The Sound of Music: Digital-y versus Analog-y

But you are missing my point: it’s not the fact that it ever was digital that’s a problem. Given a band limited signal you can get back that exact signal, no artifacts, no noise from grain boundaries in the vinyl or mag tape, etc., no added tape bias, or hiss, etc. And with digital you’ll have more of your original signal intact (within the limits of the bandlimiting). Whether 2D or 3D if the signal is bandlimited then the digital can contain all of the information in the original signal. In the 3D world the sampling is often coarce and things like right angles are important so bandlimiting can be a constraint. In audio large excursions (impulses) are limited by the bandwidth of the analog hardware and hence are already bandlimited, digitizing them and then converting back to analog can, in theory loose nothing. Doing anything in the analog form will loose info.

I’m not claiming that typical digital systems operate perfectly, far from it, but their failures come from a different place than the digitization. They come from the analog limits of the digital hardware: noise and jitter both of which can be made better with better analog equipment.

One of the faults of digital is it allows a lot of lossless manipulation and people are tempted to do that, often those manipulations hurt the quality of the audio more than limiting yourself to the things than can be done well in analog, but that’s a problem with the temperament of the humans doing the editing, not digital vs. analog.

My only point in my first message on this thread is it’s not digitization that’s a problem: it can lose nothing with a bandlimited input. At the fundamental levels that’s the wrong place to look for problems with digital hardware vs. analog hardware.

1 Like

Thanks for breaking it down into terms that even a simpleton like me can understand. I appreciate your help as always.
Now how would you personally describe Analog-y and Digital-y? It would be interesting to hear how your expertise tilts those terms one way or the other.
I never thought about the mixing bench / engineer adding his or her flavor to the dish either :thinking:

Dude, missed you at RMAF.

I’ve heard pretty convincing demos with different groups (one time a pianist another time a trio (drum, sax?, ?)) where you could listen live and then walk down the hall to hear the music being played back after having been digitized (with DSD) and converted back to analog. There are differences brought about by recording (whether analog or digital) since there are only so many mics and so many speakers, etc. but the character of the sound was the same, whether it was live or not wasn’t at all obvious with your eyes closed. Note that neither sounded like recordings you buy. Hearing master tapes and/or master DSD recordings is amazing and far from what we have at home.

Personally I’ve noticed that people tend to play analog back at higher levels than digital and say “See, analog is better.” I don’t hear it that way with a good digital system that’s playing back at the same level: needle drops can sound pretty darned good. I understand how this comes about: In the past most digital systems I heard weren’t pleasant to listen to at those high levels, but they are getting much better these days.

An ironic thing is that the DSD recordings I’ve heard of live performances are truer to the original performance that the results you get after being released on vinyl, CD or SACD. There are steps along the way to consumer releases that modify the sound and the steps/paths are almost always different from media to media.

I wonder what the file size / format is on the DSD recording and if the size / format could be maintained somehow even if it took on a different file type / playback program? Would the final playback sound “better”? Technically you are not limited to the amount of info in the file except by the current media processing and playback limits.

Indeed, the problem is not the method of recording - digital really does produce the same as its input within a specified bandwidth, assuming a perfect implementation.
It’s the implementation that’s the issue (most often on the analogue end of things), but it’s often next to impossible for us as end user to be able to make a meaningful direct comparison without having control of the source selection and mastering, apart from the one mentioned exception of taking a really good vinyl master and then digitising it with the very best kit available, and comparing input to output.
Our argument is with the recording industry, not the equipment designers :slight_smile:
.

Reading the Neil Young/Feel The Music Book. I started a Topic here if anyone has had a chance to read it yet. Ties into all the discussions here I think.

Feel The Music

Digitizing an analogue signal does introduce quantization noise. There are noise shaping techniques, which I once briefly understood, which can move this noise out of the audible region. If I look at spectrograms of 24/96 PCM I can generally see a band at about 70-80khz which I believe is shaped quantization noise. It is at very low volume and so far above the conventionally accepted limitations of human hearing that it is unlikely to have any perceptible effect.

Yes, I could have worded things more carefully, but I was talking about the limits of what can be done and better hardware can lower the quantization noise in digital below than the inherent noise in the necessary analog parts of any recording process (e.g. resistor noise) and is usually much lower than tape hiss, noise from the grains in vinyl, etc. (Which can also be lowered by better hardware.) DSD moves the quantization noise much higher than the audio band (tho DSD64 is right on the edge, DSD128 and DSD256 can do better than 24 bit PCM quantization noise over the audio band.)

2 Likes

I’ve yet to find the definitive explanation why, but DSD can (depending on design) match and exceed analog in ways that PCM, even at 24/352khz cannot. I’ve heard things like the temporal resolution of standard sample rates is too low, filters are too sharp too cheaply designed, not great low level linearity, too much noise from complex electronic components, all being responsible for the harder, more flat, and sometimes grainy sound of PCM.

To my ears, DSD is where digital and analog comparisons become less relevant because it just sounds exactly like the source. Beyond that discussions about “analog” qualities are mostly about added distortion or bass.

1 Like

I think that’s a key finding.

Many years ago I already heard self made CD recordings of an audio genius which didn’t pass the usual stages consumer media go through. They sounded miles superior to anything on disc I heard so far, even if they were standard resolution.

It seems high end vinyl releases (e.g. a 45 RPM Acousticsounds recording) at least are quite near the lacquer…Steve Hoffman published such a comparison on his forum and reported it quite on one level. This for sure is different with inferior vinyl releases.

An assumption could be, that during the long denied losses within the digital manufacturing process of consumer media (remember the story of lossless digital copying), the problems introduced (don’t know if it’s jitter, noise, whatever) are more seriously affecting the sound (affecting timing, ambience, openness) than those problems introduced into a vinyl manufacturing process (which just seems to put a kind of more or less strong haze on the performance).

If this makes sense or not, the part of your statement, that the digital production process is responsible for a noticeable degradation is also a statement of some leading digital/vinyl mastering engineers.

I guess the high end R2R listeners would probably originate their opinion of a superiority of early tape copies with a less lossy production process of those tapes, too.

An interesting article within one of Cookies mailings.

It among others tells that digital conversion (external, not within a DAC) is more harmful than an add. analog step when releasing a 16/44 recording on SACD (not sure why this should be done), that it’s harmful when having to downconvert a higher resolution recording for release on CD (and how maths plays a role in order to choose the better recording resolution for that purpose as also Ted often mentions this maths influence) and that an analog step inbetween an otherwise pure digital processing can even improve the sound quality (the 2xHD label also does so generally). Interesting, as this also limits dynamic range to the one of tape…seems to play a lesser role?

https://dsd-guide.com/analog-transfers-vs-digital-conversions-what-does-it-say-about-files-youre-listening#.Xb4fPy-1Kf0

1 Like

I have a good deal of experience with analog to digital conversion and some experience with mastering off of tape. Cookie is absolutely right that down converting from 24/96 or DSD to 16/44.1 causes a large loss of quality (as also the original AD conversion).

Most mastering engineers I know do 90% of their processing on analog equipment, for both digital and analog sources. They consider the extra AD/DA conversions to be worth it to use good quality analog tools.

What I’m not sure of is whether upsampling 16/44.1 to DSD digitally is inferior to playing a 16/44.1 file through analog and converting with another AD converter to DSD. That is certainly the way to do it if you have good analog processors that you want to use, or need to add some flavor to the digital file.

Those losses in downconversion (e.g. 24/96 to CD format) make me wonder how some hear a superiority of disc drives … but probably just compared to the redbook files streamed, not the hires files. I guess generalization in many regards is not good most of the time.

That’s my guess. Maybe for equivalent formats like CD vs FLAC or SACD vs DSD files.

Ted has stated elsewhere that it is possible to resample to any arbitrary Freq but the quality of the conversion process depends on th precision of the maths (if understand correctly). In the majority of implementation s an integer change (eg 48 to 96) the chances of getting a good job is much higher.
44.1 to 48 or 96 is not an integer conversion so I guess unless you have a theoretically perfect conversion algorithm you may be better using an interim analogue stage to avoid “bad math”.

Yes that’s what I also understood.

To convert 96k to 44.1k you need to upsample by 147 (to 14.112MHz), do a great brickwall filter job with a cutoff a little lower than 22.05k and then downsample by 320 (and possibly dither to 16 bits). And yes you need to have reasonably good precision on your filter coefficients (say 30 bits) and use about 60 bit precision math. It used to be that a lot of such convertions used sloppy math and or sloppy filters, but now iZotope filters (just as an example) do well and are available in many players and editing systems.

2 Likes

So the perceived need for an analogue stage is less than it used to be and should only decrease further?
Thanks for answering :slight_smile:
.

It’s still been my experience that doing analog editing and using DSD as the “tape recorder” is the cleanest sound.

3 Likes