R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 [All]   Go Down

Author Topic: 44k vs. 96k - a new discovery  (Read 15710 times)

Ivo

  • Full Member
  • ***
  • Offline Offline
  • Posts: 169
44k vs. 96k - a new discovery
« on: October 13, 2004, 06:33:01 PM »

Dan,

may I ask about something which is rather musical than technical (although you suggest us to remain rather technical here), but still, even the convertors are mainly for making music ...Smile I know there are already miles of theoretical discussions on this topic, but my question and sharing come from a different, rather practical angle.
At times I had Mytek, I made thorough comparisons between identical recordings made in 44 and 96 and found I like 96 k somehow always more (more vivid detail as if - even after converting them back to 44 k).
I stayed in that conclusion firmly even after getting your great convertors and kept recording only in 96 k. Today (after getting an "esoteric" hint from Ted Nightshade) I tried again to compare and again found that there definitely IS a sound difference between 96 and 44 recordings of the same sources. But to my surprise I somehow was NOT that sure what I actually prefer now. I felt as if 96 k sounds a bit "sharper" , more "crisp" and also a tiny bit more "hollow" as if, whereas 44k sounded more round, smooth, warm , kind of "analog", more "friendly" I would say (not having those bright fingertips though). Although I did not want to admit that openly to myself (after advocating 96k for so long time), I felt that now I definitely prefer 44 k recording as more pleasant to my ears. What is your feeling and experience using these two frequencies for musical recording ? May it happen that the results may be convertor dependent..i.e. some convertor may sound better in 44, other in 96 or these frequencies have some general qualities which must be considered when making a choice ?
Logged
Ivo

VELVET MASTERING
www.velvetmastering.com

SAVITA MUSIC
www.savita.cz

zztim

  • Newbie
  • *
  • Offline Offline
  • Posts: 1
Re: 44k vs. 96k - a new discovery
« Reply #1 on: October 14, 2004, 12:43:35 PM »

hello, I've got the same surprise last year. I used MOTU I/Os to record guitars and drums using different sample rates. My subjective feeling of this records were better in the case of 44k. Instead of 96k. But, there were some difficulties in mixing & mastering in 44k. The presence of sound was not enough good. I solved these problems using the analog mixing.

 
Logged

Bob Olhsson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3968
Re: 44k vs. 96k - a new discovery
« Reply #2 on: October 14, 2004, 11:49:52 PM »

The more I work with digital audio, the more I realize that most generalizations make little sense.

Dan wrote something very significant in his argument against ultra high sample rates. As you increase sample rate, you increase the requirement for precision. This is a generalization that not only makes sense but explains why in some cases working at a lower sample rate might well create less distortion than using the same tools at a higher sample rate.

ted nightshade

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1272
Re: 44k vs. 96k - a new discovery
« Reply #3 on: October 18, 2004, 11:15:50 AM »

For the record, since I don't see it mentioned in his post, Ivo is using the Lavry Blue these days and that's what he used in his latest 44.1 vs. 96 experiment.
Logged
Ted Nightshade aka Cowan

There's a sex industry too.
Or maybe you prefer home cookin'?

Barry Hufker

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 8228
Re: 44k vs. 96k - a new discovery
« Reply #4 on: October 23, 2004, 02:50:59 PM »

I wonder whether the difference is 96k versus 44.1 or whether it is the implementation of antialiasing filters.  According to a presentation by Rupert Neve, there is only one A/D chip (a brand new one and he didn't mention the maker or model) that actually trades a little bit of flat response (down1dB at 20k) for much better antialiasing.  He explained that chip makers have put the emphasis on marketing a ruler flat frequency response rather than aliasing.  So would the aliasing be of the same quality for 96 and 44.1?  I don't know.  Would it be quality antialiasing to begin with?  I don't know.

I do know that very few hardware and software manufacturers actually make good sample rate converters.  You can find interesting information here: www.audioease.com/Pages/BarbaBatch4/Barba4SRCTest.html
And while Audio Ease does want to sell products and so who is to say if you can really trust their "research," I have done Audio Ease's test myself using Barbabatch and Peak.  The results were as posted.  So, I think there is a lot more to consider than just sampling rates.
Logged

ammitsboel

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1300
Re: 44k vs. 96k - a new discovery
« Reply #5 on: October 23, 2004, 04:02:47 PM »

Barry Hufker wrote on Sat, 23 October 2004 19:50

I wonder whether the difference is 96k versus 44.1 or whether it is the implementation of antialiasing filters.  According to a presentation by Rupert Neve, there is only one A/D chip (a brand new one and he didn't mention the maker or model) that actually trades a little bit of flat response (down1dB at 20k) for much better antialiasing.  He explained that chip makers have put the emphasis on marketing a ruler flat frequency response rather than aliasing.  So would the aliasing be of the same quality for 96 and 44.1?  I don't know.  Would it be quality antialiasing to begin with?  I don't know.


Depending on the filter type and components, some can really screw up the meaning by the music signal you put into them, and sometimes more than aliasing would do.
I think that 44.1 vs. 96 has a lot to do with the filters.
Maybe also the presison of the recieve/send circuits in our converters? has anybody done a test with Jitter and different sample rates?


Best Regards,
Logged
"The male brain is designed for ecstasy" -Dr. Harvey "Gizmo" Rosenberg

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: 44k vs. 96k - a new discovery
« Reply #6 on: October 23, 2004, 09:15:19 PM »

Barry Hufker wrote on Sat, 23 October 2004 19:50

I wonder whether the difference is 96k versus 44.1 or whether it is the implementation of antialiasing filters.  According to a presentation by Rupert Neve, there is only one A/D chip (a brand new one and he didn't mention the maker or model) that actually trades a little bit of flat response (down1dB at 20k) for much better antialiasing.....




I wonder how old is the Neve presentation. I would guess it is ten years old (or more). That anti aliasing filter problem was very serious some years ago, when the requirement to protect from aliasing was real tough. In the days before AD's were designed with oversampling front ends, one had to attempt and pass as near to 20KHz as possible (flat response) and than attenuate with a very high slope filter - very many poles, up to 22.050KHz.
That was a 2 KHz transition and took tons of opamps and precision parts, and the compromises were painful. And to add "insult to injury", phase non linearity at above 10KHz was always there...

But with oversampling front ends, AD anti aliasing became easier. In fact just a X2 oversampling makes the transition band about 20KHz (a factor of 10). Todays AD's IC's are almost all of the sigma delta types and they operate at 64Xfs to 512Xfs, and the anti alias filter problem is long gone... You can have a 3 pole filter at 40Khz yield 120dB protection with next to zero phase linearity impact. So it is not what it used to be.

You can still have problems in the digital decimation filter yielding that -1dB at 20KHz (for 44.1KHz), but it is not an anti ailaising filter issue.

BR
Dan Lavry
Logged

Barry Hufker

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 8228
Re: 44k vs. 96k - a new discovery
« Reply #7 on: October 23, 2004, 11:45:34 PM »

Dan,

I head the 4-year undergraduate program in Audio Production at Webster University.  We invited Rupert Neve as a guest speaker.  I say this only to give you the circumstances under which Mr. Neve came to our campus this past spring to speak.  It was at that time (April or May of this year, I forget exactly), that Mr. Neve made the statement I attributed to him.  He even said this new chip was just coming out.

I was a witness to this statement, but this is not my area of expertise so I can't clarify it further.  But Mr. Neve spoke of it as an immediate (at that moment) development.

Barry
Logged

Joe Bryan

  • Newbie
  • *
  • Offline Offline
  • Posts: 19
Re: 44k vs. 96k - a new discovery
« Reply #8 on: October 24, 2004, 02:45:09 AM »

It's very possible that the DC blocking filter in the converters is tracking the sample rate. I.e. as you go up in sample rate, the DC blocking filter's cutoff rises (1 Hz at 48k, 2 Hz at 96k, and 4 Hz at 192k).

The increasing phase shift from the DC blocking filter smears the low-end, and causes low-end transients (like kick) to loose their punch and transparancy. To me, it sounds like a window sliding up the spectrum, you loose low-end as you gain high-end.

This isn't a slam on high sampling rates, because there are some good converters that do this correctly, and these *do* sound better at higher rates despite technical arguments to the contrary (ears are the reference, not theory).

Also, in DSP systems, higher sampling rates do have advantages as well as disadvantages (higher processing load, more disk space/fewer tracks, etc.). Specifically, well designed DSP filters will sound better at higher sampling rates because their responses can more closely approach the ideal analog response (in technical terms, the s-plane to Z-plane warping is more linear). Conversely, poorly designed filters may sound worse, especially at low frequencies because of data precision issues.

Futhermore, at higher sampling rates, analog anti-aliasing and anti-imaging filter requirements are less stringent (allowing less phase shift in the main audio band), and system latency can be reduced so digital processors can be used in real-time monitoring situations (especially vocal tracking).

-Joe

Joe Bryan
VP Engineering
Universal Audio
Logged

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 44k vs. 96k - a new discovery
« Reply #9 on: October 24, 2004, 02:58:11 PM »

Joe Bryan wrote on Sun, 24 October 2004 07:45

Specifically, well designed DSP filters will sound better at higher sampling rates because their responses can more closely approach the ideal analog response (in technical terms, the s-plane to Z-plane warping is more linear).


Not all filters perform better at higher sampling rates.  Some filters have greater accumulative error by using excess taps to perform the process.

Quote:

Futhermore, at higher sampling rates, analog anti-aliasing and anti-imaging filter requirements are less stringent (allowing less phase shift in the main audio band),


Not true.  Almost all anti-aliasing and anti-imaging filters that you are speaking of are digital and linear phase, so no phase shift issues are improved by higher sampling rates.  The analog filters in any oversampling filter these days have phase shift so far below our audible phase discernability that this is clearly not a factor.  Further, most high sample rate converters use the same analog anti-aliasing/anti-imaging filters as their lower sample rate brethren.

Indeed, the anti-aliasing and anti-imaging filters are less stringent at the higher sampling rates, but this only pertains to what you call "poorly designed filters."  Not all filters have to suffer from the problems of the poorly designed ones, and the better designed ones will perform the same at the lower rate as at the higher rate.  

Quote:

and system latency can be reduced so digital processors can be used in real-time monitoring situations (especially vocal tracking).


For the most part system latency is improved with higher sample rates, though there are some situations where latency takes a hit as sample rates increase.

Nika.
Logged
"Digital Audio Explained" now available on sale.

Click above for sample chapter, table of contents, and more.

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 44k vs. 96k - a new discovery
« Reply #10 on: October 24, 2004, 03:08:38 PM »

Barry Hufker wrote on Sat, 23 October 2004 19:50

I wonder whether the difference is 96k versus 44.1 or whether it is the implementation of antialiasing filters.  According to a presentation by Rupert Neve, there is only one A/D chip (a brand new one and he didn't mention the maker or model) that actually trades a little bit of flat response (down1dB at 20k) for much better antialiasing.  He explained that chip makers have put the emphasis on marketing a ruler flat frequency response rather than aliasing.  So would the aliasing be of the same quality for 96 and 44.1?  I don't know.  Would it be quality antialiasing to begin with?  I don't know.


Barry,

Good questions.  There is indeed a tendency for chip manufacturers to "borrow" a little bit of bandwidth above the Nyquist Frequency for their filter roll-off, knowing that the aliasing will still only enter the data stream above the audible range.  Take, for example, a 48kS/s sampling system and an ear that can only hear up to 20kHz.  We can do our filtering from 20kHz to 24kHz, but this requires a very steep filter which means a lot of filter coefficients and thus more on-chip RAM and greater processing abilities.  The alternative is that we could have our filter roll-off between 20kHz and 28kHz, knowing that the 4kHz of added aliasing is going to end up between 20kHz and 24kHz, above the audible range anyway.  

This approach is fairly common.

What you are pointing at is a very noticeable reality that simply because current converters DON'T sound the same between their 44.1kHz and 96kHz sample rates does not inherently mean that they CAN'T sound the same.  In the end, a converter designer does not actually have to sacrifice EITHER the flat response OR the aliasing.  They just have to spend a few extra cents on enough DSP power to do the job right.  If they do then we can all spend a lot less money on post-processing DSP because we can keep the material at the lower rate and get the same quality.

I thinik the chip you may be speaking about is the AKM AK5394A.  It is indeed a very good chip.

Nika.
Logged
"Digital Audio Explained" now available on sale.

Click above for sample chapter, table of contents, and more.

Barry Hufker

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 8228
Re: 44k vs. 96k - a new discovery
« Reply #11 on: October 24, 2004, 05:39:30 PM »

Nika,

I think you explained Mr. Neve's point exactly in a way I never could.  Thank you!

I will look into the chip you suggest in the hopes of understanding the technical aspects of all this a little better.

Barry
Logged

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: 44k vs. 96k - a new discovery
« Reply #12 on: October 24, 2004, 07:10:51 PM »

[quote title=Nika Aldrich wrote on Sun, 24 October 2004 15:08]
Barry Hufker wrote on Sat, 23 October 2004 19:50

\\

Good questions.  There is indeed a tendency for chip manufacturers to "borrow" a little bit of bandwidth above the Nyquist Frequency for their filter roll-off, knowing that the aliasing will still only enter the data stream above the audible range.





Would you explain your reasoning, Nika, since the alias product is a difference product, which means, for example, a 1 kHz alias if the signal is 1 kHz above the Nyquist limit.

BK
Logged
There are two kinds of fools,
One says-this is old and therefore good.
The other says-this is new and therefore better."

No trees were killed in the sending of this message. However a large number of
electrons were terribly inconvenienced.

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 44k vs. 96k - a new discovery
« Reply #13 on: October 24, 2004, 08:11:41 PM »

bobkatz wrote on Mon, 25 October 2004 00:10

Would you explain your reasoning, Nika, since the alias product is a difference product, which means, for example, a 1 kHz alias if the signal is 1 kHz above the Nyquist limit.

BK


Bob,

The alias product is a difference product from the SAMPLING frequency, not the NYQUIST frequency.  Aliasing occurs at the absolute value of the difference of the multiple of the sampling frequency.  Therefore, in a 48kS/s sampling system, a 25kHz signal will alias at 23kHz.  A 47kHz or 49kHz signal will alias at 1kHz.  A 193kHz signal will also alias at 1kHz.  This is very rudimentary sampling theory.

We think of aliasing, therefore, as "mirroring" around the Nyquist frequency, or "folding back" for any frequencies up to the sampling frequency.  Any book on the subject of sampling will confirm this.

Nika.
Logged
"Digital Audio Explained" now available on sale.

Click above for sample chapter, table of contents, and more.

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: 44k vs. 96k - a new discovery
« Reply #14 on: October 24, 2004, 10:48:35 PM »

Thanks for the clarification!

BK
Logged
There are two kinds of fools,
One says-this is old and therefore good.
The other says-this is new and therefore better."

No trees were killed in the sending of this message. However a large number of
electrons were terribly inconvenienced.

dvuckovic

  • Newbie
  • *
  • Offline Offline
  • Posts: 26
Re: 44k vs. 96k - a new discovery
« Reply #15 on: October 24, 2004, 11:41:38 PM »

Nika Aldrich wrote on Sun, 24 October 2004 21:08

Barry Hufker wrote on Sat, 23 October 2004 19:50






I thinik the chip you may be speaking about is the AKM AK5394A.  It is indeed a very good chip.

Nika.


Correct me if I'm wrong, but this could as well be the new Texas Instruments chip.

As far as I know it it supposed to be much better than the trusty-old AKM found in so much audio gear(originally intended for the A/D in medical instruments)
Logged
Dejan Vuckovic

Studio Barba

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 44k vs. 96k - a new discovery
« Reply #16 on: October 24, 2004, 11:44:10 PM »

dvuckovic wrote on Mon, 25 October 2004 04:41

Nika Aldrich wrote on Sun, 24 October 2004 21:08

Barry Hufker wrote on Sat, 23 October 2004 19:50






I thinik the chip you may be speaking about is the AKM AK5394A.  It is indeed a very good chip.

Nika.


Correct me if I'm wrong, but this could as well be the new Texas Instruments chip.

As far as I know it it supposed to be much better than the trusty-old AKM found in so much audio gear(originally intended for the A/D in medical instruments)


The 5394 is a relatively new chip that was designed exclusively for audio purposes, as is indicated by its filtering and noiseshaping curves.  AKM has had several chips in the past, but the 5394 is getting a tremendous amount of respect in design circles.

The other three companies (BB, AD, and Crystal) all have competing hardware and new chip advances as well, but I believe the quote being referred to is the AKM chip.

Nika.
Logged
"Digital Audio Explained" now available on sale.

Click above for sample chapter, table of contents, and more.

Joe Bryan

  • Newbie
  • *
  • Offline Offline
  • Posts: 19
Re: 44k vs. 96k - a new discovery
« Reply #17 on: October 25, 2004, 01:38:28 AM »

Nika Aldrich wrote on Sun, 24 October 2004 11:58

Joe Bryan wrote on Sun, 24 October 2004 07:45

Specifically, well designed DSP filters will sound better at higher sampling rates because their responses can more closely approach the ideal analog response (in technical terms, the s-plane to Z-plane warping is more linear).


Not all filters perform better at higher sampling rates.  Some filters have greater accumulative error by using excess taps to perform the process.


Noise accumulation is a function of the rounding and/or dither modes used during accumulation in the filter, and as I said, any well designed filter will take this into account. Noise buildup is not as big of an issue as Z-plane warping, which causes increasing deviations from analog (s-plane) transfer functions in the upper octave(s).

In many cases, this makes it impossible to reproduce the exact transfer function without upsampling. For example, it's impossible to recreate the exact amplitude and phase characteristics of most high-order (>2) analog filters in digital without upsampling. A perfect example is the Pultec EQ. Any DSP process that claims to be a Pultec but doesn't upsample isn't matching the response correctly.

Nika Aldrich wrote on Sun, 24 October 2004 11:58

Quote:

Futhermore, at higher sampling rates, analog anti-aliasing and anti-imaging filter requirements are less stringent (allowing less phase shift in the main audio band),


Not true.  Almost all anti-aliasing and anti-imaging filters that you are speaking of are digital and linear phase, so no phase shift issues are improved by higher sampling rates.  The analog filters in any oversampling filter these days have phase shift so far below our audible phase discernability that this is clearly not a factor.  Further, most high sample rate converters use the same analog anti-aliasing/anti-imaging filters as their lower sample rate brethren.

Indeed, the anti-aliasing and anti-imaging filters are less stringent at the higher sampling rates, but this only pertains to what you call "poorly designed filters."  Not all filters have to suffer from the problems of the poorly designed ones, and the better designed ones will perform the same at the lower rate as at the higher rate.


I was referring to the analog filters, not the digital filters.

The digital filters only do part of the work. For a typical sigma-delta ADC, the analog anti-aliasing filter is trivial because it only needs to attenuate signals above the MHz range.

However, the analog anti-imaging filter for a DAC is not trivial. In today's sigma-delta converters, there is a lot of out-of-band energy that cannot be removed by the digital filter because it's ahead of the signal-delta DAC's modulator. The modulator's noise shaping shifts the quantization noise out of the main audio band, but this energy can only be removed by the analog filter.

It is this filter that benefits the most from higher sampling rates. The higher transition band provided by higher sampling rates allows for much less phase shift in the primary audio band while attenuating the high-freq energy that wreaks havock in the analog output amps. This has a major impact on sound quality, especially transparency and transient response.

Nika Aldrich wrote on Sun, 24 October 2004 11:58

Quote:

and system latency can be reduced so digital processors can be used in real-time monitoring situations (especially vocal tracking).


For the most part system latency is improved with higher sample rates, though there are some situations where latency takes a hit as sample rates increase.

Nika.


I can't think of any examples where this is true unless you're referring to up/down SRC when the processing is oversampled but not the analog conversion, could you provide some?

-Joe
Logged

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 44k vs. 96k - a new discovery
« Reply #18 on: October 26, 2004, 11:29:02 AM »

Joe Bryan wrote on Mon, 25 October 2004 06:38

In many cases, this makes it impossible to reproduce the exact transfer function without upsampling. For example, it's impossible to recreate the exact amplitude and phase characteristics of most high-order (>2) analog filters in digital without upsampling. A perfect example is the Pultec EQ. Any DSP process that claims to be a Pultec but doesn't upsample isn't matching the response correctly.


There are a lot of EQs out there that successfully do proper DSP without upsampling.  The Sony Oxford is a classic example.

Quote:

I was referring to the analog filters, not the digital filters.

The digital filters only do part of the work. For a typical sigma-delta ADC, the analog anti-aliasing filter is trivial because it only needs to attenuate signals above the MHz range.

However, the analog anti-imaging filter for a DAC is not trivial. In today's sigma-delta converters, there is a lot of out-of-band energy that cannot be removed by the digital filter because it's ahead of the signal-delta DAC's modulator. The modulator's noise shaping shifts the quantization noise out of the main audio band, but this energy can only be removed by the analog filter.


Sure, but the roll-off of that analog filter is so far out of band that increasing it further is unnecessary with respect to phase shift in the audible range.

Quote:

It is this filter that benefits the most from higher sampling rates. The higher transition band provided by higher sampling rates allows for much less phase shift in the primary audio band while attenuating the high-freq energy that wreaks havock in the analog output amps. This has a major impact on sound quality, especially transparency and transient response.


On a typical DSM based DAC the oversampling brings the excess noise so far above the audible range that a very gentle filter can be used, just like on the A/D side.  I see no grounds for claiming that making the filter even more gentle is better on audio band material - especially if the current system provides no audible phase distortion because the filter is already gentle enough.

Quote:

I can't think of any examples where this is true unless you're referring to up/down SRC when the processing is oversampled but not the analog conversion, could you provide some?


There are plugins that, when run at higher sampling rates, have to down sample in order to do the processing and then upsample at the end of the process.  

Nika.
Logged
"Digital Audio Explained" now available on sale.

Click above for sample chapter, table of contents, and more.

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: 44k vs. 96k - a new discovery
« Reply #19 on: October 26, 2004, 03:33:22 PM »

Barry Hufker wrote on Sun, 24 October 2004 04:45

Dan,

I head the 4-year undergraduate program in Audio Production at Webster University.  We invited Rupert Neve as a guest speaker.  I say this only to give you the circumstances under which Mr. Neve came to our campus this past spring to speak.  It was at that time (April or May of this year, I forget exactly), that Mr. Neve made the statement I attributed to him.  He even said this new chip was just coming out.

I was a witness to this statement, but this is not my area of expertise so I can't clarify it further.  But Mr. Neve spoke of it as an immediate (at that moment) development.

Barry



Hello Barry,

There is probably some misunderstanding here of some sort.
Again, the -1dB at 20KHz problem is an old thing from an anti aliasing filter standpoint. The new converters are way passed such problems.

Technically speaking, the -1dB at 20KHz when operating at say 44.1KHz, is a very real issue, but is due to the digital decimation filter, not the analog anti aliasing filter.

Again, the reason for my statement is due to the increased rate operation of the front end (the modulator).

I can not tell the reason for the misunderstanding. I do however stand behind my statements, which are just technical facts.

BR
Dan Lavry

Logged

Barry Hufker

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 8228
Re: 44k vs. 96k - a new discovery
« Reply #20 on: October 26, 2004, 05:41:43 PM »

Dan,

I am not an expert on this subject by any means and I don't dispute what you have said.  There probably is a misunderstanding on my part as to the full meaning of Mr. Neve's statement, but I think Nika summarized pretty well what I understood Mr. Neve to mean.

I can't elaborate further or else I would.  And I don't want to misrepresent Mr. Neve.  So maybe we'll have to let it go with the ground we've already covered.

Barry
Logged

Joe Bryan

  • Newbie
  • *
  • Offline Offline
  • Posts: 19
Re: 44k vs. 96k - a new discovery
« Reply #21 on: October 27, 2004, 02:11:42 AM »

Nika Aldrich wrote on Tue, 26 October 2004 08:29

Joe Bryan wrote on Mon, 25 October 2004 06:38

In many cases, this makes it impossible to reproduce the exact transfer function without upsampling. For example, it's impossible to recreate the exact amplitude and phase characteristics of most high-order (>2) analog filters in digital without upsampling. A perfect example is the Pultec EQ. Any DSP process that claims to be a Pultec but doesn't upsample isn't matching the response correctly.


There are a lot of EQs out there that successfully do proper DSP without upsampling.  The Sony Oxford is a classic example.


Yes, there are plenty of EQs that can produce good filter responses without upsampling, but that does not mean they can produce every desireable filter response. As I said, there are analog filters that cannot be reproduced digitally without upsampling.

Nika Aldrich

Quote:

I was referring to the analog filters, not the digital filters.

The digital filters only do part of the work. For a typical sigma-delta ADC, the analog anti-aliasing filter is trivial because it only needs to attenuate signals above the MHz range.

However, the analog anti-imaging filter for a DAC is not trivial. In today's sigma-delta converters, there is a lot of out-of-band energy that cannot be removed by the digital filter because it's ahead of the signal-delta DAC's modulator. The modulator's noise shaping shifts the quantization noise out of the main audio band, but this energy can only be removed by the analog filter.


Sure, but the roll-off of that analog filter is so far out of band that increasing it further is unnecessary with respect to phase shift in the audible range.


It's not "so far out" as you say. The one plot you won't see on any of the DAC spec sheets is the out of band noise caused by the modulator's noise shaping. This rises quite rapidly above the digital filter's stop band, and requires some non trivial analog filtering to remove it. The analog filter must be set as low as possible to remove this noise, and this requires juggling greater in-band phase error vs. greater out-of-band noise levels.

Nika Aldrich

Quote:

It is this filter that benefits the most from higher sampling rates. The higher transition band provided by higher sampling rates allows for much less phase shift in the primary audio band while attenuating the high-freq energy that wreaks havock in the analog output amps. This has a major impact on sound quality, especially transparency and transient response.


On a typical DSM based DAC the oversampling brings the excess noise so far above the audible range that a very gentle filter can be used, just like on the A/D side.  I see no grounds for claiming that making the filter even more gentle is better on audio band material - especially if the current system provides no audible phase distortion because the filter is already gentle enough.


See above. Unless my ears and test equipment have been deceiving me all these years, it does make a difference.

Nika Aldrich

Quote:

I can't think of any examples where this is true unless you're referring to up/down SRC when the processing is oversampled but not the analog conversion, could you provide some?


There are plugins that, when run at higher sampling rates, have to down sample in order to do the processing and then upsample at the end of the process.


Ah, yes. Forgot about those. We don't usually have to deal with that problem because we still haven't run out of cycles on our DSP. Wink

Sometimes downsampling is done when the process doesn't benefit from the higher audio bandwidth, and the excess cycles are unneccessary. Digital reverb is one common example, and because of the time-smearing nature of reverb, the added delays from the SRC are trivial. This type of reverb would still benefit from lower-latency, high sample rate processing in a monitoring environment as long as the direct path wasn't downsampled.

Another example is decimated linear phase filters, which have high intrinsic delays (like any linear phase process). These would never be used in a monitoring situation to begin with.

The only other case where downsampling is necessary is when a process can't run at full speed without overloading the system. No algo designer wants to do this, they're forced to do it to get the process to run, even if it compromises quality.

-Joe
Logged

ted nightshade

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1272
Re: 44k vs. 96k - a new discovery
« Reply #22 on: November 12, 2004, 07:55:48 PM »

Since Ivo is using the Lavry Blue, maybe Dan can explain just what it is about that specific design that causes the difference in sound between 44.1 and 96 in that specific instance?

Logged
Ted Nightshade aka Cowan

There's a sex industry too.
Or maybe you prefer home cookin'?

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: 44k vs. 96k - a new discovery
« Reply #23 on: November 12, 2004, 08:04:49 PM »

ted nightshade wrote on Fri, 12 November 2004 19:55

Since Ivo is using the Lavry Blue, maybe Dan can explain just what it is about that specific design that causes the difference in sound between 44.1 and 96 in that specific instance?




Different filters yield different sound.

BK
Logged
There are two kinds of fools,
One says-this is old and therefore good.
The other says-this is new and therefore better."

No trees were killed in the sending of this message. However a large number of
electrons were terribly inconvenienced.

Bob Olhsson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3968
Re: 44k vs. 96k - a new discovery
« Reply #24 on: November 13, 2004, 10:25:14 AM »

I wouldn't assume this must be the new A to D converter rather than something downstream that his previous converter was masking.

Schallfeldnebel

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 816
Re: 44k vs. 96k - a new discovery
« Reply #25 on: November 13, 2004, 11:23:31 AM »

Was the AK 5394A not the first chip with 192K possibility? Genex uses this chip in their GXA 85 AD converter for the PCM side, not for the DSD side. This is a multibit delta sigma chip. Weiss engineering uses this chip in the new ADC 2.

The chip Mr. Neve spoke about in his lecture, probably is a cousin from the Crystal 5397, which is used by DAD, Digital Audio Denmark in their 2402 and 2408 converters, and this converter chip is not new at all, it came together with the wellknown Crystal 5396 which Daniel Weiss used for a very short moment in the ADC1 mark 2.

Both chips get extremely hot. The 5397 is a chip, and DAD can explain it better themselves, see www.digitalaudio.dk ,which uses different filtering, and therefore reduces A.I.D., which stands for aliasing intermodulaton distortion. Because the chip gets so hot, I heard the DAD boys from Denmark speak about, another manufacturer, (Texas instruments ????) was interested in producing such chip, using less power and have a lower heat dissipation.

Maybe that is the chip Mr. Neve was speaking about.

Erik Sikkema

Logged
Bill Mueller:"Only very recently, has the availability of cheap consumer based gear popularized the concept of a rank amateur as an audio engineer. Unfortunately, this has also degraded the reputation of the audio engineer to the lowest level in its history. A sad thing indeed for those of us professionals."

Brandon Schexnayder

  • Newbie
  • *
  • Offline Offline
  • Posts: 3
Re: 44k vs. 96k - a new discovery
« Reply #26 on: November 15, 2004, 05:03:15 PM »

Can I just ask a very naive question...as I am still quite new to the recording world:

Is it possible that there may be a much more basic reason for a subjective preference at 44kHz than the converters-- for instance an ingrained preference for audio at this sampling rate, simply due to what we have become so accustomed to?

I don't mean to negate that technology may have been substandard, I just genuinely wonder if there is more of an X factor at play here.
Logged
Brandon M. Schexnayder
New Orleans, LA...err now Nashville, TN

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: 44k vs. 96k - a new discovery
« Reply #27 on: November 15, 2004, 07:43:14 PM »

ted nightshade wrote on Sat, 13 November 2004 00:55

Since Ivo is using the Lavry Blue, maybe Dan can explain just what it is about that specific design that causes the difference in sound between 44.1 and 96 in that specific instance?




1. It is NOT the analog. It is NOT the clocks.

2. The biggest single difference in the AD (44.1KHz to 96KHz) is the DIGITAL decimation filter.

3. But some of the difference may also be in the DA side. At the DA the biggest difference is the DIGITAL up sampling filter.

Best Regards
Dan Lavry
Logged

ted nightshade

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1272
Re: 44k vs. 96k - a new discovery
« Reply #28 on: November 16, 2004, 10:35:48 AM »

Thanks Dan! You couldn't be clearer.
Logged
Ted Nightshade aka Cowan

There's a sex industry too.
Or maybe you prefer home cookin'?

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 44k vs. 96k - a new discovery
« Reply #29 on: November 16, 2004, 10:41:41 AM »

Schexnayder wrote on Mon, 15 November 2004 17:03

Is it possible that there may be a much more basic reason for a subjective preference at 44kHz than the converters-- for instance an ingrained preference for audio at this sampling rate, simply due to what we have become so accustomed to?


First we have to ascertain whether or not there is truly a difference in the audible results of the sampling rates.  So far everything is point at the audible difference being a manifestation of implementation - that if we design the circuits "right" then the difference goes away.  If that is the case then there would be no "preference."

Make sense?

Nika
Logged
"Digital Audio Explained" now available on sale.

Click above for sample chapter, table of contents, and more.
Pages: 1 2 [All]   Go Up
 

Site Hosted By Ashdown Technologies, Inc.

Page created in 0.076 seconds with 17 queries.