Izotope SRC

[barrows]

Actually, it is a mind-boggling subject. :-) There are very few DAC's that either do no upsampling of their own (NOS-type DAC's) or let you turn the upsampling off. Therefore, as Jud said, what you hear will always be the result of an interaction of the two upsampling/filtering processes. The only way to know what this effect is, is to listen. There are people who claim that this filter "stacking" is not a good thing. I have a Resonessence Concero which allows you to switch off all internal upsampling/ filtering so I am able to evaluate upsampling within my music player without additional filter interactions in the DAC. As you said, an interesting subject.

 

Axiom: the Concero uses an ESS DAC, right? If so you cannot turn off all the oversampling, even with the ESS' OSF set to "off" the DAC chip still does another oversampling pass which is not user defeatable. But, i suspect that this pass occurs at such a high rate already, that what you hear when the OSF is "off" is just the effect of software oversampling and its filters. It is hard to figure this stuff out with the ESS chips, as the data available on how they work internally is incomplete (even on the data sheets if you can get your hands on one), but a few hours on Internet searching will yield much information.

[Axiom05]

Axiom: the Concero uses an ESS DAC, right? If so you cannot turn off all the oversampling, even with the ESS' OSF set to "off" the DAC chip still does another oversampling pass which is not user defeatable. But, i suspect that this pass occurs at such a high rate already, that what you hear when the OSF is "off" is just the effect of software oversampling and its filters. It is hard to figure this stuff out with the ESS chips, as the data available on how they work internally is incomplete (even on the data sheets if you can get your hands on one), but a few hours on Internet searching will yield much information.

 

You may be correct, I really don't have a great understanding on how these DAC's operate internally. I thought that the actual D to A conversion takes place at much higher sample rates, e.g., MHz, but this is different from what we usually refer to as "upsampling" and this may be what you are referring to. All I know is that the Concero (ES9023 DAC) has three filter choices: 1) no filter (blue light) where data is passed bit-perfect to DAC (re. Mark Mallinson, 6Moons review), 2) 4X-upsampling with the IIR filter (magenta light) and 3) 4X-upsampling with "apodizing" filter (magenta light). The two 4X-upsampling filters are implemented in a FPGA, not on chip. If I perform upsampling in the music player, the light on the Concero is blue irrespective of which filter has been previously selected. With upsampling off in the music player, I can select among the three filter choices in the Concero. If there is additional upsampling within the DAC itself, there must not be any high-pass filter as this is done by the 4X-upsampling process, otherwise you could not take advantage of the lack of pre-ringing in the IIR filter. The "no-filter" choice on the Concero is not available on the Mirus/Invicta. If I have this information wrong or mixed up, I welcome someone to straighten me out.

 

Cheers,

Bob

[Neil]

To up sample or to not up sample? I have an Emotiva XDA2 and see most people seem to select power of 2 upsampling in A+ preferences. My DAC is capable of 192k should I select max upsampling?

[wwaldmanfan]

To up sample or to not up sample? I have an Emotiva XDA2 and see most people seem to select power of 2 upsampling in A+ preferences. My DAC is capable of 192k should I select max upsampling?

 

I am one of the least technically-minded people here, but I will attempt to explain the rationale behind this. If you think of a "bit perfect" sample as a square wave, power of two upsampling multiplies it geometrically, while preserving the shape. The "in-between" settings, i.e. redbook 44.1kHz upsampled to 192, or even just 48kHz, for that matter, rather than 88.2 or 176.4kHz forces the DAC to "interpolate" the missing data, by "filling in the gaps" in the square wave, as it is not a mathematically even step up. Whether or not you can hear these subtleties is entirely subjective. The best way to decide is trial-and-error with your own ears.

While I use power-of-two upsampling with A+, I set the steepness at 3 or 4, which is practically nothing. For me, it just sounds more realistic. Then again, i don't have a clue if my DAC upsamples internally, or not. I asked David at Peachtree, and he said "No". Then again, who knows if he is a marketing guy or a techie.

[Jud]

To up sample or to not up sample? I have an Emotiva XDA2 and see most people seem to select power of 2 upsampling in A+ preferences. My DAC is capable of 192k should I select max upsampling?

 

Best answer is also simplest: Try each (none, max, power of 2), see if you can hear a difference and if so which you like best.

[Neil]

Ok so for the last hour I have been sliding izotope ( sounds like it should be illegal) parameters and wondered if anyone can give me a head start as to what I should be listening to! For example what should I expect to hear when I change steepness, cut off, pre ringing etc. ie if I change steepness one way it makes the music more muffled and the other way sounds too bright...etc. just trying to get my head around what to expect. And what to listen for.....

 

current settings

10

1164000

1.04

200

0.65

 

equipment

new Mac mini, A+ newest version ,Emotiva XDA 2, emotiva XPA 2 , statement speakers.

[PGRPGRPG]

Ok so for the last hour I have been sliding izotope ( sounds like it should be illegal) parameters and wondered if anyone can give me a head start as to what I should be listening to! For example what should I expect to hear when I change steepness, cut off, pre ringing etc. ie if I change steepness one way it makes the music more muffled and the other way sounds too bright...etc. just trying to get my head around what to expect. And what to listen for.....

 

current settings

10

1164000

1.04

200

0.65

 

equipment

new Mac mini, A+ newest version ,Emotiva XDA 2, emotiva XPA 2 , statement speakers.

 

+1 with the above question from Neil.

Difficult for newbies to understand the trade-in & trade-off of these parameters...

Thank you by advance, if possible in "readeable" wording... ;-)

[Savage]

Currently using NOS Metrum Octave DAC & BADA

Steepness: 4

Filter Length: 2,000,000

Cut-Off: 1.03

Anti-Aliasing: 200

Pre-ring: 1.00

 

I don't think I'm at my final resting place with these settings but they are dialed in fairly well for my set-up. Having an NOS dac it's great to have this, I can get a traditional over-sampling sound, very bold and full with these settings. Or, switch back to NOS and get more delicacy and nuance. It's like having 2 Dacs...nothing to complain about there.

 

The only setting I haven't spent time with is the Filter Length Max. For this, I've just done what Jud has done

[wwaldmanfan]

Ok so for the last hour I have been sliding izotope ( sounds like it should be illegal) parameters and wondered if anyone can give me a head start as to what I should be listening to! For example what should I expect to hear when I change steepness, cut off, pre ringing etc. ie if I change steepness one way it makes the music more muffled and the other way sounds too bright...etc. just trying to get my head around what to expect. And what to listen for.....

 

current settings

10

1164000

1.04

200

0.65

 

equipment

new Mac mini, A+ newest version ,Emotiva XDA 2, emotiva XPA 2 , statement speakers.

 

 

As we know, this all is dependent on how your DAC functions, along with the rest of your system. And, naturally, your own personal taste. I do agree with you that higher steepness muddies the sound, and lower steepness gives the instruments and vocals better separation, because low steepness is altering the original music to a lesser degree. Upsampling for its own sake, is not de rigueur, just because you can.

 

From what I understand, the settings that make an audible difference are the steepness, the Nyquist cutoff, and the pre-ringing. However, the pre-ringing is proportional to the steepness, so if your steepness is in the single digits, adjusting the pre-ring will have a lesser effect.

 

As far as the filter max goes, a higher setting will tax your computer's processor more. I can't hear a difference on that one, but, I believe Damien had recommended 1M, which is right in the middle, so I go with his rec on that. The anti-alias might as well be set at max (200), but I don't hear a difference there, either. Then again, I don't have a $40,000 system (I heard one yesterday at my audio dealer: top-of-the-line Vandersteens, Ayre DAC and preamp, and Rogue tube power amp--it was a humbling experience. What was funny was the dealer was running A+ v1.5.4, and when I suggested he install the brand new upgrade to v1.5.5., even his eyes bugged out after he heard it!

 

These are the iZotope setting I use:

3

1,000,000

1.00 (default)

200

1.0 (default)

 

Like Jud says, you should play with it, and decide what you like for yourself by trial-and-error, but nobody said that futzing with this stuff was going to make your music sound better, it just might drive you nuts instead ;>)

[alcarp]

I have similar settings to the last 2 posts but find that when I upsample I lose some body, warmth and weight to the sound - it sounds a bit thin and bright.

 

What settings affect this?

[Savage]

I have similar settings to the last 2 posts but find that when I upsample I lose some body, warmth and weight to the sound - it sounds a bit thin and bright.

 

What settings affect this?

 

It could be Steepness, cut-off but, I would look especially at pre-ring. 1.00 default is a good place to start...things got bad for me quickly if I changed that on my set-up.

[Geoffrey Armstrong]

I read through this thread with interest at the weekend and really appreciate all the knowledgeable contributions here.

I've been meaning to get a handle on those iZotope settings for sometime and this encouraged me to experiment a bit.

Now I love A+ and it's definitely one of my favourite players. I'm one of those guys who likes to experiment with all the different players though and knowing there is one other player out there that's also licensed iZotope I thought I would also give it a try. That player is, of course, Fidelia. I'll try not to go into too much detail about Fidelia as it has it's own thread here.

 

Fidelia has some annoying quirks that I think might put people off it. It does also have some really interesting features though. One of those features is a choice of pre-sets for the iZotope re-sampler. There is a basic version of Fidelia and also an advanced version. If you want to keep it with the advanced features which include iZotope you have to buy the advanced license as well as the basic license. I think this prompting to add several licenses or start several trials can be annoying.

Anyway, it's possible to run the advanced features in trial mode along with the Basic player, either purchased or also in trial mode.

Another "feature" of Fidelia is its sound quality. In the Fidelia thread it is referred to variously as "clear" and sounding more "solid state" like. Compared to a Krell whereas the other players were compared to tube amps from Conrad Johnson or Audio Research. I found this interesting because after the Munich show, one of the Avantgarde guys also mentioned Fidelia as his preferred player, because "it does nothing to the original recording".

There seems to be a general consensus then, amongst those who've tried it, that Fidelia is essentially neutral and I felt this might help me even more to get a fix on what those iZotope settings are doing.

As I mentioned already it has a choice of pre-sets for iZotope's SRC. You can use these as starting points for your own custom settings or just use the pre-set you prefer. Unfortunately you can't save your own custom pre-sets. This has been mentioned here as a feature that would be great to have in A+ as well.

Since I'm using the Metrum Hex DAC, I started with custom settings from someone who posted here who's using a Metrum Octave. This had a very nice "presentation" of the music; but I thought I would then try playing with the pre-sets Fidelia supplied.

After trying a few of these I found myself going between two of the pre-sets. Both of these have a filter steepness of 79 and pre-ringing of 1.0. The only difference between them is the cut-off, which for one of these two pre-sets is 0.95 and on the other it's 1.0.

They are referred to as:

Steep, linear phase (cut-off = 1.0)

Steep, no aliasing (cut-off = 0.95)

The pre-set with the 0.95 cut-off seems to emphasis leading edges. This gives the music more momentum and bounce than any of the other settings I tried. It definitely has "good PRAT". Symbols are a little splashy though. Listening to Cassandra Wilson's Poet from Thunderbird, I also felt her voice didn't have enough body accompanying it. I was fortunate enough to be sitting just a few feet in front of Cassandra at Ronnie Scott's Club in London at the beginning of July, so I feel I know her voice well enough now.

Switch to the pre-set with all the same settings apart from the cut-off now being 1.0 and Cassandra's voice sounded just right and very natural. The symbols were now more "washy" than splashy, which I think better suits the context of this song. There was still a groove; but one that was slightly slowed compared to the previous setting and which I also feel suits this song better.

Comparing these two pre-sets on other music such as Tomattito Soy flamenco, the guitar also had the right amount of body and the "speed" felt just right. Flamenco was really gorgeous with this setting.

Similarly on Cal Tjader's Doxie from the Monterey Concerts; again switching between these two settings always lead to the same results. Everything was a little thinner brighter and faster sounding with the cut-off of 0.95 then the Steep, linear phase setting with a cut-off of 1.0, added the additional body and the groove felt right.

Now what was interesting for me was what would happen if I applied these same settings within A+? Would I hear the same qualities and differences between these two settings? You bet! …and that's the interesting thing. A+ certainly sounds difference to Fidelia. It is more like switching to a tube from a (typical) solid state amp. Everything seems richer and more organic; but the essential qualities and differences between these two pre-sets remain.

I do feel Fidelia's greater perceived neutrality helped me get a better handle on what these pre-sets are doing though. I say "perceived neutrality", because neutrality is not so easy to determine. If you're used to the sound of A+, you might hear Fidelia as being "clinical". The brain adjusts and accepts the sound you've got used to as "right". Switch from Fidelia to A+ and A+ might seem coloured by comparison.

Anyway, I'll be really interested in what the knowledgeable folks here might have to say about these two pre-sets. Do my descriptions jibe with what you'd expect to hear at these settings? If you were to tweak these a little what would you change and by how much?

Thanks,

geoff

[Savage]

A nice thorough write up Geoff,

It would be great to have the presets in A+ too, a good idea.

 

I will give your settings a try a little later today with the Octave and post back.

[Geoffrey Armstrong]

A nice thorough write up Geoff,

It would be great to have the presets in A+ too, a good idea.

 

I will give your settings a try a little later today with the Octave and post back.

 

Great! Looking forward to your findings.

 

geoff

[bdiament]

Hi wwaldmanfan,

 

I am one of the least technically-minded people here, but I will attempt to explain the rationale behind this. If you think of a "bit perfect" sample as a square wave, power of two upsampling multiplies it geometrically, while preserving the shape. The "in-between" settings, i.e. redbook 44.1kHz upsampled to 192, or even just 48kHz, for that matter, rather than 88.2 or 176.4kHz forces the DAC to "interpolate" the missing data, by "filling in the gaps" in the square wave, as it is not a mathematically even step up. Whether or not you can hear these subtleties is entirely subjective. The best way to decide is trial-and-error with your own ears.

While I use power-of-two upsampling with A+, I set the steepness at 3 or 4, which is practically nothing. For me, it just sounds more realistic. Then again, i don't have a clue if my DAC upsamples internally, or not. I asked David at Peachtree, and he said "No". Then again, who knows if he is a marketing guy or a techie.

 

Please forgive me but I believe this is not correct.

While the early days of sample rate conversion may have seen benefit with integer (i.e., "even multiple") conversion -- and this is true with some of the not-so-great conversion algorithms of today -- things have come a long, long way in the past several years.

 

In my experience, the better algorithms (and I personally rank iZotope's 64-bit SRC by Alexey Lukin as the most transparent I've yet heard), can handle the more complex math involved in finding the (higher) *common* multiple. They will, to my ears, create far more transparent results, even at non-integer conversion (e.g. 192k to 44.1k or vice versa), than other algorithms can with the simpler math involved in integer conversion.

 

So, again, this depends completely on the sample rate conversion algorithm and is not at all, in my experience a function of integer vs. non-integer conversion alone. Some algorithms are going to be a bit more benign when asked to do the simpler math. To my ears, other algorithms simply do not "care".

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[Superdad]

Hi

 

In my experience, the better algorithms (and I personally rank iZotope's 64-bit SRC by Alexey Lukin as the most transparent I've yet heard), can handle the more complex math involved in finding the (higher) *common* multiple. They will, to my ears, create far more transparent results, even at non-integer conversion (e.g. 192k to 44.1k or vice versa), than other algorithms can with the simpler math involved in integer conversion.

 

Hi Barry. Thank you again for your insights based on real experience. Always appreciated.

 

However, allow me to explain why I think that for A+ users, power of two conversion is the better sounding choice:

A while ago, prompted I recall by your suggestion that offline SRC is better--and that that is how you perforn all your interpolations--I downloaded a trial of Sample Manager and did a few experiments. Indeed, files that I pre-converted (using the exact same iZtope settings as I had been in Audirvana) sounded just a bit better than realtime upsampling of the same track. And I attribute that to the computer not having to do as much at playback time.

 

So I continued the experiment. To date, in Audivana, I have found that power of two upsampling sounds better than going to the max (which for my current USB input is 192KHz). But with offline pre-conversion with Sample Manager, this was not the case! Indeed, with offline conversion, 192 sounded just a little bit better (as opposed to just a little bit worse as with A+ realtime) than 176.4--still with Sample Manager.

Again, I attribute this to the work done offline versus realtime.

 

I hope the above helps people reconcile your experience with their own.

 

Cheers,

Alex

[bdiament]

Hi Alex,

 

Hi Barry. Thank you again for your insights based on real experience. Always appreciated.

 

However, allow me to explain why I think that for A+ users, power of two conversion is the better sounding choice:

A while ago, prompted I recall by your suggestion that offline SRC is better--and that that is how you perforn all your interpolations--I downloaded a trial of Sample Manager and did a few experiments. Indeed, files that I pre-converted (using the exact same iZtope settings as I had been in Audirvana) sounded just a bit better than realtime upsampling of the same track. And I attribute that to the computer not having to do as much at playback time.

 

So I continued the experiment. To date, in Audivana, I have found that power of two upsampling sounds better than going to the max (which for my current USB input is 192KHz). But with offline pre-conversion with Sample Manager, this was not the case! Indeed, with offline conversion, 192 sounded just a little bit better (as opposed to just a little bit worse as with A+ realtime) than 176.4--still with Sample Manager.

Again, I attribute this to the work done offline versus realtime.

 

I hope the above helps people reconcile your experience with their own.

 

Cheers,

Alex

 

If I understand you correctly, you are saying that real time conversion with Audirvana (which I believe may use iZotope's SRC) sounds better to you when it is integer, as opposed to non-integer. Is that right?

 

If so, it may come down to how each listener defines "better". If one *likes* a certain result more, that could be better. My own preference is to do a direct comparison with the unprocessed original and I'll call the one that sounds more like that original "better".

 

It may also be that when the algorithm is used in real time in that application, there is less "overhead" with integer conversion. As I don't use real time conversion or that particular application, I don't know how I'd hear it. That particular combination may well simply do better with integer conversion. I do know that my experience with iZotope's algorithm, in the applications I do use, is that is just doesn't care -- but that is just me, with the apps I have, on my computer, listening to my system. Clearly, different folks will hear it differently.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

 

P.S. Please forgive what may be a memory lapse on my part but wasn't it you who told me a while back that you did not hear a difference between the different samples at different sample rates on the Soundkeeper Format Comparison page? If it was, I'm confused. (Please understand, nothing at all against you my friend -- I'm just confused.)

[Jud]

Hi wwaldmanfan,

 

 

 

Please forgive me but I believe this is not correct.

While the early days of sample rate conversion may have seen benefit with integer (i.e., "even multiple") conversion -- and this is true with some of the not-so-great conversion algorithms of today -- things have come a long, long way in the past several years.

 

In my experience, the better algorithms (and I personally rank iZotope's 64-bit SRC by Alexey Lukin as the most transparent I've yet heard), can handle the more complex math involved in finding the (higher) *common* multiple. They will, to my ears, create far more transparent results, even at non-integer conversion (e.g. 192k to 44.1k or vice versa), than other algorithms can with the simpler math involved in integer conversion.

 

So, again, this depends completely on the sample rate conversion algorithm and is not at all, in my experience a function of integer vs. non-integer conversion alone. Some algorithms are going to be a bit more benign when asked to do the simpler math. To my ears, other algorithms simply do not "care".

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

 

I hasten to say I have not done any significant listening for myself to hear which of the rates (in my case, 352.8 or 384kHz) sounds better, if either. I just wanted to note that Miska mentioned, if I remember correctly, that having a *rational* conversion factor, as opposed to an integer, was what was important, and that the conversion factor between the 44.1 and 48 classes of rates was rational.

[Superdad]

Hi Alex,

 

If I understand you correctly, you are saying that real time conversion with Audirvana (which I believe may use iZotope's SRC) sounds better to you when it is integer, as opposed to non-integer. Is that right?

 

Thanks for your reply Barry. Yes, Audirvana Plus uses the full iZotope engine with full control over SRC parameters (Steepness, Cut-off, Max. length, Final attenuation depth, Pre/post ringing balance, and sample rate). It offers this only for realtime playback processing. And yes, I find that it sounds better when keeping the output sample rate a direct multiple of the original file.

 

If so, it may come down to how each listener defines "better". If one *likes* a certain result more, that could be better. My own preference is to do a direct comparison with the unprocessed original and I'll call the one that sounds more like that original "better".

 

Well, but in your case the original is the high-rate version and you are down-converting--correct? For a home playback user we are taking Redbook files and going up. Listening to the original (which unlike most users I can truly do because I have an NOS DAC as opposed to one with a filter doing internal up/oversampling) does not sound better because then I am listening to all the aliasing.

 

It may also be that when the algorithm is used in real time in that application, there is less "overhead" with integer conversion. As I don't use real time conversion or that particular application, I don't know how I'd hear it. That particular combination may well simply do better with integer conversion. I do know that my experience with iZotope's algorithm, in the applications I do use, is that is just doesn't care -- but that is just me, with the apps I have, on my computer, listening to my system. Clearly, different folks will hear it differently.

 

Yes, that was exactly my point. That perhaps the reason realtime non-integer conversion does not sound as good is that the software is doing even more work. And as proof I offered (in post 291 above) the results I got from comparing off-line and realtime integer and non-integer conversions.

 

P.S. Please forgive what may be a memory lapse on my part but wasn't it you who told me a while back that you did not hear a difference between the different samples at different sample rates on the Soundkeeper Format Comparison page? If it was, I'm confused. (Please understand, nothing at all against you my friend -- I'm just confused.)

 

Oh dear, there are two Alex's very active here at CA. SandyK is the other one. Perhaps it was he. (Though I would not think that HE would report not hearing differences ;-)).

[Boris75]

When I used my Audiophilleo AP2-Hegel HD10 combo, I upsampled everything to 24/192 as it gave the best sound. Now with my Benchmark DAC2 I play everything native as I cannot discern any difference between files played natively and upsampled to 24/192 with iZotope in Audirvana, presumably because the internal upsampling within the DAC2 is high-quality.

[bdiament]

Hi Alex,

 

Thanks for your reply Barry. Yes, Audirvana Plus uses the full iZotope engine with full control over SRC parameters (Steepness, Cut-off, Max. length, Final attenuation depth, Pre/post ringing balance, and sample rate). It offers this only for realtime playback processing. And yes, I find that it sounds better when keeping the output sample rate a direct multiple of the original file.

 

Okay, thanks. Glad I understood correctly.

 

 

Well, but in your case the original is the high-rate version and you are down-converting--correct? For a home playback user we are taking Redbook files and going up. Listening to the original (which unlike most users I can truly do because I have an NOS DAC as opposed to one with a filter doing internal up/oversampling) does not sound better because then I am listening to all the aliasing.

 

I have also compared converting the sample rate upward as an off-line process vs. having software do it in real time. While I hear *some* benefits (perhaps due to the gentler filtering requirements on playback), I still prefer to listen to files natively.

 

(I would add that I differentiate between upsampling, as performed by the filters in a DAC and converting the sample rate upward before the DAC. In this case, I am referring to converting the sample rate upward, regardless of anything the filter in the DAC is doing.)

 

 

Yes, that was exactly my point. That perhaps the reason realtime non-integer conversion does not sound as good is that the software is doing even more work. And as proof I offered (in post 291 above) the results I got from comparing off-line and realtime integer and non-integer conversions.

 

I would not argue with whatever you prefer. However, neither am I prepared to draw the same conclusion. First, there is no reference for comparison -- only what the listener prefers. (Again, I'd never argue with a preference but for myself, I'd only determine which is the more transparent via comparison with a reference.)

 

Also, what is true for one software application in combination with a given DAC *and* the settings selected by the user, may well be true for that combination but I'm not so sure it will be universally true -- even for the same listener.

 

 

Oh dear, there are two Alex's very active here at CA. SandyK is the other one. Perhaps it was he. (Though I would not think that HE would report not hearing differences ;-)).

 

My mistake... for which, in fact, I'm glad because that eliminates what I would have found confusing. ;-}

I know it isn't the other Alex, so either there is a third Alex or I'm not remembering the right person -- the who said they could not differentiate between the different samples on the Format Comparison page. (I would have found it confusing for a given listener to not hear differences between sample rates in one test and yet still hear differences in different *types* of conversion at the *same* sample rate.)

 

In the end, I only know the combinations I've tried and while I'm fortunate in that there have been a good many, that certainly isn't all of them.

 

Best regards,

Barry

Soundkeeper Recordings

The Soundkeeper | Audio, Music, Recording, Playback (new blog!)

Barry Diament Audio

[Superdad]

I have also compared converting the sample rate upward as an off-line process vs. having software do it in real time. While I hear *some* benefits (perhaps due to the gentler filtering requirements on playback), I still prefer to listen to files natively.

 

(I would add that I differentiate between upsampling, as performed by the filters in a DAC and converting the sample rate upward before the DAC. In this case, I am referring to converting the sample rate upward, regardless of anything the filter in the DAC is doing.)

 

As you say, the choice you are making is just between the interpolation/filter algorithms of you DAC versus doing it in software (which, for most DACs you need to upsample to 384kHz to really push aside their internal filter).

So not than I am advocating listening to straight NOS Redbook, but such is the only way to actually be "native" to sample rate.

 

BTW, I enjoyed your latest post on your new blog. Keep those coming!

 

Cheers,

AJC

[bdiament]

Hi Alex,

 

As you say, the choice you are making is just between the interpolation/filter algorithms of you DAC versus doing it in software (which, for most DACs you need to upsample to 384kHz to really push aside their internal filter).

So not than I am advocating listening to straight NOS Redbook, but such is the only way to actually be "native" to sample rate.

 

BTW, I enjoyed your latest post on your new blog. Keep those coming!

 

Cheers,

AJC

 

Actually, when I say I differentiate between upsampling, as performed by the filters in a DAC and converting the sample rate upward before the DAC, I mean to say I take these to be fundamentally different processes and not simply a difference between hardware and software (the former essentially being software anyway).

 

Upsampling as performed by a filter does not change the data from the file on the way to the DAC chip. That is, the DAC receives the data contained in the original file. With sample rate conversion, the DAC receives altered (converted) data. Two *very* different animals as far as I'm concerned.

 

Thank you for your kind and much appreciated feedback on the latest blog post. It is a new adventure and one I'm already enjoying very much.

 

Best regards,

Barry

Soundkeeper Recordings

The Soundkeeper | Audio, Music, Recording, Playback

Barry Diament Audio

[Superdad]

Upsampling as performed by a filter does not change the data from the file on the way to the DAC chip. That is, the DAC receives the data contained in the original file. With sample rate conversion, the DAC receives altered (converted) data. Two *very* different animals as far as I'm concerned.

 

Sorry Barry, I am not following you here. If you upsample/oversample/interpolate/whatever at all, then you have a filter with coefficients and the original data is altered. Do it in software, do it in a DAC's separate DF chip, do it in the DAC chip's built-in filters as part of the delta-sigma modulation--it's all an alteration, and the choice comes down to the parameters and sonic results/trade-offs of each. Am I missing something? (very likely, so feel free to correct/enlighten me)

[Miska]

Upsampling as performed by a filter does not change the data from the file on the way to the DAC chip. That is, the DAC receives the data contained in the original file. With sample rate conversion, the DAC receives altered (converted) data. Two *very* different animals as far as I'm concerned.

 

"DAC" here can mean three different things:

1) A device, box that contains bunch of electronic components, connected to a computer or some other source

2) A chip that is inside (1) and takes digital data in and converts it to analog form

3) The actual conversion stage that converts digital symbols from digital to a signal in analog domain

 

Now any modern chip (2) has at least three functions inside:

1) Digital up-sampling/oversampling/interpolation filter

2) Delta-sigma modulator, converting output from (1) to high-rate low-bit output for (3)

3) Actual conversion stage converting output of (2) to analog domain

 

Of these functions, (1) and (2) can be moved to be performed in a computer software to a varying extent depending on the particular device. Regardless of where (1) and (2) are performed, however, function (3) inside DAC chip never sees the original data you have in a file in case of PCM. If your files contain DSD then it is much more likely that it goes straight to (3) - one of the leading ideas behind DSD.

 

It would be nice if you could express in more detail what exactly you mean by upsampling not altering data while sample rate conversion doing it? To my knowledge it can only be the case if both operate under certain carefully selected strict conditions in order to meet this particular criteria...