Izotope SRC

[Miska]

My thought was this might be like the 44.1 apodizing ADC filter case: Perhaps less ringing or chance of aliasing to start with.

 

I'm not sure I follow... Aliasing is only issue for oversampling in certain circumstances, and that doesn't depend on the apodizing behavior. Most likely apodizing will have less aliasing and digital images due to necessity.

 

If ADC has less ringing/aliasing or is apodizing, then the situation is most likely not affected at all regardless if the oversampling filter is apodizing or not...

 

If ADC let aliases through already, there's nothing much you can do about it later, it is already mixed with the true content.

[Superdad]

Dearest Barrows and Miska:

While I greatly appreciate your explanations, I feel that you make it more complicated than it really is. Jud and I (and probably many others here) are familiar with the main parameters of an oversampling filter--and through long experimentation with the 5 essential SRC filter controls afforded us by iZotope, we have a handle on the effect various settings have--both sonically and visually.

 

 

My questions were with regards to so-called "apodizing" filters, and the notion that they could--when used during playback--reduce the ringing energy embedded in the recording (due to the filters involved in the sample-rate-down-conversion the signal goes through to arrive at 16/44.1). You both talked about the effect and the pros/cons of apodizing and other filter characteristics, but it was Jud who cued me to look again at Ayre's whitepaper which reminded me that in this realm, "apodizing" just means to start the filter roll-off at a frequency less than Nyquist. Woopie-do. Must have been the marketing guys at Meridian who decided to make a big deal of it and call it an "apodizing" filter. Wish they had just called it an "early roll-off" filter 'cause my auto-spell keeps trying to change apodizing to anodizing!

 

Again, I appreciate the illustrations and exposition on the pros and cons of adjusting the amount, shape, and pre/post balance of ringing energy. I am sure it will be helpful to new readers.

 

And I'm with you Miska:

It sure seems like trying to get rid of ringing energy embedded in the recording (due to unavoidable SRC filters--at least for low sample rate material)--simply by rolling off the whole signal early--is throwing out the baby with the bath water. I have never liked cut-off factors of less than 1.0 (I'm at 1.03-1.04)--they pad down the top-end life, though perhaps I should try it will some poor recordings.

 

 

Ayre's "listen" filter is a minimum phase/slow roll off filter (not apodising), as such it is designed to eliminate all pre-ringing, and reduce all post ringing to just one cycle. The trade off is that it has a slight droop in high frequency response, and it allows alias energy through. Ayre's "measure" filter is minimum phase (no pre ring) but with a steeper roll off, allowing for more extended high frequency response, and better suppression of alias products, but lots more (post) ringing energy.

 

Barrows:

Unless the Ayre "Listen" filter is anodizing (rolling off early), why would you say that its minimum-phase and shallow slope settings result in a droop in high-frequency response? As you say, the slow roll-of is letting more aliasing through up higher (though it is also causing less ringing energy of its own).

You are consistent in saying that their steeper roll-off "Measure" filter allows for "more extended high frequency response," but that leaves me with the same confusion.

Update: I just looked closely at the graphs and text of Atkinson's measurements and he clearly says/shows the "Listen" filter rolling off earlier than the "Measure" one. So by definition, Ayre's "Listen" filter is "apodizing"--just not very much.

 

As we all have discovered, the best sounding (okay, most "pleasing"--I don't want to start another truth/accurate/subjective/objective debate!) filter design is a combination of trade-offs, and the ideal settings for each DAC (and source/target sample rates) are likely different. That said, some generalizations are possible. Most critically, in my (and other's) experience, reducing the filter's steepness always seems to sound better simply because it reduces ringing energy overall, and the extra aliasing artifacts that are allowed to stay as a consequence is not a hard trade off to make for the resulting immediacy--regardless of where on the pre/post side you desire to balance the remaining ringing.

 

One thing that surprises me is that Damien has Audirvana's default Steepness setting for iZotope set at 150! I'm at 7 (and found some DACs that sound best at 21-24), and Jud is way down at 3. (Per the iZotope designer, filter order is the setting times 4; so e.g. my 7 is a 28th-order filter; 150 would be a 600th-order filter!) Lot's of people have thanked me for suggesting they dial the 150 way down into the 20s and below.

[Jud]

If ADC has less ringing/aliasing or is apodizing, then the situation is most likely not affected at all regardless if the oversampling filter is apodizing or not...

 

 

Right, which is what I thought the situation might be with hi res - why should there be ringing or aliasing? Or is there commonly ringing, for example, in hi res files?

 

The situation is different for me than you, because you know what you are doing, while I am just fooling around. :-)

 

So because an apodizing filter involves me making more settings that I understand even less well than a non-apodizing filter, and thus I am even less sure about harming the sound in some way, where I don't have to try to use one, that may be the better course. (That or use a filter made by someone who understands what he is doing. Know anybody like that? :-D)

[Jud]

Alex (Superdad) - Can't speak for Miska, nor pretend to really understand much about filters (without the math or scope measurements, we're just trying to use vague concepts without real checks). There were two points at which I read your understanding of what Miska said as different from mine; don't know who's right (if either!).

 

I thought Miska said one could have both extended highs and less overall ringing with good filter algorithm design.

 

I also thought he said any non-apodizing filter must allow through all the ringing from the recording. The only way to reduce that, as I'm understanding it, is with an apodizing filter. That in turn appears to necessitate a cutoff at less than Nyquist, though I could easily be wrong about that.

[barrows]

Right, which is what I thought the situation might be with hi res - why should there be ringing or aliasing? Or is there commonly ringing, for example, in hi res files?

 

The situation is different for me than you, because you know what you are doing, while I am just fooling around. :-)

 

So because an apodizing filter involves me making more settings that I understand even less well than a non-apodizing filter, and thus I am even less sure about harming the sound in some way, where I don't have to try to use one, that may be the better course. (That or use a filter made by someone who understands what he is doing. Know anybody like that? :-D)

 

Jud, and voila, you have hit on what may be the real, biggest advantage of higher sample rates. I know some feel the increased bandwidth is a factor, I really doubt that frequencies way above our ability to hear matter. But, the ability to use filters which do not produce audible artifacts is what matters. Hence I feel their is great value in high res versions of any analog source, despite the fact that some of those sources may not even have any musical content above 20kHz.

As to Damien's recommendation of steep filter slopes, most engineers seem to believe that 100% suppression of alias energy is very important, while many who design filters based on subjective listening observations generally seem to prefer to allow some alias energy through in favor of reduced ringing.

[Superdad]

Alex (Superdad) - Can't speak for Miska, nor pretend to really understand much about filters (without the math or scope measurements, we're just trying to use vague concepts without real checks). There were two points at which I read your understanding of what Miska said as different from mine; don't know who's right (if either!).

 

I thought Miska said one could have both extended highs and less overall ringing with good filter algorithm design.

 

I also thought he said any non-apodizing filter must allow through all the ringing from the recording. The only way to reduce that, as I'm understanding it, is with an apodizing filter. That in turn appears to necessitate a cutoff at less than Nyquist, though I could easily be wrong about that.

 

 

Jud: Don't be so modest, I think you understand more than you realize.

 

My main point was to be clear that an "apodizing" filter is just one that starts at a frequency below Nyquist (the way it is sometimes thrown about I was at first unsure). You already know what those sound like because you have played with iZotope cut-off settings below 1.0 (I recall you trying .98--or maybe that was a pre-ring setting you tried--and ending up at around 1.03).

 

Extending highs and less overall ringing? Sure, you have that too with your shallow Steepness setting. But yes, you let a bit more aliasing in higher up--and folded back down--so you can't get too shallow.

 

Again you are correct about how some are trying to avoid letting the A/D ringing through (really its SRC filter causing it when down-sampling to Redbook) by starting the DAC's filter earlier than Nyquist. But it seems to me (and apparently to Miska as well) that trying to do that is not a world different than trying to conventionally EQ out some other nasties embedded in the recording--albeit with a much higher-order filter and in the digital domain.

 

 

So far, nothing I have have read anywhere has discussed other filter parameters (used in designing h/w or s/w audio digital filters) besides the same five that A+'s iZotope affords us control of. Slope, cut-off, length (i.e. number of taps), maximum attenuation ("anti-aliasing" iZotope's terms), and balance between linear and minimum phase.

 

Like a lot of use here, I am not an engineer (I just fraternize with a few and try to keep learning), so I am certain to be leaving out a number of other aspects of filter design. Of course fields other than low-frequency digital audio have their whole worlds of maths and filter topologies. But for audio, I'd love to know more about other possible and common coefficients used to create digital filters.

[Miska]

Right, which is what I thought the situation might be with hi res - why should there be ringing or aliasing? Or is there commonly ringing, for example, in hi res files?

 

Yes there is, how much the ADC has HF roll-off vs ringing at higher rates depends on the chip. But usually ringing at 2x rates is quite close to 1x rates, because manufacturers want to expand the HF response close to 44.1/48k. At 4x rates ADC chips typically start rolling off early and have less ringing.

 

In case of mastering stage conversion from 4x (or higher) to 2x rates, the amount of ringing is usually the same as for 1x rates to keep the extended bandwidth.

 

Of course, all else being equal, every time you double the sampling rate, length of the ringing (in time) is half and the frequency of the ringing is twice higher.

 

So as a summary, filter that is apodizing at 1x rates tends to become non-apodizing at 4x rates in regards to ADC chip filters.

[Miska]

As to Damien's recommendation of steep filter slopes, most engineers seem to believe that 100% suppression of alias energy is very important, while many who design filters based on subjective listening observations generally seem to prefer to allow some alias energy through in favor of reduced ringing.

 

And I think the real art is in finding a design method that minimizes energy of ultrasonic images (these are not really aliases), while at the same time minimizing ringing, plus keeping the top octave untouched... IOW, getting as close as possible to the impossible.

 

Rest is about finding nice parameters for the design. Plus how the filter applied (recursive, single-pass, etc)...

 

(in my earlier posting you can see response of the min-phase filter I personally use for listening)

[Superdad]

Miska:

 

It has been a long time since I visited your web site. Reading it just now I was surprised to see that you offer an OS X version. Downloading the trial now!

 

I am intrigued by what you refer to as delta-sigma modulators for upsampling to 1-bit/24.576mHz. How does one feed that to a DAC? Or like a DSD's lower rate bitstream, does one just run it through a low-pass filter of some kind--no "DAC" needed?

 

And for more traditional upsampling I see you offer a wide range of filters which you have no doubt carefully tuned.

I like that you include a convlover that accepts DRC files from other apps. I've been working up a plan to try out REW and Acourate.

 

I'm confused about your Network Audio Adapter. Is that hardware or software? And why is it listed as system requirement only for the Mac OS X version HQPlayer? Please explain more about that.

 

Sorry to be off topic everyone (I bet there is a whole other thread devoted to HQPlayer here on this very forum, right? Guess I should go look.)

I'm just now getting very curious to listen to HQPlayer. You all know me as a big fan of the performance and custom control afforded by Audirvana, so Miska's s/w has a high bar to reach. But I am very open minded and the feature list for the player makes it clear that his is a VERY serious and capable product. I'm not into multi-channel or DSD at this time, but even without launching it there looks to plenty else of interest to me.

 

Goodnight,

ALEX

[Miska]

My main point was to be clear that an "apodizing" filter is just one that starts at a frequency below Nyquist (the way it is sometimes thrown about I was at first unsure).

 

That definition doesn't cover it yet. It is a filter that filters a filter. So, filter A is apodizing if result of filtering impulse response of filter B results in impulse response of A. Thus it depends on all parameters rather than just the corner frequency.

[Miska]

It has been a long time since I visited your web site. Reading it just now I was surprised to see that you offer an OS X version. Downloading the trial now!

 

Just as heads-up, it currently supports only network audio adapter output ("NAA"). CoreAudio support didn't make it into 3.0 release schedule, but it's planned.

 

I am intrigued by what you refer to as delta-sigma modulators for upsampling to 1-bit/24.576mHz. How does one feed that to a DAC? Or like a DSD's lower rate bitstream, does one just run it through a low-pass filter of some kind--no "DAC" needed?

 

It is essentially "DSD512", but at 48k base rate. So it can be converted the same way as DSD, but since it's higher frequency you don't need as steep low-pass filter or alternatively you get much less out-of-band noise. DSD512 should work with a DAC like Yulong DA8 (to be verified) or with a suitable DIY converter. Of course the "real DSD512" is also supported at 22.5792 MHz.

 

I'm confused about your Network Audio Adapter. Is that hardware or software? And why is it listed as system requirement only for the Mac OS X version HQPlayer? Please explain more about that.

 

It is a minimal Linux installation on PC or ARM-based platform like CAPS/FitPC/Alix/SolidRun CuBox with a small server process that sits between network and audio device(s), like USB DAC or built-in S/PDIF of CuBox, or I2S of some other ARM platforms. I offer the necessary server process "networkaudiod", while the Linux installation (Debian Wheezy) needs to be done separately.

 

HQPlayer can use these as audio output devices, just like any local device. Output to those "networked DACs" is included in all platforms in addition to local playback, but on Mac it is currently the only option... Good side is that it bypasses CoreAudio completely.

[amit]

Hello ,Is the discussion here apply only to upsampling or downsampling can benefit also from the suggestions given here?

[Miska]

Hello ,Is the discussion here apply only to upsampling or downsampling can benefit also from the suggestions given here?

 

For downsampling (decimation) you need to pay special attention to filter parameters, because any leakage will cause unsuppressed frequencies above target Nyquist frequency to alias (fold down) to audio band and get mixed with the real content. Changing/removing filter ringing from previous stage is not an issue because it will "always" get replaced in downsampling.

[Superdad]

Hello ,Is the discussion here apply only to upsampling or downsampling can benefit also from the suggestions given here?

 

Wellcome to the CA forums Amit!

Whether upsampling or downsamplimg, a digital filter is, by definition, imposed on the signal. And the same issues of slope, cut-off, and ringing exist. In fact, the iZotope sample rate converter that this thread concerns is more often used in the music production side--as part of the firm of the same name's popular iZotope RX waveform editing/processing suite--for down-converting. (It is available as a trial download and is fun to play with because you can graphically see the effects of imposing different SRC parameters, and if you do it with a pulse waveform file you can even see the ringing change with an individual impulse.)

 

One question you do bring to mind is if Audirvana Plus uses the iZotope engine to automatically downsample if you try to play a track recorded at a higher sample rate than your DAC accepts. My DAC's input accepts up to 192kHz streams, and I never tried to play hi- res material with A+ into a DAC that couldn't accept it. I'm sure some else here will answer that--and I'll try it later for myself with another DAC I have lying around.

[Superdad]

Just as heads-up, it currently supports only network audio adapter output ("NAA"). CoreAudio support didn't make it into 3.0 release schedule, but it's planned. snip.........

...........

........HQPlayer can use these as audio output devices, just like any local device. Output to those "networked DACs" is included in all platforms in addition to local playback, but on Mac it is currently the only option... Good side is that it bypasses CoreAudio completely.

 

Well, Debian is available in a build for Mac Intel platforms (heck, there is a Power PC version--I could revive and put to use one of my old G4 minis!), so I do have devices available to make a NAA for HQPlayer.

 

I am more intrigued by your sigma-delta upsampling to feed PCM files to DSD 512-capable DACs. If it will work with Sonore's forthcoming DSD board (which does not use any conventional DAC chip at all)--either in their Rendu or in their USB converter--that would be VERY unique! 100% DAC-chipless PCM playback--or rather a pure software DAC.

Something new to daydream about...

[JohanS]

One thing that surprises me is that Damien has Audirvana's default Steepness setting for iZotope set at 150! I'm at 7 (and found some DACs that sound best at 21-24), and Jud is way down at 3. (Per the iZotope designer, filter order is the setting times 4; so e.g. my 7 is a 28th-order filter; 150 would be a 600th-order filter!) Lot's of people have thanked me for suggesting they dial the 150 way down into the 20s and below.

Indeed Superdad, here is another THANK YOU for pointing out the Steepness setting to me. I'm currently down to 24, and that pleases my ears a a lot!

[kurb1980]

With my DAC I find that steepness at 3 cutoff 1.02 and pre-ringing 1.0 in combo with my DAC's filter yields a ideal sound I can listen to for now! I have experiments with steepness with 1 thru 20, cutoff 1.0 thru 1.02, and pre-ringing 0.0 thru 1.0! Captured all my graph settings and sound profile changes in my note pad.

 

44.1kHz test tone upsampled to 176.4kHz

[JohanS]

(partial repost from the A+ 1.6 wishlist thread)

 

My thoughts about upsampling have changed since I read comments of Jud, Superdad and others in this and other threads about the SRC Izotope upsampling in A+. I choose 'Power of 2 upsampling only (2x, 4x, ...). After altering the advanced parameters I'm hearing an easily discernible sound quality improvement! No doubt the output is not 'bitperfect' to the original 16bit/44.1kHz waveform. But this waveform on the CD is certainly not resembling the original analogue waveform!! It is a sample itself. Probably the resulting upsampled waveform does fill in the gaps and is more like the original analogue waveform than its lower resolution digital nephew on CD. Most importantly, my ears are more pleased while listening to the music that is upsampled by the Izotope upsampling algorithm.

 

 

With regard to the advanced parameter I'll briefly sum up the results.

 

 

Steepness: I'm on 24 now.

To my ears a grand piano sounds clean and direct with a lower setting and sounds like there's a curtain between me and the piano with a higher (150) setting. A setting of 3 (Jud's setting) was a bit too much. Sound became sometimes a bit 'harsh', so I gave the 'twenty-something' setting of Superdad a try. Music sounds just more relaxed now.

 

 

Filter max length: 1.600.000

I think I heard less of the acoustical environment with a very low setting (reverb was a bit muffled?); as the 2.000.000 gave some hick-ups I just put it on 1.600.000 with no 'acoustical' reason. If someone has a suggestion: feel welcome!

 

 

Cutoff freq.: I'm on 1,05 and consider 1,04

Now, this thing is really something! I strongly encourage everybody to try a setting just above '1,00'. It broadens the soundstage (wider / deeper) and better separation of instruments. I guess that's because more second, third, ... order harmonics come through (as a setting higher than 1,00 lifts the cutoff frequency above 22 kHz if I've understood correctly). However 1,06 was too much, the sound became smeared (instruments start to lose there focused position within the soundstage).

 

 

Anti-aliasing: I kept 200, never touched it.

 

 

Pre-ringing: I have yet to figure this one out. I'm on 0,7 now.

I hear a difference on some recordings between 0 and 1, but I cannot describe it properly sonically. I choose 0,7 based on Superdad's advice (0,65) and Damien's preference (0,73). Thought 0,7 was nice somewhat in the middle (maybe I should try 0,69 :-)).

 

 

Anyway, in the end it's all about enjoying the MUSIC and not about figures, filters and difficult technical stuff ;-).

[PhilipAC]

Hi everyone

 

Am I right in assuming that these iZotope settings have no effect if I use "DSD conversion to PCM" with a DSF file and listen on PCM 176.4 as produced by A+?

 

Thanks

[elcorso]

Hi everyone

 

Am I right in assuming that these iZotope settings have no effect if I use "DSD conversion to PCM" with a DSF file and listen on PCM 176.4 as produced by A+?

 

Thanks

 

Yes they have it:

 

Up sampling in PCM

 

Down sampling from DSD to PCM. (it's on A+ user manual)

 

Roch

[Jud]

(partial repost from the A+ 1.6 wishlist thread)

 

My thoughts about upsampling have changed since I read comments of Jud, Superdad and others in this and other threads about the SRC Izotope upsampling in A+. I choose 'Power of 2 upsampling only (2x, 4x, ...). After altering the advanced parameters I'm hearing an easily discernible sound quality improvement! No doubt the output is not 'bitperfect' to the original 16bit/44.1kHz waveform. But this waveform on the CD is certainly not resembling the original analogue waveform!! It is a sample itself. Probably the resulting upsampled waveform does fill in the gaps and is more like the original analogue waveform than its lower resolution digital nephew on CD. Most importantly, my ears are more pleased while listening to the music that is upsampled by the Izotope upsampling algorithm.

 

Anyway, in the end it's all about enjoying the MUSIC and not about figures, filters and difficult technical stuff ;-).

 

Well, you're right, it's about the music. Still it is worthwhile to say a couple of words about the technical side, so far as I think I understand it (relying as always on those more knowledgeable to correct significant errors).

 

- Bit perfect: There are very, very few DACs that can send a bit perfect signal to the D/A conversion stage - the Metrum DACs, the Audio Note kit DACs, PeterSt's Phasure DAC (but only if it is not used as intended, at least with Redbook material), perhaps a couple of others. If you do send a bit perfect Redbook signal to the D/A conversion stage, you will get distortion at fairly high measured levels. Nevertheless, many people like this sound, so there are purchasers for these DACs.

 

But most of the DACs in the world use chips that "oversample" or "upsample" or "interpolate" (same thing, different words - the last is most precisely correct), usually by 8x or more, before doing D/A conversion. So the issue isn't whether one should be bit perfect, since almost no DACs are capable of it and you get distortion anyway. The issue is whether to do the upsampling in the DAC or in software prior to the DAC, or partially in both. Since most DACs have a max input of 4x rates, 176.4/192, but internally use 8x rates, 352.8/384, it will usually be the case for those of us using iZotope that upsampling will take place both in software and the DAC. Thus if you have a Redbook source, you might use iZotope to upsample to 176.4, then your DAC internally upsamples to 352.8.

 

- "Fill in the gaps":

 

Probably the resulting upsampled waveform does fill in the gaps and is more like the original analogue waveform than its lower resolution digital nephew on CD.

 

That is what most people intuitively believe would be the result of upsampling, but it isn't correct. What Nyquist, whose name appears in the iZotope settings, proved mathematically is that as long as you are sampling at more than twice the highest "frequency of interest" - in the case of digital audio, that you are sampling at more than twice the upper limit of human hearing, 20kHz - you can perfectly reconstruct the wave you were sampling. Perfectly. No such thing as "filling in the gaps" in the waveform, the waveform is already fully filled in at Redbook sampling rates.

 

So why upsample? Because the perfect reconstruction depends on perfect filtering. Perfect filtering requires two things in short supply in the real world, instantaneous response (not merely brief, but no time at all) and infinite time. What happens when our necessarily imperfect filters meet the 44.1k sample rate? The sorts of things we've talked about in this very thread in terms of trying to design our own filters, such as aliasing and ringing. Higher sample rates allow filter designs that produce less of these problems while not having to cut off as much of the high frequency response. For example, all other things being equal, a less steep filter will produce less ringing. That is why the chips in almost all DACs upsample to at least 8x rates - to allow their built-in filters to produce a better response.

 

When we upsample with iZotope, what we are trying to do is produce even less of this nastiness - aliasing, ringing, etc. - than the filters built in to our DACs, while maintaining good high frequency response. We have the advantage of iZotope's excellent filtering algorithms, the additional processing power and speed a modern CPU provides over a typical DAC chip, and the fact that we are willing to carefully listen. We have the disadvantages (most of us, anyway) of knowing nearly nothing about filter design and not having good lab measuring equipment to check our results.

[VandyMan]

Probably a dumb question, but is there any reason to take your speakers crossover slope into consideration when setting the steepness in iZotope?

[pl_svn]

OT: you all bad guys (including me) remember and pay close attention in the future at keeping anything that begins with "iz", "up" or "ov" in this specific thread only, or... =:-p

 

sorry, but... OMG!!!

(can't remember: God created cops on... which day was it? ;-)

 

end OT

[Jud]

Johan, for a little discussion on a very elementary level (the only level I know!) about the various settings and what they do, have a look at Semi-Customized DAC, Part IV: Photos and Filters - Blogs - Computer Audiophile and the sixth comment, from Superdad, correcting an error I made in my discussion of the steepness setting.

 

By the way, with regard to the iZotope settings in the blog post: I'm trying a cut-off frequency of .99. Preliminary results are positive. There are audible differences, but whether they are for the good or otherwise is often not easy to tell. What makes me think this may be a positive difference is listening to two tracks from Jakob Dylan's first solo album, Seeing Things: "Everybody Pays As They Go," and "This End of the Telescope." With cut-off of 1.02 and .99, listen to the sustain on the guitars - is it the right amount, too little or too much? Also listen to vocals, both lead and background. Is there enough "body," as well as "air"? (While so far I think the results are positive, this could also easily be power of suggestion after the discussion about apodizing filters.)

 

My settings likely will not work for everyone. I'm able to avoid the internal DAC chip's 8x interpolation filter entirely. Other system differences, such as in speakers, will likely also make a difference to the results.

[Jud]

Probably a dumb question, but is there any reason to take your speakers crossover slope into consideration when setting the steepness in iZotope?

 

Don't know about steepness and crossover slope, but I think the pre-ringing setting and the phase characteristics of the speaker's crossover could be related. E.g., Vandersteen's crossovers are made to work with the "time aligned" nature of the speakers, and a setting other than linear phase, to the extent it is "dispersive" (the time through the filter changes with frequency) could, I speculate, mess up time alignment.