Izotope SRC

[Superdad]

Thanks. Can you confirm that izotope and the filter settings is in force only when up sampling is chosen.

 

Yes, the filter is only there as part of the iZotope convolution engine. When upsamling is off you are forced to listen the full effect of your DAC's digital filters. The higher you upsample in s/w, the more you negate your DAC's built-in DF. If you can upsample to 384kHz (if the unit can accept that), you generally are at the native rate of the DAC chip, so the filter is all on the computer. Less than that and it's a combo of h/w and s/w filters--unless you have an NOS DAC.

 

So it comes down to whether you can dial in an upsampling filter with A+/iZotope that delivers a more real and captivating sound than you get from just your DAC (i.e. with s/w upsampling off).

 

But as I detailed in a different thread, recently I had an experience turning a very good s-d DAC into a word class wonder with iZotope tuning--with CDs 2x to just 88.2kHz!

 

Sow's ears won't become silk purses, but since less expensive models most often rely on the chip maker's cheap (or just poorly designed) digital filters, those units may show great leaps if other aspects of the circuits (power supply, clocks, output stage) had some design care and quality.

[Viola]

Any settings for the NadM51 dac?

[jamesg11]

Any settings for the NadM51 dac?

 

Any settings for Mytek dac?!!

[Jeff65]

[ATTACH=CONFIG]5063[/ATTACH]

 

Why do these graphs all show the filter at 22 Khz? When the sampling rate is 88.2 kHz isn't the Nyquist frequency 44.1 kHz?

[Jeff65]

Never mind the last post. I understand now.

 

The filter needs to remove artefacts created above the original sampling rate.

[Jud]

With my NOS/filterless DAC, running either "Power of 2 oversampling" or "Oversampling 2x only", I am currently at:

 

Steepness: 7 dB

Filter max length: 1,150,000 Samples

Cutoff freq.: 1.04 x Nyquist

Anti-aliasing: 200

Pre-ringing: 0.65

 

Last week, into a sigma-delta DAC prototype (which we chose the least bad sounding stock internal filter setting as 'filter-off' was not possible), we preferred Steepness of 21 and filter max length of 1,450,000, with all else the same as above. What a difference that made--even though we were only able to run 88.4kHz into it.

 

After the results from the S-D prototype, I may take another run at settings for my NOS DAC as I really have not tried anything new since 1.4.6. And 1.4.9.x betas are certainly a game changer.

 

Hope that helps you. Best to use your absolutely most realistic instrument discs (and a vocal harmony too) to make the adjustments. Much easier to tune to make piano and percussion sound spooky real than to try to get the gestalt of a average recording just right. Once you get your best stuff making you shake your head, the rest of your collection will sound better too! Except for the majority of recordings which may still shout at you and make you sad. Just the way it is.

 

I've only played with pre-ringing thus far; haven't yet had the desire or time to explore the other modifications. But still, even just with that one parameter, it's very, very interesting how different the filter in your DAC can be from what comes through iZotope alone.

 

The Bifrost, a really fine DAC for its price, has a max input sample rate of 176.4/192 and does its D/A conversion at 352.8/384, so there's always at least one round of internal 2x upsampling. There, after a fair amount of close listening (including well-done recordings of someone I'd heard live the week before, performing on an acoustic instrument), I settled on a pre-ringing setting of .61. Anything higher made things too "dry," as I put it to alcarp. More specifically, the acoustic instrument I heard in concert and in recordings was a viola da gamba, and going higher than .61 made it sound too smooth and linear, no "rosin on the bow," as if it were an electrically amplified instrument rather than an acoustic one. Lower than .61 made things too "sloppy," by which I mean too reverberant in general, and too difficult to locate individual instruments in the acoustic space. Those folks who've gone swimming in indoor pools - remember all the echoing, and how it was pretty impossible to pick out where individual noises were coming from? So think of acoustics not nearly on that extreme a level, but just a taste of it.

 

Last night for the first time I was able to spend some time listening to different pre-ringing settings with a new DAC. This DAC is capable of 352.8/384 input, which is where it does its D/A conversion. So no in-DAC upsampling is required when iZotope upsamples to those input rates, which I have it set to do (power of 2). Though I had to keep the volume down (lovely wife sleeping upstairs), I'm nevertheless fairly confident in the results, because of the time spent and the quality of the recordings. (One of them was Mapleshade Records' Diplogenesis, two acoustic jazz guitarists playing compositions that highlight the sounds of the guitars rather than pyrotechnic technique, so you can get a really good fix on whether what's coming from your system sounds like acoustic guitars. Another was the 24/96 download of Officium by Jan Garbarek and the Hilliard Ensemble, recorded in a monastery.)

 

The criteria were pretty much the same:

 

- Going toward 1.00, at what point were things too "dry"? Where was the reverberation from the body of the acoustic guitar too de-emphasized, leaving only the bare string sound? Where was there too little sense in Officium that the venue was a monastery rather than a regular recording studio?

 

- Going toward .5 (I didn't bother with anything lower - with the Bifrost things got way too sloppy down below .5), where was I simply adding echo and "spreading out" the sounds of individual instruments, making them slightly more difficult to locate in the soundstage, while not making them sound any more natural? (E.g., where did the guitars have as much reverberant "body sound" as guitars do in real life, but no more?)

 

But the conclusion was very different: The pre-ringing setting I've settled on for the time being with my current DAC is .97. (I know, the default is 1.00 and Damien has suggested .6 as an alternative; here I am with settings of .97 and .61, just a hundredth or so off. Why? Don't know, just gotta be that little bit different I guess. And I fancy, though I could easily be deluding myself, that I actually can tell the difference.) Which leads me to wonder about the Bifrost's internal filtering and how heavily it cuts down pre-ringing, for example.

[BobP63]

I may take another run at settings for my NOS DAC as I really have not tried anything new since 1.4.6. And 1.4.9.x betas are certainly a game changer.

 

I look forward to your experimentation of filter settings within 1.4.9.x, for now I'm enjoying your current settings choosing to only up sampling 44.1 and 48 kHz by 2X

 

Thanks again.

[wwaldmanfan]

This subject is extremely technical for me, I just want to get the best sound out of my system.

My DAC (Peachtree DAC-iT) is limited to 44.1, 48, and 96kHz, and does not upsample internally, AFAIK. I've played around with the different iZotope settings, but don't hear a big difference. Anybody with the same DAC willing to give me some suggestions?

[Viola]

Since the Nad M51 takes the incoming signal and upsamples it to 35bit 844khz, is there any advise at what could be better(SRC settings)to do with redbook format? I am playing natively all my hi res. I am only concern with the best way to play CDs on audirvana with NadM51.

 

Thanks guys!

[Jud]

This subject is extremely technical for me, I just want to get the best sound out of my system.

My DAC (Peachtree DAC-iT) is limited to 44.1, 48, and 96kHz, and does not upsample internally, AFAIK. I've played around with the different iZotope settings, but don't hear a big difference. Anybody with the same DAC willing to give me some suggestions?

 

It does upsample internally. (Just about all DACs do, except for a mere handful.) The DAC-It uses an ESS SABRE chip (ES9023) that oversamples by a high amount. I don't remember the exact numbers, but when I looked at the datasheet for the very similar ES9022 it oversampled something like 50x for 44.1/48 rates and 25x for 88.2/96. So if you might get one "round" of iZotope upsampling with 25x in-DAC upsampling, it's not surprising at all that iZotope makes very little audible difference to you.

[wwaldmanfan]

It does upsample internally. (Just about all DACs do, except for a mere handful.) The DAC-It uses an ESS SABRE chip (ES9023) that oversamples by a high amount. I don't remember the exact numbers, but when I looked at the datasheet for the very similar ES9022 it oversampled something like 50x for 44.1/48 rates and 25x for 88.2/96. So if you might get one "round" of iZotope upsampling with 25x in-DAC upsampling, it's not surprising at all that iZotope makes very little audible difference to you.

 

Thank you for your response. I had telephoned David at Peachtree, and he claimed the unit plays files at their native rate. And, the chip does not even support 88.2 kHz. If you try to play a 88.2 kHz file, the input defaults to 44.1 kHz.

What do I know? I think I will switch off software upsampling in Audirvana for now. Like I said, I really can't hear a difference, and I have a pretty good ear. Thanks again.

[Superdad]

I've only played with pre-ringing thus far; haven't yet had the desire or time to explore the other modifications.

 

Hi Jud:

While the rest of your post certainly jives with what I hear when adjusting pre-ring (good descriptions and musical choices BTW!), I am really surprised that you have not yet ventured into adjustments of Steepness, Cut-off, and Filter length.

 

While the nature of the sonic attributes those parameters adjust are different, I find them (especially Steepness) no less important than the pre-ring setting.

 

Steepness is the one to start with, but cut-off is worthwhile (though I am sure you will adjust in a VERY narrow +/- range centered about 1.0 and end up at a one-hundreth step).

 

Exact Max-length is not critical, I'll be surprised if you can hear a change of less than 50,000-100,000. First just get a feel for that one by swinging very wide, then bracket in progressively closer. Ideal filter length setting does seem to depend upon what Steepness you end up on.

 

If you are sensitive to 0.01 pre-ring adjustments, then you will have NO trouble hearing and tuning the other settings. Sadly, the Steepness setting only has granularity to whole numbers, as I would really like a half-step on that one.

 

Have fun and enjoy your new DAC!

 

ALEX

[Jud]

Hi Jud:

While the rest of your post certainly jives with what I hear when adjusting pre-ring (good descriptions and musical choices BTW!), I am really surprised that you have not yet ventured into adjustments of Steepness, Cut-off, and Filter length.

 

While the nature of the sonic attributes those parameters adjust are different, I find them (especially Steepness) no less important than the pre-ring setting.

 

Steepness is the one to start with, but cut-off is worthwhile (though I am sure you will adjust in a VERY narrow +/- range centered about 1.0 and end up at a one-hundreth step).

 

Exact Max-length is not critical, I'll be surprised if you can hear a change of less than 50,000-100,000. First just get a feel for that one by swinging very wide, then bracket in progressively closer. Ideal filter length setting does seem to depend upon what Steepness you end up on.

 

If you are sensitive to 0.01 pre-ring adjustments, then you will have NO trouble hearing and tuning the other settings. Sadly, the Steepness setting only has granularity to whole numbers, as I would really like a half-step on that one.

 

Have fun and enjoy your new DAC!

 

ALEX

 

Hey Alex.

 

There are a few reasons I haven't done anything yet with the other settings.

 

- I haven't had time to listen and confirm the new pre-ringing setting sounds best yet, let alone going on to the other adjustments.

 

- Adjusting multiple settings means interaction (at least I think it does). I don't want to go floundering randomly into that territory. So I feel I need to take some time to study what each setting does by itself and how it affects the others. From reading about pre-ringing, I think this may require both some searching to find sources and some time to digest.

 

- I'm so pleased with the sound that I'm in no hurry.

[Superdad]

Hey Alex.

 

There are a few reasons I haven't done anything yet with the other settings.

 

- I haven't had time to listen and confirm the new pre-ringing setting sounds best yet, let alone going on to the other adjustments.

 

- Adjusting multiple settings means interaction (at least I think it does). I don't want to go floundering randomly into that territory. So I feel I need to take some time to study what each setting does by itself and how it affects the others. From reading about pre-ringing, I think this may require both some searching to find sources and some time to digest.

 

- I'm so pleased with the sound that I'm in no hurry.

 

No sweat Jud. I'm not trying to push you (okay, maybe a little;-)). Certainly understand wanting to have time to listen (I never have enough time to just relax with the music!) , and I am glad you are enjoying the sound you are getting.

 

The only real interaction I found that required a couple of steps of iteration is between Steepness and Max filter length. Even that was not wildly critical (well Steepness is), and I think you will find it all pretty easy. And you will be excited when you get it just right. Be sure to include some very natural piano amongst your circuit of test tracks when setting.

 

I don't think you will find yourself questioning your pre-ring setting at all as you set the others. Personally I found the experience to be a lot like dialing in to Nth-degree the four main tonearm adjustments of a vinyl rig. For the most part, once you get one setting right, you can move on to the others.

 

Cheers,

Alex

[BobP63]

Superdad, since we're talking adjustments anything new on your setup for 1.4.9.x? I'm being a bit lazy and want to leverage off of your ear for now ;-)

[Superdad]

Superdad, since we're talking adjustments anything new on your setup for 1.4.9.x? I'm being a bit lazy and want to leverage off of your ear for now ;-)

 

Nothing new yet. I've been out of town. Will listen and tweak a little later this week. But new family dog may alter that plan!

[Jud]

 

- I haven't had time to listen and confirm the new pre-ringing setting sounds best yet, let alone going on to the other adjustments.

 

So then: Had the chance to listen at normal volume. As you might expect, this had the effect of allowing more reverberance (sound coming up out of silence earlier and taking longer to decay). Listened to quite a few songs, including most of The Gathering, a collection from William Ackerman's Imaginary Road Studios. If the name sounds familiar, it's because he was the guy behind Windham Hill. Lots of natural piano and guitar. So I changed again, but just to .98 this time, and that's where I'm reasonably sure it'll stay.

 

One song in particular, Rudy Perrone's The Prophet, sets off a foreground guitar theme against a background guitar accompaniment, and the rhythm of the whole thing turns on the foreground notes starting and ending in time with the background, which has fewer notes played with more sustain. So if the decay of the background notes is off even a little, it just becomes a tangled mess. With pre-ringing at .98, the rhythm was perfect.

[jhwalker]

After re-reading this thread, I decided to spend some time today rather scientifically testing the effect of the various isotope SRC options in Audirvana Plus.

 

 

Following the good advice of other members of the forum, I chose a well-recorded piano title with which I was very familiar (the Zenph "re-performance" recording of "Rachmaninoff Plays Rachmaninoff"). I well-acquainted with the sound of a quality piano, having majored in music back in the antediluvian days so this was a good choice for me. I plan to re-visit these tests with other sources later on, including voice (I was a vocal performance major), string quartets, jazz piano trios, etc.

 

 

Starting with the "Pre-ringing" slider, I started at 0 / Min. Phase - after all, we don't *want* pre-ringing, do we? I observed the piano sounded a bit "dead" and muffled with this setting, so began by moving the slider all the way to the right (i.e., 1 / linear) - wow, I could really hear the pre-ringing! The piano sound got very congested, especially during complex passages - you could really hear all the ringing, but the piano *did* sound quite a bit more "live" (i.e., less muffled), so I decided to compromise (for now) on 0.5, which seemed to minimize the ringing I was hearing, while still preserving some of the treble energy I liked in the 1.0 setting.

 

 

Next, I moved to the "Cutoff freq" slider. Again, I started on the left at the lowest setting that made "sense" to me (i.e., 0.9 of Nyquist, or around 19.8kHz) - I can't see starting the rolloff much below that, though my hearing only tests up to about 17k these days In any case, I could hear a distinct lack of "sparkle" with the cutoff this low, so I moved it about as high as makes "sense" to me; i.e., about 1.10 - I know we need to filter out the ultrasonic frequencies to avoid foldback artifacts in the audible range, so can't run the cutoff frequency *too* high. I finally settled on 0.99 as my cutoff - again, a balance between the high frequency rolloff I was hearing at the lower setting vs. the artifacts I had begun to hear at the higher setting.

 

 

Finally, I worked with the steepness setting. I couldn't hear the differences as clearly here, but finally settled on 24dB - I reasoned I wanted a steeper curve, since I was starting the rolloff relatively high, at 0.99.

 

 

Finally, I listened to the new settings vs. turning off oversampling entirely. The new settings were clearly much better / clearer, with a very live but controlled sound vs. the virtually lifeless sound with no upampling. I wrote down the new settings so I could easily reproduce them later - definitely an opportunity for Damien to add SRC presets to the interface

 

 

Just to make sure I had everything set optimally, I verified my oversampling settings (I only oversample 44.1k and 48k recordings, with the reasoning that recordings originally recorded at a higher rate do not "need" oversampling), and finally glanced at the Audirvana front panel . . . and realized I had been working all along with a 24/88.2k recording, with no oversampling at all!

 

 

And now you know the rest of the story - it's very easy to fool yourself. I *very clearly* heard the changes I was expecting, though there was no change at all between any of the playback sessions

 

 

Just thought you all might like to have the opportunity for a little levity at my expense

[Jud]

After re-reading this thread, I decided to spend some time today rather scientifically testing the effect of the various isotope SRC options in Audirvana Plus.

 

 

Following the good advice of other members of the forum, I chose a well-recorded piano title with which I was very familiar (the Zenph "re-performance" recording of "Rachmaninoff Plays Rachmaninoff"). I well-acquainted with the sound of a quality piano, having majored in music back in the antediluvian days so this was a good choice for me. I plan to re-visit these tests with other sources later on, including voice (I was a vocal performance major), string quartets, jazz piano trios, etc.

 

 

Starting with the "Pre-ringing" slider, I started at 0 / Min. Phase - after all, we don't *want* pre-ringing, do we? I observed the piano sounded a bit "dead" and muffled with this setting, so began by moving the slider all the way to the right (i.e., 1 / linear) - wow, I could really hear the pre-ringing! The piano sound got very congested, especially during complex passages - you could really hear all the ringing, but the piano *did* sound quite a bit more "live" (i.e., less muffled), so I decided to compromise (for now) on 0.5, which seemed to minimize the ringing I was hearing, while still preserving some of the treble energy I liked in the 1.0 setting.

 

* * *

 

Just to make sure I had everything set optimally, I verified my oversampling settings (I only oversample 44.1k and 48k recordings, with the reasoning that recordings originally recorded at a higher rate do not "need" oversampling), and finally glanced at the Audirvana front panel . . . and realized I had been working all along with a 24/88.2k recording, with no oversampling at all!

 

 

And now you know the rest of the story - it's very easy to fool yourself. I *very clearly* heard the changes I was expecting, though there was no change at all between any of the playback sessions

 

 

Just thought you all might like to have the opportunity for a little levity at my expense

 

I thought something was very strange when you described your pre-ringing settings. Zero is *minimum phase*, which means *maximum* pre-ringing. This is going to sound not at all "dead," but more like the instruments were recorded in a tiled room or indoor pool. I found anything below .5 uselessly "sloppy" for this adjustment, but that was with a Bifrost or no in-DAC filter, and your DAC's filtering will likely differ. 1.00 is *linear phase*, which is not zero pre-ringing as you might have thought, but *equal* pre- and post-ringing.

 

To see what the settings will do in terms of pre-ringing and phase, Google terms like "iZotope" and "pre-ringing," which will give you the iZotope setting instructions. Then read the filter discussion at the Resonessence web site (Digital Filters | Resonessence) to get some idea of how these pre-ringing and phase settings affect the sound. I know Superdad thinks we should do this all by ear - just joking! - but even though the map is not the territory, I find the map does help give me some sense of direction about what I'm doing with these adjustments.

[jhwalker]

I thought something was very strange when you described your pre-ringing settings. Zero is *minimum phase*, which means *maximum* pre-ringing.

 

Are you sure about that? I've been reading up quite a bit on this before I began experimenting and everything I've read indicates the key benefit of minimum phase filters is that *all* the ringing comes *after* the original impulse, so that there is *no* pre-ringing at all. This comes at the expense of phase distortion, but pre-ringing is minimized / eliminated.

 

Or perhaps we're just using different terminology?

[Jud]

Are you sure about that? I've been reading up quite a bit on this before I began experimenting and everything I've read indicates the key benefit of minimum phase filters is that *all* the ringing comes *after* the original impulse, so that there is *no* pre-ringing at all. This comes at the expense of phase distortion, but pre-ringing is minimized / eliminated.

 

Or perhaps we're just using different terminology?

 

Sorry, I was confused/confusing. It's *post* ringing that is maximized by minimum phase. (The energy that would go into pre-ringing can be moved but not eliminated.) So minimum phase sounds even more echo-y and "live" than linear phase.

[jhwalker]

Sorry, I was confused/confusing. It's *post* ringing that is maximized by minimum phase. (The energy that would go into pre-ringing can be moved but not eliminated.) So minimum phase sounds even more echo-y and "live" than linear phase.

 

OK, now we're on the same page! So my thought is that live sound *never* has pre-ringing (how could it?) and that post-ringing is more like real-life acoustics (plus it's mostly obscured by the actual sound); thus my preference for minimum phase filters.

 

That said, I'm fascinated by the whole thing and am fairly obsessed by finding the "best" settings for resampling 44.1 and 48 recordings! This will be enough to occupy me for months

[Jud]

OK, now we're on the same page! So my thought is that live sound *never* has pre-ringing (how could it?) and that post-ringing is more like real-life acoustics (plus it's mostly obscured by the actual sound); thus my preference for minimum phase filters.

 

The ringing we're talking about is all artificial. It's caused by the filtering, not the recording. So when you push all that energy into post-ringing, it's like cranking up reverb to artificially high levels. Not only that, but minimum phase filters have "dispersion." This means the timing between input to the filter and its output varies with frequency. You can imagine what this does to the image stability of an instrument like a piano. Pile lots of reverb on top of that, and you get the effect I described earlier in the thread, where it becomes harder to locate individual instruments/vocalists and the whole soundstage becomes diffuse. Adjusting the slider to higher numbers gets closer to linear phase, which is non-dispersive so imaging is better; and it cuts down on the artificial post-ringing (admittedly at the cost of adding that energy to pre-ringing). It's no accident that the default is 1.00, linear phase.

 

So: You have artificial ringing, which you cannot minimize with the pre-ringing adjustment, you can only move it around. (I'm guessing the number represents the proportion of pre-ringing to post-ringing, with the pre-ringing amount as the numerator and the post-ringing amount as the denominator. So 0 is all post-ringing; 1.00 is equal amounts of pre- and post-ringing; .5 would be 1/3 pre-ringing, 2/3 post-ringing. Would be grateful for any corrections or additional explanation re this or anything else I say here.)

 

If you cannot minimize ringing with the pre-ringing setting, how can you minimize it? By adjusting steepness. The less steep the filter, the less ringing. So why not just minimize the filter's steepness as much as possible? Because then it won't do its job of filtering - or at least it won't filter very much until quite a way above whatever's set as the cut-off frequency. This creates more risk of aliasing ("folding") from higher ultrasonic frequencies, though a less steep filter is less subject to aliasing overall than a steeper one. It also potentially lets more ultrasonic energy through your tweeters, though I don't know how concerned we have to be about that.

 

You could have a gentler filter to minimize ringing and avoid the problems described above by setting the cut-off lower. But if you want to do that and not impact audible frequency response, then you want to push the sample rate as high as you can. When you get to high enough sample rates, anything far enough from the cutoff frequency to fold back into the audible band would be filtered out, even by a relatively gently sloping filter.

 

That said, I'm fascinated by the whole thing and am fairly obsessed by finding the "best" settings for resampling 44.1 and 48 recordings! This will be enough to occupy me for months

 

Months, eh? Filter designers have been at this for years, and they actually know something, unlike most of us. Just sayin'.

 

P.S. This is what I meant when I commented earlier in the thread that all this stuff is related. And I still have no idea what max filter length does in terms of affecting the sound. As mentioned above, would be very grateful for explanations, or corrections of any misstatements, by someone knowledgeable.

[vk80]

Found a little bit of 'nirvana' now that Audirvana supports Audio Unit plugins: Basically emulate the Ayre filters that eliminate pre-ringing and minimize post-ringing. It comes at the expense of upper frequency loss, but that's where the Audio Units come into play; Load up an EQ plugin and gently raise 16k-20k bands to your liking. Bliss (for me at least).

[Superdad]

P.S. This is what I meant when I commented earlier in the thread that all this stuff is related. And I still have no idea what max filter length does in terms of affecting the sound. As mentioned above, would be very grateful for explanations, or corrections of any misstatements, by someone knowledgeable.

 

I'll try to get my engineer friend onto this board to help clarify (he lives and breaths this stuff, including filter design), but in the meantime I can unhelpfully tell you that iZotope's max filter length adjustment varies the number of FIR filter coefficients.

 

While some (including Damien) have said that if you have the computing power, keeping the setting at the max of 2,000,000 is best. But my friend gave me a very technical (at least for me) explanation as to why this is not the case. And while my goldfish memory can not regurgitate his theory, I can tell you that it is easy to hear the adjustment, and that you are all most likely to find an ideal somewhere between 1,100,000 and 1,600,000. It is slightly interactive with where you like the Steepness. I found that the shallower the steepness, the lower I preferred the filter sample max length. For a DAC that likes a 7dB steepness, I ended up at 1,150,000. For a sigma-delta DAC that sounded best at 24dB, we ended up around 1,550,000.