Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: From 44.1 to 96, which is the best resampler? (Read 67104 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

From 44.1 to 96, which is the best resampler?

Reply #75 –
Quote from: Arnold B. Krueger on
Quote
Anther example would be frequency inversion: Flip the sign of every 2nd sample. This will turn the spectrum upside down.  It's still a linear operation.

Again, you really need to revisit what for me was second-semester electronics, and basic signal analysis.

You guys are using different meanings of the word "linear". In DSP, "linear system" often means "linear and shift-invariant system". The systems given by SebastianG are linear, but not shift-invariant.

From 44.1 to 96, which is the best resampler?

Reply #76 –
Quote from: Alexey Lukin on
You guys are using different meanings of the word "linear". In DSP, "linear system" often means "linear and shift-invariant system". The systems given by SebastianG are linear, but not shift-invariant.
This discussion goes very nicely, don´t you all think that?    I wonder if anyone actually downloaded the files I presented - and not only for measuring frequency response. Alexey, you mentioned on another thread here a (indeed not a very scientific) study which described the same effects I´m hearing when using different settings of the iZotope SRC. I count myself to be one of the persons who can hear different settings. BTW, let me congratulate you on developing the most perfect resampler. It is an awsome piece of work.

However, I would love to know the reason why this software SRC contains the ability to tune the resampling characteristics? I ask because until iZotope SRC emerged all other software resamplers weren´t configurable at all. Who was interested in this? Forgive me asking this, but I´m really curious.
marlene-d.blogspot.com

From 44.1 to 96, which is the best resampler?

Reply #77 –
I'm not sure how long the resampling modes have been there, but the Unix program SoX has had all kinds of configurable resampling options for a while now. I'm not certain, but it's possible they predated iZotope's.

I don't use anything else for resampling at the moment. I have no idea why people would used closed alternatives.

From 44.1 to 96, which is the best resampler?

Reply #78 –
You cannot please all users with the same SRC filter. Some prefer absolutely no aliasing, others prefer softer filters and less ringing, etc. That's why iZotope provides a choice. It is also educative to look at corresponding filter frequency responses and project how big the distortions can be.

From 44.1 to 96, which is the best resampler?

Reply #79 –
Quote from: Alexey Lukin on
You cannot please all users with the same SRC filter. Some prefer absolutely no aliasing, others prefer softer filters and less ringing, etc. That's why iZotope provides a choice. It is also educative to look at corresponding filter frequency responses and project how big the distortions can be.
Then it is a good thing that iZotope offers that much options. Because I prefer less ringing and more aliasing/imaging. 


Quote from: Canar on
I'm not sure how long the resampling modes have been there, but the Unix program SoX has had all kinds of configurable resampling options for a while now. I'm not certain, but it's possible they predated iZotope's.

I don't use anything else for resampling at the moment. I have no idea why people would used closed alternatives.
I didn´t know that. Thanks for the info. Sadly I´m not into using Unix.
marlene-d.blogspot.com
 

From 44.1 to 96, which is the best resampler?

Reply #80 –
Quote from: Cavaille on
until iZotope SRC emerged all other software resamplers weren´t configurable at all. Who was interested in this?
The iZotope flexibility is impressive indeed. Some other (and older) resamplers did offer some choice in filter steepness, like my good old Sonic Solutions DAW. That was already in the early 90's  Their user manual gave a nice explanation about the options:
Quote
From the Sonic Solutions user manual:
The heart of the sample rate conversion process is the low pass filter, which is designed to reject aliases (down conversion) or images (up conversion) — both of which are detrimental to sound quality. The filter’s response above cutoff must be very steep if it is to be effective at rejecting these artifacts.
Very steep filters can introduce audible distortions, however. Any filter design for sample rate conversion therefore involves tradeoffs among a number of factors: aliasing/imaging, inband/stopband ripple, reduced gain at very high frequencies, and time smear. Each of these factors has effects that can vary with the sonic characteristics of the material: sometimes it may be preferrable, for example, to allow a small amount of aliasing in order to preserve other aspects of signal integrity.
To provide the most flexibility in tailoring the conversion process to different types of program material, HD SRC offers three different filter choices: Steep, Gentle and Gentlest. The Steep option is a Nyquist filter optimized for maximum alias/image rejection. The other options trade a limited amount of aliasing/imaging for reductions in other distortions. Your choice of a filter curve for a given rate conversion operation should be based on careful audition.
[/size]BTW I haven't managed to ABX your testfiles (yet). Hopefully some other users will try too. If the differences are audible, it shouldn't be too difficult to ABX them, right ?

From 44.1 to 96, which is the best resampler?

Reply #81 –
Quote from: Kees de Visser on
The iZotope flexibility is impressive indeed. Some other (and older) resamplers did offer some choice in filter steepness, like my good old Sonic Solutions DAW. That was already in the early 90's  ... BTW I haven't managed to ABX your testfiles (yet). Hopefully some other users will try too. If the differences are audible, it shouldn't be too difficult to ABX them, right ?
The differences are subtle, but yes, they are audible. Though I can´t post any comments like this: "My girlfriend ran from the kitchen because suddenly it sounded better".    My boyfriend isn´t interested in this matter at all.  He just loves his turntable...
marlene-d.blogspot.com

From 44.1 to 96, which is the best resampler?

Reply #82 –
Quote from: Alexey Lukin on
Quote from: Arnold B. Krueger on

Quote
Anther example would be frequency inversion: Flip the sign of every 2nd sample. This will turn the spectrum upside down.  It's still a linear operation.

Again, you really need to revisit what for me was second-semester electronics, and basic signal analysis.

You guys are using different meanings of the word "linear". In DSP, "linear system" often means "linear and shift-invariant system". The systems given by SebastianG are linear, but not shift-invariant.

Exactly.  It's probably only a terminology issue.  Well, when I mean LTI (linear and time invariant) I say LTI and not "linear".

Also, I'm not saying "aliasing is linear" because it doesn't make much sense. But the sampling process -- using that definition of linearity that is the "L" in LTI system -- is of course a linear process.  Talking about time invariance here is hardly possible since the input is time continuous (or at a higher sampling rate than the output) and the output is time discrete.  So, the sampling process doesn't even qualify as something that can be checked for being LTI or not.

Quote from: Arnold B. Krueger on
Believe it or not, audio takes place in both the amplitude and time (frequency) domains

That looks almost like an insult.

Cheers!
SG

From 44.1 to 96, which is the best resampler?

Reply #83 –
Quote from: Arnold B. Krueger on
The definition of nonlinear that I provided is absolutely orthodox and classical.
Well, this comes down to how you define non linear. "New frequency components appearing" might be a common property of common non-linear systems, but it's not the "classical definition".

The classical definition is even correctly quoted at the start of the relevant wikipedia page...
http://en.wikipedia.org/wiki/Linear
...which is basically what SebG said. If it satisfies those equations, it's linear. If it doesn't, it's non-linear.


For example, if zero stuffing to increase the sample rate is non linear, then it should be possible to find two signals, x and y, where:

zero_stuff(x) + zero_stuff(y)

is not the same as

zero_stuff(x+y)


If the samples of the signal x are x1, x2, x3, ...  and the samples of the signal y are y1, y2, y3, ...
then either way (i.e. zero stuff then add, or add then zero stuff) you're going to get these output samples:
x1+y1, 0, x2+y2, 0, x3+y3, 0

Either way round, you get the same result - and it doesn't matter what the input samples are. QED.

(I'm not a mathematician. It would be more complicated to prove if you started with two different sample rates since the "+" part would be more difficult to define in the digital domain - however, f(x+y)=f(x)+f(y) would still be true).


Does the time domain transfer function have powers of x other than 1? No, but it has conditionals/limits and/or deltas which mean it's not LTI.

Cheers,
David.

From 44.1 to 96, which is the best resampler?

Reply #84 –
Quote from: Alexey Lukin on
You cannot please all users with the same SRC filter. Some prefer absolutely no aliasing, others prefer softer filters and less ringing, etc.

May be for low sampling rates (though I think the source material is also a contributing factor as to which filter configuration is preferable).

But it seems unlikely to be true at CD rates.  If it were, since CDs are usually mastered from something greater than 44.1k, they would have to be sold in two variants: a low ringing version and a low aliasing version.

  -bandpass

From 44.1 to 96, which is the best resampler?

Reply #85 –
Quote from: bandpass on
But it seems unlikely to be true at CD rates.  If it were, since CDs are usually mastered from something greater than 44.1k, they would have to be sold in two variants: a low ringing version and a low aliasing version.
Not necessarily. Since this sonic differences are subtle, most people won´t even notice them. In many recordings however you can see different resampling algorithms. For example recent TELARC recordings doesn´t have a decreasing frequency fall-off at high frequencies, they seem to be linear (together with dithering noise). But these releases come from DSD recordings and DSD conversion to 16/44.1 is very simple and don´t use a resampler to my knowledge at all.

With... let´s say the newest Madonna release you have a soft frequency cut-off which suggests a not so steep filter was used. My guess is that this is used regulary used today by mastering engineers, if they know it or not. I assume that those mastering engineers knows about ringing and are using it to their liking. But that´s only a guess.
marlene-d.blogspot.com

From 44.1 to 96, which is the best resampler?

Reply #86 –
Well, some CD players already have a selection of several oversampling filters 

From 44.1 to 96, which is the best resampler?

Reply #87 –
Quote from: Alexey Lukin on
Well, some CD players already have a selection of several oversampling filters 
Do you mean the different filter-settings Sony used several years ago for their CD player? I´ve had one of these players - but not for that reason - and if I recall my memory correctly the differences between the filter settings were subtle (maybe the same as with different resampling configurations). But to my knowledge the Sony were not doing any upsampling.... but then, filtering doesn´t apply to resampling only.

Since Sony doesn´t use this anymore I guess most people couldn´t hear it or didn´t use it at all. A lot of guessing again... 
marlene-d.blogspot.com

From 44.1 to 96, which is the best resampler?

Reply #88 –
Quote from: Cavaille on
But to my knowledge the Sony were not doing any upsampling....

Every audio DAC is using resampling. What other kind of filters could be selectable?

From 44.1 to 96, which is the best resampler?

Reply #89 –
Quote from: Alexey Lukin on
Every audio DAC is using resampling. What other kind of filters could be selectable?
They do?? That´s new to me. Or do you mean that oversampling is not so different from resampling? I don´t know how these filter settings worked. I´ve had four different settings. According to a magazine I read years back, the first filter used a normal, steep brickwall filtering, the second a spline algorithm (much like pioneers "legato link"), the third emulated old D/A-converters with a lot of post-ringing and no pre-ringing, the fourth filter was a combination of filter 1 & 2. But I didn´t use them, I connected the player to my amp digitally anyway.
marlene-d.blogspot.com

From 44.1 to 96, which is the best resampler?

Reply #90 –
Yes, oversampling is the same as resampling. All converters are using it nowadays.

From 44.1 to 96, which is the best resampler?

Reply #91 –
Quote from: Alexey Lukin on
Yes, oversampling is the same as resampling. All converters are using it nowadays.
Ok, that makes sense. I did a bit of reading, now I know the differences between resampling & oversampling - tiny differences. Even the earliest CD players used oversampling (4x) on 16 Bit converters. Then upsampling in CD players is just another marketing gag. How nice to have the possibility of doing it yourself with the desired configuration in iZotope SRC.    A bit time consuming though...
marlene-d.blogspot.com

From 44.1 to 96, which is the best resampler?

Reply #92 –
Quote from: Alexey Lukin on
Yes, oversampling is the same as resampling. All converters are using it nowadays.
So in light of the OP's question, what is your opinion about the (practical/theoretical) differences in the analog output between:
- playing a 44.1kHz source through a DAC that runs at 44.1kHz clock and uses internal oversampling, and
- SRC'ing the 44.1 kHz source to a higher (say 96kHz) rate and playing it through the same DAC but runs at that higher rate, still using oversampling.

From 44.1 to 96, which is the best resampler?

Reply #93 –
Theoretically it is possible to achieve better quality in the second case, because the software SRC conversion can be easily superior to the DAC conversion. However, audibility of differences is under question, considering that they are probably ultrasonic.

From 44.1 to 96, which is the best resampler?

Reply #94 –
Quote from: Kees de Visser on
Quote from: Alexey Lukin on
Yes, oversampling is the same as resampling. All converters are using it nowadays.


So in light of the OP's question, what is your opinion about the (practical/theoretical) differences in the analog output between:
- playing a 44.1kHz source through a DAC that runs at 44.1kHz clock and uses internal oversampling, and
- SRC'ing the 44.1 kHz source to a higher (say 96kHz) rate and playing it through the same DAC but runs at that higher rate, still using oversampling.


...which, if I understand the marketing jargon correctly, is the same as asking about oversampling vs upsampling, respectively.  Both of which are involve resampling.  (Please correct me if I'm wrong).

From 44.1 to 96, which is the best resampler?

Reply #95 –
Quote from: krabapple on
...which, if I understand the marketing jargon correctly, is the same as asking about oversampling vs upsampling, respectively.  Both of which are involve resampling.  (Please correct me if I'm wrong).

The terms "upsampling" and "downsampling" are -- strictly speaking -- only describing the process of inserting zeros or throwing samples away respectively. There's no anti imaging/aliasing filtering involved. Check out the dspguru FAQ on decimation and interpolation for the terminology.

Cheers!
SG

From 44.1 to 96, which is the best resampler?

Reply #96 –
Quote from: SebastianG on
Quote from: krabapple on

...which, if I understand the marketing jargon correctly, is the same as asking about oversampling vs upsampling, respectively.  Both of which are involve resampling.  (Please correct me if I'm wrong).

The terms "upsampling" and "downsampling" are -- strictly speaking -- only describing the process of inserting zeros or throwing samples away respectively. There's no anti imaging/aliasing filtering involved. Check out the dspguru FAQ on decimation and interpolation for the terminology.

Cheers!
SG


I get what you're saying, but I'm addressing the way CDPs are marketed  --  internal 'oversampling'  of CD play has been standard since at least the early 90s, but some 'high end' players market themselves as 'upsampling' as well.  I presume the only difference is that in the latter, the output is from a (for example) 96 kHz or higher SR DAC,  I hasten to add that I don't expect this to make a hair of  difference, except to gullible wallets.

From 44.1 to 96, which is the best resampler?

Reply #97 –
Hi I'm new to the Forum

Oversampling is a term used in the IC industry to refer to sigma delta DACs and ADCs.

The  correct term for what is being discussed is Interpolating.

http://datasheets.maxim-ic.com/en/ds/MAX5556-MAX5559.pdf

From 44.1 to 96, which is the best resampler?

Reply #98 –
A perusal of the datasheets eg  http://datasheets.maxim-ic.com/en/ds/MAX9850.pdf  and  http://focus.ti.com/lit/ds/symlink/pcm4104.pdf  . shows that although they use interpolation, they still have sharp digital low pass filters.

If a cd player manufacturer is advertising "upsampling" they may be refering to the use of interpolation followed by a gentle slope digital low pass filter eg http://www.arcam.co.uk/downloads/upsampling%20multidac.pdf

From 44.1 to 96, which is the best resampler?

Reply #99 –
Quote from: SebastianG on
The terms "upsampling" and "downsampling" are -- strictly speaking -- only describing the process of inserting zeros or throwing samples away respectively. There's no anti imaging/aliasing filtering involved. Check out the dspguru FAQ on decimation and interpolation for the terminology.

Quote
Loosely speaking, "decimation" is the process of reducing the sampling rate. In practice, this usually implies lowpass-filtering a signal, then throwing away some of its samples.

"Downsampling" is a more specific term which refers to just the process of throwing away samples, without the lowpass filtering operation. Throughout this FAQ, though, we'll just use the term "decimation" loosely, sometimes to mean "downsampling".
That's exactly how MATLAB defines it, though I'm sure I was taught the other way around - i.e. downsampling is LPF+decimiation.

I think anyone who says there is a useful well understood difference in meaning between upsampling / oversampling / resampling is being rather optimistic. "Interpolation" is even worse!

Cheers,
David.