Izotope SRC

[Michael Cullinan]

Listening to the 13, .91, .3 sounds super natural , organic out of my speakers. Sweet!

[junker]

Listening to the 13, .91, .3 sounds super natural , organic out of my speakers. Sweet!

 

Awesome. Thanks for the confirmation. I'm currently using the almost identical 14, 0.9, 0.3 filter just to make sure the signal above 22.05KHz is down 40dB. Definitely sounds great to me too.

 

I was wondering about how the ultra-sonic noise above the Nyquist frequency comes back into the audible range... So, currently these settings are -40db at Nyquist, and are down to -110dB at Nyquist + (22.05-20 KHz). I think this is where is would start showing up in the audible range because the noise starts to increase coming back from 22.05 at -40 to -110 at 20KHz in the audible range. And would be even lower <20KHz. Does that seem correct?

[junker]

14 / 0.9 / 0.3 Settings & Filtered Pulse

 

 

[levandier]

Awesome. Thanks for the confirmation. I'm currently using the almost identical 14, 0.9, 0.3 filter just to make sure the signal above 22.05KHz is down 40dB. Definitely sounds great to me too.

 

I was wondering about how the ultra-sonic noise above the Nyquist frequency comes back into the audible range... So, currently these settings are -40db at Nyquist, and are down to -110dB at Nyquist + (22.05-20 KHz). I think this is where is would start showing up in the audible range because the noise starts to increase coming back from 22.05 at -40 to -110 at 20KHz in the audible range. And would be even lower <20KHz. Does that seem correct?

 

My understanding is that aliasing is a problem with downsampling where "unwanted energy finds its way into the audio band" because of sampling at less than 2x the rate of the highest frequency content. Aliasing isn't an issue with upsampling, but rather "mirror like" imaging (unwanted ultrasonic energy) and hence why the red portion of the SRC filter doesn't fold back below nyquist when upsampling like it does when downsampling in izotope. Here's a link to a great paper by Dan Lavry on Sampling, Oversampling, Imaging and Aliasing:

 

http://lavryengineering.com/pdfs/lavry-sampling-oversampling-imaging-aliasing.pdf

[Miska]

I was wondering about how the ultra-sonic noise above the Nyquist frequency comes back into the audible range...

 

Through intermodulation...

[Jud]

Regarding different filters for CD vs. hi-res:

 

Barry Diament as well as Keith Johnson have said they feel "4x" rates (176.4 & 192kHz) are where things start to sound real. Johnson and the folks at Spectral built a CD player around certain principles, one being that CD playback will sound more like hi-res playback if "group delay" is minimized.

 

In turn, the Resonessence folks have some very nice info on their web site talking about group delay and several other aspects of filters: Digital Filters | Resonessence

 

Hope this helps folks looking for the best filters for various situations.

[VandyMan]

I don't understand DAC filters too well, but won't most DACs apply a filter to remove ultra-sonic noise before conversion, even if not up sampling? If that is the case, should we use a low steepness so that the DAC has more data to work with for its own filter?

[MemoryPlayer]

Dear Superdad & levandier,

What are your suggestions for iZotope settings using a dCS Elgar Plus at 176.4kHz?

I'm using a MacBook Pro with 16GB and 512GB SSD and Weiss INT 202 Firewire driving the Elgar with two AES/EBU digital adapters, no digital cables at all.

Thanks,

MP

[Jud]

I don't understand DAC filters too well, but won't most DACs apply a filter to remove ultra-sonic noise before conversion, even if not up sampling? If that is the case, should we use a low steepness so that the DAC has more data to work with for its own filter?

 

The conversion is done by a low-pass filter that removes the ultrasonic noise and leaves the music. So it's the filter itself that does what we think of as the D/A conversion. And of course this must be done regardless of upsampling. Upsampling allows the filter's action to be more gentle, which helps avoid audible distortion. But upsampling itself is not perfect, so as always there are trade-offs.

[levandier]

Regarding different filters for CD vs. hi-res:

 

Barry Diament as well as Keith Johnson have said they feel "4x" rates (176.4 & 192kHz) are where things start to sound real. Johnson and the folks at Spectral built a CD player around certain principles, one being that CD playback will sound more like hi-res playback if "group delay" is minimized.

 

In turn, the Resonessence folks have some very nice info on their web site talking about group delay and several other aspects of filters: Digital Filters | Resonessence

 

Hope this helps folks looking for the best filters for various situations.

 

Thanks Jud! I enjoyed reading the Resonessence piece; it's a really nice overview. I'm going to give their apodizing filter a try tonight after the kids go to bed. I estimate: steepness 27 (to get around -96db at 22050 as they describe), cutoff .907 (to get -6db at 20khz as in the graph) and preringing 1, linear phase.

[Michael Cullinan]

The guys who make the Concero are releasing an update soon that will allow upsampling to go to 384hz. More fun! Allowing DXD playback.

[VandyMan]

Thanks Jud. I understand the claims for up-sampling in general, but when additional params like steepness are added in, I get a little lost. Since my DAC is going to do low-pass filtering no matter what (I think it does, anyway), I'm unclear what the final curve would look like. If I use the A+ setting we are discussing, would the output after D/A conversion still look like the graphs posted here or would the low-pass filter in the DAC cause the result to be very different?

[Superdad]

Junker: I just tried your 14 / 0.9 / 0.3 settings. They don't work for me at all!

 

One cut I use is the very demanding St. James Infirmary from Louis Armstrong "Satchmo Plays King Oliver." It has very directional brass, distinct and natural cymbals, clarinet, voice, and then at end everything at once very loudly but with spatial precision. The other cut I used this morning was from Brian Bromberg's "Wood" CD (tons of a very large, plucked acoustic bass, nice piano, and then delicate cymbals that come in).

 

With both of the above tracks, your settings resulted (for my DAC only of course) in everything becoming a wooly jumble. Brass was splayed everywhere, Louis sounded like he had marbles in his mouth (well, more than usual)--it was all a big mushy wall of sound.

 

The biggest culprit for me was the pre-ring at 0.3. Just putting that back to 0.65 helped a ton.

So for the moment I am still back at 7.5 / 1.04 / 0.65 (with Filter Max. Length at 1,100,000)

 

But I am ALWAYS happy to try new things. Thanks,

ALEX

[pl_svn]

So for the moment I am still back at 7.5 / 1.04 / 0.65 (with Filter Max. Length at 1,100,000)

 

... could you, please, check if on your Mini too 7,5 reverts to 7 after quitting/relaunching Audirvana? :-/

[Michael Cullinan]

That's shocking. I thought it was one of the best yet. Weird that it would be so different. Going to check when I get home.

[junker]

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

 

Had very good sound with the -40 db filter, but will try to do a careful comparison to the steeper -110 filter to see if I hear anything strange.

 

... could you, please, check if on your Mini too 7,5 reverts to 7 after quitting/relaunching Audirvana? :-/

[Superdad]

... could you, please, check if on your Mini too 7,5 reverts to 7 after quitting/relaunching Audirvana? :-/

 

Wow, you are right! I had not quit A+ in a day or two (since moving Steepness to 7.5 from 7), so I did not notice. Pre-ring stays at 0.65 after quitting though. So it is just the Steepness that goes back to a whole number. I bet Damien never even considered that some tweaker (oh that sounds bad) would want a fraction of a step for Steepness.

[pl_svn]

Wow, you are right!

 

so it happens both on systems that use "." or "," as decimal separator? I hoped it was just the usual bug when localizing "Made in US" apps (iZotope, not A+) :-/

[Jud]

Thanks Jud. I understand the claims for up-sampling in general, but when additional params like steepness are added in, I get a little lost. Since my DAC is going to do low-pass filtering no matter what (I think it does, anyway), I'm unclear what the final curve would look like. If I use the A+ setting we are discussing, would the output after D/A conversion still look like the graphs posted here or would the low-pass filter in the DAC cause the result to be very different?

 

Yes, your DAC - all DACs - must do low-pass filtering or you would get the unfiltered digital signal, which would sound something like a Niagara Falls of harsh static. (Anyone whose DAC has "lost lock" on a digital signal input can testify that this noise is unpleasant enough to make you bolt out of your chair and rush over to the stereo to turn it off immediately.)

 

The final curve should look as much like the analog input as possible, hopefully. One thing that upsampling should do is keep aliasing artifacts out of that final curve. If you Google for Stereophile, DACs, and aliasing, I'm hoping (though I haven't had a chance to do it myself) that you will find some test graphs from Stereophile DAC reviews showing what aliasing artifacts look like on an oscilloscope trace.

[Miska]

I don't understand DAC filters too well, but won't most DACs apply a filter to remove ultra-sonic noise before conversion, even if not up sampling? If that is the case, should we use a low steepness so that the DAC has more data to work with for its own filter?

 

When you upsample, you move the Nyquist (fs/2) frequency up. DAC's digital filters will leave the newly created space between old and new Nyquist frequency mostly alone, since the DAC doesn't know that it is not being used... (it cannot tell difference between true 24/192 hires content and RedBook upsampled to 24/192)

 

If you use more leaky upsampling filter than the DAC's filter, the DAC will also become more leaky. If you use less leaky filter, the DAC will also become less leaky...

 

DAC's filter will deal only with frequency content above Nyquist (fs/2) of the incoming sample rate. Thus, more you upsample, higher you push the impact of DAC's built-in filter.

 

Of course DAC should always also have an analog filter. Cut-off frequency of this analog filter is almost always fixed. And for modern delta-sigma DACs it is usually low order, typically 2nd order filter.

 

Higher you manage to push the digital image frequencies, more the analog filter will be able to remove those.

 

RedBook content has digital images at 22.05 - 44.1 kHz frequencies, when upsampled to 192 kHz (assuming perfect attenuation) the digital images will only appear at 169.95 - 192 kHz frequencies (and higher).

[VandyMan]

Thanks guys, really helpful.

 

If you use more leaky upsampling

 

Miska, what do you mean by "leaky" ?

 

The final curve should look as much like the analog input as possible, hopefully.

 

Jud, I meant the impulse response graphs and the graphs showing the high-frequency roll-off.

[vk80]

so it happens both on systems that use "." or "," as decimal separator? I hoped it was just the usual bug when localizing "Made in US" apps (iZotope, not A+) :-/

 

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.

[pl_svn]

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+.

 

or, thinking about it, being 1db the smallest audible difference... makes perfect sense only integer values are accepted

 

so the bug is A+ allowing to enter decimals that iZotope doesn't use/save/remember

[Miska]

Miska, what do you mean by "leaky" ?

 

By leaky I mean filter that let's any content between original and new Nyquist frequency. For example for 44.1 -> 176.4 conversion that means frequency band between 22.05 and 88.2 kHz. IOW, non-perfect stop band attenuation.

 

Different filters leak different amounts with varying frequency characteristics.

[Superdad]

or, thinking about it, being 1db the smallest audible difference... makes perfect sense only integer values are accepted

 

so the bug is A+ allowing to enter decimals that iZotope doesn't use/save/remember

 

Sorry, I don't agree. That setting sets the slope (I think it is or should be "X dB per octave" but it may actually be steeper than that based on the curves being posted). So at some distance from the point of inflection, a half-dB per octave (or quicker depending upon what the number actually means) will be much more.

Besides, I find the steps 7.0, 7.5, 8.0 to be quite audible.