Audio Asylum Thread Printer
Get a view of an entire thread on one page
|For Sale Ads|
In Reply to: RE: Mahler 3: Need Help with High Rez Recording Versions posted by Mel on July 10, 2017 at 20:21:23
And I think it really illustrates why you just can't release a recording as a one-shot experience with no editing - you HAVE to edit! About 26 seconds in, one of the horn players misses one of his notes - it's not THAT audible, but it's there. Just imagine listening to this mistake repeatedly, anticipating that it's going to be coming up, and losing your focus on the music in the meantime. Not good.
But OTOH, it does raise the same question which Mel raised (way down below in this thread): if you have to edit anyway, why not just record the work in DXD from the get-go, without all this conversion back and forth. Sure, I know the argument that the actual amount of time taken up by the edits is not that large. But Jared has told us that some movements of his previous releases contain a VERY LARGE number of edits (and there's nothing unusual about that - it's just the normal way classical studio recordings are made these days - and even the so-called "live" performances are a patched-together result of a series of performances, edited together).
So, how about it, Jared? I know you and Tom are believers in DSD, but might we hope that, at some point, we might have a DXD-sourced recording or two from Channel Classics? ;-)
The answer to your "why not just record in DXD in the first place" is there is no such thing as a direct DXD (or for that matter any PCM) recording method or hardware. Not at least with any A/D converter made in the last fifteen years used by any recording production outfit. The last really good direct PCM (20 bits on a really good day) A/D converter made was the Pacific Microsonics Model One/Two. These two products have been out of manufacture by at least fifteen years, but can be heard used today (by one of its inventors) by Keith O. Johnson on Reference Recordings.
PCM recording, including DXD (352.8KHz/24 PCM) is today accomplished by A/D conversion performed by Delta-Sigma modulation (a 1-bit or multibit DSD like Pulse Density Modulation process), then converted on-the-fly in the converter box to any sample rate PCM supported. The disadvantage of recording in DXD in the A/D converter are two:
1- The digital processing power available within the A/D converter is not as much as available in a reasonable PC, thereby limiting the robustness of the filter and conversion algorithms used. Plus, offline conversion in post does not have to be in realtime, further adding to the conversion choices.
2- This first PDM to PCM conversion, which is a lossy process, can not be undone in the future. Since native DSD content will increase in value in time as the format becomes the industry standard, the future value would be undermined.
If this sounds self serving, I'm happy to demonstrate to anyone visiting Boston the degradation of even one DSD > DXD conversion. You can experience this yourself with the DSD256 and converted DXD files available on NativeDSD. It helps of course to perform the demonstration with surround files, where the localization cues are much more recognizable, and using a DAW like Pyramix to facilitate A/B switching on the fly.
Until DSD post processing becomes available and commonplace, the best practice for complex large scale recordings is to record in DSD, post process in DXD, then offer deliverables in all playable formats. When DSD processing is made available, the original DSD masters will be available for reprocessing, and the DXD edited masters discarded.
. . . and I think that some of your claims (DXD edited masters will be discarded - really?) will be news to companies like 2L! ;-)
Check out the "Original Source" column on the far right:
Yes, I'm afraid that's to 2L's unfortunate disadvantage if they actually track the DXD output from their converters, and not the DSD doing an offline post processing conversion to DXD.
Morton has always been a DXD proponent, through the days of using their DAD AX24 and AX32 A/D converters, to their current use of Merging's Hours. All use Delta-Sigma modulators as front ends (the actual analog to digital conversion element), and employ onboard realtime 352.8KHz 24 bit converters and decimation filters for the DXD conversion, if that output format is chosen. This results in a lossy second generation output.
I have no knowledge of 2L's workflow, but I do hope they actually track in the fastest bit rate DSD (DSD256 in the case of Horus), then do an offline DXD conversion using Pyramix's excellent Aprodizing conversion algorithms, keeping the original DSD for potential future use. Any other approach makes little sense; throwing out the original higher definition PDM (DSD) format in the A/D converter box to keep the working editing format (DXD) output, where you could have both for just the cost of storage.
As I mentioned in my previous post, if you create two projects in Pyramix (a DXD and DSD) and load a DSD file in each, you can switch between each on the fly. Listening to the original DSD, then the same DSD file converted on the fly to DXD, and played out through the DXD Project Mixer, you immediately sense the added sharpness and reduction of some spaciousness details inherent in PCM. Glad to demonstrate to anyone passing through Boston.
Thanks again Chris
Of course you're going to encounter some deterioration on the converted file!
But actually, at this level of resolution (either DSD256 or DXD) I'd be surprised if the deterioration IS really audible and not just a product of expectations and/or psychology. But whatever. I guess I'll just have to pay you a visit the next time I'm in Boston (a city I've missed so far in my life) - I'd love to hear those Sound Labs! ;-)
Same as occurs within an A/D converter outputting DXD instead of DSD, via the conversion within the converter. You're always welcome to visit Chris and judge for yourself :)
I apologize for cutting and pasting - this is from John Siau of Benchmark Digital:Modern PCM sigma-delta converters produce much lower error signals than 1-bit sigma-delta DSD converters. The errors in the DSD system are due to the 1-bit quantization that occurs in 1-bit sigma-delta DSD converters. Multi-bit PCM sigma-delta converters can be fully dithered and do not suffer from this un-dithered truncation. Every added bit reduces the noise signal by 6 dB. A 4-bit sigma delta converter is 24 dB quieter than a 1-bit sigma-delta DSD converter. Right from the start, 1-bit DSD signals have higher losses than multi-bit PCM signals.John wrote all this in 2015. Have things changed in the meantime?
Conversion from 1-bit DSD to multi-bit PCM is a lossless process inside the audio band. The only thing that is removed is the out-of-band noise above the Nyquist limit of the PCM system. Nothing else is lost. . .
Conversion from multi-bit PCM to 1-bit DSD is always a lossy process. The loss is due to the 1-bit truncation. This truncation introduces a very large ultrasonic error signal that makes the ultrasonic region unusable for audio.
Processing a 1-bit signal to create a 1-bit signal is also always a lossy process. A volume control is one of the simplest processes in a multi-bit PCM system, but it creates a large error signal when applied in a 1-bit DSD system. The same is true for any other 1-bit to 1-bit DSP process. The lossy part of these DSP processes is the quantization back to 1-bit. Cascaded 1-bit truncation processes can rapidly degrade the audio quality. For this reason, DSD is always processed as multi-bit PCM.
Any DSP process applied to a 1-bit signal produces a multi-bit signal. No loss of information occurs until this is quantized back to a 1-bit signal. Why incur the loss by going back to a 1-bit signal after the processing chain?
All practical DSD systems require some sort of DSP processing (gain control, mixing, filtering, etc.) and all of these processes produce multi-bit PCM results. Taking these lossless multi-bit results and adding loss by truncating them back to a 1-bit DSD signal makes absolutely no sense. DSD complicates the processing and adds unnecessary losses to the signal path. DSD does not simplify the signal path. There is absolutely no truth to the marketing hype that claims that 1-bit DSD is a simpler system than multi-bit PCM. The exact opposite is true.
Nothing has changed since 2015, and there's no such thing as a "PCM Sigma-Delta Converter". PCM is an acronym for Pulse Code(ed) Modulation, and PDM for Pulse Density Modulation, also referred to as PWM (Pulse Width Modulation).
You'd have to ask John what he's talking about, but as I read it, there's two possibilities:
1- He's substituting the acronym PCM for the actual PDM (two very different formats), in which case his argument makes some sense.
2- He's conflating two independent processes and calling that a "PCM sigma-delta" converter (which coincidentally are the separate processes inside a pro A/D converter). The first process, (on the A/D side), is the Delta-Sigma modulation of the audio signal; producing either multiple parallel Pulse Density Modulation (PDM) bitstreams (multi-bit PDM, which appears to be John's inaccurately naming multi-bit PCM), or much less frequently a 1-bit PDM bitstream (DSD as with the Grimm A/D Converter). If PCM is desired/required, then these PDM bitstreams (1 to 8 typically) are then followed by low pass filtering (decimation) and conversion to parallel X-bit wide Pulse Code Modulation binary words. That latter process is most certainly a lossy process. Once decimation filtered, you can no longer reconstruct the original PDM (DSD if 1-bit) bit stream.
But so what? My original statement was, and remains that all acoustic audio recordings are made using A/D converters, and the vast majority of those A/D converters are all front ended with Delta-Sigma modulators, producing an actual DSD 1-bit bitstream (as with the Grimm A/D), or from two to eight parallel (multibit) bit streams of Pulse Density Modulation. And, any conversion to any other format for whatever reason (typically post process sweetening and/or conversion to a deliverable format) is a post process from the original A/D PDM bitstream(s) conversion, and is lossy. So, if you can archive the original A/D conversion before performing the PDM to PCM/whatever format, why not do so for future processing possibilities? And especially, if you have the option, why not do those format conversions offline using more powerful conversion algorithms with more powerful processors than are available in the A/D converter box?
Seems self evident to me. But on the other hand, I don't have products to sell, or customers to convince.
But, as I read it, I think that your differences with what John wrote seem to lie in what each of you thinks is lost when going from DSD to PCM.
When John says:Conversion from 1-bit DSD to multi-bit PCM is a lossless process inside the audio band. The only thing that is removed is the out-of-band noise above the Nyquist limit of the PCM system. Nothing else is lost. . .And you say:
Conversion from multi-bit PCM to 1-bit DSD is always a lossy process. The loss is due to the 1-bit truncation. This truncation introduces a very large ultrasonic error signal that makes the ultrasonic region unusable for audio.If PCM is desired/required, then these PDM bitstreams (1 to 8 typically) are then followed by low pass filtering (decimation) and conversion to parallel X-bit wide Pulse Code Modulation binary words. That latter process is most certainly a lossy process. Once decimation filtered, you can no longer reconstruct the original PDM (DSD if 1-bit) bit stream.
So. . . as I read this, I conclude that you're definitely removing SOMETHING in going from DSD to PCM. But what is it? Is it the out-of-band noise and nothing else, as John seems to suggest? I mean, in a way, you guys are kind of saying the same thing. Right?
Chris, you're asking a great question that has been discussed with many thousands of words. I'll give you the short answer, and you can research the underlying discussions. I'd suggest if you're interested, the many postings at Computer Audiophile. Do a search on Miska, the provider of HQ Player, particularly on the DAC Forum.
The short answer to your concluding question is yes, the decimation filter has to eliminate the out of band noise, and in doing so, injects phase artifacts (unfortunitly required by the steepness of the filter) that render the resulting PCM words not convertible back to the original PDM/DSD bitstream(s). Therefore, it's a lossy process.
It's the effects of the necessary decimation filter required in the DSD > PCM conversion to preclude folding back into the audio band any energy above half the sampling rate of the converted PCM result. Since the vast majority of energy in any PDM (DSD) bitstream lies above the audio band, being almost all the energy approaching the bit rate frequency, the decimation filter has to be very steep approaching the upper frequency of the audio band of interest. The decimation filter shaping, with its subsequent phase distortions and nonlinearities (necessary to achieve the required filtering shape), make the reconstruction process impossible back to the original bitstream. Therefore, it's lossy, and in my experience quite perceivable.
A PDM (DSD if 1-bit) bitstream is almost entirely uncorrelated white noise. It's a useable A/D converting format through the process of noise shifting (an activity of the Delta-Sigma Modulator), where the uncorrelated noise is shifted to above the audio band, then filtered out when the audio content is retrieved in playing. Since PDM/DSD is already an analog bitstream, it simply requires low pass filtering to retrieve the audio content (there's many ways to do this, which I won't go into here, but the filter has to be either very steep (not good) if the audio is to be retrieved at the original bit rate, or the bit rate upconverted to a much higher bit rate, and a gentler shaped filter employed).
But for digital post production processing, since PDM/DSD is an analog representation of changing input analog signal levels, it contains no digital values ala PCM. It only contains analog signal level changes represented by the density of bits in the bit stream. Since no values are represented (actually, the density of bits is directly proportional to percent of modulation, which in turn CAN be translated to absolute signal level values in the conversion to PCM), the bit stream must be converted to a value based format (PCM) to be further digitally processed.
So, back to my original point; if as a recording producer you can archive the A/D converters original front end Delta-Sigma modulation bitstream(s), in the form of a 1-bit DSD bit stream(s) at the original recorded bit rate, why not do that? Then you can convert to whatever required/desired 2nd generation format meeting the needs/desires of the producing engineer/label, and have as backup the original PDM/DSD content for future possibilities.
2L's Morten Lindberg, as well as Bert van der Wolf of Northstar/Turtle/Challenge Records (among others) say they prefer the sound quality of DXD, and so record in that format (they say). That's their prerogative and choice, but it is a 2nd generation conversion.
I'm not sure I agree with everything you're setting forth in your post, but I don't want to get into a multi-day discussion of PCM vs DSD - that argument has been going on for a couple of decades now! ;-)
BTW, I've seen some of those discussions with Miska, but I'm sure there are many more which would be useful to follow.
I grew up with a performance of Messiah (Sargent? Sir Thomas?) replete with a horn flub in the Hallelujah Chorus. It became normative. Nothing since sounds right.
I think that happened on the Barbirolli Mahler Fifth - the LP album was release that way (i.e., with the missing horn part - unfortunately, I've forgotten exactly where it was). But when EMI was preparing one of the reissues, they allowed the horn player to come back into the studio and record his missing notes, which the engineers then edited back into the original recording! ;-)
Try here (all DXD albeit not Channel Classics). BTW does this look like it has a commercial future to you? It's in its third year. As I doubt that anyone can hear any difference between DXD and 24/192 I will remain sceptical:
. . . there doesn't seem to be a single title there which is in multi-channel. Not that I would get too excited about Michaela Petri (oops! I guess it's Michala Petri - have I been pronouncing her name wrong all these years?) and her recorder anyway. ;-)
But in any case, I appreciate the thought! And I'm not sure I could tell the difference between 96K and DXD myself! . . . angels dancing on the head of a pin. . .
So, other than that, Mrs. Lincoln, how did you enjoy the play?
Down below I brought up the comparison of multi mike recordings vs the old 3 mike recordings. Also the old recordings were done analog with tube mikes and tube recorders. Remember my listening is 2 channel. I sampled a bunch of Ivan Fischer Channel recordings on Tidal. Mahler, Wagner, Bruckner, and Stravinsky. There definitely is a house sound. I then listened to a Mercury Anatal Dorati recording mainly of Stravinsky. I much prefered the Mercury. More presence,greater dynamics and much better hall ambience. The channel sounds laid back with much less dynamics. Remember I spent my recording career using all analog tube equipment and when I did record classical orchestras I used the 3 mike Decca tree arrangement. Just my opinions. Also I don't necessarily believe it is a digital thing. More of an engineering thing. I have heard some great digital. Professor Johnsons Reference recordings for instance. Also Brothers in arms.
. . . the dynamics and hall ambience. I'm sure that at least part of this impression is accounted for by the two-channel vs. five-channel systems each of us is experiencing the performances through. Was the Mercury Stravinsky recording you listened to the Dorati/LSO Firebird? The general consensus is that that's one of the best recordings Mercury ever made. (BTW, I think I have about 75% of all the Mercury recordings which were issued on CD or SACD, plus a couple of the Plangent-process hi-rez downloads which derived from the later volumes of the big Mercury boxes, such as the Tchaikovsky Suites with Dorati and the New Philharmonia Orchestra, the Paray/Detroit Organ Symphony, etc.)
All other things being equal however, yes, I do prefer more minimally microphoned recordings, and in that sense I agree with you. Minimally microphoned multi-channel recordings, such as the ones on the Nishimura label (NLA unfortunately), have been produced - but even here, not all of these recordings hit the sonic bulll's eye! OTOH, I have to say that multi-microphoned recordings have gotten better and better over the years, and I generally find the ones made in the last decade or so to be quite enjoyable.
Post a Message!
This post is made possible by the generous support of people like you and our sponsors: