[aos]

Upsampling 16/44 would sound better because of massive jitter reduction that is achieved in asynchronous sample rate conversion (jitter reduction is well documented in any such chip's datasheets) and also perhaps because of gentler filters with higher corner frequencies and more linear phase response (as sacd lover mentioned). You don't really need upsampling to achieve these benefits but from what I know DACs that do that without upsampling are rare and very expensive. With upsampling chips you can do that for far less money than manually designing a low jitter reclocking. As for filters, sometimes they are simply omitted. That makes signal looks horrible on measuring equipment but some people say it sounds better without them.

 

[sacd lover]

I have read many of the arguments why 24/96 or 24/192 should be better than sacd.Before I heard sacd I was rooting for 24/192 based on these arguments. The 24/96, 24/192 look good on paper, but when I actually listened to the two, sacd sounded better, atleast to me. 24/96 dvd-a and upsampled cd still sound like very good cds to me; sacd sounds different, analog-like is the common descriptor, and thats what I hear. I dont know why this is so and I offered some counter theories I have read as to possibly why. We didnt know what jitter was at the beginning of the digital era, so there may be more digital distortions that are undiscovered as of now that sacd minimizes. In any event, sacd does something better than pcm that my ears prefer.

 

[NathanJM]

Yeah reading all this makes SACD look like it should sound horrible. But to my ears SACD sounds much better than redbook.

I know in statistics, the best estimator is not always the one that returns the closest estimate of the true value (MVUE), but one that gets close with the least amount of variance (MLE). Perhaps we're encountering something like this here. Perhaps SACD DOESNT get as close to real sound as redbook or dvd-a is capable of, but perhaps is more true in what it CAN do and so sounds better on the whole?

just some thoughts, 'cause SACD DEFINITELY blows away any redbook cd ive ever listened to (on average. of course some outstanding redbook sound about as good and sometimes better at points to my ears as a lower end sacd). I read that article JaZZ references and I was completely shocked to hear about the horrific breakdown beyond 8000Hz. Personally, i havent noticed anything like this (or at least nothing major enough to stick out in my mind). I know it isnt an issue of my ability to hear higher frequencies--i've been tested to 22,500.

I think I would have to conclude that there is something else at play here, and that these measurements arent telling the whole story.

Now, I am going to re-listen to my SACDs and listen even more closely for these mentioned failings of DSD.

 

[JaZZ]

Quote:

Originally posted by sacd lover It supposedly takes a 350khz sampling rate to make the low pass filter gentle enough to completely eliminate the ringing seen as a consequence of the decimation filter, and best illustrated by the 22khz brickwall filter with 16/44 pcm digital you noted. The nyquist frequency for 350khz would 175 khz. The audioband only needs to extend to a minimum of 20khz. They can then use a very gentle filter roll off and prevent the ringing of the steep slope filters.

Of course with a sample rate of e.g. 96 kHz there is indeed some ringing left, as well as with 192 kHz (to a minor degree because of the flatter and less abrupt filter slope), but it's most likely of no meaning for the sound, because it doesn't affect the audio band: with signals below 20 kHz the ~45 kHz filter resonance isn't excited anymore. Which on the other hand is the case with redbook, where a 16 kHz signal excites the 22 kHz filter resonance and decays with 16 kHz. So I really doubt the need for 350 kHz*... but of course nobody has ever compared 192 to 350 kHz PCM by listening.

Quote:

Originally posted by aos Upsampling 16/44 would sound better because of massive jitter reduction that is achieved in asynchronous sample rate conversion (jitter reduction is well documented in any such chip's datasheets) and also perhaps because of gentler filters with higher corner frequencies and more linear phase response (as sacd lover mentioned).

I don't know about the jitter reduction (some say upsampling increases jitter), but one thing upsampling can't do is allow a gentler filter. It does allow a gentler analog filter (the same applies to common oversampling which most modern DACs are equipped with), and this may be a minor sonic advantage because of the coloring potential of electrical components, but together digital (= over-/upsampling) and analog filter form a normal antialiasing filter with all its shortcomings in terms of phase distortion and ringing.

Quote:

Originally posted by NathanJM Yeah reading all this makes SACD look like it should sould horrible. But to my ears SACD sounds much better than redbook.

Yes, it sounds better.

I never heard DVD-A though.

JaZZ

________________________________________________

* Not to forget: the recording equipment – especially the microphones – will always be an essential (and uncalculable) part of the low-pass filter, so the hope that 350 kHz sampling will cure the ringing by sheer bandwidth extension is probably an illusion anyway.

 

[sacd lover]

Jazz, Wadia had something they called a digimaster filter. The whole purpose of this filter was to emphasize time domain behavoir. Correct time domain behavoir leads to an impulse response with less pre and post ringing, the downside being early high frequency roll off. Wadia claims the absence of ringing results in a more natural and easy to listen to presentation. Increasing the sampling rate moves these impulse cycles so close together ringing is eliminated. Look in any stereophile test page of a cd player; you always see the stairstep waveform with clearly visible ringing.Example: stereophile vol24 no4 page 166 figure 5. This is the impulse response and the ringing is due to the filter. Wadias method curtailed the ringing but allows ultrsonic digital images of the audio data to bleed through and contaminate the analog output signal; so its a trade off. Having a sufficiently high sampling rate solves this without the tradeoff. 350khz was the guess at the minimum sampling rate to avoid ringing; sacds sample rate is clearly more than adequate while pcm at 96/192 is not. The ringing is present on the entire waveform at all times not just near the cutoff AND IT DOES EFFECT THE AUDIOBAND. A filter that starts at 45khz will still ring , just not as much, and this is the point, higher sampling = less ringing, and at some point no ringing. Again this is theory, but I believe a viable one. Note analog will show no ringing because its continuous. This 350khz sampling rate would move the impulse cycles so close as to be continuous. This wears me out, I need a beer; rootbeer that is, a&W in a glass bottle.

 

[JaZZ]

SACD Lover...

...I've always been impressed by the Wadia philosophy and the square waves that really look like square waves, other than with normal CD players. I'm glad you mention Wadia, because it's an example how to (nearly) eliminate the ringing just by applying a gentle filter and renouncing a flat frequency response up to 20 kHz – and this with 44.1 kHz!

I once had a Wadia X32 at home for auditioning, and I liked its very focussed and unglaring sound, but missed some brilliance after all. So I decided for the Theta Pro basic II.

Unfortunately (

) I'm not American and thus not a Stereophile reader. I'm not sure about your descriptions of the measuring diagrams: «stairstep waveform with clearly visible ringing» is almost a paradox – a low-pass filtered signal with the thus induced ringing can show no staresteps anymore...

Quote:

Note analog will show no ringing because it's continuous.

Well, it's continuous and thus needs no filter, but it's filtered anyway: by the bandwidth restriction of the microphones, the analog tape, the cutting process and finally the pick-up/tone-arm/vinyl system forming a quite sharp filter resonance.

 

[aos]

>>I don't know about the jitter reduction (some say upsampling increases jitter),

I haven't heard that before. One of the reasons for existence of these chips is to reduce jitter. Datasheets are chock full of measurements to show that, as well as theory behind it. And even if
that were not the case, you can still make sure that the jitter is low by using a low jitter clock for the output clock generator - since the conversion is asynchronous, and you can make output clock to be (derived from) external oscillator which you can set to be as low jitter as you can buy.

>>but one thing upsampling can't do is allow a gentler filter. It does allow a gentler analog filter

Yes, it allows gentler analog filter. It is possible to have a DAC without either analog or digital filter (like the early version of my portable DAC). With upsampling digital filter is part of the process so it can't be avoided. Some people say that any digital filter - even the best ones - screw up the sound.

 

[JaZZ]

Quote:

Originally posted by aos One of the reasons for existence of these chips is to reduce jitter. Datasheets are chock full of measurements to show that, as well as theory behind it.

You may be right. But can we say that jitter reduction is the cause for the characteristic upsampling sound? I think no. There have been lots of efforts to eliminate jitter in common players, and nobody mentioned outstanding sound with these, at least not a characteristic one AFAIK.

Quote:

It is possible to have a DAC without either analog or digital filter (like the early version of my portable DAC). With upsampling digital filter is part of the process so it can't be avoided. Some people say that any digital filter - even the best ones - screw up the sound.

Are you alluding to the Audio Note concept? Why did you get away from your former filterless DAC? Actually this one would be heavily exposed to aliasing, right? But I guess it's much less harmful than one thinks because the music on CD is low-pass filtered before A/D conversion anyway... or what is your experience in this concern?

 

[aos]

>>There have been lots of efforts to eliminate jitter in common players, and nobody mentioned outstanding sound with these, at least not a characteristic one AFAIK.

I thought it was the opposite? Lot of people upgraded their clocks to low jitter ones and reported substantial changes in sound. I myself did one such modification and was quite happy with the results. But other than clock upgrade, where benefits depend on particular chips used and their configuration, the actual real reclocking that yields guaranteed low jitter at the D/A chip itself is a fairly complex proposition since you have to build a custom PLL from basic parts and have an elastic memory buffer to allow asynchronous reading. I have one such DIY DAC (March Heilingers at al.) and the sound is just marvellous. Modded SACD775 is not bad but this reclocked Redbook was better. But from what I know very few designs of this type exist - I am aware of two DIY designs and in commercial world I heard Wadia was doing that. Anyhow these asynchronous sample rate converters achieve jitter reduction in a different way - first they upsample to a very high frequency (tens of MHz I think) and then they downsample to target frequency. Note that these chips can also DOWNSAMPLE or even do 1:1 (i.e. 44 to 44 kHz), still with same low jitter benefits. Typically jitter reduction in all these digital audio chips (not just sample rate converters) starts only above audio range, while many people claim that the jitter is much more dangerous at low frequencies. By using very high intermediate frequency they are somehow able to achieve similar jitter rejection curves as you can see with real reclockers (which start rejecting it even below audio range, i.e. at several Hz). I really need to read the theory behind it. It would be very interesting to try different sample rates as that would give some clue as to what makes them sound better (or different).

>> Are you alluding to the Audio Note concept? Why did you get away from your former filterless DAC? Actually this one would be heavily exposed to aliasing, right? But I guess it's much less harmful than one thinks because the music on CD is low-pass filtered before A/D conversion anyway... or what is your experience in this concern?

Yes, most people mention Audio Note in this respect. My concern with portable DAC was power consumption so I had to omit digital filter. And I still do, but I have added an analog filter now. And yes, if you look at the signal in time or frequency domain, you can definitely see artifacts in the ultrasonic range since obviously the 2 pole analog filter is not all that steep. However it sounds good regardless - which is probably Audio Note's selling point. One concern here is that non-attenuated artifacts can damage some speakers and amplifiers. I remember reading that SACD uses 50kHz or so filters to get rid of their own artifacts (as their process-related noise rises dramatically with frequency) beucase of these concerns. So I added analog filter for a piece of mind, although anything sold today and advertised for 24/96 or more has to handle wider than audio bandwidth anyway.

 

[JaZZ]

Quote:

Originally posted by aos I thought it was the opposite? Lot of people upgraded their clocks to low jitter ones and reported substantial changes in sound.

I didn't mean to deny that jitter reduction can make the sound better, just that it hasn't produced the same (common) signature known from upsampling DACs.

 

[aos]

>> I didn't mean to deny that jitter reduction can make the sound better, just that it hasn't produced the same (common) signature known from upsampling DACs.

Really? That is really intriguing, can you point me to any discussions on that matter, i.e. comparing low jitter redbook with upsampled redbook?

 

[daniel422]

Quote:

Note analog will show no ringing because it's continuous. -------------------------------------------------------------------------------- Well, it's continuous and thus needs no filter, but it's filtered anyway: by the bandwidth restriction of the microphones, the analog tape, the cutting process and finally the pick-up/tone-arm/vinyl system forming a quite sharp filter resonance.

Don't forget the speakers! Perhaps the ultimate LPF in the chain!

 

[JaZZ]

Quote:

Originally posted by aos Really? That is really intriguing, can you point me to any discussions on that matter, i.e. comparing low jitter redbook with upsampled redbook?

...just don't take my statement too seriously... it's more an assumption because I never read things like «wide soundstage», «less digital sound» or «reminding of hi-rez»... attributes that come to mind from my own upsampling experience and which I read from other people's comments as well. I'm quite sure that upsampling creates a characteristic sonic signature, and I didn't read the same about jitter reduction so far.


Daniel...

...speakers, yes, of course!

And maybe even headphones...

 

[aos]

My experience from de-jittering is that it creates more natural sound, more analog and less digital if you will, improves focus of the imaging, and that it adds depth to the soundstage (so you are suddenly aware of the size of recording space as well as depth-wise position of the performers). Big improvement was for string instruments if I remember. I haven't heard any upsampling DACs yet, that's my next project (i.e. design and make one) since so many people speak highly of them.

 

[Wodgy]

It's worth mentioning that synchronous upsamplers also reduce jitter, though not necessarily as much as a dedicated asynchronous upsampler.