Oversampling DAC vs NON Oversampling DAC

[usernaim250]

"The suggestion that oversampling reduces jitter doesn't make sense to me."

 

Whether or not it does make sense to you, or whether it works in practice as promised, the explanation given by Robert Adams of Analog Devices (http://www.theaudiocritic.com/back_issues/The_Audio_Critic_21_r.pdf) makes it perfectly clear that it works in theory. The same material has been presented at AES and to my knowledge no one disputes its theoretical rectitude.

 

As far as whether it works, the measurements of the Benchmark are certainly fantabulous and back up a claim of being jitter-impervious. The sound--well any comments on the Benchmark or any other ASRC DAC are by nature about the whole DAC, not just ASRC. Whether or not it is "perfect" or "as good as it gets," or matches one's taste, there is little doubt that the Benchmark offers a degree of sonic performance that represents genuine and substantial advance over what preceded it on the market at the $1000 price point.

 

[bdiament]

Hi usernaim250,

 

Thank you for the link. I'll have to read the article.

 

Since my initial post on the subject, I have realized that changing the sample rate IS in fact used to reduce jitter in some USB DACs.

 

From my perspective, this is not necessarily a positive trade, to put it mildly. I've yet to hear on-the-fly sample rate conversion that does not brighten and harden the sound of the original source. (Even much of the off-line SRC I've heard has the same problem and it does not have to perform its function in real-time.)

 

Note with regard to that last, I am not talking about oversampling in the DAC.

 

Thanks again for the link. I look forward to reading the article.

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com

 

 

 

 

 

 

[idiot_savant]

I think there should be some clarification on DACs and OS/NOS/Upsampling DACS-

As I understand it:

You can have a NOS DAC - this has no digital filtering or analog filtering, so will have images around the sample rate, making them sensitive to amplifier/speaker matching due to IM issues

You can have an OS DAC - this has a digital filter, removing (hopefully) the images, which should be less sensitive to amplifiers/speakers, but may be more sensitive to jitter

Then you can have an "upsampling" DAC - this probably has an ASRC, which trades between better measured jitter performance against doing poorer filtering vs the OS DAC.

 

Am I a million miles away with my categories?

 

[audiozorro]

Here is another link that discusses NOS, OS and Upsampling DACs:

 

http://www.audioholics.com/education/audio-formats-technology/upsampling-vs-oversampling-for-digital-audio/upsampling-vs-oversampling-for-digital-audio-page-2

 

Whatever choice you think you want to make, at the end of the day it usually comes down to what sounds better, which a generally determined by superior implementation.

 

Thus I believe that these discussions while interesting, are futile since the best NOS, OS and Upsampling DACs have one thing in common, they sound excellent beyond anything else out there. It kind of reminds me of past debates such as which amp sounds better, the best tube amp in the world or the best solid state amp in the world? And what happens when a solid state amp sounds more tube-like and a tube amp sounds more like solid state? Endless debates often ensue with only the possibility that folks may agree to accept the some differences but neither is clearly better.

 

Another easy example for many to explore is trying to determine the best headphones, since they are usually affordable audio components (excluding perhaps the pricey Stax electrostatic headphones). Which is the best, the best AKG, the best Grado, or the best Sennheiser? You could buy all three and it would cost you less than one tenth the price of the best or most expensive DAC. If you're like me you may find that they all sound excellent and different and you may have preferences based on the music you're listening to.

 

You want another example with a clear cut winner? Which sounds better iTunes or Amarra? I'm still waiting for the first poster to claim that iTunes sounds better.

 

 

[idiot_savant]

I'd absolutely agree that the technology/implementation issues must be less important than the final outcome of a DAC - how good it sounds. However, in this case, the technology that is used will influence how it sounds due to interactions with the source material and other components in the chain. For example, a listener with whose favourite material has very little HF content, listening through an amp and speaker that handle HF content very well with very little IM distortion may find a NOS DAC to sound very good - however, put that same DAC into a different amplifier, or change the material being listened to, and the outcome can be very different. From a technical perspective, anything being listened to has already been through a filter ( in the ADC ), which is expecting a reconstruction filter upon playback.

 

I'm sure Amarra is an excellent product, but have heard no valid explanations as yet as to why it should sound any different to iTunes ( assuming volume controls are at full scale, etc )

 

[usernaim250]

"You want another example with a clear cut winner? Which sounds better iTunes or Amarra? I'm still waiting for the first poster to claim that iTunes sounds better."

 

And just what does that tell us? That if you pay $1500 you are unlikely to admit you were duped? That those skeptical of Amarra's claims are unlikely to pay the coin to test it? Or even to bother verifying what they understand to be impossible (that is, different sound with bit-transparent software).

 

And by the way, IIRC CA DID have a poster who said Amarra was excellent--but so was iTunes and that he couldn't pick which was better.

 

 

 

[bdiament]

Hi usernaim250,

 

No one need be "duped" because they can audition the demo of Amarra without spending a cent.

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com

 

 

[usernaim250]

I didn't know about the free trial but people can indeed be duped (actually dupe themselves) whether or not they spend a cent!

 

I'm also deeply suspicious in cases where every report is that A is better than B or that A is the same as B, but none that B is better than A.

 

We know from great experience that whenever there is an audible difference, that differences in taste, hearing, what have you will mean that some people prefer each option. I know several people, for instance, who prefer an Audio Technica AT95 ($40) to a Koetsu Rosewood ($4000). People who prefer old speakers from the 70s to the best of today's high end. Cassette to CD. A Naim Nait ($1400 integrated) to 252/300 ($20k of separates), single ended triodes to what have you, and on and on. CD to vinyl and vice versa. Tubes to transistors, and vice versa. 78s to LP, and vice versa. In audio as in life, many people prefer hot dogs to prime rib, or cookies to cake.

 

So if EVERYONE who claims a difference says that A is better than B (assuming both are reasonable options, which iTunes is), I suspect that they are unduly influenced by non-cognitive factors and that there actually is no or vanishingly little difference. If there actually were an audible difference, at least a few people would prefer the "worse" one, for a variety of reasons.

 

 

 

[KMASCII]

I just couldn't help but say something about the Amarra website. I found the website and read some of the literature there. All I can say is wow... I'm surprised they didn't mention that their software could scientifically altering the molecular structure of audio signals. That would have been cool to see in print. lol

 

To certain extent, I do have to agree with you anim250. People will find emotional attachment to certain equipment, and I'm no different. But I do find myself migrating among different headphones and sources depending on my what I feel like listening to as well.

 

[bdiament]

Hi naimuser250,

 

If you have made your mind up without actually making the comparison, I won't try to change it.

 

However, at least in the situation where one has access to the masters, comparison between playback scenarios can have a valid reference upon which to base conclusions.

 

Playback either sounds like the master or it doesn't. The degree to which it might not will vary with different scenarios. I have found the same situation when evaluating sample rate algorithms or dither/noise shaping algorithms for the mastering room: some folks will "like" the results from certain applications but liking the results or not liking them are separate from how the results compare with the unprocessed original.

 

Skepticism can be a good thing. Actually hearing the items in question can be too.

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com

 

P.S. I should add that I use many applications in my recording and mastering work. Much to my surprise, when I thought all decent apps will sound the same, when I compared a simple operation in all of the programs, I found they all do indeed produce different sounding results. And Sonic Studio's soundblade (Amarra is based on the same code) does indeed sound better, more open and more transparent than anything else in my experience.

 

Further, the hardware associated with Sonic's products is OEM'd from Metric Halo. I have used MH interfaces for many years now and their ULN-8 (basis for Sonic's 304, 305 and Model 4) is easily the most transparent digital hardware I've ever heard.

 

Just my perspective.

bd

 

[usernaim250]

I don't know what makes you think I have made up my mind.

 

What I was commenting on was the logic that would conclude Amarra is better based on a few reports on the internet, which I believe is spurious. For one thing, whether it is new masterings of records or new equipment, people have a strong tendency to report improvement with whatever's new. Often they double back later.

 

The opinion of a mastering engineer like yourself who does have access to masters and can aim to test transparency rather than preference carries a more weight. Nonetheless, unless your tests were blind, I choose to take it with a grain of salt.

 

I haven't heard Amarra so I don't know if it affects the sound when not using its DSP (i.e. just passing bits without alteration). Since I am not a potential customer, I probably won't. I am competent to say that their blurb does not inspire confidence, but I know many if not most blurbs are worthless even in cases where the product is worthy. I also believe that if Amarra does what they say, then soon someone else will do it cheaper and better, sooner rather than later. $1500 is not something I will pay for software with no intrinsic or lasting value. If I were a mastering engineer whose livelihood depended upon it, and with the facilities to confidently test, the value equation might change. But as a consumer, no way.

 

At some point this summer I do plan to set aside time to do blind testing of Foobar, MediaMonkey, itunes on XP, and itunes on a Mac. Maybe I'll even get to it:) I'll report on it of course.

 

[CG]

Personally, I have no problem believing that Amarra has the potential to sound better than iTunes on a Mac. There's other software out there, Play for one, that sounds better, too. Not just a little, either.

 

iTunes is a great data base...

 

Whether or not Amarra sounds better to you in your system is a judgment call on your part. Same for whether it's worth the price they are asking.

 

[bdiament]

It was the tone of your post that made me believe your mind was already made up. I'm glad to hear I was mistaken.

 

I am in full agreement with you regarding conclusions about *anything* that is drawn from what someone on the Internet says. I've often said one can choose anything, whether an audio product or something else, and find an Internet report *somewhere* that says it is "the best".

 

There is no substitute for direct, empirical experience. And yes, my assistant does the switching when I compare things, so I am not aware of whether I'm hearing A or B until after the tests are complete. Not that I have the same perspective as you regarding sighted tests (I feel it very much depends on the particular listener) but just for that added bit of fun, I prefer having my assistant do the switching so I can concentrate on the listening.

 

As to the ultimate value of Amarra or any other product, this of course is a personal call. Whether someone will "do it cheaper and better" remains to be seen. I use four different applications when I master because each does something the others don't, however, while soundBlade is considerably more expensive than the others, I also have yet to hear its sonic equal... at least so far.

 

I would be interested in the results of your listening tests, when and if you do them.

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com

 

[hifitubes]

I'd like to hear any listening test results.

 

I'm planning on doing some myself with itunes, J. River, XXhighend, and Foobar.

 

[Sar]

In D-to-A conversion since you are going from periodic, discrete data to a continuous signal, it is inherent that you have to fill in the gap with fabricated information. WIthout upsampling you are leaving it to the conversion process to decide how to join the dots. With upsampling, you guide the process by inserting samples which constrain the conversion to what you consider more realistic. In either case you are fabricating missing information but if you do it intelligently you can expect the end result to be more constrained towards realism.

 

[gordguide]

" ... You want another example with a clear cut winner? Which sounds better iTunes or Amarra? I'm still waiting for the first poster to claim that iTunes sounds better. ..."

 

iTunes (or more precisely, Quicktime, since iTunes is a front-end for Quicktime) playing back a given file resolution, let's say 16/44.1 as an example, with appropriate hardware setting (16/44.1/0dBFS) and transmitting it via the digital domain to a NOS DAC at 16/44.1 should sound just fine ... there is no modification of the file whatsoever unless the DAC does something before the D/A converter.

 

Yet, there are playback regimens and hardware choices we can make that perform modification of the digital data that are said to sound better. Some of those involve SRC via Quicktime (insofar as QuickTime can access routines made available by Core Audio, which, by the way, has no inherent limit in sample rate).

 

Alternately, you could change the backend for iTunes from QuickTime to a 3rd party application, such as via Pure Music, or eliminate iTunes and QuickTime both from the equation by just using the library of iTunes (in reality, use just the MacOS File structure) and have a 3rd party player and backend, such as Fidelia.

 

It's hardly surprising that the different approaches offer different results, yet if you were to ask each to perform the routine I outlined earlier ... play a 16/44.1 file unchanged ... they all should sound identical.

 

So, clearly, there is nothing inherently "wrong" with iTunes. The differences only appear when we modify the file, and to what extent, and via which routines.

 

Since I don't think anyone here would argue that all DACs sound the same I don't find it difficult to believe that a Software implementation involving data alteration would vary as well.

 

Perhaps it's nit-picking as well, but if you don't like how iTunes does this, it's QuickTime that's the culprit, and even then, more specifically it's the default (Apple-provided, Core Audio compliant) routines accessed by QuickTime rather than anything inherent in QuickTime itself.

 

As for that aspect, Apple has no intention of discouraging the robust audio software ecosystem that has evolved on the Mac. Consider that it's entirely possible to add, say, a licensed version iZotope 64-bit SRC to QuickTime's available routines, but Apple leaves that to 3rd party providers instead.

 

It's shooting yourself in the foot when you render your software developers impotent, and Apple has charted a very careful course here, going as far as it dares with iApps, etc.

 

Instead they decided to provide a robust environment for innovation in audio processing via the implementation of Core Audio system-wide.

 

That we might choose a 3rd party solution for SRC is hardly a surprise.

 

[gordguide]

RE: the link at http://src.infinitewave.ca provided by bdiament on Tue, 06/23/2009 - 16:24 in the comment entitled:

Favorite SRC algorithms

 

I found a comparison between Apple's SRC to iZotrope's interesting.

 

Using Apple Core Audio (Leopard ... same as Snow Leopard's) and your choice of iZotrope/Steep/No Alias) and running through the various signal examples, Apple's comes up looking pretty good.

 

The filters are close ... impulse response and phase are identical, and transition and passband seems even closer to ideal than iZotrope's.

 

The major differences show up in the 1kHz tone and the sweep results.

 

In both cases, although Apple's comes up looking less than idea, they are not onerous, with Apple's sweep response showing artifacts at the -130~150dB range, and the -0.1dBFS artifacts arriving at a peak of about -160dB with no evidence of the -60dbFS tone.

 

It becomes more interesting if you compare Core Audio-Leopard to other routines available.

 

Certainly from the results there is not much evidence that Apple's default routines offer poor performance ... far from it.

 

How that reconciles with what people hear is another issue entirely. Interesting link none the less.

 

[bdiament]

Hi gordguide,

 

"...The major differences show up in the 1kHz tone and the sweep results...."

 

"...In both cases, although Apple's comes up looking less than idea, they are not onerous, with Apple's sweep response showing artifacts at the -130~150dB range, and the -0.1dBFS artifacts arriving at a peak of about -160dB with no evidence of the -60dbFS tone.

 

It becomes more interesting if you compare Core Audio-Leopard to other routines available.

 

Certainly from the results there is not much evidence that Apple's default routines offer poor performance ... far from it.

 

How that reconciles with what people hear is another issue entirely. Interesting link none the less.

 

 

And therein lies the key. In my view, the graphs alone are close to meaningless. Where they take on significance is in listening to the results created by various algorithms, in direct comparison with the unconverted original, while looking at the graphs.

 

What I've found is that what seems "inaudible" from the graphs alone, corresponds with something (I'm not exactly sure what) that is quite audible.

 

Where the graphs suggest Leopard's SRC is relatively benign, the listening tests say something quite different - in view of what can be achieved by the best.

 

Just my perspective, of course.

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com

 

 

[Codifus]

.....worrying about bit-perfection within the DAC is superfluous. If it sounds good, then it probably is whatever is happening.

 

I thought the bit perfect issue was about delivering bit perfect data to the DAC. No?

 

CD

 

[gordguide]

" ... I thought the bit perfect issue was about delivering bit perfect data to the DAC. No? ..."

 

In a word, no (or perhaps "not really").

 

When we're playing back digital audio, we start from a certain point. As consumers, that basically means something like a CD.

 

Given the original is a CD, then the idea of "Bit Perfect" centres around validity of the data we copy to a hard drive. Assuming we can compare the data in a CD versus our copy on the hard drive, and assuming we have a bit-for-bit copy of the audio data itself (ie AIFF or WAV file) then we are assured that our process is error-free ... the transfer didn't alter the data stored on the optical disk.

 

We could take that further ... we could, for example, want a bit-for-bit transfer to the DAC for real-time playback. Again, if we can do this, then we are assured of the ability to make an error-free transfer to the D/A converter.

 

But, is bit perfect absolutely necessary? No, it's not. For example, we could use two identical computers sharing the same equipment to record a live event. Unless some very unlikely happy accidents were to happen, the two versions won't be bit-for-bit identical to each other. Yet, they are both accurate representations of the musical performance; we could choose either one as the basis for our storage and/or further processing.

 

When we process audio for real-time playback, for the most part, we alter the data along the way. Upsampling definitely alters the data, for example. Simply changing the bit depth without altering the musical data (by padding the extra bits with zeros) won't be bit-for-bit identical; the 24-bit version won't pass a checksum compared to the 16-bit version. Yet we have not fundamentally changed the musical data encoded in each version.

 

The ability to make a bit perfect copy is important, as it can validate our process to a certain extent, but it's not the Holy Grail of digital audio storage and reproduction. Indeed, we normally seek to alter the data in high quality playback.

 

"Bit Perfect" can be important ... say, while comparing a lossless codec's ability to reduce file size of an AIFF file. But it's not something that absolutely must be maintained when we seek to improve playback sound quality.

 

As for using the CD as a source, it becomes clear when you realize that the CD is far from a bit-perfect representation of the original performance, since at a minimum it was probably recorded at a higher resolution. So we have to place a point in the process of * performance - mastering - final product - storage (where the CD arrives) - playback * to use as our "bit-perfect" reference. Changing data is not necessarily bad or unwanted. In fact, we are very unlikely to own a commercial disk where the data has not been significantly altered before we got our hands on it.

 

Perhaps borrowing a medical maxim might be helpful ... "first, do no harm". If you can do that, then bit-for-bit becomes unimportant.