Izotope SRC

[Superdad]

I will say this much, for whatever it may be worth: Especially with modern NOS DACs, it is very possible whatever artifacts are created are not unpleasant. It is likely also true that having relief from the sonic "signature" of 8x oversampling makes listening to an NOS DAC a pleasure.

 

VERY true Jud! My DAC engineer friend (who is also a programmer and chip designer for LSI Logic) constantly regals me about how much and why the filters in most DACs (and DAC chip sets) are the musical downfall of the world. And the trouble is, a digital filter of some sort MUST be used in EVERY sigma-delta DAC-chip-based DAC.

Of course almost all DACs are sigma-delta now because new R2R ladder DAC chips are not being designed and those that exist are not 24-bit (BB1704 being about the best of what is available unless you go the wild and expensive routes like MSB or TotalDAC and built your own R2R resistor network DAC, or gang some chips that were not meant as DACs and create glue-logic to make them work--as Metrum has).

 

The naturalness of NOS (coupled with a great output stage!) is so nice, the artifacts were never a bother. Until you hear them banished with nice upsampling s/w which makes things so easy for the DAC. But again, NOS DACs that can take high bit rates are, sadly, still few and far between.

[pl_svn]

tough question, I guess: what do artifacts sound like? :-/

[Jud]

tough question, I guess: what do artifacts sound like? :-/

 

As I said to a friend in a PM, they cannot be terribly evident or you would know immediately the DAC wasn't very good.

 

I think PeterSt may have said here or in the Phasure forums that he had an Audio Note NOS DAC for quite some time before a "sonic signature" became evident to him and he began thinking of designing his own DAC.

 

I have not answered your question directly, though, so perhaps someone with greater expertise can say something more definitive?

[elcorso]

tough question, I guess: what do artifacts sound like? :-/

 

Like MUSIC...!

 

The proof is a very easy one: When you play well recorded music it sound like (music) heaven. But you play a nasty one it sound like nasty (music).

 

Then, where are the supposed 'artifacts'?

 

Regarding Superdad@ "...or gang some chips that were not meant as DACs and create glue-logic to make them work--as Metrum has...". The only thing I can tell you is, some years ago were in my home some Brazilian pop quartet played music, and they me ask for an 'old style' matchbox, well, one of the gays played the matchbox as a percussion instrument, it was lovely and if you listen to the recording it sounded like a percussion instrument. Those that were no present there always ask me what is this lovely instrument... And the 'glue'? Well, the matching of this four Brazilian players was fantastic, that was the glue!

 

I'm sorry for the OP if out of thread.

 

Roch

[pl_svn]

so... these artifacts are just something that can happen, not something that in some form is (almost) always there, right?

 

as it was :-p with LPs: there may be some dust or a scratch somewhere ;-)

[Superdad]

so... these artifacts are just something that can happen, not something that in some form is (almost) always there, right?

 

as it was :-p with LPs: there may be some dust or a scratch somewhere ;-)

 

We are using the term "artifacts" as shorthand for high frequency digital distortions whose lower harmonics end up in the audible range. This is a shortcoming of filterless NOS DACs when used for 16/44 digital audio. But the methods to avoid the aliasing artifact shortcoming (especially poorly done digital filters) have their own set of shortcomings. Sometimes the cure is worse than the disease. And the disease is caused by the 16bit/44.1kHz CD music delivery standard adopted so long ago. High bit rates mostly eliminate the problems (for NOS and conventional DACs) by significantly raising the frequency at which filters need to start.

 

I tried to find so graphs for easy visualization, but instead came across one of the clearest non-technical posts speaking to the issue of NOS DAC "artifacts." Hope this makes things clearer for you: NOS DAC - Marketing BS?

[Superdad]

Like MUSIC...!

 

Regarding Superdad@ "...or gang some chips that were not meant as DACs and create glue-logic to make them work--as Metrum has...". The only thing I can tell you is, some years ago were in my home some Brazilian pop quartet played music, and they me ask for an 'old style' matchbox, well, one of the gays played the matchbox as a percussion instrument, it was lovely and if you listen to the recording it sounded like a percussion instrument. Those that were no present there always ask me what is this lovely instrument... And the 'glue'? Well, the matching of this four Brazilian players was fantastic, that was the glue!

 

Roch

 

Please, I want it to be clear that I did not in the least mean my reference to the innovative NOS path taken by Metrum to be dismissive. Quite the contrary. He has worked hard to develop a new method of getting an array of chips (my engineer friend explained what he uses and how the glue works, but that is not important here) to perform a bit like a multi-bit R2R ladder DAC, and as I recall, his DAC chips drive the outputs directly with no additional output stage! I've not yet heard one, but I read lots of raves, so my hat's off to Mr. Cees Ruijtenberg.

[pl_svn]

as I recall, his DAC chips drive the outputs directly with no additional output stage! I've not yet heard one, but I read lots of raves, so my hat's off to Mr. Cees Ruijtenberg.

 

correct: up to the Octave there's even no output stage. the new top of the line model, the Hex, instead has output tranformers

(no idea why: but your engineer friend does for sure)

 

you definitely should at least try one ;-)

 

 

thanks for pointing me to that thread =:-)

I'll stay wrong and keep my belief that measures don't mean (almost) a damn. as when a guy who makes tube amplifiers showed me how bad my poor Fatman was measuring and I hardly could resist telling him his almost perfect, on the bench, amplifiers sounded... mortadella! :-p

[elcorso]

correct: up to the Octave there's even no output stage. the new top of the line model, the Hex, instead has output tranformers

(no idea why: but your engineer friend does for sure)

 

you definitely should at least try one ;-)

 

 

thanks for pointing me to that thread =:-)

I'll stay wrong and keep my belief that measures don't mean (almost) a damn. as when a guy who makes tube amplifiers showed me how bad my poor Fatman was measuring and I hardly could resist telling him his almost perfect, on the bench, amplifiers sounded... mortadella! :-p

 

For my understanding no output stage is the best 'output stage". Transformers must be very good to be designed as an output (matching) stage. As an example, a few very good Tube amplifiers are 'transformerless': The output tubes are directly coupled to your speakers input. I never tried one of this since I'm on ELS speakers, with some kind of heavy load, mainly on high frequency, then there should be a matching transformer to couple to them. Of course, the better the transformer is, the better the SQ. Mr. Cees Ruijtenberg found a way to utilize his industrial DAC chips in 'cascade' and those been, with the resulting impedance, a matching as an 'output stage': I figure that less steps can give a better SQ, but I strictly don't know. The only thing I know is the lovely, sweet and direct, then non veiled sound I'm getting now. And I forgot to tell you, my unit has only a week in burn in, and is rarity, achieving 24/192.

 

Maybe the new model, the Hex, couldn't be used that way, since they have 16 DAC chips in cascade per channel, and I guess the resulting impedance is not what they wanted, then they placed a transformer.

 

Like you, I don't believe in measurements, or don't want to rely totally on them, since in my long music listening life I don't get what I wanted relaying strictly on them. For example, if you read The Wine Spectator and choose your wines for their recommendations, you are not guarantee you would get those you like better. They are a lot of people trying to duplicate an Stradivarius violin, they looks and sounds like, but...

 

Again, I'm sorry for the OP.

 

My iZotope choice (of today), 2X Redbook up sampling:

 

Steepness: 1

 

Filter max. length: 500,000 (UT-Gandi, PCM recording)

Filter max. length: 2,000,000 (Solo Violin Sonatas, DSD recording)

 

Cutoff freq.: 1.05

 

Anti aliasing: 200.0

 

Pre-ringing: 0

 

 

Roch

[Jud]

I played around with a single iZotope SRC setting yesterday, pre-ringing, and it was revelatory. Not in the sense of hearing wonderful things I never had before, but rather in the sense of beginning to become acquainted with the way sound changes due to this setting, and by extension due to other filter setting variations, not only in iZotope, but in nearly all DACs. (Of course we have a heavy representation of NOS DAC listeners in this thread, who own DACs whose sound does not depend on upsampling filters.) It is striking to realize we absolutely lack a common vocabulary to communicate the way the digital equipment the vast majority of us are using affects the sound. We are really stuck back in the days of a frequency response centered dictionary - "dark," "bright," no top end, too much top end, too little bass response, etc. The effects of the filter changes made by the pre-ringing adjustment have virtually nothing to do with any such descriptions. I do not have time at the moment, but hope to be able to comment a little further at some point in this thread about language to more accurately describe these sorts of changes.

 

Another point that came home to me while I was listening and testing: I first tested with modern rock music (Richard Thompson's new album Electric, which I think is wonderful, by the way). Then I used a DSD of classical music, the first part of Jordi Savall's version of Bach's sixth Brandenburg Concerto. A couple of months ago I had the privilege of hearing Maestro Savall in concert from about 20 feet away without electronic amplification. While the pre-ringing adjustment I came up with from listening to Richard Thompson was different by only a hair from the final adjustment I settled on after listening to the Savall (.6 with Richard Thompson, .61 after Savall), it did impress me that I was doing two very different things in listening to these two artists. With Richard Thompson, I was aiming for some sort of Platonic ideal of what sounded best or most accurate. With the Savall, I had an actual sound in mind that I could try to meet. And very importantly, I could not. I knew that my system was falling short of the real sound of Savall playing, and that all I could do with this particular adjustment was to try to get as near as the system's limits would allow.

 

There are just so many areas where we are thinking, talking, and working orthogonally to reality - the words we use to talk about the sounds our systems produce; the sound quality goals we aim for when we listen and tweak; the sonic changes our systems are capable of versus those necessary to sound like the real thing.

 

Pretty heavy stuff from just moving a slider around on a preference panel! ;-)

[levandier]

I've been playing around with izotope some more (it's quite addictive!) and I have a couple of observations. First, I noticed that the ringing frequency isn't exactly at nyquist but depends on the cutoff frequency. The lower the cutoff frequency, the lower the ringing frequency (sorry if this was obvious already to everyone and I'm just figuring it out now).

 

Second, I was thinking more about what kind of effect the image energy has on the audible spectrum when I noticed some graphs in the Metrum Acoustics Octave thread:http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/metrum-acoustics-nos-mini-dac-octave-10867/index3.html#post151984

 

So I tried upsampling a 20khz test tone to 88.2khz using various settings suggested in this thread with interesting results:

 

steepness 1, cutoff 1.05, preringing 0 (very similar output to an NOS DAC):

 

 

 

steepness 7, cutoff 1.04, preringing .65 (like a nice compromise between NOS and non-NOS):

 

 

 

steepness 24, cutoff .89, preringing .65:

 

 

 

steepness 27, cutoff .907, preringing 1

 

 

 

 

As always, what's most important is how it sounds, but another interesting data point nevertheless....

[pl_svn]

We are really stuck back in the days of a frequency response centered dictionary - "dark," "bright," no top end, too much top end, too little bass response, etc.

 

... not me: a couple of posts back I wrote about a certain tube amplifier sounding... mortadella! :-p

(I should have said "industrial mortadella" as good mortadella is delicious)

 

very good point, though ;-)

[Jud]

I've been playing around with izotope some more (it's quite addictive!) and I have a couple of observations. First, I noticed that the ringing frequency isn't exactly at nyquist but depends on the cutoff frequency. The lower the cutoff frequency, the lower the ringing frequency (sorry if this was obvious already to everyone and I'm just figuring it out now).

 

Second, I was thinking more about what kind of effect the image energy has on the audible spectrum when I noticed some graphs in the Metrum Acoustics Octave thread:http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/metrum-acoustics-nos-mini-dac-octave-10867/index3.html#post151984

 

So I tried upsampling a 20khz test tone to 88.2khz using various settings suggested in this thread with interesting results:

 

steepness 1, cutoff 1.05, preringing 0 (very similar output to an NOS DAC):

 

[ATTACH=CONFIG]4698[/ATTACH]

 

 

steepness 7, cutoff 1.04, preringing .65 (like a nice compromise between NOS and non-NOS):

 

[ATTACH=CONFIG]4699[/ATTACH]

 

 

steepness 24, cutoff .89, preringing .65:

 

[ATTACH=CONFIG]4700[/ATTACH]

 

 

steepness 27, cutoff .907, preringing 1

 

[ATTACH=CONFIG]4701[/ATTACH]

 

 

 

As always, what's most important is how it sounds, but another interesting data point nevertheless....

 

What would be nice for comparison (and would eliminate having to look back at the linked Metrum Octave thread to get a better idea what is going on) is to show graphs of three items: (1) a test tone produced at Redbook resolution; (2) same as (1) with 2x or 4x interpolation; (3) the same test tone, produced originally at the resolution to which (2) was interpolated. So, for example: (1) test tone on Redbook; (2) test tone on Redbook interpolated to 176.4kHz; (3) test tone at 176.4kHz resolution.

 

Regarding how it sounds, a couple of other comparisons would be useful. You've already got the sort of "worst case" scenario, 20kHz, the test tone closest to Nyquist (for Redbook resolution) anyone is likely to be able to hear. Other test tones I can think of to graph at the same settings would be 15kHz (highest tones from cymbals, piccolo, triangle, so the highest frequencies likely to show any audible change when listening to music); 2048Hz (right in between conversational speech frequencies and higher tones contributing to intelligibility, like fricatives); and 27Hz (just below lowest standard piano key - let's see what happens at the other frequency extreme).

[pl_svn]

show graphs of three items: (1) a test tone produced at Redbook resolution; (2) same as (1) with 2x or 4x interpolation; (3) the same test tone, produced originally at the resolution to which (2) was interpolated.

 

seriously: is this test tones approach anything more than a very simplified attempt at evaluating something way more complex? :-/

 

I mean: if I want to show how good my Tivoli Model Two sounds... all I have to do is play some electronic music. but if I play, instead, just a baroque solo cello... =:-/ (not even thinking about a full orchestra :-p )

[levandier]

What would be nice for comparison (and would eliminate having to look back at the linked Metrum Octave thread to get a better idea what is going on) is to show graphs of three items: (1) a test tone produced at Redbook resolution; (2) same as (1) with 2x or 4x interpolation; (3) the same test tone, produced originally at the resolution to which (2) was interpolated. So, for example: (1) test tone on Redbook; (2) test tone on Redbook interpolated to 176.4kHz; (3) test tone at 176.4kHz resolution.

 

Regarding how it sounds, a couple of other comparisons would be useful. You've already got the sort of "worst case" scenario, 20kHz, the test tone closest to Nyquist (for Redbook resolution) anyone is likely to be able to hear. Other test tones I can think of to graph at the same settings would be 15kHz (highest tones from cymbals, piccolo, triangle, so the highest frequencies likely to show any audible change when listening to music); 2048Hz (right in between conversational speech frequencies and higher tones contributing to intelligibility, like fricatives); and 27Hz (just below lowest standard piano key - let's see what happens at the other frequency extreme).

 

For 15khz and steepness 1, cutoff 1.05 and preringing 0:

 

 

 

For the other 3 sets of settings:

 

 

 

For 2048hz and 27hz, all the settings have enough attenuation at 42khz+ to avoid ripples like in the first graph.

 

seriously: is this test tones approach anything more than a very simplified attempt at evaluating something way more complex? :-/

 

I mean: if I want to show how good my Tivoli Model Two sounds... all I have to do is play some electronic music. but if I play, instead, just a baroque solo cello... =:-/ (not even thinking about a full orchestra :-p )

 

I agree, the proof's in the listening, but I find these tones helpful in understanding how the various izotope parameters affect the sound. For instance, as Jud points out above, 15khz is the highest tone for cymbals, triangle etc.; it would therefore be prudent to design an izotope filter that has sufficient attenuation at 29khz to avoid beats at 15khz and below where it's likely to noticeably affect the sound....

[Freann]

For 15khz and steepness 1, cutoff 1.05 and preringing 0:

 

[ATTACH=CONFIG]4702[/ATTACH]

 

 

I’d like to see what different pre-ringing settings does with all other setting are unchanged.

Like Jud I have a hard time grasping exactly how it influences the sound. I gravitate towards lower settings than 0.6, but I’m not sure what is “right” quite yet.

[elcorso]

--------------------------------

---------------------

Pretty heavy stuff from just moving a slider around on a preference panel! ;-)

 

A simple slider setting and we are closer to the real thing...

 

I could say, my friend: "We are all welcome to the incredible world of digital music playback!"...

 

But (again), thanks Dr. Alexey Lukin & Damien Plisson for those 'sliders'.

 

Roch

 

PS/ On my non NOS DACs the filter settings are very different from those on the Metrum NOS, but absolutely necessary (for my ears & system under A+).

[Freann]

pl_svn: we are not measuring music – we are measuring machines. Measurements are good for understanding how the machine works, one aspect at a time.

 

Playing around with the pre-ringing slider it’s apparent – at least to me – that time is more important than frequency response. The hifi industry has been obsessing about frequency response graphs for the last... 60 years? I’m not saying they are unimportant. It’s time (...) for new ways of visualizing timing accuracy in hifi kit. Especially for cables...

[pl_svn]

pl_svn: we are not measuring music – we are measuring machines. Measurements are good for understanding how the machine works, one aspect at a time.

 

Playing around with the pre-ringing slider it’s apparent – at least to me – that time is more important than frequency response. The hifi industry has been obsessing about frequency response graphs for the last... 60 years? I’m not saying they are unimportant. It’s time (...) for new ways of visualizing timing accuracy in hifi kit. Especially for cables...

 

... right what I was saying: those test only show what happens in frequency response which, I believe, is "the least" important aspect of what our "machines for music" do ;-)

[Freann]

... right what I was saying: those test only show what happens in frequency response which, I believe, is "the least" important aspect of what our "machines for music" do ;-)

 

Problem is – you can’t separate “time” from frequency... They are different aspects of the same thing.

[Freann]

Put another way – pre-ringing never occurs in real life. It only happens in “digital” systems. Pre-ringing lets you know what is going to happen – before it happens. Almost like time-travel... That is why it is so hard to understand it’s influence on sound reproduction.

[pl_svn]

There are just so many areas where we are thinking, talking, and working orthogonally to reality - the words we use to talk about the sounds our systems produce

 

... I just recalled an italian writer's (and gourmet) depiction of a certain wine who (also) said: "... and traces of a fox' wet tail" :-p

[Michael Cullinan]

Listening at

steepness 1, cutoff 1.05, preringing 0 (very similar output to an NOS DAC)

 

To me sounds really good. I see the zig zag graph but have no idea what that means. Linear seems to make the waves more equal and consistent. Maybe real music is unequal and that's why is seems more listenable to me. It comes down to more of a feeling than actual sound study. Maybe just the way I listen.

[Superdad]

I played around with a single iZotope SRC setting yesterday, pre-ringing, and it was revelatory. ....

 

...the sonic changes our systems are capable of versus those necessary to sound like the real thing.

 

Pretty heavy stuff from just moving a slider around on a preference panel! ;-)

 

It was great to wake up this morning to another thoughtful post from you Jud!

And thanks for the thumbs-up on the new R. Thompson album. I have most all of his catalog from about 1983 forward, and I enjoy both his acoustic and electric sides. My wife hates his voice though, so I sneak onto her playlist other people performing his great songs. BTW, "My Soul, My Soul" and "First Breath" (from Front Parlour Ballads and The Old Kit Bag) are good demo and test tracks. I ran into Mr. Thompson at a coffee shop in Santa Monica a few years ago. He lives near where I grew up, and next door to Harry Shearer I think.

 

Anyway, I am glad that you have invested the time to really grok what the pre-ring setting can do for/against the music. As you learned and pointed out, the finer adjustments--while very significant to getting to that ah-ha magic spot--are best heard not by listening "at the sound," but by listening to see which setting gets you closer to that "close your eyes and see the musicians" headspace, the place where the instruments and voices seem most real.

And while yes, it is an overall, does the wall of sound create a semi-real space thing, I also find that piano, brass, cymbals and vocals are quite helpful in bracketing around the various filter adjustments. A very real-sounding piano from a stereo is a spooky thing.

Come to think of it, the instrumental qualities were adjusted more by the settings other than pre-ring. Pre-ring for me really was about the overall focus, though brass (on the Armstrong "St. James Infirmary") really made it easy for me to "see" that change in focus.

 

I look forward to hearing about your experience with the other iZotope parameters.

Be well,

ALEX

[orgel]

The only thing I can tell you is, some years ago were in my home some Brazilian pop quartet played music, and they me ask for an 'old style' matchbox, well, one of the gays played the matchbox as a percussion instrument, it was lovely and if you listen to the recording it sounded like a percussion instrument.

 

Completely off topic, but Gene Krupa does something similar in the excellent film Ball of Fire (directed by Howard Hawks, with Barbara Stanwyck and Gary Cooper). Here it is on YouTube:

 

 

(It's in the encore/reprise at the end of the clip.)

 

--David