Topic: Sample rates

Basic question, please help.

I was playing around with sample rates. I drive active speakers through RME ADI 2, using Tidal.

-the audio midi interface on my Macbook gives an option to sample at up to 768 kHz on the RME output

-Tidal max is 192 kHz

When I use the 768 kHz setting I naively thought that it would give higher quality sound but it's terrible.

When switching the sample rate back to 192 kHz, the sound is as it should be

Can some please tell me - is it best to match the sample rate on the Macbook output to the sample rate of that particular track in Tidal (ranges from 44.1 to 192)?

Why is this - is the DAC trying to oversample if its rate is higher, which degrades the audio through the speakers?

Thanks.

2 (edited by waedi 2024-08-17 01:26:57)

Re: Sample rates

Not the DAC but the Mac is changing the sample rate of the music file to fit the output setting.
The output setting (the sample rate set for the interface) is best when matching the music file, as you found out.

M1-Sequoia, Madiface Pro, Digiface USB, Babyface silver and blue

3 (edited by KaiS 2024-08-17 06:16:15)

Re: Sample rates

1. Resampling - up or down - is a destructive process that changes the audio.

Sample rate is not “the higher the better”, but playing a file at its native rate keeps its integrity.


2. A computer’s software based resampling doesn’t necessarily use the best algorithm available.
It’s more like offering the functionality with as low processing power as possible.


3. ADI-2 offers all it”s functionality unlimited up to 192 kHz.
Above available DSP-power needs to cut down on the features, see ADI-2 DAC’s manual page 16.

This might have been the major point and is the biggest sound change from 192 to 768 kHz sample rate:
EQ, Loudness function and other DSP processings are switched off.
Only if you anyway didn’t use them it makes no difference for ADI-2.


4. Personally I do hear a bettering sound difference (not night and day) going from 44.1 to 96 kHz.
This is with ADI-2/4 Pro SE and a high end analog turntable connected - listening to analog vinyl discs.

Further up the sound doe not improve or change any more for me.
Still, for that purpose I do use 192 kHz - why not if it’s there smile


5. When streaming - Tidal - I use the automatic switching based on the audio track’s SR, my combination of iPhone / RME ADI-2 offers.

Re: Sample rates

waedi wrote:

Not the DAC but the Mac is changing the sample rate of the music file to fit the output setting.
The output setting (the sample rate set for the interface) is best when matching the music file, as you found out.

Thanks for your answer. To be clear, I can set the sample rate on the Midi interface on the Mac, and the RME will mirror it.

Re: Sample rates

KaiS wrote:

1. Resampling - up or down - is a destructive process that changes the audio.

Sample rate is not “the higher the better”, but playing a file at its native rate keeps its integrity.


2. A computer’s software based resampling doesn’t necessarily use the best algorithm available.
It’s more like offering the functionality with as low processing power as possible.


3. ADI-2 offers all it”s functionality unlimited up to 192 kHz.
Above available DSP-power needs to cut down on the features, see ADI-2 DAC’s manual page 16.

This might have been the major point and is the biggest sound change from 192 to 768 kHz sample rate:
EQ, Loudness function and other DSP processings are switched off.
Only if you anyway didn’t use them it makes no difference for ADI-2.


4. Personally I do hear a bettering sound difference (not night and day) going from 44.1 to 96 kHz.
This is with ADI-2/4 Pro SE and a high end analog turntable connected - listening to analog vinyl discs.

Further up the sound doe not improve or change any more for me.
Still, for that purpose I do use 192 kHz - why not if it’s there smile


5. When streaming - Tidal - I use the automatic switching based on the audio track’s SR, my combination of iPhone / RME ADI-2 offers.

Thanks for your answer.

I run my set up in a way which audiophiles would disapprove of! Using the RME's default Loudness function gives remarkable sound on my active ATCs as they are well known for really only coming alive at higher SPLs. When oversampling, you state that all DSP in RME is off, and that's why the ATCs go back to sounding very flat.

Out of interest, if oversample the output at 768kHz on a 192kHz Tidal stream but at much higher volume where the ATCs start to work, would I hear any degradation in sound?

Re: Sample rates

waedi wrote:

To be clear, I can set the sample rate on the Midi interface on the Mac, and the RME will mirror it.


The "Audio-Midi-Setup" application in the MacOS, there is a place to select the sample rate.
It's the same as in the RME Fireface Settings, those two are the same setting.
You can use one or the other program to change the sample rate.
Follow what Kai says.

M1-Sequoia, Madiface Pro, Digiface USB, Babyface silver and blue

7 (edited by KaiS 2024-08-17 21:56:48)

Re: Sample rates

MrBalham wrote:

[I run my set up in a way which audiophiles would disapprove of! Using the RME's default Loudness function gives remarkable sound on my active ATCs as they are well known for really only coming alive at higher SPLs. When oversampling, you state that all DSP in RME is off, and that's why the ATCs go back to sounding very flat.

Out of interest, if oversample the output at 768kHz on a 192kHz Tidal stream but at much higher volume where the ATCs start to work, would I hear any degradation in sound?

ATCs if I recall correctly, are studio monitors.
They are intended to sound “flat”, in the sense of neutral, kind of.
If they would sound hyped in any way, the resulting productions made on them would sound boring, as they would project a wrong picture of the recording.

In studios the standard listening level is 80-85 dBSPL, that’s quite loud, too loud for most domestic users.

So, using ADI-2’s loudness for adaption is a valid way to go.
If set properly, Loudness even goes partly or fully out of the way when dialing up louder.
Adjust it to taste.


According SRs:
I don’t see a point in using anything above 192 kHz, to the contrary there’s no advantage, only downsides.
Upsampling 192 kHz to 768 kHz using a minor quality algorithm doesn’t make sense, the result is a degradation.

The reason I explained in my posting above.

Re: Sample rates

A lot of lies. Dear God I have no idea where to start or if to even chime in. Considering the rotten responses I have gotten the last few times I gave you guys valuable information, suffice to say, there are levels to this game, and so far you have failed. Upsampling is a thing. There are usually only advantages, not downsides.  Please do not take anything waedi or KaIS write serious. Always work with the best quality, products and resolution available if you can. You are welcome MrBalham.

9 (edited by Muffin 2024-08-27 17:37:19)

Re: Sample rates

torbenscharling wrote:

A lot of lies. Dear God I have no idea where to start or if to even chime in. Considering the rotten responses I have gotten the last few times I gave you guys valuable information, suffice to say, there are levels to this game, and so far you have failed. Upsampling is a thing. There are usually only advantages, not downsides.  Please do not take anything waedi or KaIS write serious. Always work with the best quality, products and resolution available if you can. You are welcome MrBalham.

What?

I take what waedi and KaIS writes far more seriously than what you write for the simple reason that they have earned it.

Re: Sample rates

torbenscharling wrote:

A lot of lies. Dear God I have no idea where to start or if to even chime in. Considering the rotten responses I have gotten the last few times I gave you guys valuable information, suffice to say, there are levels to this game, and so far you have failed. Upsampling is a thing. There are usually only advantages, not downsides.  Please do not take anything waedi or KaIS write serious. Always work with the best quality, products and resolution available if you can. You are welcome MrBalham.

Hi
could you please provide comprehensible technical facts especially related to how ADI 2 handles PCM signals for the fools?

Thanks!

Re: Sample rates

torbenscharling wrote:

A lot of lies. Dear God I have no idea where to start or if to even chime in. Considering the rotten responses I have gotten the last few times I gave you guys valuable information, suffice to say, there are levels to this game, and so far you have failed. Upsampling is a thing. There are usually only advantages, not downsides.  Please do not take anything waedi or KaIS write serious. Always work with the best quality, products and resolution available if you can. You are welcome MrBalham.

As the saying goes: "What goes around, comes around."

The tone you're using here is completely unacceptable.

KaiS appears to have a depth of knowledge that might exceed your ability to comprehend. Insulting others due to a potential learning difficulty is far from courteous.

If you've lost the ability to self-reflect and can no longer communicate with others appropriately, it might be best for you to stay away from online forums. That would benefit both you and the other participants here.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

12

Re: Sample rates

torbenscharling wrote:

A lot of lies. Dear God I have no idea where to start or if to even chime in. Considering the rotten responses I have gotten the last few times I gave you guys valuable information

For the record: you only got backlash when you wrote a stupid thread about RME interfaces having 'crippled ports' at higher sample rates. This was and is complete nonsense that no one understood. You seem to live in a parallel world with this kind of thinking.

torbenscharling wrote:

Upsampling is a thing.

In some very specific cases, yes. But not in general as you claim here.

torbenscharling wrote:

There are usually only advantages, not downsides.

Sure. That's why the whole world constantly upsamples. Not.

torbenscharling wrote:

Please do not take anything waedi or KaIS write serious.

Why should anyone take your words serious the way you write them? Apart from that here comes the parallel world again, claiming an experienced and reputable sound enginieer and studio owner to have no clue what he says. Mind blowing!

Regards
Matthias Carstens
RME

13 (edited by beat8000 2024-08-28 22:46:37)

Re: Sample rates

I have tested a lot of upsamplers and resamplers
To my opinion e.g. PGGB-RT has a really good sound but it takes about 14 seconds on my computer to recalculate all 17 tracks of one CD for 768 kHz:

https://mobidrive.com/sharelink/i/YmVhd … XRt0OqemIR

Additionally a rather high input for Analog is displayed  if I use a rate higher than 192 kHz

https://mobidrive.com/sharelink/i/YmVhd … KuklzQIw8t

I think there should be no Analog Input

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

14

Re: Sample rates

It's in the manual: High frequency noise at higher sample rates, measurable but not audible.

Regards
Matthias Carstens
RME

Re: Sample rates

beat8000 wrote:

I have tested a lot of upsamplers and resamplers
To my opinion e.g. PGGB-RT has a really good sound but it takes about 14 seconds on my computer to recalculate all 17 tracks of one CD for 768 kHz:

https://mobidrive.com/sharelink/i/YmVhd … XRt0OqemIR

Upsampling is just another form of digital filtering to avoid antialiasing on systems that have poor DA.

If audio signal is sampled 44.1 kHz, no upsampling, resampling, digital filtering is not able to add any information to signal. It may only better or worse mimic ideal filter which has to cut off frequencies above 22.05 kHz to avoid aliasing.

And there is no ideal solution for that case by definition. There are always compromises by definition, one cannot beat mathematical reality.  One may prefer one method, the other another, but none of them is perfect, none of them is "right".

16 (edited by beat8000 2024-08-29 09:43:06)

Re: Sample rates

Kubrak wrote:

If audio signal is sampled 44.1 kHz, no upsampling, resampling, digital filtering is not able to add any information to signal. It may only better or worse mimic ideal filter which has to cut off frequencies above 22.05 kHz to avoid aliasing.

The original CD in my example above is recorded with 96 kHz (please see the picture link above)
To my opinion there seems to be a sound improvement if I compare the result of PGGB-RT upsampling with the original or e.g. with the SRC-Resampler
According to the Remastero website PGGB-RT is not the standard on the fly resampler as e.g. SRC but a near ideal resampler

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

17 (edited by beat8000 2024-08-29 09:42:07)

Re: Sample rates

MC wrote:

It's in the manual: High frequency noise at higher sample rates, measurable but not audible.

Ok, I have checked the manual

As is common in professional Digital Audio Workstations, the level meters of the ADI-2 Pro are
band limited to 40 kHz, so do not show the excessive noise levels of 768 kHz and DSD, but
everything within the audio range and a bit above.

I can't hear this effect shown as Analog Input during the USB playback
However the Analog level on the display is higher than my USB 1/2 output level at 768 kHz wink

https://mobidrive.com/sharelink/i/YmVhd … KuklzQIw8t

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

18 (edited by KaiS 2024-08-29 09:56:55)

Re: Sample rates

Kubrak wrote:

Upsampling is just another form of digital filtering to avoid antialiasing on systems that have poor DA.

If audio signal is sampled 44.1 kHz, no upsampling, resampling, digital filtering is not able to add any information to signal. It may only better or worse mimic ideal filter which has to cut off frequencies above 22.05 kHz to avoid aliasing.

Not to forget–
The great majority of DAC-chips do exactly this for DA-conversion:

Internal upsampling to a way higher sample rate, plus calculation of a digital filter, then output the result on that very high sample rate.
This is the reason why there is a choice of reconstruction filters e.g. inside ADI-2.
Preliminary upsampling just dublcates some of that process.

Kubrak wrote:

And there is no ideal solution for that case by definition. There are always compromises by definition, one cannot beat mathematical reality.  One may prefer one method, the other another, but none of them is perfect, none of them is "right".

That‘s exactly the point:

DA-conversion and it‘s filtering is a compromise between:
• Frequency response linearity of the upmost range.
• impulse response precision and shape.
• Aliasing artifacts inside the audible range.

The optimization is in the range between so called “Sharp” filters on one side, that optimize freq. response and anti-aliasing, on cost of impulse respone –
and on the other side the NOS filters that optimize impuls response on cost of audible high frequency loss and aliasing artifacts.


IMO the best compromise is located in between:
Slow Filter (on AKM DACs) and SD-LD and Brickwall (on ESS chip based DACs).
They offer good and natural shaped impulses (no pre-ringing, which doesn’t exist in nature), while sacrificing a little bit of upmost treble and allow a minor amout of aliasing.

The result is cleaner and more detailed, with better separation of high frequency instruments like hihats and shakers or other percussion instruments.
The ambience, room sound definition benefits either from that.

19

Re: Sample rates

beat8000 wrote:
MC wrote:

It's in the manual: High frequency noise at higher sample rates, measurable but not audible.

Ok, I have checked the manual

Note that this chapter only talks about 192 kHz because DigiCheck is limited to that, and is used to show that there is no change in the audible range. At 384 and 768 kHz the high frequency noise gets much worse.

As is common in professional Digital Audio Workstations, the level meters of the ADI-2 Pro are
band limited to 40 kHz, so do not show the excessive noise levels of 768 kHz and DSD, but
everything within the audio range and a bit above.

The AD-side shows any frequency within the current sample rate range and is not limited.

Regards
Matthias Carstens
RME

Re: Sample rates

beat8000 wrote:

The original CD in my example above is recorded with 96 kHz (please see the picture link above)
To my opinion there seems to be a sound improvement if I compare the result of PGGB-RT upsampling with the original or e.g. with the SRC-Resampler

OK it does not change anything on principle. Upsampling does not make the signal anything better. It is just kind of filter. One has certain compromises another yet another compromises.

Re: Sample rates

Kubrak wrote:

Upsampling does not make the signal anything better. It is just kind of filter. One has certain compromises another yet another compromises.

Ok, to my understanding the steps on the curves are much smaller e.g. with the best settings of the PGGB-RT and near to the ideal.
Additionally there might be less wrong interpolations.

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

22

Re: Sample rates

There are no steps at all. The output of the SRC runs through the same FS/2 filter as always, so similar to a DAC output no steps can be seen or measured.

Regards
Matthias Carstens
RME

Re: Sample rates

MC wrote:

There are no steps at all. The output of the SRC runs through the same FS/2 filter as always, so similar to a DAC output no steps can be seen or measured.

As far as I know I'm always using the ADI 2 Pro with the SRC switched off

Additionally I'm meaning the steps as part of a frequency curve as shown below

https://www.open-end-music.com/filedata … 1684725226

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

24 (edited by ramses 2024-08-29 13:15:13)

Re: Sample rates

beat8000 wrote:
MC wrote:

There are no steps at all. The output of the SRC runs through the same FS/2 filter as always, so similar to a DAC output no steps can be seen or measured.

As far as I know I'm always using the ADI 2 Pro with the SRC switched off

Additionally I'm meaning the steps as part of a frequency curve as shown below

https://www.open-end-music.com/filedata … 1684725226

The link cannot be opened, please use a cloud platform that does not require registration.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

25 (edited by beat8000 2024-08-29 13:47:06)

Re: Sample rates

Ok, here is the link:

https://mobidrive.com/sharelink/i/YmVhd … uwp3mF5uSP

I have just found a similar frequency curve with steps also at the end of page 77 in the ADI 2 Pro manual

Apart from that I just want to listen to my favorite films and music with the ADI 2 Pro in the best possible way.
If I have done something wrong with the PGGB-RT and e.g. the switched off ADI 2 Pro SRC or other settings I would like to correct it. wink

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

Re: Sample rates

You do nothing wrong. Just waste your time and resources.

There is no way to "improve" something that cannot be improved in this case. Yes, it could help to some extent if using DA that has costed 20 USD. It could help a bit if using DA that has costed 100 USD. But we speak about mastering grade DA.

But fine, if that works for you. Just great. You are the only one who can decide what you like the best.

I understand, that smooth digital graph of signal upscaled to "hires" looks better than graph of crude digital signal before upscaling.... But the "magic" happens in DA process and the both signals are smooth at the end of DA.

That "magic" may be done better or worse. That is what makes DA equipment better or worse. But there is no miracle to be done to "improve" original digital source. There is always some kind of compromise. One cannot avoid it. The information lost by sampling is lost forever, it cannot be magically regained by upsampling. What happens in between samples is just sort of "qualified guess" (of course based on math).

27 (edited by beat8000 2024-08-29 16:24:52)

Re: Sample rates

Ok, I think the comparison between e.g. the original and the upsampled version is harder if I turn PGGB-RT on than off

Switching PGGB-RT off is done immediately but switching upsampling to 768 kHz on again takes about 15 seconds for one track on my computer

I think it would be almost unusable if I had to wait 15 seconds after each track, but fortunately it is gapless because the upsampling of the next track will be executed in time before the next track begins.

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

28 (edited by Kubrak 2024-08-29 16:48:59)

Re: Sample rates

As I said, if you like upsampled sound more than original, than that is just fine. It is your choice what you like better.

As KaiS explained good DAs do upsampling on chip as part of digital filtering the output signal anyway... One even may choose different filters... So, everything is more related to taste and what one needs. There is always compromise, one cannot have it all perfect. That analog perfectness has been lost way back in digital domain. One cannot regain it completely back. One may touch it one way or another.

Analog and digital are two rather different beasts. One may transform one to another, but they live in quite diffent worlds. Even mathematically those worlds are very, very different and have quite different set of rules. In fact they have only very few things in common.

Re: Sample rates

KaiS wrote:

IMO the best compromise is located in between:
Slow Filter (on AKM DACs) and SD-LD and Brickwall (on ESS chip based DACs).
They offer good and natural shaped impulses (no pre-ringing, which doesn’t exist in nature), while sacrificing a little bit of upmost treble and allow a minor amout of aliasing.

The result is cleaner and more detailed, with better separation of high frequency instruments like hihats and shakers or other percussion instruments.
The ambience, room sound definition benefits either from that.

If I got you correctly (based on an ADI 2 DAC FS w/ ESS chip), the Brickwall filter would be your favorite.

If I compare the impulse responses diagrams of the manual on page 58 - Brickwall and Sharp on page 57, I can't see any (audible) difference - they look pretty much the same to me. But the frequency response of the sharp filter reaches 20 kHz without any loss, while Brickwall already lost -4.5 dB (beginning at ~ 19 kHz). Therefore, I doubt, that Brickwall could be audibly different compared to the sharp filter.

30 (edited by KaiS 2024-08-29 21:41:00)

Re: Sample rates

Kubrak wrote:

What happens in between samples is just sort of "qualified guess" (of course based on math).

The typical DA‘s or SRC’s reconstruction filter removes everything, really everything between single samples.

So, not even a guess what could have been there.

If the filtering is imperfect, some noise is generated, even in the lower audio band.
This noises are the aliasing artifacts, strongest with no filter at all (like the so called NOS filter).

People might like this artifacts, in this case it‘s an “improvement“
But it’s no reconstruction of the original, more like an effect comparable to what tube amp harmonics (distortions) do.
Only that aliasing is completely disharmonic.

31 (edited by KaiS 2024-08-29 21:38:49)

Re: Sample rates

user317 wrote:
KaiS wrote:

IMO the best compromise is located in between:
Slow Filter (on AKM DACs) and SD-LD and Brickwall (on ESS chip based DACs).
They offer good and natural shaped impulses (no pre-ringing, which doesn’t exist in nature), while sacrificing a little bit of upmost treble and allow a minor amout of aliasing.

The result is cleaner and more detailed, with better separation of high frequency instruments like hihats and shakers or other percussion instruments.
The ambience, room sound definition benefits either from that.

If I got you correctly (based on an ADI 2 DAC FS w/ ESS chip), the Brickwall filter would be your favorite.

If I compare the impulse responses diagrams of the manual on page 58 - Brickwall and Sharp on page 57, I can't see any (audible) difference - they look pretty much the same to me. But the frequency response of the sharp filter reaches 20 kHz without any loss, while Brickwall already lost -4.5 dB (beginning at ~ 19 kHz). Therefore, I doubt, that Brickwall could be audibly different compared to the sharp filter.

I’m on ADI-2/4 Pro SE, and SD-LD is my favorite for this ESS chip model.
Manual pages 88, 89.
I only lately stumbled over the Brickwall filter and kind of like it too.
It does sound different than the Sharp filter to me.
Still the SD-LD sounds more “open”.

The whole thing is only relevant at 44.1 and a little less at 48 kHz sample rate, just to bring this fact into mind again.

32 (edited by Kubrak 2024-08-29 22:31:32)

Re: Sample rates

KaiS wrote:
Kubrak wrote:

What happens in between samples is just sort of "qualified guess" (of course based on math).

The typical DA‘s or SRC’s reconstruction filter removes everything, really everything between single samples.

So, not even a guess what could have been there.

If the filtering is imperfect, some noise is generated, even in the lower audio band.
This noises are the aliasing artifacts, strongest with no filter at all (like the so called NOS filter).

People might like this artifacts, in this case it‘s an “improvement“
But it’s no reconstruction of the original, more like an effect comparable to what tube amp harmonics (distortions) do.
Only that aliasing is completely disharmonic.

I meant it in the way that reconstruction is not in real world perfect. So, what happens in between samples is "guess". Nobody knows what was in source analog signal in between samples. There is only theoretical assumption what should have been there. And also, filter on AD side is not ideal. Plus there is always at least subtle jitter and so on.

OK I used qualified guess instead of reconstruction... But is there a difference between those two in real physical world? But still, things work and it is part of "magic".

33

Re: Sample rates

beat8000 wrote:

Ok, here is the link:

https://mobidrive.com/sharelink/i/YmVhd … uwp3mF5uSP

I have just found a similar frequency curve with steps also at the end of page 77 in the ADI 2 Pro manual

That scope pic shows a 'signal curve', if at all. And no steps at all.

Maybe time to throw in Monty:

https://www.youtube.com/watch?v=cIQ9IXSUzuM

And as this thread is about sample rates here is Dan on this subject:

https://www.youtube.com/watch?v=-jCwIsT0X8M

Regards
Matthias Carstens
RME

34 (edited by beat8000 2024-08-30 12:15:59)

Re: Sample rates

MC wrote:

That scope pic shows a 'signal curve', if at all. And no steps at all.

Maybe time to throw in Monty:

To my opinion stairsteps were shown in the video at 4:20.

https://mobidrive.com/sharelink/i/YmVhd … 7y0fBytDzW

However, as far as I know, Monty says that these steps do not exist in the digital waveform between samples because these informations are undefined.

Additionally, with the minimum samples in each half-waveform, a perfect sine wave can be created.

In this case, I think why send so much informations when just a few samples are enough? wink

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

35 (edited by Kubrak 2024-08-30 12:49:49)

Re: Sample rates

beat8000 wrote:

Additionally, with the minimum samples in each half-waveform, a perfect sine wave can be created.

In this case, I think why send so much informations when just a few samples are enough? wink

That is what we try to explain you. Upsampling does not bring any new information. It is not going to improve signal. Especially for mastering grade DA. Hi quality upsampling might help in case one uses very poor DA hardware.

36 (edited by beat8000 2024-08-30 13:07:48)

Re: Sample rates

I mean, it seems that at least one sample per half wave is enough to reconstruct a perfect sine wave depending on the frequency of the original signal

I'm not sure if this is also true for complex signals as mentioned in the second video above

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

Re: Sample rates

Yes, it is true for any signal composed of frequencies up to half of sampling frequency. It is mathematically prooven, so there is no question about it.

There was that nice example with square signal. Ideal square signal has frequencies going indefinitely high.

38 (edited by beat8000 2024-08-30 13:30:51)

Re: Sample rates

Yes, for one sine curve it is clear for me
However what happens if there might be a lot of sine curves e.g. for a complex signal at the same time
How is it possible to assign the samples to the correct sine wave?

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

39 (edited by Kubrak 2024-08-30 14:33:33)

Re: Sample rates

It shows that part of video with square signal. What goes in goes after reconstruction out.....

I haven't seen the mathematical proof of that it works. But I doubt I would be able to comprehend the math behind.... And I had lots of math lessons including digital domain math and I like math....

Simply, it works just fine. Yes, it is rather hard to believe, but it evidently works just fine. When touching digital one cannot use most of experience from analog. As I said earlies, digital domain and analog domain are very different beasts/worlds. They have its own rules and its own math.

For example analog world has nothing like aliasing. That is unique to digital. Similar with dithering....

But if you do not trust it works, you may make a test. Get complex signal in analog, make A/D at 48 kHz, make D/A, feed signal to A/D and compare first version of signal afrer A/D and second version after A/D. They should be almost the same.

Or you could compare signals in analog. The best would be to invert phase of one of them and add them. If everything works right, they should cancel and the result would be just residual noise. But one has to adjust signals for delay made by AD and DA so that they are time alligned.

40 (edited by beat8000 2024-08-30 15:02:49)

Re: Sample rates

For my CD comparison, I'm not entirely sure if the 96 KHz original or the upsampled version to 768 kHz is better.
The original seems to be quieter, e.g. during lots of small pulses and also with opposing melodies, e.g. one at 1/4 notes and the other at 1/32 notes.
However I find especially the bass is much darker and punchier on the upsampled version.
In my opinion, the resolution seems also higher and the small pulses more precise.

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

41 (edited by waedi 2024-08-30 15:27:52)

Re: Sample rates

beat8000 wrote:

For my CD comparison, I'm not entirely sure if the 96 KHz original or the upsampled version to 768 kHz is better.
The original seems to be quieter, e.g. during lots of small pulses and also with opposing melodies, e.g. one at 1/4 notes and the other at 1/32 notes.
However I find especially the bass is much darker and punchier on the upsampled version.
In my opinion, the resolution seems also higher and the small pulses more precise.

Put both versions into a DAW and make the null-test.
It will turn out the two files will cancel out each other and your impression was wishful thinking.
Also use the free plugin Youlean loudness meter for measure peak dB and LUFS.

M1-Sequoia, Madiface Pro, Digiface USB, Babyface silver and blue

42 (edited by unpluggged 2024-08-30 15:45:18)

Re: Sample rates

beat8000 wrote:

Nobody knows what was in source analog signal in between samples. There is only theoretical assumption what should have been there.

OMG.

Sampling theorem, does this name ring any bells? Or, perhaps, Nyquist-Shannon? Or even Kotelnikov?

Fireface UCX II + ARC USB > ADI-2 Pro FS R BE > Neumann KH 750 DSP + MA 1 > KH 120 A

43 (edited by beat8000 2024-08-30 16:06:38)

Re: Sample rates

unpluggged wrote:
beat8000 wrote:

Nobody knows what was in source analog signal in between samples. There is only theoretical assumption what should have been there.

OMG.

Sampling theorem, does this name ring any bells? Or, perhaps, Nyquist-Shannon? Or even Kotelnikov?

I think I haven't written what you have quoted wink

According to my check the quote above is from @Kubrak

Win10 Pro, ADI-2 Pro, Basis 1, Adam A3X; RL 906; Grace M902B, Glockenklang Bugatti, Strauss SE-NF-3

Re: Sample rates

Yes I have written that. And I stand behind that. As nobody knows if analog signal sampled has fulfilled Nyquist-Shannon rule or not. It should, but ....

And one may have source that goes beyond 20kHz and sample it 44kHz with not so steep filter.... But yes, if things are done properly, everything works just fine.

Re: Sample rates

KaiS wrote:

I’m on ADI-2/4 Pro SE, and SD-LD is my favorite for this ESS chip model.
Manual pages 88, 89.
I only lately stumbled over the Brickwall filter and kind of like it too.
It does sound different than the Sharp filter to me.
Still the SD-LD sounds more “open”.

The whole thing is only relevant at 44.1 and a little less at 48 kHz sample rate, just to bring this fact into mind again.

Thanks for your notes, Kai. Adi 2 DAC FS @ESS doesn't provide a SD-LD filter but a Brickwall filter, which shows different behavior as your SD-LD.
Do you by any chance have a Youtube link which shows best the differences? I would be interested. I would compare Brickwall with Sharp here - but I doubt I could hear any difference because they look pretty similar.

46 (edited by KaiS 2024-08-30 21:55:20)

Re: Sample rates

user317 wrote:
KaiS wrote:

I’m on ADI-2/4 Pro SE, and SD-LD is my favorite for this ESS chip model.
Manual pages 88, 89.
I only lately stumbled over the Brickwall filter and kind of like it too.
It does sound different than the Sharp filter to me.
Still the SD-LD sounds more “open”.

The whole thing is only relevant at 44.1 and a little less at 48 kHz sample rate, just to bring this fact into mind again.

Thanks for your notes, Kai. Adi 2 DAC FS @ESS doesn't provide a SD-LD filter but a Brickwall filter, which shows different behavior as your SD-LD.
Do you by any chance have a Youtube link which shows best the differences? I would be interested. I would compare Brickwall with Sharp here - but I doubt I could hear any difference because they look pretty similar.

ADI-2/4 Pro SE and ADI-2 DAC ESS version seem to be different in this regard.
ADI-2/4 Pro SE, which is only available with ESS DAC chip, offers both SD-LD and Brickwall.

if you want to hear the difference ESS Sharp vs Brickwall why not just listen to it on your own device?!
I’d even say, without pre-Justice, go through all the filters available and find out what suits your taste best.

Use 44.1 kHz tracks, and don’t expect night and day differences.

47

Re: Sample rates

waedi wrote:
beat8000 wrote:

For my CD comparison, I'm not entirely sure if the 96 KHz original or the upsampled version to 768 kHz is better.
The original seems to be quieter, e.g. during lots of small pulses and also with opposing melodies, e.g. one at 1/4 notes and the other at 1/32 notes.
However I find especially the bass is much darker and punchier on the upsampled version.
In my opinion, the resolution seems also higher and the small pulses more precise.

Put both versions into a DAW and make the null-test.

You can't put two files with totally different sample rates into a DAW. Unless one of them is automatically downsampled, which makes the comparison invalid.

waedi wrote:

It will turn out the two files will cancel out each other

No matter how you downsample this will not happen. The upsampling alone already introduce slight phase shifts over frequency, which does not allow simple null tests anymore. You could do a better test comparing them with Deltawave, which can be set to ignore (compensate) inaudible phase shifts.

Regards
Matthias Carstens
RME

Re: Sample rates

KaiS wrote:

IMO the best compromise is located in between:
Slow Filter (on AKM DACs) and SD-LD and Brickwall (on ESS chip based DACs).
They offer good and natural shaped impulses (no pre-ringing, which doesn’t exist in nature), while sacrificing a little bit of upmost treble and allow a minor amout of aliasing.

The result is cleaner and more detailed, with better separation of high frequency instruments like hihats and shakers or other percussion instruments.
The ambience, room sound definition benefits either from that.

Kai, did you perform some similar comparisons with the filters in the Qudelix-5K?
Do you have a recommendation for the Qudelix-5K also?
That would be nice. Thanks.

Re: Sample rates

Yes, there is something called “Hybrid…” or so that I liked best.
I don’t have the unit with me and it’s not in the manual, so I can’t look for the full name.

Re: Sample rates

Thank you very much. It's called "Hybrid fast roll-off".
I will try it. Have a nice weekend.