Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[pschelbert]

Hi
in foobar2000 there are two nice components for re sampling.

Resampler-V:
there is in foobar2000 a component which is called
Resampler-V
that one can do SSRC or SOX, passband, stopban attenuation, 3 Phase setting: linear. minimum, intermediate
Great is that you can see the plot of the frequency response and pulse response.
Wish there: Continuous phase adjust (not just linear, intermediate, minimum)

Resampler (SOX) mod2:
There is another one, which has only
-Quality
-Passband
-Phase (but continuous from linear to minimum)


Application:

To circumvent the internal DAC-filter.
If you play a CD at 44.1kHz, the DAC does set his internal filter to 44.1kHz. It may be that you do not like the filter (because its linear, not step enough , to steep etc.), or want to try a filter with another phase response.

How to do:
Upsample the signal for example to 192kHz with an upsampler which allows to select frequency and phase response (from linear to minimum).
Play the file. The DAC is setting his internal filter now for 192kHz, which has now as good as no effect in the range of up to 20kHz. The upsample-filter will by far dominate.
Result: The DAC internal filter is circumvented!!

Nicest solution would be: Integrate Resampler-V into JRiver MC22 and add continuous phase adjust to it.

Peter

[marko]

Hi
in foobar2000 there are two nice components for re sampling.

Resampler-V:
there is in foobar2000 a component which is called
Resampler-V
that one can do SSRC or SOX, passband, stopban attenuation, 3 Phase setting: linear. minimum, intermediate
Great is that you can see the plot of the frequency response and pulse response.
Wish there: Continuous phase adjust (not just linear, intermediate, minimum)

Resampler (SOX) mod2:
There is another one, which has only
-Quality
-Passband
-Phase (but continuous from linear to minimum)


Application:

To circumvent the internal DAC-filter.
If you play a CD at 44.1kHz, the DAC does set his internal filter to 44.1kHz. It may be that you do not like the filter (because its linear, not step enough , to steep etc.), or want to try a filter with another phase response.

How to do:
Upsample the signal for example to 192kHz with an upsampler which allows to select frequency and phase response (from linear to minimum).
Play the file. The DAC is setting his internal filter now for 192kHz, which has now as good as no effect in the range of up to 20kHz. The upsample-filter will by far dominate.
Result: The DAC internal filter is circumvented!!

Nicest solution would be: Integrate Resampler-V into JRiver MC22 and add continuous phase adjust to it.

Peter

I don't mean to be rude, so apologise in advance if this is taken that way, but... seriously?

I just read that whole post and understood not one single part of it.
Do the people who do understand it actually enjoy listening to music?
What I mean is, do you hear it, and it makes you happy because it reminds you of your first girlfriend, or of the time you saw them live at whatever stadium with all your mates... or...

Do you hear it and it makes you happy because you can tell it has continuous phase adjust, and if you could tell it was just linear, it make you mad?
Is this what SOX is all about?

I'm just a bit bemused and clearly way out of my depth here. Still, you should try reading some of this stuff from the outside looking in. It's mad.

-marko.

[kstuart]

This will do the job of getting rid of the DAC-internal filter (mostly)

The upsampling in the recent Schiit multibit DACs is better than anything you can do on PC, so you only want to bypass the DAC if it has poor resampling.

[JimH]

The upsampling in the recent Schiit multibit DACs is better than anything you can do on PC, so you only want to bypass the DAC if it has poor resampling.

That's a bold statement.

[kstuart]

That's a bold statement.

From their web site:
"... time- and frequency-domain optimized digital filter with a true closed-form solution. This means it retains all the original samples, performing a true interpolation. This digital filter gives you the best of both NOS (all original samples retained) and upsampling (easier filtering of out-of-band noise) designs."

Longer explanation from the designer:
"The below are the claims of the digital filter:

1. The filter is absolutely proprietary.

2. The development tools and coefficient calculator to derive the above filters are also proprietary.

3. The math involved in developing the filter and calculating has a closed form solution. It is not an approximation, as all other filters I have studied (most, if not all of them). Therefore, all of the original samples are output. This could be referred to fairly as bit perfect; what comes in goes out.

4. Oversimplified, however essentially correct: The filter is also time domain optimized which means the phase info in the original samples are averaged in the time domain with the filter generated interpolated samples to for corrected minimum phase shift as a function of frequency from DC to the percentage of nyquist - in our case .968. Time domain is well defined at DC - the playback device behaves as a window fan at DC - it either blows (in phase) or sucks (out). It is our time domain optimization that gives the uncanny sonic hologram that only Thetas and Schiit Multibits do. (It also allows the filter to disappear. Has to be heard to understand.) Since lower frequency wavelengths are measured in tens of feet, placement in image gets increasingly wrong as a function of decreasing frequency in non time domain optimized recordings - these keep the listener's ability to hear the venue - not to mention the sum of all of the phase errors in the microphones, mixing boards, eq, etc on the record side. An absolute phase switch is of little to no value in a non time domain optimized, stochastic time domain replay system. It makes a huge difference with an Schiit Yggdrasil DAC.

5. This is combined with a frequency domain optimization which does not otherwise affect the phase optimization. The 0.968 of nyquist also gives us a small advantage that none of the off-the shelf FIR filters (0.907) provide: frequency response out to 21.344KHz, 42.688KHz, 85.3776KHz, and 170.5772KHz bandwidth for native 1,2,4, and 8x 44.1KHz SR multiple recordings - the 48KHz table is 23.232, 46.464, 92.868, and 185.856KHz respectively for 1,2,4, and 8x. This was the portion of the filter that had the divide by zero problem which John Lediaev worked out in 1983, to combine with #4 above AND retain the original samples.

This is what the competition offers:

Frequency domain optimization FIR filters with Parks-McClellan optimization. The development tools for these types of filters can be downloaded for a price range of free to $300 on the internet. Parks-McClellan is the goto filter optimization for audio design. These filters are derived with no closed form math; only successive approximation. The original samples are lost. The output is approximated. An educated guess. This optimization is ubiquitous in the front end of delta sigma dacs as well as standalone digital filters. While there is no inherent phase shift within Parks-McClellan filters, there is no optimization of phase either. The listener is left with what remains from the mixing boards, transducers, brick-wall filters, etc which can and usually do destroy proper phase/position information. Finally, it is processor efficient and economical to implement. Read cheap.

Any avoidance of the Parks-McClellan pablum requires a lot of original DSP work. Am I a prophet who received the tablets from God or some other high-end audio drivel. Hell, no. I was the producer and director of this project and worked with Dave Kerstetter (hardware-software), John Lediaev (Math), Tom Lippiat (DSP Code), Warren Goldman (Coefficient Generator and development tools) for a total of 15 or so man years. These folks either taught math at The University of Iowa, Computer Science at Carnegie-Mellon University, worked at think tanks like the Rand Corporation – you get the idea. We did this for no money - What we all had in common was that we loved audio. All other audio pros were interested in Parks-McClellan and pointed and laughed at us. That's the way it happened.

It was worth it, every hour, day, and year. So go for it if you want. For what it is, it is not a lot of money."

FYI, this guy has been doing this longer than anyone - he designed the first separate DAC product, roughly 30 years ago.

[BillT]

I've got a bridge to sell you.....

[kstuart]

An audiophile friend who makes a hobby out of comparing DACs, and who uses JRiver exclusively, said the following after comparing the Schiit Bifrost Multibit DAC's internal filtering to JRiver MC's resampling (note that this was with the older SSRC of course):

"Schiit Filter >> JRiver upsampling to 192kHz. Upsampling closes the entire stage. Sense of soundwave propagation from instruments, especially bass (long wave lengths), separation, layering, positioning, angles, they get become discombobulated and simplified without the Schiit Filter. The sense of space is really amazing with the Filter. "

Of course, this is subjective and anecdotal.

[pschelbert]

The upsampling in the recent Schiit multibit DACs is better than anything you can do on PC, so you only want to bypass the DAC if it has poor resampling.

Hi

it not about upsampling. Its about to change the internal DAC-reconstruction filter.
The described trick let you do that.
You move the DAC-filter out to higher frequncies and insert the DAC-filter extrnally.
However with an external filter you can do eveything a Schiit does and much more with even more precision (64bit floating point!).
The sky is the limit...

Okay the question is if its really get better than what is in the DAC.

Peter

[blgentry]

Application:

To circumvent the internal DAC-filter.
If you play a CD at 44.1kHz, the DAC does set his internal filter to 44.1kHz. It may be that you do not like the filter (because its linear, not step enough , to steep etc.), or want to try a filter with another phase response.

I think you're probably misinformed about how modern DACs work.  Do you use a very old DAC?  Or one of the 2 or 3 or 4 modern DACs that use multi-bit DAC chips?  If you're using  a modern DAC that is not multi-bit, you can't "circumvent the filter", because there isn't one!  Modern DACs are Delta Sigma, so the steep filter you are describing is not required.

Brian.

[kstuart]

Hi

it not about upsampling. Its about to change the internal DAC-reconstruction filter.
The described trick let you do that.
You move the DAC-filter out to higher frequncies and insert the DAC-filter extrnally.
However with an external filter you can do eveything a Schiit does and much more with even more precision (64bit floating point!).
The sky is the limit...

Okay the question is if its really get better than what is in the DAC.

Peter

With an external filter you cannot "do everything a Schiit does and with even more precision".

The whole point of the Schiit filter is that it is the only filter that has perfect precision.  It is not an approximation.   Read the description above.

[pschelbert]

I think you're probably misinformed about how modern DACs work.  Do you use a very old DAC?  Or one of the 2 or 3 or 4 modern DACs that use multi-bit DAC chips?  If you're using  a modern DAC that is not multi-bit, you can't "circumvent the filter", because there isn't one!  Modern DACs are Delta Sigma, so the steep filter you are describing is not required.

Brian.

I use a sigma-Delta DAC, RME Fireface UFX.
It does upsampling additionally internally. There are steep digital filters used. Its an urgent need in any DAC to avoid aliasing (see details of the chip).
The trick shifts this filter to a higher frequency. The edge-effects are then much lower. The introduced filter with a lower edge frequency does dominate, hence the design of how it's done is then up to you. May be its not better then the built in ...

Peter

[pschelbert]

With an external filter you cannot "do everything a Schiit does and with even more precision".

The whole point of the Schiit filter is that it is the only filter that has perfect precision.  It is not an approximation.   Read the description above.

Hi

there is nothing like a perfect filter. Its all a compromise. Steepness versus ringing versus lin-phase or mixed- or minimumphase.
Also there is no perfect precision. A real implementation has always some precision restriction. However it does not hurt, as the DAC-chip is the limiting factor anyway in all these implementations. On the digital side the precision is way higher than what is possible in the analog domain.


Peter

[blgentry]

I use a sigma-Delta DAC, RME Fireface UFX.
It does upsampling additionally internally. There are steep digital filters used. Its an urgent need in any DAC to avoid aliasing (see details of the chip).

I think you're misinformed.  A Delta Sigma DAC changes multi-bit input into an internal Delta Sigma format, which is usually 1 bit or as many as 4 or 5 bits.  This is at a very very high frequency.  So any filter used at the output will be at a super duper high frequency.

Maybe I'm wrong.  I'm open to seeing evidence that a modern DAC uses a steep filter in the audible band (around 20kHz).  I'm pretty sure that none do, but show me one and I'll gladly tell you that I was wrong.

Brian.

[pschelbert]

Hi

yes thats right, but for the conversion to this superhigh frequency, antialiasing must happen, with a sharp steep filter.

[pschelbert]

I think you're misinformed.  A Delta Sigma DAC changes multi-bit input into an internal Delta Sigma format, which is usually 1 bit or as many as 4 or 5 bits.  This is at a very very high frequency.  So any filter used at the output will be at a super duper high frequency.

Maybe I'm wrong.  I'm open to seeing evidence that a modern DAC uses a steep filter in the audible band (around 20kHz).  I'm pretty sure that none do, but show me one and I'll gladly tell you that I was wrong.

Brian.

Hi Brian

thats the DA of RME UFX. ex BurrBrown, now TI

here you go:
http://www.ti.com/lit/ds/symlink/pcm4104.pdf
***************************************
Datasheet excerpt:
The    PCM4104    features    delta-sigma    architecture,
employing   a   high-performance   multi-level   modulator
combined  with  a  switched  capacitor  output  filter.

Linear Phase, 8x Oversampling Digital
Interpolation  Filter
*********************

See page 11 and page 13 for filter and blockdiagram

Peter

[kstuart]

Does upsampling 44 to 96 make sense in this rendition of sox on JR?

It totally depends on which DAC you are using.

If the DAC internally upsamples everything to 96khz, and if the DAC's internal SRC is worse than SOX, then yes.

If the DAC internally upsamples, but the DAC's internal SRC is better than SOX (such as Schiit's patented technology), then no.

If the DAC does not internally upsample, then definitely no.

[sorepinky]

Truly appreciate the awesome contributions of a digital genius here like kstuart.  An expertise acquired entirely from marketing literature is an invaluable asset to be envied and admired by all.

[kstuart]

Truly appreciate the awesome contributions of a digital genius here like kstuart.  An expertise acquired entirely from marketing literature is an invaluable asset to be envied and admired by all.

I have professional qualifications in both "digital" (your terminology) and audio.

You are likely using something right now that I helped design.

And now I am waiting for a comment that is not an Ad Hominem Fallacy:

https://en.wikipedia.org/wiki/Ad_hominem

[sorepinky]

And now I am waiting for a comment that is not an Ad Hominem Fallacy:
https://en.wikipedia.org/wiki/Ad_hominem

Given the credibility of some of your other source material, it isn't entirely surprising that the Wikipedia page commences with "This article's factual accuracy is disputed".

This thread is about SoX resampling.

For example, the 9018 ESS Sabre DAC supposedly takes all PCM sources, no matter what their sample rate, and upsamples to 84.672MHz.  How can MC's upsampling from say 44.1kHz using SoX's algorithms to 176.4kHz make an audible difference compared to using SSRC's algorithms, when the DAC will upsample 480 times higher to 84.672MHz anyway?

I only ask because after a few weeks listening, I kind of think with SoX the sound is pretty good.

[JimH]

kstuart,
Please don't market anything here.

sorepinky,
Please don't provoke him.

[pschelbert]

Given the credibility of some of your other source material, it isn't entirely surprising that the Wikipedia page commences with "This article's factual accuracy is disputed".

This thread is about SoX resampling.

For example, the 9018 ESS Sabre DAC supposedly takes all PCM sources, no matter what their sample rate, and upsamples to 84.672MHz.  How can MC's upsampling from say 44.1kHz using SoX's algorithms to 176.4kHz make an audible difference compared to using SSRC's algorithms, when the DAC will upsample 480 times higher to 84.672MHz anyway?

I only ask because after a few weeks listening, I kind of think with SoX the sound is pretty good.

Hi

all the Sigma-Delta DAC do convert the PCM to a very fast 1 or 4-5bit serial bitstream. They need antialising filters. As I said in this thread, this antialising filter can be moved out to higher frequency (2-4x only in most DACS, limit often is 192kHz sample-rate). The point is, if you want to get rid (mostly) of the effect of this filter in the audible range, you upsample with SOX or other tool and introduce your own antialising filter which introduces the changes you want or prefer (amplitude and phase, time smearing. There is no way to do any "perfect" filter though. There are just theoretical limits (up to now nobody jumped over his shadow...). Its just a trade-off. And yes its not clear if you can do any better than the professionals like AD, TI, AKM, ESS...
The nice thing its just software, no soldering or hardware changes needed.
SOX is for sure a great algorithm.

Peter

[blgentry]

all the Sigma-Delta DAC do convert the PCM to a very fast 1 or 4-5bit serial bitstream. They need antialising filters. As I said in this thread, this antialising filter can be moved out to higher frequency (2-4x only in most DACS, limit often is 192kHz sample-rate).

This isn't the place to discuss it, but you're just wrong.  Antialiasing filters far above the audible band aren't audible.  That's why oversampling was invented!  But I won't post about this here any more if you won't.  If you want to discuss it, maybe we should start another thread.

Brian.

[BillT]

I thought over sampling was invented to improve the performance of DACs.

Back in prehistory and the start of CD, the original Philips idea was to use a 14 bit system, which was entirely adequate for domestic audio and easier to produce. Sony wanted a bit of wiggle room with a 16 bit system. In the end a 16 bit system was developed between them, but could be used with lower resolution DACs.

Sony could make 16 bit DACS, Philips could only make 14 bit DACs but they had clever engineers who worked out an oversampling system which could deliver 16 bit performance from a 14 bit DAC.

The largely pointless bigger number race has continued ever since.

[blgentry]

When I said "oversampling" I meant converting the input sample rate (44.1kHz for CD) to a higher sample rate like 88.2 or 176.4kHz.  That technique (upsampling or overampling) was invented to shift the aliasing frequency FAR outside of the audible band, which allows a more gentle low pass filter at frequency way outside the range of human hearing.

Brian.

[pschelbert]

This isn't the place to discuss it, but you're just wrong.  Antialiasing filters far above the audible band aren't audible.  That's why oversampling was invented!  But I won't post about this here any more if you won't.  If you want to discuss it, maybe we should start another thread.

Brian.

Hi

if you upsample from CD (44.1kHz) you need an antialisingfilter just to cut at 22050Hz. This is a steep filter which must not damage anything in the audible range 20-20000Hz (passband). So thsi filter is very close to the audible range. There are no doubt other filters in the DAC which are way highter than audible range.
Such a filter is implemented in SOX, which in turn is adjustable (its in the SOX command line version).
I understand its only for experts, may be not really adequate here, and I understand it will not be soon implemented.

Peter

[blgentry]

if you upsample from CD (44.1kHz) you need an antialisingfilter just to cut at 22050Hz.

Incorrect.  If you do straight DA conversion at 44.1kHz you need that filter.  If you *upsample* to 88.2 or 174.6, then your filter needs to be at half of the *upsampled* frequency.  So 44.1kHz or 88.2kHz respectively.  Which are far outside the audible band.

Again, this is why upsampling was invented in DACs.  So that the filter would be outside the audible band.

All modern D/S DACs use a filter that's much higher than 22.05 kHz.

Brian.

[pschelbert]

Incorrect.  If you do straight DA conversion at 44.1kHz you need that filter.  If you *upsample* to 88.2 or 174.6, then your filter needs to be at half of the *upsampled* frequency.  So 44.1kHz or 88.2kHz respectively.  Which are far outside the audible band.

Again, this is why upsampling was invented in DACs.  So that the filter would be outside the audible band.

All modern D/S DACs use a filter that's much higher than 22.05 kHz.

Brian.

no, as CD has alising from 22050Hz on, if you upsample you generate just multiples (inserting zeros between samples). This has to be filtered before DA conversion, in the digital domain which was the idea to do it in the upsampled DACs. The analog filter can then be very relaxed. Just consult a DSP-book, then you see the effect of just upsampling without filtering. You can do that ss well with software Soundforge, just insert zeros or insert zeros and filter. Observe the spectrum..

Peter

[blgentry]

no, as CD has alising from 22050Hz on, if you upsample you generate just multiples (inserting zeros between samples).

That's not how upsampling works.  You don't just "insert zeros".  Upsampling generates NEW SAMPLES between the existing samples.  If you upsample from 44.1kHz to 88.2kHz, you generate a new sample in between each existing pair of samples.  If you upsample from 44.1kHz to 176.4kHz, you generate 3 new samples in between each existing pair of samples.

That's exactly what SoX and other SRC algorithms do when they convert from a lower sampling rate to a higher one.  Correspondingly, the aliasing noise is shifted UP to 1/2 of the new sampling rate.

Question for you:  Why do you think upsampling is desirable?  It's entire purpose inside of DACs is to shift the aliasing noise up.  Then you can use a more gentle filter at a very high frequency that is not audible.  If this were not true, why would we want to upsample in the first place?

Brian.

[flac.rules]

Question for you:  Why do you think upsampling is desirable?  It's entire purpose inside of DACs is to shift the aliasing noise up.  Then you can use a more gentle filter at a very high frequency that is not audible.  If this were not true, why would we want to upsample in the first place?

Brian.

Is it? It also reduces noise, which I would say is an important purpose of oversampling.

[blgentry]

^ What noise? Aliasing noise?

Brian.

[dtc]

A couple of informal comments.

In the early days, Phillips wanted to increase the sound quality of its CD players. There are many reports that they did this by putting in 3 extra samples between the 44.1KHz samples and those extra samples were is fact simply zeros. I do not know the details of what they did, but since there are so many reports of them adding 0's I would not dismiss it out of hand. In general we think of upsampling as some sort of interpolation, but there is at least some though that Philips did it differently.

We should distinguish between early  upsampling and todays DACs. Early upsampling CD players did upsample to 96KHZ and 192KHz, for example. I still have one that has a switch for 96 or 192. That was definitely a way to use a less sharp filter and to move the filter noise to higher frequencies.  When these CD players were also used with external sources  or when stand alone DACs used these techniques, they also used the internal  clock and thus were a way to lower jitter associated with the input.

Most modern DACs use Sigma Delta techniques to upsample to extremely high sample rates in the MHz range using 1 bit or a few bits. Some DACs upsample internally to some intermediate frequency like 192 KHz or 384 KHz before doing to the Sigma Delta process. Sox can be used to bypass that process, if you think Sox is better at it than the DAC. But that also requires knowing if the DAC does this intermediate upsampling and, if it does, what sample rate it upsamples to internally. For example, using Sox to upsample to 192 KHz when the DAC upsamples to 352 KHz would probably hurt the process. This depends entirely on the particular DAC. Likewise, upsampling to 192 KHz may have little or no effect if the DAC does the Sigma Delta process on the original 44.1 KHz signal.

Discussing moving the filtering to higher frequencies is appropriate to DACs that upsample to 96 KHz and 192 KHz, for example. However, it is only part of the process with Sigma Delta DACs, since the noise shaping of those DACs is often not a simple slow row off low pass filter. It is often a finely detailed low pass filter and many DACs offer different versions of those filters. The filters definitely move the side effects of the filter to higher frequencies but they also deal with other effects, like time smearing.

The discussion above about the time smearing algorithms that Schiit uses should not be simply dismissed as marketing hype. They claim to be using very different techniques than most other manufacturers and they have good credentials for their claims. That does not mean they are the best that can every be done, but they should not be dismissed. Their DACs with the new techniques are receiving very good reviews by a lot of users.

Bottom line, for me, is that this topic is worthwhile and it is also more complicated than most people realize. The Sox upsampling may or may not be useful depending on what the DAC does internally and the whole Sigma Delta process and the subsequent filters may be far more important than the initial software upsampling.

[dtc]

^ What noise? Aliasing noise?

Brian.

When I see this discussed it is usually in terms of quantization noise.

[blgentry]

When I see this discussed it is usually in terms of quantization noise.

Isn't quantization noise a function of the number of bits per sample?  Not of the sampling rate?  So upsampling should have no effect on quantization noise.  I'm pretty sure that's right.

Brian.

[Hendrik]

Isn't quantization noise a function of the number of bits per sample?  Not of the sampling rate?  So upsampling should have no effect on quantization noise.  I'm pretty sure that's right.

If the algorithm is really smart it can move the noise into higher frequencies by using noise shaping, which is only possible if you have a higher sample rate.

[flac.rules]

^ What noise? Aliasing noise?

Brian.

Quantization noise, as others have stated (in conjunction with noise shaping). (not that I am necessarily defending the usefulness or quality of the DAC mentioned in the thread)

[BillT]

If the algorithm is really smart it can move the noise into higher frequencies by using noise shaping, which is only possible if you have a higher sample rate.

Doesn't that only affect noise in the ADC-DAC system? In real world recordings noise will be dominated by acoustic or pre-amp noise i.e. noise generated before the ADC, so noise shaping is going to be rather academic, especially if you use 18 bits or more.

[flac.rules]

Doesn't that only affect noise in the ADC-DAC system? In real world recordings noise will be dominated by acoustic or pre-amp noise i.e. noise generated before the ADC, so noise shaping is going to be rather academic, especially if you use 18 bits or more.

I am sure many will argue that having a high quality DAC in general is pretty academic when it comes to the final playback quality.

[pschelbert]

A couple of informal comments.

In the early days, Phillips wanted to increase the sound quality of its CD players. There are many reports that they did this by putting in 3 extra samples between the 44.1KHz samples and those extra samples were is fact simply zeros. I do not know the details of what they did, but since there are so many reports of them adding 0's I would not dismiss it out of hand. In general we think of upsampling as some sort of interpolation, but there is at least some though that Philips did it differently.

We should distinguish between early  upsampling and todays DACs. Early upsampling CD players did upsample to 96KHZ and 192KHz, for example. I still have one that has a switch for 96 or 192. That was definitely a way to use a less sharp filter and to move the filter noise to higher frequencies.  When these CD players were also used with external sources  or when stand alone DACs used these techniques, they also used the internal  clock and thus were a way to lower jitter associated with the input.

Most modern DACs use Sigma Delta techniques to upsample to extremely high sample rates in the MHz range using 1 bit or a few bits. Some DACs upsample internally to some intermediate frequency like 192 KHz or 384 KHz before doing to the Sigma Delta process. Sox can be used to bypass that process, if you think Sox is better at it than the DAC. But that also requires knowing if the DAC does this intermediate upsampling and, if it does, what sample rate it upsamples to internally. For example, using Sox to upsample to 192 KHz when the DAC upsamples to 352 KHz would probably hurt the process. This depends entirely on the particular DAC. Likewise, upsampling to 192 KHz may have little or no effect if the DAC does the Sigma Delta process on the original 44.1 KHz signal.

Discussing moving the filtering to higher frequencies is appropriate to DACs that upsample to 96 KHz and 192 KHz, for example. However, it is only part of the process with Sigma Delta DACs, since the noise shaping of those DACs is often not a simple slow row off low pass filter. It is often a finely detailed low pass filter and many DACs offer different versions of those filters. The filters definitely move the side effects of the filter to higher frequencies but they also deal with other effects, like time smearing.

The discussion above about the time smearing algorithms that Schiit uses should not be simply dismissed as marketing hype. They claim to be using very different techniques than most other manufacturers and they have good credentials for their claims. That does not mean they are the best that can every be done, but they should not be dismissed. Their DACs with the new techniques are receiving very good reviews by a lot of users.

Bottom line, for me, is that this topic is worthwhile and it is also more complicated than most people realize. The Sox upsampling may or may not be useful depending on what the DAC does internally and the whole Sigma Delta process and the subsequent filters may be far more important than the initial software upsampling.

Hi

if you insert zeros and then go through an interpolation filter (convolution) or you calculate the intermediate samples in a different way, is an implementation issue. Mathematics behind is the same.

What you need in any case is an antialisingfilter in the digital domain, for any upsampling to remove the alias frequencies. This is mandatory in any type of upsampling, no matter how the converter otherwise works. If you upsample a CD, the antialiasfilter must cut off to lets say -100dB to -130dB above 22050Hz.

Peter

[pschelbert]

I am sure many will argue that having a high quality DAC in general is pretty academic when it comes to the final playback quality.

However its true if you have a DAC from one of the big fours (TI, AKM, ESS, AD) you have not to worry, they know what they do and its execellent performance of all these chips. They filter correct and some have even user selectable filters (minphase, linear, sharp cut off, slow cut-off etc.)

Peter

[kstuart]

I am sure many will argue that having a high quality DAC in general is pretty academic when it comes to the final playback quality.

That would be fine if we listened to words, but in listening to music in 2016, there is more of a difference between DACs than any other part of the audio chain.  (And this difference is no longer a matter of pricing.)

However its true if you have a DAC from one of the big fours (TI, AKM, ESS, AD) you have not to worry, they know what they do and its execellent performance of all these chips.

The chips are fine as far as they go, the implementation can make a significant difference.

People seem to have a blind trust that someone selling a device has a wide and deep knowledge of electronics and also has not cut any corners in order to increase profit margin (cf Bay Bridge bolts, Challenger o-rings, etc.).

[blgentry]

If you upsample a CD, the antialiasfilter must cut off to lets say -100dB to -130dB above 22050Hz.

As far as I know, this is incorrect.  Increasing the sampling frequency, by doing Sample Rate Conversion, increases the aliasing frequency to 1/2 of the new sampling frequency.  So if you upsample (oversample, SRC) 44.1kHz audio to 352.8 kHz, the aliasing frequency is now 352.8/2 = 176.4 kHz.  Not 22.05kHz.  Notice that I said "sample rate conversion".  This is not the same as simply inserting zeros.

If I'm incorrect, please provide a reference to back up your claim.

Brian.

[pschelbert]

In order to do your sample rate conversion, first you need to antialias at 22050Hz. The resulting upsampled spectra has then the alias at 96kHz for example if your samplerate is 192kHz.

here:
https://www.google.ch/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&uact=8&ved=0ahUKEwiAzZyrpKHOAhXFC8AKHZnZAaMQFgguMAE&url=http%3A%2F%2Fwww.cppsim.com%2FBasicCommLectures%2Flec10.pdf&usg=AFQjCNGj64cXeo1Db4AnuTU7lo8d9xMwHg

or here:
http://melodi.ee.washington.edu/courses/ee518/notes/lec9.pdf

or here:
http://src.infinitewave.ca/help.html

Peter

[flac.rules]

That would be fine if we listened to words, but in listening to music in 2016, there is more of a difference between DACs than any other part of the audio chain.  (And this difference is no longer a matter of pricing.)

The chips are fine as far as they go, the implementation can make a significant difference.

People seem to have a blind trust that someone selling a device has a wide and deep knowledge of electronics and also has not cut any corners in order to increase profit margin (cf Bay Bridge bolts, Challenger o-rings, etc.).

I am not sure what you mean? Your posts are words, my posts are words, It seems like everyone here is talking about perceivable differences in music reproduction. That fact does't change because we use words to communicate. Anyway, its is hard to find any scientific backing for such a claim, differences between reasonable DAC designes are pretty small, especially when compared to things like loudspeakers, which clearly are a more important part of the chain. But this might be drifting a bit of topic, so I will concentrate on the technical aspects the digital domain and DA-conversion, as I see that as the threads main topic.

[pschelbert]

I am not sure what you mean? Your posts are words, my posts are words, It seems like everyone here is talking about perceivable differences in music reproduction. That fact does't change because we use words to communicate. Anyway, its is hard to find any scientific backing for such a claim, differences between reasonable DAC designes are pretty small, especially when compared to things like loudspeakers, which clearly are a more important part of the chain.

Hi

I made some listening test with several users in a german forum. Two tests actually. Now I am doing a third test. Result up to now:
Difference to be heard is very difficult, even in a mediocre ultra cheap DAC (around USD50). Measurements are not really good of that DAC, however obvioulsy HighEnd users cannot distinguish!
A third test is going on with really high priced DAC. Lets see.
Conclusion up to now: buy a professional DAC, one which is used in studios, like RME, Lynx, Focusrite, Motu, Apogee etc.
Or go with one of the Japanese brands for reasonable price, which know how to do electronics (Yamaha, Sony, Teac, Marantz, Denon etc.)
Invest your money otherwise in Loudspeakers and headphones of high quality (not costly ones but good ones!): Stax, Sennheiser, Beyer, AKG

I agree, the difference in sound reproduction is significant in loudspeakers and headphones, while DAC does a minimal change, as far as you have a good, even rather cheap DAC.
Driver is key: ASIO works great in Windows (Mac probably kernel is okay)

There are some strange tube DAC out there, highly priced, yet bad performance, so beware...
Get the datasheet with frequency response, THD, IMD, SPDR etc. if you get it... if not stay away

Peter

[blgentry]

In order to do your sample rate conversion, first you need to antialias at 22050Hz.

Hmm, OK.  That makes some sense I suppose.

The links you provided are kind of hard to follow, but let's set that aside.

All of your writing about this, is all about the digital lowpass filter that's applied before sample rate conversion then.  But any DAC that upsamples (which all modern ones do) is doing the same thing.  So all of your discussion is about filter parameters of an alternative low pass filter implemented in software as part of SRC.

I'm kinda scratching my head as to why you think this matters.  Particularly when you go on to say that in your tests people can't tell the difference between DACs.  You seem to be arguing two sides:  One that says all DACs are the same and another that says that minor changes in a filter (that would normally be inside the DAC) are audible.

Well, if nothing else, I think I learned something.  Though I'm still confused by absolute details of SRC when upsampling.  I thought new samples were calculated by successive approximation.  At least one of the links talks about inserting zeros and then low pass filtering the result. Which seems all kinds of wrong.

EDIT:  I take it back.  Anti-alias filtering before Sample Rate Conversion (up) makes no sense.  Why in the world would DAC manufacturers have implemented SRC in a DAC if it *still* had to have a steep filter in the audible band?  Why not just use the digital brick wall filter at 22.05 and be done with it?  There would be no reason to upsample in the first place.  I stand by my statement until further proof is presented.

Brian.

[pschelbert]

Hi

there are different ways of doing upsampling. one is inserting zeros and then convolute with a filter to do the approximation, others are calculating with spline or other fitting tools the samples. How SOX for example does it, I do not know.

Upsampling makes the analog reconstruction filter easy. Thats it. The hard job is done by digital filtering then a relaxed analog filter does the reconstruction (remove of the stairs and the noise).
There is a lot of theory and explanation for that topic of AD and DA conversion in the links I provided and elsewhere (a traditional book  on this is "Oppenhein Schafer, discrete time signal processing", the classic book).
See: second edition 1998/1999, page210: Oversampling and DA Conversion

If you do that in the chip or externally with software, it is more or less the same result.

Sound of DAC (prelim results):
Thats just the result of a blind test versus the very real original sound file.
People in a blind test got everything wrong, they could not tell which file is right not even hear a difference of two different files , one the original, one the file through the DAC.

If there is a difference I hear from people in forums always "you can hear a sound difference even if you cannot measure".
What the test gets clear: you cannot hear a lot, even if the soundfile is rather different. You can measure much more and much better than anybody can hear.
This is valid for DAC. A reasonable good DAC is by far better than anybody can hear.

Loudspeaker, headphone, microphone are a much more complicated world, its 3D, so not as easy to characterize....

Peter