Thankfully mobile phones and other portables are not (yet) driving the 2-channel DAC (and ADC) chip industry to the point where quality devices are no longer economically viable. If you look at the top offerings by Burr Brown, Crystal, and Analog Devices you'll see that these are particularly unsuited to the architectures you sketched.
The cost of these devices is also an order of magnitude above that of the ones used in mobile applications.

But one has to wonder how large the relevant market indeed is, and whether these devices are not created more for fun/love/honour than for financial reasons.
After all DACs and ADCs made on bipolar and BiCMOS processes have been killed stone dead by the arrival of delta-sigma implementations on commodity CMOS,
in the early 90s.

This is more or less niche stuff, and if you look at the turnovers of AD, Wolfson, ... you see that this niche cannot be very significant in the grand scheme of things.
But then again, niches are there to be filled, not?




As the output of a properly constructed DAC is not stepped at all this can't be it, right?

So what then about a bit of harmonic distortion, a bit of intermodulation, and a dollop of frequency-dependent saturation? All things that may make the
sound more agreeable in a domestic replay environment.

Hi all,

Thank you Martin for the quote, I am honored and humbled.

One thing I often repeat in threads about high res on various fora is something I don't see mentioned elsewhere (outside of some private exchanges between myself and a number of colleagues).
To wit, a good many of the converters spec'd for 4x rates (i.e. 176.4k and 192k) actually sound worse at these rates than they do at 2x rates (i.e. 88.2k and 96k). I attribute this to the significantly increased demands on clocking accuracy and on analog stage performance at wide bandwidth. To date, despite the huge number of DACs on the market claiming say, 192k, I've heard only a very small group that seem to be able to deliver on the format's potential.

And in my experience, that potential is great. With the better converters, to my ears, a threshold is crossed and for the very first time in my experience, I'm getting back the sound of my mic feed. This is something I've never experienced in the past, regardless of the price or format (analog or digital) of the recording/playback device. Unfortunately, all too many of the devices on the market don't achieve the performance necessary to really show this.

Also, it does not seem to be stated often enough that the advantage of delivering such wide bandwidth is not merely to deliver sine waves to bats and dogs. Research done by KEF decades ago showed the effects of bandwidth on time response. It is the aural equivalent of achieving more accurate focus and the effect, as far as I can hear it, is not subtle.

Combined with longer word lengths (i.e. 24-bit) the result is to my ears an appreciably larger step up from 24/96 (which does not cross the threshold) than 24/96 is from 16/44.

If I've learned anything in being in this pursuit of audio for the past decades it is that different folks have different sensitivities to different aspects of sound. As such, it doesn't really surprise me that to some, there is no great benefit (if any) in higher res. Just as to some, there are no great differences between different cables or from the use of vibration isolation or clean AC, etc. etc. Happily, for those of use who recognize the differences, they are there to be enjoyed. Those to whom there are no differences can benefit from the cost savings in assembling their systems.

By the way, for those wondering about 24-bit vs. 16-bit, I would suggest taking a good 16-bit digital recording and copying it, dropping the level on the copy by say, 30-40 dB (the range where many harmonics and spatial cues might be encoded). That is, take a recording that peaks near 0 dBFS, copy it at (let's say -36, to see what losing ~6 bits does), the make a copy of the copy, raising the level by the same amount it was dropped for the first copy. Now compare the result with the original. Ideally, the source recording will be rich in instrumental harmonics and spatial cues (something from Keith Johnson would work). When I try this, I find the cellos start to sound more than a little like kazoos. This is exactly what I hear with low level information in 16-bit recordings - even the best of them, especially when a high res version, created from the same mastering, is available for direct comparison. It is almost like the lights in the performance space have been dimmed, a cloud injected into the room's atmosphere and all the instruments replaced with cheap knock-offs of themselves. Switching back to the high res restores the lights, clarity of the air and the quality of fine instruments.

All just my perspective, of course.d

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

Remember that FM radio (BBC) has already come through a 14/10 NICAM ADC/DAC conversion before it is transmitted as an analogue signal - and at only 32kbs sample rate. So if you like FM quality this means automatically that you're happy with less than 16-bits and a sample rate of only 32kbps (15kHz bandwidth). http://www.audiomisc.co....CMandNICAM/History.html gives a nice potted history of FM in the UK.

Hi Togil,



I'm not sure if your post was addressed to me, so please forgive me if it was not.

To the three points you raised, I would say:
1. Roll off in an amplifier (or any other device, whether solid state or "vacuum state", to my ears, compromises performance. But, we all do hear differently and have different sensitivities to different aspects of sound.
2. From my perspective, FM radio at its best is like cassette at its best.
3. While pan potted recordings allow folks (even those with very poor hearing) to point to where a given instrument is, I submit this is not the same as discerning the size or the room and character of the materials from which it is constructed.

What I find striking about the best DACs is not how close CD seems to be getting to high res but how hopelessly behind high res even the best CD remains.

As I mentioned in item #1, my experience has been that different folks have different sensitivities to different aspects of sound. Hence, different responses do not surprise me. I can only report on my own and read that of others with interest.

Regarding the gear I'm using:
My converters (A-D and D-A) are the Metric Halo ULN-8.
Digital files are played in various applications (soundBlade, Peak, Wave Editor, etc. all the way to iTunes) on a Mac computer feeding the DAC via Firewire.
Monitoring is via Magnepan 3.7 speakers, running full range, supplemented by a pair of Outlaw LMF-1 EX subwoofers crossed over at 30 Hz.
Response at the listening position is +/- 1.5 dB with no EQ or anything else between the amps and speakers other than the Nordost cabling I use.
The room is fully treated acoustically and is fed by dedicated electrical lines.
All components, including the speakers and subs are "afloat" on my own Hip Joints roller bearing design for vibration isolation.
The rest of the gear sits in my own design Enjoyyourshelf racks where each item is also "afloat" on its own air bearing, to provide vertical isolation to the horizontal and rotational isolation provided by the Hip Joints.

I use this system in my mastering work (and to evaluate client mixes, etc.). It is extraordinarily good at getting out of the way and letting me hear the source, for better or for worse.

There are a few photos here and here.
There are also some downloadable samples for format comparison between 24/192, 24/96 and 16/44 on the Soundkeeper site here.

On the system in my listening room/studio, the format comparison samples reveal just how much is compromised as soon as one goes down to 24/96, much less 16/44.
Note, however, as I've said before, my experience has been that many DACs do not perform very well at the 4x rates (e.g. 192k), so the comparison samples can be used to evaluate DACs as well. For my ears, with the ULN-8, the differences are such that I hear a great jump upward in quality going from 24/96 to 24/192, than I do going from 16/44 to 24/96. This is the reason why I say a threshold is crossed when the 4x rates are properly done.

All, just my perspective, of course.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

How were these various versions generated?

Have you tried Magnepan 1.7s , they only go up to 22kHz and should therefore not show up the difference between 44 and 192 ...

By the way, from Wiki
480 man-hours of listening tests conducted at the London AES convention in 1980 by Laurie Finchman of KEF concluded that subjects could not distinguish a 20 kHz band limited version of a test signal from the original played back on equipment capable of reproducing sound up to 40 kHz.

Edited by user 15 January 2012 14:31:27(UTC)  | Reason: Not specified

Hi Werner,



Each version is generated in the best way I know how to create it.

The 24/192 represents the original recording.

The best way I know how to create 24/96 is from a 24/192 source, with sample rate converted using iZotope's 64-bit SRC algorithm (to my ears, by far the most faithful to the original of the dozen or more SRC algorithms in my toolbox).

The best way I know how to create 16/44 is from a 24/192 source, with sample rate converted using iZotope's 64-bit SRC and word length reduction provided by iZotope's MBIT+ dither/noise shaping algorithm (also to my ears, by far the most faithful to the original of the dozen or more dither/noise shaping algorithms in my toolbox).

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

I don't hear 12khz any more, possibly not 10. But on my speakers, which are supposidly good out to 60khz, I can as easily hear the difference between a violin sound that is bandwidth limited to 15khz as opposed to one which is not as I did 30 years ago when I first got the LP containing the recording and could hear 15khz. I attribute this to effects outside the audible bandwidth having an effect on things inside it. Timeing effects or whatever; it changes something.

Perhaps you "shouldnt" hear the difference as you go up in sample frequency but the benefits will include moving the brickwall filter further away from the audio band. I believe this is less of an issue as the industry gets better at filter design. See the developments such as the Meridian and DCS Apodising filters. Listen to what happens when you put a supertweeter on a system. And read the article in the Critic (I dont have the copy to hand) on research around our ability to hear higher than traditionally expected frequencies.

I agree with you. Unless you separate the testing of the quality of the replay DAC operating at different sample rates, from the testing of the intrinsic quality of the source material at different sample rates, then you haven't really "proved" anything. These 2 concepts, though seemingly similar, are two completely different things; and I think this is where a lot of the confusion about how low-res compares to high-res creeps in ...

Frank