plenty of DACs are running at 24 bits without a bit of dither in sight. Also, all up/oversampling chips are DSP chips.
Daniel Espley
This post is made possible by the generous support of people like you and our sponsors: