Offline Upsampling

[kumakuma]

This is contrasted to the hotwater on demand heaters, where they crank the heating elements to heat the water "on the fly" or on demand. This would be hardware upsampling.

 

On the fly upsampling can also be done in software. For example, some of us here use JRiver to convert everything to DSD on the fly.

[Bill Lord]

On the fly upsampling can also be done in software. For example, some of us here use JRiver to convert everything to DSD on the fly.

 

Thanks. I left that out.

 

So three options

 

1- Upsample via Hardware in realtime

2- Upsample via Software in realtime

3- Upsample via Software offline

 

I supposed you'd have to record 1 & 2 with a high end recording device and compare it to 3.

 

Did anyone read this article?

 

Upsampling vs. Oversampling for Digital Audio | Audioholics

 

Here are the summary points from the article.

 

 

What Does This All Mean - Will it Sound Better?

 

 

So the question remains whether upsampling or oversampling actually make music sound 'better'. How much do we need? We have seen the main motivation behind oversampling and how it allows us to use simpler digital and analog filters as well as helping us with quantization noise. The effects of upsampling are greatly debated. While it is true that upsampling does help us in attenuating the amount of jitter caused by sampling errors and an inaccurate clock, whether this jitter is audible or not is a point of contention. There is no doubt that wide bit words and super-high sampling rates that are touted by the latest products are largely marketing. Oversampling has been around for a very long time and has been used extensively in audio products to not only improve sound quality through 'better' filtering but to make these same products much cheaper. Upsampling, on the other hand, is relatively newer and debated greatly. The effects of upsampling are no doubt overstated. By carefully designing the sampler, ADC, digital processing path, and oversampling DAC, the upsampling and asynchronous rate transfer can, in my opinion, be avoided.

 

 

The Purists Point of View

 

 

There are basically two points of view regarding this upsampling an oversampling. The audio 'purists' want no additional processing on their signal and want whatever comes in from the source to come out as analog. They talk about zero oversampling DACs and such that are completely filter free both in the analog and digital domain. That is one extreme that some may argue is the purest since it avoids any digital artifacts and it's quality relies on human perception by arguing that the human ear in itself acts as a brickwall filter after 20 kHz. Whenever we get into debates of human perception, the math and theory go out the window. Does it sound better without all the digital processing and filtering even with the image of the signal sitting just past fs/2? The energy past 22.05kHz is still present and you are still sending it to the speaker's tweeter. How will the tweeter react to such out-of-band frequencies that are present? Furthermore, sending such a signal that is not limited in bandwidth could cause stability problems with wide-bandwidth amplifiers that have a high unity-gain crossing. The overall system's signal-to-noise- ratio will be adversely affected as well. The DAC will also introduce frequency spurs all over the place. If we don't filter them at all, what will their presence do to the sound? It's a complicated problem and such a minimalist approach could introduce more non-linearities and negative effects, more so than the digital processing ever would.

[acousticsguru]

The problem is, I've heard good and terrible sound from equipment based on either school of thought. Personally, I'm of the school who believes that it would add to the user's peace of mind if it could be confirmed that a better-sounding result were invariably based on a puristic approach. But I'm a purist as well as a sceptic: unable to talk myself into hearing what ratio tells me to. In case of doubt, I'll trust my ears, lean back and wait for someone to prove what doesn't seem to make sense. I have enough engineer friends who occasionally forget that science is nothing but work in progress, and that it is ultimately unscientific to claim something can't be because we lack proof.

 

In short, the attitude that's always baffled me is the one primarily exhibited by engineering know-alls: no, I will not hear a difference because there can't be one (and they're the ones who dare call "self-fulfilling prophecy" on those who do hear a difference).

 

I've always liked the dCS approach to upsampling (they practically invented both concept and term). They never made any claim whatsoever. They said, look, we're offering this expensive tool (that we built with another purpose in mind) that really doesn't do anything, at least nothing we know of that should make any audible difference one way or another. Now listen…

 

Greetings from Switzerland, David.

[copy_of_a]

- Possibly, if alphaD will improved, will necessity in additional settings. Possible it will noise shaping, possible other. Need think about it.

(...)

- Possible, idea about mid variant between linear and minimal phase filter will get continuation.

many thanks, Yuri!

 

 

Converting from 16 bit

For conversion from 16 bit need more intensity of dithering. In the article we can see why.

Just checked your dither a bit more in detail. It seems your dither is close to but not quite standard "TPDF 2LBS p2p" dither. There's an anomaly when upconverting from 16bit to 24bit (even with the lowest dither level): the noise level of your dither is still applied to levels at around -120db. From a technical standpoint with regard to "pure dither" this should not happen (with both of my PSP Audioware Dither plugins as well as with iZotrope and finally with Steinberg Wavelab's internal dither set to standard "TPDF 2LBS p2p" the noise level stays below -140db in 24bit). Now, -120db is of course beyond the audible range and the bit-meter shows no dropouts or so. But I wonder why your dither is acting that way?

[ALRAINBOW]

For me it's about me , this topic is abiut ones likes not waht everyone should.

So it's all good for someone some where's I guess.

Purist is good approach but as so many love upsmpling there just might be something to it.

[Bill Lord]

The problem is, I've heard good and terrible sound from equipment based on either school of thought. Personally, I'm of the school who believes that it would add to the user's peace of mind if it could be confirmed that a better-sounding result were invariably based on a puristic approach. But I'm a purist as well as a sceptic: unable to talk myself into hearing what ratio tells me to. In case of doubt, I'll trust my ears, lean back and wait for someone to prove what doesn't seem to make sense. I have enough engineer friends who occasionally forget that science is nothing but work in progress, and that it is ultimately unscientific to claim something can't be because we lack proof.

 

In short, the attitude that's always baffled me is the one primarily exhibited by engineering know-alls: no, I will not hear a difference because there can't be one (and they're the ones who dare call "self-fulfilling prophecy" on those who do hear a difference).

 

I've always liked the dCS approach to upsampling (they practically invented both concept and term). They never made any claim whatsoever. They said, look, we're offering this expensive tool (that we built with another purpose in mind) that really doesn't do anything, at least nothing we know of that should make any audible difference one way or another. Now listen…

 

Greetings from Switzerland, David.

 

David can you comment on where to find more about dCS's upsampling?

[acousticsguru]

David can you comment on where to find more about dCS's upsampling?

 

Technically? There really isn't much that I know of except the famous caveat that I believe even made it into the manual (I'd have to check), and the white papers by e.g. Mike Story that could be downloaded from their website before it was rebuilt and migrated (the links have since disappeared - a pity, I liked the clearcut explanation on the difference between upsampling and oversampling there, for example). dCS upsamplers are based on a studio DDC, the dCS 972 - so much for the original "purpose": all they purportedly tried to achieve was to build a sample rate converter for recording studios/mastering facilities that could be daisy-chained if needed (changing sample rates back and forth) without audibly affecting the sound quality (lifting the hood reveals they see quite a need for computing horse power for their interpolation filters to do the deed, but the "how" has been kept a secret). A hilarious aside is that my contact at dCS told me it never crossed anyone's mind at dCS that a sample rate converter could be used in a home audio system (= what for?) until there was demand from […] customers (mind you, it wasn't a benign adjective like "audiophile" he used, but I forget and would hate to misquote…) in, if memory serves, Asia. You'd have to contact dCS, the customer support is very friendly and helpful, although I doubt they'll give away (m)any technical secrets…

 

Greetings from Switzerland, David.

[Superdad]

Did anyone read this article?

 

Upsampling vs. Oversampling for Digital Audio | Audioholics

 

Here are the summary points from the article.

 

There is a lot wrong with that article. Not the least of which is it was written at a time when "upsampling" required us of the terrible sounding asynchronous sample rate converters. These days, upsampling and oversampling are essentially the same thing. They have always just been forms of interpolation, it is just that oversampling is always done at multiples of the original rate. With modern s/w or h/w at high rates, the distinction is moot. What matters much more is the quality (parameters chosen) of the digital filter used, the type and quality of dither used (if any), and the ability of the DAC's input to accept the data at a rate that bypasses its own, resource-constrained interpolation filters.

 

Also, the article is wrong to claim that NOS DACs don't produce imaging artifacts down in the audible band. They sure do and they are often quite nasty. I say this as a person who has been using a NOS PCM1704K R2R ladder DAC for many years. But I up/over-sample with s/w to feed it at 352.8 or 384KHz.

[Jud]

There is a lot wrong with that article. Not the least of which is it was written at a time when "upsampling" required us of the terrible sounding asynchronous sample rate converters. These days, upsampling and oversampling are essentially the same thing. They have always just been forms of interpolation, it is just that oversampling is always done at multiples of the original rate. With modern s/w or h/w at high rates, the distinction is moot. What matters much more is the quality (parameters chosen) of the digital filter used, the type and quality of dither used (if any), and the ability of the DAC's input to accept the data at a rate that bypasses its own, resource-constrained interpolation filters.

 

Also, the article is wrong to claim that NOS DACs don't produce imaging artifacts down in the audible band. They sure do and they are often quite nasty. I say this as a person who has been using a NOS PCM1704K R2R ladder DAC for many years. But I up/over-sample with s/w to feed it at 352.8 or 384KHz.

 

The quoted portion of the Audioholics article is not great. There is no DAC that is completely "filterless." Believe me, you don't want to hear the unfiltered bitstream. I have, and it sounds like a waterfall of harsh static - a loud, obnoxious "white noise" sound, nothing like music. The digital bitstream, which consists of samples of the music taken 44,100 times per second, *must* be filtered to obtain music. The NOS DACs whose marketing claims no filtering do actually have filters (except the Phasure, which relies on filtering in the PC). For example, the Audio Note kits whose marketing says they have no filtering actually have their circuitry configured so there is a transformer that acts as the analog filter. This primitive filtering, with no upsampling/oversampling/interpolation, results in distortion that some people hear as a "warm" or "exciting" sound. This distortion is so evident that within a couple of years of the origin of the CD player, the industry had settled on a standard of 8x oversampling for DAC chips (44.1 interpolated to 352.8; 48 interpolated to 384) that is still in effect today.

 

There are only a tiny handful of NOS DACs made (Audio Note, Phasure, 47 Labs, Metrum Acoustics, perhaps a very few others). If you don't have one of these, there is a choice of three places to do the 8x interpolation: (a) "on the fly" in the PC; (b) "on the fly" in the DAC; © offline in the PC. If done "on the fly" in the DAC, most DACs accomplish this in three rounds of doubling, 44.1 -> 88.2 -> 176.4 -> 352.8. Also these days, all but this same tiny handful are what are called "sigma-delta" DACs, which "sigma-delta modulate" the input into a DSD-like format at a sample rate anywhere from a few mHz to tens of mHz, and there is the same choice of where to do this part of the conversion process.

 

The choice is also affected by the input rate your DAC allows. If your DAC takes a maximum 96kHz sample rate, then most of the interpolation (from 88.2 to 352.8kHz) and all of the modulation will be done in the DAC, even if you do, for example, interpolate in the PC to 88.2kHz. Or if your DAC has an ESS chip and a max input of 192kHz, then even if you interpolate in the PC to 192kHz, one step of the interpolation (from 192 to 384kHz) and all of the modulation up to 40(!) mHz or so will be done inside the ESS chip. On the other hand, if you have a DAC that allows DSD input, then all of the interpolation and most or all of the modulation can be done in the PC. So how much any external interpolation or modulation will affect the sound is very DAC dependent.

 

The two potential advantages of offline interpolation are these:

 

- You can do it without causing any increased activity or resource use in the PC or DAC while the music is playing; and

 

- The filtering (which must be applied after interpolation) can be more sophisticated, because it can take much more time compared to on the fly interpolation and filtering, which needs to happen quickly enough that you don't notice a pause during playback, or if you do it's brief enough you don't mind it.

[audiventory]

Is the premise here, to upsample offline, is that Audio files converted to the maximum audio sample rate supported by the DAC improves the sound quality vice letting the hardware do it in realtime?

 

Because the offline upsampling software may have audible effects that are less harmful than the hardware upsampling?

 

My way to try and digest this is, the analaogy is -thinking of your hot water heater in your house. It has hot water ready to go, just turn the faucet. This is just like offline upsampling. Both are inefficient, but both deliver to it's intended purpose.

 

This is contrasted to the hotwater on demand heaters, where they crank the heating elements to heat the water "on the fly" or on demand. This would be hardware upsampling.

 

Help me out, this is neat if it really makes a difference.

 

Hi Bill,

 

Sample rate conversion algorithm don’t improve quality. It only allow work DAC in «light» (I call it «optimal») mode - without additional processing in hardware.

 

Result oversampling here is improved sound.

 

But it is not guaranted result, of course.

 

From technical point of view no difference between software and hardware: both execute programs (algorithms).

 

1. Real-time algorithm limited in CPU resources.

 

You need provide constant big throughput of water. How you can heat (max temperature) water depend on power of your heater.

Less noise/distortions = more computer power

 

 

2. For better suppressing of conversion aliases (noise) used filters that demand big number of calculations.

Less noise = more calculations

 

 

3. Big number of calculations lead to accumulating of errors of quantization (noise due rounding errors). Thus better use float point formats, that has slow math.

Less noise = more slow calculation

4. Hardware can be limited in calculation precision (only integer, as example).

Hardware resources (FPGA, as example) can have limited logical elements number for realization of big algorithms.

Also how can be released algorithm depend on builtin math single of chip processor and its clock rate, available memory, its organizing (cache size, speed transfer form main memory to cache/registers,…).

From my experience moving PC algorithm to single of chip processor or FPGA is hard work with some compromises.

 

Hardware decisions: you have limited LEGO’s details

 

 

5. PC’s processors has great abilities for all these things. Here you can write software and almost don’t worry how it will linked with hardware resources.

However (for hardware algorithms too) when you give a little improving to filter (as example, better 5 … 10 dB aliases suppression) you can get great increasing of number of calculations.

Especially for values of feature that already HiEnd processing.

 

For HiEnd: Small improving = add many calculations

Best regards,

Yuri

[audiventory]

Just checked your dither a bit more in detail. It seems your dither is close to but not quite standard "TPDF 2LBS p2p" dither. There's an anomaly when upconverting from 16bit to 24bit (even with the lowest dither level): the noise level of your dither is still applied to levels at around -120db. From a technical standpoint with regard to "pure dither" this should not happen (with both of my PSP Audioware Dither plugins as well as with iZotrope and finally with Steinberg Wavelab's internal dither set to standard "TPDF 2LBS p2p" the noise level stays below -140db in 24bit). Now, -120db is of course beyond the audible range and the bit-meter shows no dropouts or so. But I wonder why your dither is acting that way?

 

Hi Copy_of_a,

 

For dithering I don’t follow some standards, only achieved result.

 

AuI’s dither was developed from zero (first generation algorithm).

 

After I get very good feedback from one my friend. He described it as «dither open more subtle details of sound».

He inspired me try dither to «from 16 bit» conversion. I added and he was glad of result.

 

Some time later I re-work dithering algorithm, where was researched influence of intensity of dithering (second generation with several improvements of the first).

 

Later I wrote the article about dithering.

 

As low level I accepted dithering intensity enough for smoothing traditional distortions for «to 16 bit» conversion.

 

After I generated 16 bit file and for «compensating» its distortions was need more high level. This level was accepted as maximal intensity.

 

I assume, difference between levels defined by place, where we apply distortion for rounding distortions eliminating:

 

1. Before converting "to 16 bit" we have correct signal.

Then we correctly round it to 16 bit.

We fix troubles before appearing.

 

2. After converting generated signal to 16 bit.

Then we try apply dither to errors during «from 16 bit» conversion.

We fix troubles after appearing.

Here need more intensity.

 

All dither intensity values appeared only from measurements, not from theory.

 

Best regards,

Yuri

[audiventory]

I have enough engineer friends who occasionally forget that science is nothing but work in progress, and that it is ultimately unscientific to claim something can't be because we lack proof.

 

In short, the attitude that's always baffled me is the one primarily exhibited by engineering know-alls: no, I will not hear a difference because there can't be one (and they're the ones who dare call "self-fulfilling prophecy" on those who do hear a difference).

 

Hi David,

 

Possible reason of difference (what you hear, but can't explain from technical point of view) is not in place that we think.

 

As example, somebody listen difference between binary identical files, or difference between FLAC and WAV.

 

Technically (comparing binary content, it is identical) file format in not reason of difference.

This I know exactly due I can repeat results of experiment in any time.

Thus reason placed in other place: how decoded, how playback, ..., subjective perception,..., non-pure experiment

 

Same things about resampling. As I noted in post above, oversampling don't improve sound in audio file, only allow DAC playback carefuly prepared sound. Here reason of sound improving after oversampling.

 

Best regards,

Yuri

[Bill Lord]

Yuri, Jud, and SuperDad, thank you, thank you for all your clear and detailed explainations. I sure hope others benefit from your responses, as it has helped me understand this technical process.

 

I am going to test and hopefully implement this process-if I can hear the difference.

[rodrigaj]

Yuri, Jud, and SuperDad, thank you, thank you for all your clear and detailed explainations. I sure hope others benefit from your responses, as it has helped me understand this technical process.

 

I am going to test and hopefully implement this process-if I can hear the difference.

 

I suggest, if you have a DSD Dac, trying the D128 setting with dither on light. Yuri's demo allows such a trial.

[dean70]

Currently testing offline up-sampling to 64 bit float/384k sample rate WAV files (the file sizes are huge!) using r8brain (free) against on-the fly up-sampling using JRiver. It is hard to pick a difference with the tracks I have tried so far.

[copy_of_a]

After I get very good feedback from one my friend. He described it as «dither open more subtle details of sound».

He inspired me try dither to «from 16 bit» conversion. I added and he was glad of result.

(...)

After I generated 16 bit file ...

Dear Yuri, many thanks for the detailed reply!

What I take from your post is that you've generated a "generic" 16bit dither file that you also use to dither to 24bit, is that correct?

So when converting from, say, 32bit floating point to 24bit integer the very same 16 bit dither-file is applied?

May I ask you if you use the same 16bit dither file also to convert from PCM to DSD?

[Jud]

Currently testing offline up-sampling to 64 bit float/384k sample rate WAV files (the file sizes are huge!) using r8brain (free) against on-the fly up-sampling using JRiver. It is hard to pick a difference with the tracks I have tried so far.

 

What is the rate you're upsampling to, and what is your DAC?

[dean70]

What is the rate you're upsampling to, and what is your DAC?

 

U/S to 384KHz through W4S DAC-2 DSDse DAC. Trying single tracks from 44.1k, 96k & 192k sources with r8brain and AuI ConverteR (free). Tested with an extreme example of 64 bit float (r8brain SRC) to match the internal JRiver DSP rate.

[audiventory]

What I take from your post is that you've generated a "generic" 16bit dither file that you also use to dither to 24bit, is that correct?

 

Hi Copy_of_a,

 

1. I generated sine 16-bit float and save it in 16-bit file without dithering.

 

We can see its spectrum on picture at page

What is dithering audio?

«Attention!!! Non-traditional case!», Source 16-bit sine generated without dithering

 

 

2. This 16-bit file was converted to 24 bit file with AuI’s dither alphaD.

Next pictures in the page show spectrums for different dither intensities.

 

 

So when converting from, say, 32bit floating point to 24bit integer the very same 16 bit dither-file is applied?

 

 

When both (input and output) bit depth more 16 bit, dithering automatically turn OFF (led can light - in list of input files may be files with different bit depths).

 

 

May I ask you if you use the same 16bit dither file also to convert from PCM to DSD?

 

Dithering AuI alpaD is same for all types of conversion. Possible in some cases dither ON during PCM to DSD conversion can give some sound improvements, like 16 to 24 bit.

 

Here need remember that most part of 16 bit stuff already dithered.

 

Currently testing offline up-sampling to 64 bit float/384k sample rate WAV files (the file sizes are huge!) using r8brain (free) against on-the fly up-sampling using JRiver. It is hard to pick a difference with the tracks I have tried so far.

 

Hi Dean70,

Using 64 bit float format of end-user file for playback is not recommended.

 

64 bit float format more match for music production. There applied big number of processing (for avoiding accumulated errors).

 

I suppose, you have inline converting 64-bit float to 32 bit integer (format DAC - I don't know: what your output settings?) in player and, possible, some processing in DAC.

 

May be DAC oversample input signal anyway.

Without reliable knowledges about algorithms, used in software and hardware, and measurements, we can assume only.

 

You can try:

 

1. Turn off any DSP in audio player (bit perfect playback).

 

2. Bit depth of converted file set equal maximal bit depth of DAC.

 

As example: If you use 24-bit DAC, better playback 24-bit file. Possibly DAC (as device) support 32 bit but chip support 24-bit only and we have inline truncating of bit depth to 24-bit (I don't know what inside your DAC, it's only assuming of course).

Here better use 24 bit instead maximal 32 bit.

 

Of course these advices and nobody don't guarantee audible improving of quality. It depend on combination software, hardware, its algorithms, mode, settings, individual perception, mood, music content, volume,…

Need consider many details.

 

Best regards,

Yuri

[dean70]

Hi Dean70,

Using 64 bit float format of end-user file for playback is not recommended.

 

64 bit float format more match for music production. There applied big number of processing (for avoiding accumulated errors).

 

I suppose, you have inline converting 64-bit float to 32 bit integer (format DAC - I don't know: what your output settings?) in player and, possible, some processing in DAC.

 

May be DAC oversample input signal anyway.

Without reliable knowledges about algorithms, used in software and hardware, and measurements, we can assume only.

 

You can try:

 

1. Turn off any DSP in audio player (bit perfect playback).

 

2. Bit depth of converted file set equal maximal bit depth of DAC.

 

As example: If you use 24-bit DAC, better playback 24-bit file. Possibly DAC (as device) support 32 bit but chip support 24-bit only and we have inline truncating of bit depth to 24-bit (I don't know what inside your DAC, it's only assuming of course).

Here better use 24 bit instead maximal 32 bit.

 

Of course these advices and nobody don't guarantee audible improving of quality. It depend on combination software, hardware, its algorithms, mode, settings, individual perception, mood, music content, volume,…

Need consider many details.

 

Best regards,

Yuri

 

Its an experiment at this stage to see what the benefits are. The DAC is Sabre32 9018 based (32 bit?) & 384k is the max PCM input rate supported. Player DSP is required for Convolution filters (phase and freq response), which is more important than any up-sampling.

[audiventory]

Its an experiment at this stage to see what the benefits are.

 

Hi Dean70,

 

Sound benefits are subjective thing. As rule, better sound some correlate with better measured parameters.

 

As example, somebody prefer vinyl. But digital audio systems has better features from technical point of view.

Somebody accuse digital stream's jitter, but is non even comparable with instability of mechanical systems.

 

Here own taste matter. Right both parties.

 

I don't deny any subjective impression. But try find its reasons.

 

In music production used amp emulators for improving «dry digital» sound. For adding «tube» sounding (additional «nice» distortions).

 

 

Therefore hardly talk here about benefits. Impossible proof anything.

 

Possibly only perform sample rate conversion with minimal (how we can) distortions.

 

I don't know what inside your DAC.

 

Maximal limit of external sample rates don't proof absent of intermediate frequencies for highest available output sample rates.

How really work the DAC know only its developers.

 

I don’t nothing claim. Only hypothesis.

 

I can’t capture real-time audio stream by software player (need special software).

I can’t measure output of DAC (need expensive enough precise measurement tool).

 

I’m only can measure parameters of software what offline convert files.

Need watch measured features and try link it with subjective perception.

 

If I haven’t result of measurements, I can only build hypothesis. Which need carefully check.

All software and hardware (except mine) for me are «black boxes».

 

Therefore anywhere I add «possibly» and «I suppose»

 

 

 

Player DSP is required for Convolution filters (phase and freq response), which is more important than any up-sampling.

 

It’s optional.

 

Convolution filters is IIR (Infinite impulse response) filters. Usually, it used only for sample rate conversion.

 

 

I don’t know what inside you player, possible there IIR or other filters or other kind of DSP used for EQ, correction or other purposes without "turn OFF" possibilities.

 

 

«Bit perfect» playback mean just transit audio file content to DAC. Without any DSP in player (EQ, plugins, ...)

 

Possibly, the player has such mode.

 

You can try set output sample rate (in settings of player) same to sample rate converted files for avoiding resampling.

 

EQ and plugins must be turned OFF.

 

Best regards,

Yuri

[mkrzych]

There are only a tiny handful of NOS DACs made (Audio Note, Phasure, 47 Labs, Metrum Acoustics, perhaps a very few others). If you don't have one of these, there is a choice of three places to do the 8x interpolation: (a) "on the fly" in the PC; (b) "on the fly" in the DAC; © offline in the PC. If done "on the fly" in the DAC, most DACs accomplish this in three rounds of doubling, 44.1 -> 88.2 -> 176.4 -> 352.8. Also these days, all but this same tiny handful are what are called "sigma-delta" DACs, which "sigma-delta modulate" the input into a DSD-like format at a sample rate anywhere from a few mHz to tens of mHz, and there is the same choice of where to do this part of the conversion process.

 

The choice is also affected by the input rate your DAC allows. If your DAC takes a maximum 96kHz sample rate, then most of the interpolation (from 88.2 to 352.8kHz) and all of the modulation will be done in the DAC, even if you do, for example, interpolate in the PC to 88.2kHz. Or if your DAC has an ESS chip and a max input of 192kHz, then even if you interpolate in the PC to 192kHz, one step of the interpolation (from 192 to 384kHz) and all of the modulation up to 40(!) mHz or so will be done inside the ESS chip. On the other hand, if you have a DAC that allows DSD input, then all of the interpolation and most or all of the modulation can be done in the PC. So how much any external interpolation or modulation will affect the sound is very DAC dependent.

 

 

So, to understand on my side this subject lets assume that the DAC has the clocking as below in the table and the maximum input accepted by USB is 24/96kHz, how to interpret it and how many steps is done for each particular rates up to 96kHz?

 

[Jud]

So, to understand on my side this subject lets assume that the DAC has the clocking as below in the table and the maximum input accepted by USB is 24/96kHz, how to interpret it and how many steps is done for each particular rates up to 96kHz?

 

[ATTACH=CONFIG]16785[/ATTACH]

 

Hi Krzysztof, please remind me what DAC you have. I think the answer in any case will likely be that you would gain fairly little advantage from sample rate conversion in the PC, online or offline, because you will simply substitute one step in the PC for one in the DAC chip. That is -

 

CD rip, no PC conversion, in the DAC you will get: 44.1 -> 88.2 -> 176.4 -> 352.8; then sigma-delta modulation to MHz rates, then conversion to analog.

 

CD rip, PC conversion 44.1 -> 88.2, in the DAC you will get: 88.2 -> 176.4 -> 352.8; then sigma-delta modulation to MHz rates, then conversion to analog.

 

CD rip, PC conversion 44.1 -> 96, in the DAC you will get: 96 -> 192 -> 384; then sigma-delta modulation to MHz rates, then conversion to analog.

 

Because of so many conversions in the DAC even if you do the first one in the PC, I think you will likely hear little difference. That is of course speculation on my part, so if you do the conversion and hear something better out of it, great.

[Bill Lord]

Hi- Here is a video on r8brain.

 

[Bill Lord]

Here is a website for comparing Sample Rate Converters:

 

SRC Comparisons