Izotope SRC

[vk80]

I downloaded the iZotope RX program and for what it's worth, I don't see where you can resample with tenths decimal. The slider only allows for whole numbers and if you double click to enter a tenths decimal value, it rounds up or down, so perhaps it's more of a iZotope limitation that A+.

 

Also, Pre-ringing only allows for tenths decimal point in iZotope RX. It will round up the hundredths decimal (e.g., 0.65 rounds up to 0.7, and 0.64 rounds down to 0.6); oddly A+ seems to allow for the hundredths decimal and does not round up or down.

 

EDIT: Hopefully this didn't cause too much confusion, but 'resample' above should be replaced with 'change steepness'.

[junker]

Okay, I decided to make a -65dB @ Nyquist filter to make sure the ultra-sonic noise is filtered more aggressively. Also, using a 0.6 pre-ring on this filter as it still doesn't have much pre-ringing. The settings are 20/0.90/0.60.

 

SuperDad just curious if this one sounds better...maybe your system is more sensitive to the >22.05KHz digital noise?

[vk80]

Finally throwing in my wildcard contribution for settings. I personally like how it sounds - natural / non-artificial. It's modeled after the Ayre settings, but taken a bit further (in my mind at least) to A) increase the peak of the pulse and bring it closer to 0 dB and B) minimize as much as possible the pre and post ringing.

 

Steepness: 1

Cutoff: 1.05

Pre-ringing: 0

 

 

At 10k there is a -0.44 dB drop (Ayre's was at -0.75)

 

 

 

At 20k there is a -3.74 dB drop (Ayre's was at -3.2)

 

 

 

Result pulse

 

[pl_svn]

Steepness: 1

Cutoff: 1.05

Pre-ringing: 0

 

with the Metrum Octave I'm getting a very natural and pleasing sound

just... pretty "bold" and... wow basses (maybe too much wow :-p)

 

tried, as usual, only with 2x oversampling of redbook (I don't have 48kHz files and I like how the Octave sounds on its own from 96 upwards)

[pl_svn]

tried a few more different tracks and... yes, very pleasing and natural but definitely too bold and too heavy basses

 

not really how I like it, but something I could live with ;-)

 

oh... and I also had to turn the volume knob down

[junker]

Just maxed out the steepness, and minimized the cut-off frequency to see the effect on a bright Pixies recording but it still sounded bright.

 

Are the advanced parameters for up-sampling even working?

[vk80]

Just maxed out the steepness, and minimized the cut-off frequency to see the effect on a bright Pixies recording but it still sounded bright.

 

Are the advanced parameters for up-sampling even working?

 

Definitely for me. If I move the cut-off frequency to 0.5 I get the 'behind a wall' effect. Treble is almost muted.

 

Apologies for the question, but are you stopping playback by hitting the stop button when you change settings? (Or even closing and reopening the program as a failsafe to ensure settings take?)

[junker]

Thanks for the reply. Usually just stopping playback from iTunes and not killing the application, but I have had to kill it a few times when A+ would play the wrong album/track from what I selected. I'll restart the computer, and maybe power cycle the DAC as well to see what happens. Everything was sounding too bright.

[vk80]

tried a few more different tracks and... yes, very pleasing and natural but definitely too bold and too heavy basses

 

not really how I like it, but something I could live with ;-)

 

oh... and I also had to turn the volume knob down

 

Totally understand. The combo of steepness and cutoff does bring down the upper treble frequencies starting at 10k which would make the bass/midrange sound a little louder. It does sound natural, but a bit bass heavy in retrospect. On the flip side, I messed around a bit and came up with the following which maintains the highs instead of suppressing them. The compromise is there's a little more post ringing, but it's not a lot. For me, I think I will use these settings for most recordings and my previous settings for those that are on the brighter/edgier side [to tame them].

 

Steepness: 7

Cutoff: 1.18

Pre-ringing: 0

 

 

[Alexey Lukin]

What settings would emulate the following?

http://www.ayre.com/pdf/Ayre_MP_White_Paper.pdf

For “the best of both worlds” the settings are: Steepness = 3, Cutoff shift = 1, Pre-ringing = 0. Actually steepness = 3 is slightly too steep, but steepness = 2 is slightly too shallow. The Ayre's filter is in the middle between those two. Here is the comparison of Steepness = 3 vs. Steepness = 2:

 

 

As a side note, Ayre's filter is not really apodizing. Apodizing filters fall to –100 dB at Nyquist, not –6 dB.

 

 

I'm curious why you estimate the steepness so low. Isn't the number supposed to represent the dB attenuation per octave? Clearly I am confused about the steepness curves!

Steepness controls the filter order (something like order = 4*steepness). However unlike IIR filters, FIR filters used in SRC do not have a fixed dB/oct rating in the transition band. You can see that the frequency response curve bends and the slope gets steeper as the frequency rises.

 

 

I think pre-ringing truncates to one decimal place as well...

No, Pre-ringing and Cutoff Shift are floating-point numbers and are only truncated on a screen, not in the algorithm. Steepness is an integer.

[Superdad]

Steepness controls the filter order (something like order = 4*steepness). However unlike IIR filters, FIR filters used in SRC do not have a fixed dB/oct rating in the transition band. You can see that the frequency response curve bends and the slope gets steeper as the frequency rises.

 

No, Pre-ringing and Cutoff Shift are floating-point numbers and are only truncated on a screen, not in the algorithm. Steepness is an integer.

 

Thanks very much for confirming that. After playing with the RX2 s/w and looking at the graphs, it became obvious to me that the steepness number is definitely not a dB/octave figure. And a DAC designer friend of mine explained it to me earlier this evening in much the same terms that you did.

 

As for the Steepness being only an integer, I am not sure I agree--at least within Audirvana +. I am pretty sure that I hear a 0.5 difference when I am bracketing around the perfect spot for my DAC.

 

I am still concerned about the source files that everyone here are using. The ones that Levandier posted a few days ago were mostly 24/96 or 32/96 and need to first be downsampled before upsampling. But the settings used for downsampling affect the graphic results of upsampling. I found this out when I could not get my curves to look the same as his--with the same upsampling settings.

(This also highlights how/why what happens in the mastering process affects what we get during playback--and validates the idea that some recordings will sound best with different playback filter characteristics.)

So maybe we should standardize test tone files--just native 16/44--before going much further.

 

Lastly, I notice that a lot of the graphs being posted here show the upsample rate of 88.2kHz (and I assume the snaps are taken after hitting the process button). Isn't that yet another variable in our kitchen taste test comparisons?

 

Goodnight all. I'll try some your newly suggested settings in the morning when my ears are fresh.

ALEX

[Alexey Lukin]

As for the Steepness being only an integer, I am not sure I agree--at least within Audirvana +. I am pretty sure that I hear a 0.5 difference when I am bracketing around the perfect spot for my DAC.

We used to have a floating-point steepness prior to 2005, but changed it to integer-only since then. Try a null test of a .5 value with nearest integer values.

If there's a demand in finer control of steepness we can consider it for future updates.

[Bob Stern]

Steepness controls the filter order (something like order = 4*steepness).

 

Is that right? If so, steepness=8 is 32nd order = 192dB/octave.

Should that be: order = steepness/4, so that steepness=8 is 2nd order = 12dB/octave?

 

 

However unlike IIR filters, FIR filters used in SRC do not have a fixed dB/oct rating in the transition band. You can see that the frequency response curve bends and the slope gets steeper as the frequency rises.

 

Very interesting. Thanks!

[vk80]

So maybe we should standardize test tone files--just native 16/44--before going much further.

 

I personally have been using lavendier's 44.1 files posted a little later in the thread. Here is that post.

 

http://www.computeraudiophile.com/f11-software/izotope-src-15352/#post211534

 

However, I just found an online generator here:

 

Online Tone Wave File Generator | Sine, Sweep, Noise & more

 

I was able to create and download a nice 16/44.1 test pulse file by making the following selections site:

 

1) Click Pulse

2) Select Sine from 'Type' drop down

3) Enter "0" for 'Level'

4) Enter "1" for 'Width'

5) Do not checkmark the 'q' box.

6) Select 44.1 from the 'kHz' drop down.

7) Click Download .wav file.

 

Here's what it looks like when downloaded (untouched).

 

 

 

Lastly, I notice that a lot of the graphs being posted here show the upsample rate of 88.2kHz (and I assume the snaps are taken after hitting the process button). Isn't that yet another variable in our kitchen taste test comparisons?

 

For the ones I've upsampled, they have all been from lavendier's 44.1 files and I've selected 88.2 to coincide with the 'Power of 2...' upsampling setting in A+.

 

Cheers.

[levandier]

For “the best of both worlds” the settings are: Steepness = 3, Cutoff shift = 1, Pre-ringing = 0. Actually steepness = 3 is slightly too steep, but steepness = 2 is slightly too shallow. The Ayre's filter is in the middle between those two. Here is the comparison of Steepness = 3 vs. Steepness = 2:

 

 

Hi Alexey, first of all, thank you so much for your fantastic SRC and dither algorithms and thank you for licensing them so they can be included in an affordable player like Audirvana!

 

I'm curious, as you are the brains behind the algorithms, do you upsample yourself, and if so, what settings do you use? Also, shouldn't dither be applied after upsampling? (I'm not sure if Audirvana does this and if so, what dither settings are used).

 

Thanks!

 

Oh, and one other quick question. My trial period ends in a couple days. Your site says that izotope rx will then go into demo mode. Does that mean I'll still be able to use izotope just to see the effect of various settings on the shape of the filter curve?

[pl_svn]

For “the best of both worlds” the settings are: Steepness = 3, Cutoff shift = 1, Pre-ringing = 0.

 

... sounds "ok" here. and I mean... just ok :-/

not "dead" but at the same time not "alive". feels as if "something" gets cut away too early

 

 

vk80's suggested Steepness: 7, Cutoff freq: 1.18, Pre-ringing: 0 sounds way more, again, "natural" and alive

and the "boldness" I felt in his first version settings is gone

 

 

(please forgive my total lack of technical speech :-p )

 

 

oh... I'm using Filter max lenght: 1.100.000 and Anti-aliasing: 200 (from Superdad settings)

[levandier]

I have one more question for Alexey. It seems from the description of what the filter length parameter does that the maximum value of 2,000,000 should give the highest quality but several Audirvana users find a lower value to sound better. Can you provide us with some insight about the filter length parameter and what the optimal value is (assuming processor power isn't an issue)?

 

Thanks!

[pl_svn]

Steepness: 7

Cutoff: 1.18

Pre-ringing: 0

 

... still a little "muted" highs (lacking "sparkle"?) but boldness is gone

very alive :-)

[Jud]

As for the Steepness being only an integer, I am not sure I agree--at least within Audirvana +. I am pretty sure that I hear a 0.5 difference when I am bracketing around the perfect spot for my DAC.

We used to have a floating-point steepness prior to 2005, but changed it to integer-only since then. Try a null test of a .5 value with nearest integer values.

If there's a demand in finer control of steepness we can consider it for future updates.

 

First, let me join others in thanking you for this excellent software. Second, with respect to finer control of steepness: As you noted earlier, duplicating the Ayre filter would involve a steepness setting between 2 and 3. So it does appear there are cases where slightly finer control may be desirable.

[vk80]

... still a little "muted" highs (lacking "sparkle"?) but boldness is gone

very alive :-)

 

See if you like these. Takes it a step further to eliminate any high frequency filtering through the redbook cap of 22k... at this point, we're getting maximum available frequency band out of 44.1 files if I understand correctly. Result for me is crystal clear highs.

 

Steepness: 4

Cutoff: 1.28

Pre-ringing: 0

 

 

 

 

@lavendier -- can you explain a little more the impact of the above settings in relation to your post the other day located here: http://www.computeraudiophile.com/f11-software/audirvana-1-4-a-14190/index36.html#post211141

 

I'm still learning, and have two specific questions:

 

1.) How can one develop a graph like those in your post as another check point in what output may look like with various iZotope settings/tests?

 

2.) Does the aliasing of frequencies past 22k have an impact to sound? To be frank, I am not sure if I can tell... plus aren't those frequencies out of the human hearing range?

[pl_svn]

... back to Superdad's Steepness: 7, Cutoff freq: 1.04, Pre-ringing: 0,65 and... these are still, so far, the best settings for my NOS Octave :-) :-) :-)

 

 

@vk80: ooops, missed your post. going to try yours right-away ;-)

[pl_svn]

Result for me is crystal clear highs.

 

 

indeed! :-)

just... may I have some of those bold basses back, please? :-p :-p :-p

 

 

but... wait: now I'm in need of a long walk in the woods to give my ears/brain some rest ;-)

 

"btw": thank you so much :-)

[levandier]

See if you like these. Takes it a step further to eliminate any high frequency filtering through the redbook cap of 22k... at this point, we're getting maximum available frequency band out of 44.1 files if I understand correctly. Result for me is crystal clear highs.

 

Steepness: 4

Cutoff: 1.28

Pre-ringing: 0

 

[ATTACH=CONFIG]4666[/ATTACH]

 

[ATTACH=CONFIG]4667[/ATTACH]

 

 

@lavendier -- can you explain a little more the impact of the above settings in relation to your post the other day located here: http://www.computeraudiophile.com/f11-software/audirvana-1-4-a-14190/index36.html#post211141

 

I'm still learning, and have two specific questions:

 

1.) How can one develop a graph like those in your post as another check point in what output may look like with various iZotope settings/tests?

 

2.) Does the aliasing of frequencies past 22k have an impact to sound? To be frank, I am not sure if I can tell... plus aren't those frequencies out of the human hearing range?

 

1) For the first two graphs, I upsampled the 44.khz pulse train test file to 88.2khz and then selected View:Spectrum Analyzer. For the third graph, I took a redbook audio file and upsampled it to 88.2khz.

 

2) That's a tough question and I don't know the answer :-) If you ask a filterless NOS DAC designer, they'll say it won't impact sound because we can't hear above 20-22khz anyway (or even lower as we get older), and so it doesn't matter that image energy isn't filtered - our ears filter it. But if you ask someone like Dan Lavry, who argues vehemently and very impolitely against NOS DAC's in his white papers, he says that image energy, even though it can't be heard directly, can still degrade sonic performance because it makes your amp, pre-amp, and speakers work harder and can lead to intermodulation distortion in the audible range. Who's right I suppose is a personal thing depending on your sensitivity to things like pre-ringing and the gear in your chain (for example, Stereophile has an intermodulation distortion test in the measurements section of their amp reviews and some amps do better than others).

[Alexey Lukin]

Is that right? If so, steepness=8 is 32nd order = 192dB/octave.

Should that be: order = steepness/4, so that steepness=8 is 2nd order = 12dB/octave?

No, it's right: the order is approximately 4 times the steepness. For FIR filters it's not about dB/oct, it's more about frequency resolution measured in Hz.

 

 

Here's what it looks like when downloaded (untouched).

[ATTACH=CONFIG]4665[/ATTACH]

The reason why you see ringing is the bandlimited interpolation in RX: it plots the expected output of the D/A converter. If you only want to see ringing present in the filter impulse response itself, temporarily change your Waveform interpolation order (found in Preferences > Display) to 1.

 

 

what settings do you use? Also, shouldn't dither be applied after upsampling? (I'm not sure if Audirvana does this and if so, what dither settings are used).

I'm usually too lazy to upsample for casual listening, but when I need to upsample I typically go with settings close to defaults. I think that different SRC settings are a matter of personal preferences and I do not have strong preference for one or another setting. That's why iZotope SRC provides all the settings to tweak.

 

 

My trial period ends in a couple days. Your site says that izotope rx will then go into demo mode. Does that mean I'll still be able to use izotope just to see the effect of various settings on the shape of the filter curve?

Yes, you should be able to open files, analyze the spectrum, change SRC parameters, inspect the curves, and save screenshots in the demo mode.

 

 

It seems from the description of what the filter length parameter does that the maximum value of 2,000,000 should give the highest quality but several Audirvana users find a lower value to sound better. Can you provide us with some insight about the filter length parameter and what the optimal value is (assuming processor power isn't an issue)?

For this parameter the higher, the better. However it only matters with “tricky” combinations of sampling rates, like 44.1k to 192k. When the max filter length is not enough, the output signal may have periodic minor crackle (usually too low to be audible, like –100 dB relative to the signal level). However if you are upsampling 2x or 4x, this should never happen, unless you are using insanely high Steepness. So, for 2x or 4x this parameter should have no effect, which can be confirmed by a null test.

 

 

As you noted earlier, duplicating the Ayre filter would involve a steepness setting between 2 and 3. So it does appear there are cases where slightly finer control may be desirable.

These settings are pretty close to one another already, but we'll consider adding more flexibility for Steepness.

 

Thanks everyone for the feedback!

[Jud]

1) For the first two graphs, I upsampled the 44.khz pulse train test file to 88.2khz and then selected View:Spectrum Analyzer. For the third graph, I took a redbook audio file and upsampled it to 88.2khz.

 

2) That's a tough question and I don't know the answer :-) If you ask a filterless NOS DAC designer, they'll say it won't impact sound because we can't hear above 20-22khz anyway (or even lower as we get older), and so it doesn't matter that image energy isn't filtered - our ears filter it. But if you ask someone like Dan Lavry, who argues vehemently and very impolitely against NOS DAC's in his white papers, he says that image energy, even though it can't be heard directly, can still degrade sonic performance because it makes your amp, pre-amp, and speakers work harder and can lead to intermodulation distortion in the audible range. Who's right I suppose is a personal thing depending on your sensitivity to things like pre-ringing and the gear in your chain (for example, Stereophile has an intermodulation distortion test in the measurements section of their amp reviews and some amps do better than others).

 

I would say the following just on the basis of reading, and heartily invite comments from those who actually know:

 

- There seems to be broad consensus that NOS DACs with 44.1kHz input produce audible artifacts through aliasing/intermodulation. This can be somewhat ameliorated by reducing filter steepness, but this can give rise to various other problems. NOS DAC builders (at least those who do not rely on upsampling of RedBook through software) have decided any artifacts or other disadvantages are exceeded by the sonic disadvantages of upsampling, which has its own unavoidable imperfections.

 

- There is less consensus whether *exclusion* of low ultrasonic harmonics of various instruments (e.g., trumpet, cymbals) detectably alters the sound of those instruments. (Do these ultrasonics intermodulate with other vibrations produced by the instruments to create audible products, for example?)

 

Edit: This last issue has implications for how "leaky," in Miska's term, an upsampling filter ought to be.