How many bits, how fast, just how much resolution is enough?

[sdolezalek]

Yuri and Jussi: Based upon the last two posts, and without asking either of you to disclose any trade secrets, is it fair to assume that how you treat these ultrasonics is a difference between using HQPlayer and Audiventory for upsampling files to DXD or DSD256+ resolutions?

[audiventory]

I would never do that, because it removes hires from the hires - slows down transients.

 

I have invested money and effort to have wide band capable system and don't want to spoil it.

 

But each to their own.

 

I can fill infrasound clearly (as vibration). But things linked with ultrasound are not so obvious for me.

 

You know, I don't consider hi-res for playback ultrasound, but as better way of applying of analog filter only.

 

May be future science works give us new knowledges about system ultrasound-body-ears-brain that we will use.

[Sal1950]

Certainly no one is going to take your RedBook away. For that matter, if like many people who've been tested you can't tell a difference between RedBook and mp3, you can save even more bandwidth. (Being perfectly serious, not specious.) But if you're in my situation, then I think it's nice to have a choice of resolutions, filtering, and modulation.

TBF most recordings aren't good enough to warrant massive resolution / bandwidth... so for that huge majority it's probably best not to spoil your enjoyment of that subject material by telling yourself it's not as good as it could have been because the file size isn't half a Gig.

I do a LOT of listening to Spotify 320 kbps HQ data rate and although not indistinguishable from RedBook or better, in most cases, with the music I love most, it's quite livable and enjoyable. I'd prefer to have the RB rate that Tidal offers but refuse to put $.02 into the pockets of Jay Z, Kanye West, and the rest of the gangster clan there.

In the end I've only reached out and purchased a half dozen albums I demo'd on Spotify to get a better quality playback. YMMV

[Maldur]

I just complete a small test in home. 1) DVDA from standalone player; 2) SACD from another standalone player; 3) SACD image from computer via Miska's DSC1 DAC. Speakers, amplifier and content is the same, and I know for sure that DVDA and SACD is made from one master. All played back in same time, relative in sync.

Listener was my daughter, a classical musician. He don't knows anything what plays and what are nuances. He just changed inputs from remote.

Results are: DVDA sounds like ordinary (poor) recording, DSC1 sounds like a live orchestra and SACD sounds like DCS1 but not so loud as DCS1 (indeed, there is small difference in levels).

DCS1 don't have any output capacitors, both standalone players are moddified to direct out from DAC chip (Cirrus and BB respectively) with capacitors (again exact the same).

When he first selected a signal from DSC1, small smile goes over the face - I think this says more than thousand words.

[audiventory]

Yuri and Jussi: Based upon the last two posts, and without asking either of you to disclose any trade secrets, is it fair to assume that how you treat these ultrasonics is a difference between using HQPlayer and Audiventory for upsampling files to DXD or DSD256+ resolutions?

 

Hi Sdolezalek,

 

For conversion either PCM to DSD or PCM to PCM (DXD is one kind of PCM) used upsampling or downsampling.

 

There used filtering anyway.

 

AuI have 2 upsampling/downsampling mode (by filter mode):

 

1. Traditional (non-optimised)

1.1. PCM to PCM: 0…[Fs/2 - transient band]. Where Fs is minimal sample rate between input and output ones.

1.2. PCM to DSD and back: 0…[20…27 kHz + transient band, depend on input and output sample rates]

Width of the band defined by noise level only, allowable into output band (in output file).

Traditional mode currently have linear phase filters only.

 

 

2. Optimised

0…[20 kHz + transient band]. Rest band is filtered.

I recommend it as «DAC helper».

Optimised mode have linear and minimal phase filters.

 

As default, used optimised mode. I consider it as more «right».

However for users, who prefer wider band, traditional mode available.

 

I suppose, optimised mode is single case, where I recommend perform hearing test before spectrum analisys :-)

[sdolezalek]

2. Optimised

0…[20 kHz + transient band]. Rest band is filtered.

I recommend it as «DAC helper».

Optimised mode have linear and minimal phase filters.

 

I suppose, optimised mode is single case, where I recommend perform hearing test before spectrum analisys :-)

 

Yuri: Yes, thank you, that is what I remembered, that in Optimized Mode, in addition to upsampling and using "better" filters you also applied a second filter to remove noise and other other information beyond the 20kHz limit.

 

I believe that when I use HQPlayer to do upsampling "live" Jussi allows the user to choose filter types and characteristics (including linear and minimal phase filters) but is not applying a second filter to cut off information above 20kHz. That would make the results of doing an upsampling from 16/44 to DSD256 different between the two programs.

 

That was what I was trying to clarify.

[gmgraves]

Can You show (or at least tell) how the filtering differs when it is feed from 16/44 or 24/44 signal?

 

You misunderstand me, I think. A 24-bit DAC's filter doesn't perform differently when processing 16-bit/44.1 KHz material than it does when it is decoding 24-bit/96 or 192 KHz material, it handles 16-bit as if it were 24-bit. With 16-bit/44.1 KHz material, the sample rate (44.1 KHz) must roll-off very steeply above 22.5 KHz to satisfy the Nyquist theorem. Such so-called "brick wall" filters can cause problems like phase shift in the audible band. When a 16-bit/44.1 KHz source (such as a CD) is played on a DAC that can handle 24-bit/96 or 192 KHz files, the sampling frequency is much higher than 44.1 KHz by twice or even 4X (in a 192 KHz capable DAC). This moves the sampling rate far outside the 22.5 KHz bandwidth of a16-bit/44.1 KHz file (such as a CD). There are several different ways to do this, but the result is that the "Q", or steepness of the filter can be much more gentle, and therefore does not introduce the phase shift and other anomalies caused by steep, brick-wall filters required to get the audio signal down by at least -20 dB by the time the sampling frequency is reached (at least 2X the "Nyquest" frequency [or top frequency being quantized], in this case 22.5 KHz). I hope this clears up the confusion. It's a bit oversimplified as an explanation, but just remember that the further the DAC's sampling rate is from the highest frequency in which one is interested (again, 22.5 KHz for Red Book CD), the better the sound. Oh, yes, one other thing. 16-bit sounds better through a 24-bit DAC than it does through a 16-bit DAC due to the filtering employed, but also because no commercial 16-bit DACs have 16-bit linearity, just as no 24-bit DAC gas 24-bit linearity. BUT a 24-bit DAC does have 16-bit linearity (at least) which lowers quantization error of 16-bit material substantially. But I never said or inferred that 16-bit through a 24-bit DAC sounds better than 24-bit through the same DAC, which I believe you think I was saying.

[mansr]

You misunderstand me, I think. A 24-bit DAC's filter doesn't perform differently when processing 16-bit/44.1 KHz material than it does when it is decoding 24-bit/96 or 192 KHz material, it handles 16-bit as if it were 24-bit. With 16-bit/44.1 KHz material, the sample rate (44.1 KHz) must roll-off very steeply above 22.5 KHz to satisfy the Nyquist theorem. Such so-called "brick wall" filters can cause problems like phase shift in the audible band. When a 16-bit/44.1 KHz source (such as a CD) is played on a DAC that can handle 24-bit/96 or 192 KHz files, the sampling frequency is much higher than 44.1 KHz by twice or even 4X (in a 192 KHz capable DAC). This moves the sampling rate far outside the 22.5 KHz bandwidth of a16-bit/44.1 KHz file (such as a CD). There are several different ways to do this, but the result is that the "Q", or steepness of the filter can be much more gentle, and therefore does not introduce the phase shift and other anomalies caused by steep, brick-wall filters required to get the audio signal down by at least -20 dB by the time the sampling frequency is reached (at least 2X the "Nyquest" frequency [or top frequency being quantized], in this case 22.5 KHz). I hope this clears up the confusion. It's a bit oversimplified as an explanation, but just remember that the further the DAC's sampling rate is from the highest frequency in which one is interested (again, 22.5 KHz for Red Book CD), the better the sound. Oh, yes, one other thing. 16-bit sounds better through a 24-bit DAC than it does through a 16-bit DAC due to the filtering employed, but also because no commercial 16-bit DACs have 16-bit linearity, just as no 24-bit DAC gas 24-bit linearity. BUT a 24-bit DAC does have 16-bit linearity (at least) which lowers quantization error of 16-bit material substantially. But I never said or inferred that 16-bit through a 24-bit DAC sounds better than 24-bit through the same DAC, which I believe you think I was saying.

Uh, I think you're confused. Any 44 kHz signal needs a sharp filter, even if the DAC is capable of higher sample rates.

[Maldur]

You misunderstand me, I think.

 

So it really was, I'm sorry again.

[esldude]

While the topic has arisen, maybe now is a good time to say something about phase. With delta-sigma DACs and digital filtering you really don't have phase issues with the great majority of DACs out there today. At least not to 20 khz you don't. So worry about phase at 44 khz or 48 khz rates is a non-issue. When it can become an issue is boutique DACs that decide to use a different filter that alters the flat response below 20 khz or has some other alleged benefit.

 

Here is an article from way back in 2004 showing that phase is fine right up to 20 khz.

 

digital

[Miska]

While the topic has arisen, maybe now is a good time to say something about phase. With delta-sigma DACs and digital filtering you really don't have phase issues with the great majority of DACs out there today. At least not to 20 khz you don't. So worry about phase at 44 khz or 48 khz rates is a non-issue. When it can become an issue is boutique DACs that decide to use a different filter that alters the flat response below 20 khz or has some other alleged benefit.

 

Majority of phase variations come from analog domain reconstruction filter following the DAC. With the transient optimized 4th order analog filter I have in my DSC1 design, phase shift at 20 kHz is somewhere around 20 degrees, fc point is at 100 kHz and group delay is parctically flat to 50 kHz. This gives optimal results at DSD256 or higher rates.

 

If a DAC chip has digital filters running only up to 352.8/384 rate, then either the filter with 100 kHz cut-off is not fully effective, or it needs to be very high order one. Alternatively the cut-off frequency needs to be moved down which will increase phase shift in the audio band. This is one of the reasons it is in many cases better to run digital filters and modulator for example to 11.3/12.2 MHz (DSD256) rate in a computer resulting in better reconstruction. Especially because many DACs have only 2nd order analog filter...

[The Computer Audiophile]

Here is an article from way back in 2004 showing that phase is fine right up to 20 khz.

digital

 

As Abraham Lincoln famously said, "I read it on the Internet, so it must be true."

[jabbr]

Uh, I think you're confused. Any 44 kHz signal needs a sharp filter, even if the DAC is capable of higher sample rates.

 

The "gentle" filter I prefer is to convert the 44 kHz signal to DSD512 then 32 tap FIR followed by a gentle analog filter. Sounds great.

[jabbr]

 

That is not hard to try, because you can take some DSD256 recording and convert it to RedBook. My ADCs are limited to DSD128 (based on TI's PCM4202 chip), but already that makes a difference.

 

I've already outlined my approach of obtaining files is as close to the native mastered format as feasible.

 

The problems with testing format resolutions against each other is that it's rather easy to have Redbook vs so-called high res files that really don't have any high res content-- and so of course the comparisons will fail to show a difference.

 

I was suggesting that, you, for example, would be in an excellent position to suggest or provide files which do show differences to your own ability to hear.

 

Starting there it might be interesting to see what percentage of people could similarly hear a difference.

 

I'm not sure that the results would change my own approach but, you know, for those folks who are into making their own decisions on the basis of A/B DBT, it might be interesting.

[mansr]

The "gentle" filter I prefer is to convert the 44 kHz signal to DSD512 then 32 tap FIR followed by a gentle analog filter. Sounds great.

The upsampling needs a sharp (digital) filter or you'll get images starting at 22.05 kHz.

[gmgraves]

Uh, I think you're confused. Any 44 kHz signal needs a sharp filter, even if the DAC is capable of higher sample rates.

 

Does that not depend on whether or not up-sampling is used on the 44.1 KHz. Don't most modern 24-bit DACs do that? I see that I failed to make that distinction, in which case, mia culpa. Thank you for pointing that out Mansr!

[esldude]

As Abraham Lincoln famously said, "I read it on the Internet, so it must be true."

 

You know, I have read that before. Who would argue with old Abe other than pro-slavery folks like Stephen Douglas? So I'll just apply this idea to the CA forum so everything will be copacetic.

[Ralf11]

Albert Einstein had a long running argument with Abe lincoln, but Heisenburg lost it somewhere

[Sal1950]

Albert Einstein had a long running argument with Abe lincoln, but Heisenburg lost it somewhere

I'd fill everyone in but I was forced to sign a NDA by the folks at Meridian (MQA "Its Not DRM, It's Not DRM").

[gmgraves]

Albert Einstein had a long running argument with Abe lincoln, but Heisenburg lost it somewhere

 

Heisenburg was confused.

[Jud]

While the topic has arisen, maybe now is a good time to say something about phase. With delta-sigma DACs and digital filtering you really don't have phase issues with the great majority of DACs out there today. At least not to 20 khz you don't. So worry about phase at 44 khz or 48 khz rates is a non-issue. When it can become an issue is boutique DACs that decide to use a different filter that alters the flat response below 20 khz or has some other alleged benefit.

 

Here is an article from way back in 2004 showing that phase is fine right up to 20 khz.

 

digital

 

Majority of phase variations come from analog domain reconstruction filter following the DAC. With the transient optimized 4th order analog filter I have in my DSC1 design, phase shift at 20 kHz is somewhere around 20 degrees, fc point is at 100 kHz and group delay is parctically flat to 50 kHz. This gives optimal results at DSD256 or higher rates.

 

If a DAC chip has digital filters running only up to 352.8/384 rate, then either the filter with 100 kHz cut-off is not fully effective, or it needs to be very high order one. Alternatively the cut-off frequency needs to be moved down which will increase phase shift in the audio band. This is one of the reasons it is in many cases better to run digital filters and modulator for example to 11.3/12.2 MHz (DSD256) rate in a computer resulting in better reconstruction. Especially because many DACs have only 2nd order analog filter...

 

...which in turn brings up something about filtering that *may* (I speculate - grossly) have something to do with my preference for linear phase filtering at the PCM interpolation stage.

 

Notice Miska mentioned flat "group delay." Group delay is how long it takes a signal to get through a filter. When group delay is flat, that means the delay is the same across frequencies.

 

A minimum phase filter has the *shortest* group delay. However, minimum phase filters tend not to have *flat* group delay. They delay some frequencies more than others, "spreading out" the time that the various frequencies come out of the filtering, and are thus called "dispersive" filters. Linear phase filters, though they have longer overall group delay, have flat group delay and are thus not dispersive.

 

The Vandersteen speakers I own are built to be what the manufacturer calls time and phase correct. The drivers are positioned so that all frequencies arrive at the listening position at roughly the same time (if the speakers are set up in the room in accordance with the extensive instructions in the manual), and the crossovers between drivers are "first order" filters, which are linear phase.

 

I speculate that linear phase filters in the DAC work with the speakers so imaging and localization of instruments/vocals is as intended by the manufacturer, whereas dispersive minimum phase filters seem, to my ears, to result in less good/solid performance in those areas.

[mansr]

Does that not depend on whether or not up-sampling is used on the 44.1 KHz. Don't most modern 24-bit DACs do that? I see that I failed to make that distinction, in which case, mia culpa. Thank you for pointing that out Mansr!

 

The filtering is part of the upsampling. If not done, the higher-rate digital signal will contain images. This filtering is actually one of the main reasons for upsampling in the first place. It's much easier to create a sharp digital filter than an analogue one. A 2x upsampled (and thus digitally filtered) signal has essentially no content in the upper half of its frequency band, making construction of an adequate analogue filter much simpler.

[jabbr]

The upsampling needs a sharp (digital) filter or you'll get images starting at 22.05 kHz.

 

Oh I suppose that's technically correct in some sense -- which is why is described the process as "gentle" not necessarily regarding a particular filter slope.

 

The point is that the final analog filter can be gentle in order to minimally affect phase

[JohnSwenson]

The Vandersteen speakers I own are built to be what the manufacturer calls time and phase correct. The drivers are positioned so that all frequencies arrive at the listening position at roughly the same time (if the speakers are set up in the room in accordance with the extensive instructions in the manual), and the crossovers between drivers are "first order" filters, which are linear phase.

 

Hi Jud,

 

something doesn't seem right here, the order of the filter doesn't determine whether it is linear phase or not, THAT is determined by the TYPE of the filter (Butterworth, Chebyshev, Bessel etc). The Bessel filter is linear phase in the pass band (but not the transition band or stop band). It does quite good in maintaining the shape of non sinusoidal waveforms, but you give up steepness in order to achieve that, the slopes are very gentle. Higher order Bessel filters are still linear phase in the pass band, but have steeper slopes which is generally a good thing given the gentle slopes in general.

 

Any of the other types are not linear phase even though they are first order, and a Bessel is always linear phase no matter what its order.

 

So for crossovers you want to use Bessel filters if you want to keep waveforms looking (sounding) good.

 

John S.

[Jud]

Hi Jud,

 

something doesn't seem right here, the order of the filter doesn't determine whether it is linear phase or not, THAT is determined by the TYPE of the filter (Butterworth, Chebyshev, Bessel etc). The Bessel filter is linear phase in the pass band (but not the transition band or stop band). It does quite good in maintaining the shape of non sinusoidal waveforms, but you give up steepness in order to achieve that, the slopes are very gentle. Higher order Bessel filters are still linear phase in the pass band, but have steeper slopes which is generally a good thing given the gentle slopes in general.

 

Any of the other types are not linear phase even though they are first order, and a Bessel is always linear phase no matter what its order.

 

So for crossovers you want to use Bessel filters if you want to keep waveforms looking (sounding) good.

 

John S.

 

Thanks John. I was under the apparently incorrect impression that linear phase was limited to first order (analog) crossover filters. But in any case, the crossover filters used in Vandersteens are linear phase.

 

 

Sent from my iPhone using Computer Audiophile