Excluding hi-res and HD, why use a DAC with processing capabilities greater than 16 / 44.1?

[Fokus]

[ATTACH=CONFIG]10938[/ATTACH]

 

[ATTACH=CONFIG]10939[/ATTACH]

 

You wouldn't expect to find this level of relative variation in impulse response between two speakers,

 

You must be joking. Either that, or you still don't get it.

[Jud]

You must be joking. Either that, or you still don't get it.

 

Hi, Fokus. The basic point I am trying to make is a quite simple one: It is incorrect to think of oversampling as an exact multiplication process. Oversampling involves filters with variations in their responses to common tests, such as the impulse response tests shown in the graphs. Do I think these graphs mean there must be an audible difference between these two filter settings greater than one would find between two speaker drivers? No, certainly not based purely on the filters' impulse responses. That's why I referred to a level of relative variation, rather than saying the absolute variation between filters is greater than that between speakers.

 

I have heard the iZotope 64-bit SRC filter at linear phase and intermediate phase settings, and whether for reasons of phase behavior, dispersion, or other factors, the two settings did sound quite different to me. Of course I am fully capable, when playing around as a complete amateur with these settings, of making something that no self-respecting designer would send out into the world. But having heard the various filters available in Miska's HQPlayer, I can say that I was very surprised by the degree of difference to the sound that the different choices made. If I was surprised, then I felt perhaps other people might also be surprised by the notion that there could be substantial differences in sound between commercially available filters, though these days it is becoming more common for both player software and DACs to offer different filter settings.

 

Re whether I "get it," unless and until I learn the math, then no, I won't get it on anything other than a shallow metaphorical level.

 

A couple of questions, if you feel like answering:

 

- Do you feel impulse response graphs such as these tell us anything useful about filters, and if so, what?

 

- Why is the ringing behavior of filters something that often seems to come up in commonly run tests, manufacturers' websites, and discussions among designers (e.g., Miska, PeterSt, Charles Hansen) on boards like these? Is there any actual importance to it? Is it a myth which the manufacturers and designers themselves believe? Or do you feel it is something in the nature of a conspiracy perpetrated on credulous buyers?

[wakibaki]

Well, just to be clear, I doubt I would engage you to vet or peer review any study or research of mine. You demonstrate an appalling lack of the objectivity needed for that task.

 

Given that though, you are rather sadly out of date in your knowledge or understanding of the issues involved. Or of the studies that have clearly provided different conclusions that what you have choosen to adopt- as your opinion.

 

Nor am I employed as a researcher paid to do your basic research for you.

 

But mostly your opinion is greatly at odds with people in the industry and hobbyists whose opinions I respect.

 

And even the most arrogant of them are not usually given to making such statements of "fact" as you did in the message I originally politely replied to.

 

Seriously, you should consider doing some more research and losing the attitude. Doesn't bother me, but it is not the way to win friends and influence people around here. Even when one has intereting things to say, as I am assured you do.

 

No skin off my back though- no amount of nastiness will get me to do your research for you or gain you the respect you desire. Or get me to engage with you in a foolish conflict of opinions.

 

If you had anything, Paul, you would provide a reference.

 

w

[wakibaki]

Since any filter existing in software can be frozen into hardware at trivial effort and cost then if any audible benefit were obtainable it would be incorporated into some DAC and the fact written up into the advertising blurb and used to sell it. Since we see no evidence of this, there is no audible benefit to be obtained.

 

Engineering is far from being all about the esoteric, mostly it is about the application of common sense to comparatively straightforward real-world issues that only require consideration in the absence of (audiophile) mumbo-jumbo.

 

If you are of the school that believes that your sighted listening impressions trump the controlled double-blind testing undertaken by DAC manufacturers, then undoubtedly you will 'hear' a difference, or think you hear a difference, if whatever it is you are doing in your skull can reasonably be described as thinking.

 

w

[k6davis]

Since any filter existing in software can be frozen into hardware at trivial effort and cost then if any audible benefit were obtainable it would be incorporated into some DAC and the fact written up into the advertising blurb and used to sell it. Since we see no evidence of this, there is no audible benefit to be obtained.

 

Engineering is far from being all about the esoteric, mostly it is about the application of common sense to comparatively straightforward real-world issues that only require consideration in the absence of (audiophile) mumbo-jumbo.

 

If you are of the school that believes that your sighted listening impressions trump the controlled double-blind testing undertaken by DAC manufacturers, then undoubtedly you will 'hear' a difference, or think you hear a difference, if whatever it is you are doing in your skull can reasonably be described as thinking.

 

w

 

You should get out more.

 

Keith

[Jud]

Since any filter existing in software can be frozen into hardware at trivial effort and cost then if any audible benefit were obtainable it would be incorporated into some DAC and the fact written up into the advertising blurb and used to sell it. Since we see no evidence of this, there is no audible benefit to be obtained.

 

w

 

There's a very good, tremendously educational thread on hi res in the forums here that's gone into some detail on filters. Highly recommended. Title talks about 192 and 48kHz sample rates.

[Paul R]

If you had anything, Paul, you would provide a reference.

 

w

Amusement. I am very mercenary and open for hire. I can do remedial manners as well as remedial math and audio 101. Good rates too.

[Sam Lord]

Jud- this is written very well indeed.-Paul

+1 Your detailed post was very helpful Jud.

+2 Another great job. Jud, for a guy who is modest about his math abilities, you possess some of the finest explanatory and logical skill I've come across. When you question something I've written, I've done it wrong!

Since any filter existing in software can be frozen into hardware at trivial effort and cost then if any audible benefit were obtainable it would be incorporated into some DAC...

You should get out more.

Indeed, and try building filters. Sorry wakibaki, but you have no understanding of the properties of basic electronic parts or electronic circuits. I suggest you start with resistors, then move on to the reactive parts. Oh wait, resistors are reactive to some degree, and that fact alone nullifies your statement.

You must be joking. Either that, or you still don't get it.

No, Jud made a very clear point, and clearly gets it.

[mayhem13]

 

; and better in the listener's PC than in the DAC.

 

This statement makes quite a few assumptions.....first being that the implemented additional algorithmic filtering is both audible and beneficial......how you apply the smoothing can have inverse effects as your earlier impulse graphs show. Seperating the component waveforms for phase correction is inherently problematic as the actual correction occurs only when the two waveforms are summed later. Again, the accuracy of the summation is a prediction done by the filter designer. You can theoretically throw filters at a waveform till infinity and produce no appreciable improvement to the final waveform after the summation. What many within the digital audio realm propose is essentially 'tweaking' the filters when what's actually needed is good ol manual phase correction......which simply isn't a reality. As with all things, there's always a compromise which leaves us with our algorithms.

 

......in regards to phase correction which I 'assumed' was the core of this particular point. I have quite a few thoughts on phase quadrature in the analog realm which again forces my logic back to analog part of the chain. The filters you propose can in fact perform an excellent job of smoothing...........and then the end user puts all that wonderfully converted digital through a speaker whose designer misaligned the phase overlapp right at 1khz....or the most sensitive part of human hearing. How's that for a kick in the pants?

[Paul R]

I think that phase correction, the way you are thinking about it, is more in the digital realm than the the "filters" in a DAC. Whether DAC filters are better done in a computer or in a piece of hardware is quite arguable. On the computer side, you have Miska and PeterST who both have done miraculous things using the CP in a computer. On the other side, you have, among others, Charlie Hanson at Ayre, whose FPGA based filters are nothing short of terrific. And several approaches in between the two extremes.

 

It isn't an argument that will be won, and indeed, may be on a path that will lead to a convergence. Easily updated filters in a DAC or building top quality DAC hardware into a computer. That sounds a lot like what we are doing today, no?

 

Now, phase manipulation - are you taking about stuff like what Dave Higginbottom is doing with his DEQX stuff? That stuff is amazing and enough to give enough the most staunch audio purist pause about DSP. I keep wanting to pull the trigger on one, but am waiting for the price to come down just a wee bit more.

 

-Paul

 

 

This statement makes quite a few assumptions.....first being that the implemented additional algorithmic filtering is both audible and beneficial......how you apply the smoothing can have inverse effects as your earlier impulse graphs show. Seperating the component waveforms for phase correction is inherently problematic as the actual correction occurs only when the two waveforms are summed later. Again, the accuracy of the summation is a prediction done by the filter designer. You can theoretically throw filters at a waveform till infinity and produce no appreciable improvement to the final waveform after the summation. What many within the digital audio realm propose is essentially 'tweaking' the filters when what's actually needed is good ol manual phase correction......which simply isn't a reality. As with all things, there's always a compromise which leaves us with our algorithms.

 

......in regards to phase correction which I 'assumed' was the core of this particular point. I have quite a few thoughts on phase quadrature in the analog realm which again forces my logic back to analog part of the chain. The filters you propose can in fact perform an excellent job of smoothing...........and then the end user puts all that wonderfully converted digital through a speaker whose designer misaligned the phase overlapp right at 1khz....or the most sensitive part of human hearing. How's that for a kick in the pants?

[mayhem13]

Unaware of Higgenbottom's work. Please share a link to something you found of interest?......as I'm interested too.

 

FYI....i was talking about phase correction filters both within a DAC chip and extraneous software.....not sure of that makes a difference to the content of your reply.

 

And I would challenge your statement that Miska or Peter have accomplished miraculous achievements in digital filtering and artifacts as the results of that work is completely shrouded in subjectivity as it stands. This isn't to not acknowledge the hard work they've contributed......but simply to point out that the work is far from proven. As you clearly understand, such 'revelations' past what industry already knows and understands would have the potential for significant gains in communications, manufacturing, aerospace, medical and every other technical field you can imagine. I would submit that if Miska and Peter have such faith in their accomplishments to date, far better to seek capital support now and further develop the technology past theoretical experimentation.

 

All things have humble beginnings.....but it's the end game that matters......

[Paul R]

Unaware of Higgenbottom's work. Please share a link to something you found of interest?......as I'm interested too.

 

DEQX High Definition Audio

 

 

FYI....i was talking about phase correction filters both within a DAC chip and extraneous software.....not sure of that makes a difference to the content of your reply.[/Quote]

 

Phase correction in the digital or analog world? Before or after D2A conversion? It all rather makes a difference.

 

 

And I would challenge your statement that Miska or Peter have accomplished miraculous achievements in digital filtering and artifacts as the results of that work is completely shrouded in subjectivity as it stands. [/Quote]

 

I do not think they would agree with you, as both of them measure everything quite objectively. However either or both of them can speak for themselves. I have no idea where you are getting the information you have to make such an assumption as this. I think your assumption is challenged even by perusing the postings of their measurements they have done here on CA over the years.

 

 

This isn't to not acknowledge the hard work they've contributed......but simply to point out that the work is far from proven. As you clearly understand, such 'revelations' past what industry already knows and understands would have the potential for significant gains in communications, manufacturing, aerospace, medical and every other technical field you can imagine. I would submit that if Miska and Peter have such faith in their accomplishments to date, far better to seek capital support now and further develop the technology past theoretical experimentation.

[/QUote]

 

I am not at all sure what "revelations" you are talking about or are referring to. They both have top of the line products available that do exactly and precisely what they say they do, using well known and well documented scientific methods. Though their results are startling, I am not sure the technology is. Same is true for some of the other folks on the system. I just used those guys as examples because they have nice tough thick skins... YMMV.

[wakibaki]

Indeed, and try building filters. Sorry wakibaki, but you have no understanding of the properties of basic electronic parts or electronic circuits. I suggest you start with resistors, then move on to the reactive parts. Oh wait, resistors are reactive to some degree, and that fact alone nullifies your statement.

 

You mean like invent, design the hardware for an asynchronous S/PDIF receiver and write the VHDL to run it? You can find the VHDL on my website...

 

 

Or design and build a tube headphone amp?

 

 

Or a TPA6120A based portable headphone amp with independent R/L digital volume control using PIC and PGA2320...

 

 

...or even a touch-sense/remote relay-switched attenuator with a unique illuminated display?

 

 

How 'bout an electrostatic headphone amplifier using Depletion MOSFETs in simulation?

 

 

w

[Superdad]

Since any filter existing in software can be frozen into hardware at trivial effort and cost then if any audible benefit were obtainable it would be incorporated into some DAC and the fact written up into the advertising blurb and used to sell it. Since we see no evidence of this, there is no audible benefit to be obtained.

 

Come on guys, this is not rocket science. The coefficients for an SRC filter to begin to equal the number of taps that can be generated with SoX or iZotope (or HQPlayer or XXHighEnd) on a computer can require an enormous FPGA--we are talking the $240 behemoths, not the modest size FPGAs that Chord, Ayre, PS Audio, etc. use. Which is not to say that you can't create and load a very nice sounding filter into a modest FPGA (see my recent report about optimizing one in s/w for loading into the FPGA of the forthcoming Bottlehead DAC-- http://www.computeraudiophile.com/f8-general-forum/do-all-dacs-sound-more-or-less-same-18877/index12.html#post300613).

 

But as far as standard DAC chips having room or creatively (i.e. best SQ) optimized filters in them? Fagetaboutit!

 

And Wakibaki, where have you been hiding for the past few years? Digital filter performance (and sound) is a lot of what stand-alone DAC designers/marketers have been talking about for a long time.

[Sam Lord]

...Sorry wakibaki, but you have no understanding of the properties of basic electronic parts or electronic circuits...

You mean like invent, design the hardware for...

I'm very sorry wakibaki, I was wrong to have made to unfounded assumptions about your abilities. I hope you accept my apology. I assume that when you said this

 

"...Since any filter existing in software can be frozen into hardware at trivial effort and cost..."

 

that you had a particular meaning, not the very general meaning that I assumed. Again, I apologize.

[Fokus]

- Do you feel impulse response graphs such as these tell us anything useful about filters, and if so, what?

 

Yes. The impulse response is the filter. Years ago, days after some magazine published a rather extensive test report on the new top-line Meridian CD player I had a clone of its allegedly revolutionary filter running at home.

 

- Why is the ringing behavior of filters something that often seems to come up in commonly run tests, manufacturers' websites, and discussions among designers (e.g., Miska, PeterSt, Charles Hansen) on boards like these? Is there any actual importance

 

Depends. When the filter acts in the commonly-accepted audible band its ringing signature is audible when you know what to listen for.

 

But when it approaches 20kHz this is no longer the case, at least when one auditions competently-designed filters (even when they are 'wrong') and the listening test is done correctly. I've seen this time and time again, with many listeners and on top class systems. I'd love to trial this with teenagers, though.

 

As for iZotope, you are aware that only a tiny subset of its parameter space yields valid filters? It is no wonder that one hears differences when some of the filters tried are just effects generators generally unworthy of the term AI or AA filter.

 

to it? Is it a myth which the manufacturers and designers themselves believe? Or do you feel it is something in the nature of a conspiracy perpetrated on credulous buyers?

 

It makes attractive ad copy, for sure. It looks spectacular, and myths are indeed very easy to build ...

[Miska]

This isn't to not acknowledge the hard work they've contributed......but simply to point out that the work is far from proven.

 

You can compare some of my filters to number of others here:

SRC Comparisons

Newest ones are from 2.9.1 version. These are optimized for shortest possible impulse response while having best possible stop-band attenuation. (and when used for oversampling, these are apodizing, so in most cases they replace the ringing from original source)

 

Lot of ADC and DAC chips have poor filters with just 75 dB stop-band attenuation and cascade designs. Even more, many DAC chips use proper filters only up to 8x oversampling and then go with S/H or linear interpolation. Just horrible.

 

HQPlayer can run a single-pass conversion for example to 512x with very high stop-band attenuation.

 

And then of course we could talk a lot about noise shaping and dithering. I doubt a DAC chip can have a proper random number generator for it's dither.

[Fokus]

You can compare some of my filters to number of others here:

SRC Comparisons

 

Ah, thanks for pointing out that yours are now up for scrutiny at SRC ... it seems that you and me follow the same strategy for setting filter requirements.

[sandyk]

Precisely. Not rocket science.

 

I read your piece. You have no blind-test data to support your assertions, you have no right to style yourself as an engineer.

 

w

Hi Fred

Alex C has never claimed to be that, however a very good friend of his , who is working closely with him, is highly respected E.E. John Swenson.

 

Kind Regards

Alex

[k6davis]

I throw up my hands. I'm completely out of my league when it comes to the science being discussed here.

 

Here's what I'd like to know. There seems to be a lot of excitement about FPGA's. Are they a more powerful platform from which to build an upsampling algorithm than a Core i7 in a PC or Mac?

[Fokus]

For executing algorithms the underlying architecture doesn't matter to the result, although the ease of implementation can obviously differ.

 

The appeal of FPGAs is that they allow anyone to build a DAC without using an actual DAC chip.

[k6davis]

For executing algorithms the underlying architecture doesn't matter to the result, although the ease of implementation can obviously differ.

 

The appeal of FPGAs is that they allow anyone to build a DAC without using an actual DAC chip.

 

Thank you. That was very helpful.

 

Keith

[Miska]

For executing algorithms the underlying architecture doesn't matter to the result, although the ease of implementation can obviously differ.

 

Well, other than from processing capabilities point of view. So certain amount of processing resources is needed to perform given algorithm. For example I calculated that HQPlayer performs roughly 3.2 billion calculations per second for DSD64 upsampling. And given nature of the algorithms, at minimum 142 MHz clock speed is needed. This scales linearly so DSD128 takes twice the amount of resources and clock speed, DSD256 four times and DSD512 eight times. So DSD512 is about 25.6 billion calculations per second and minimum 1.136 GHz clock speed.

[Miska]

For fun, I calculated that it would require roughly an Altera Stratix V FPGA to perform those calculations and that about 2000€/piece, so more that 6x as much compared to Core i7 4770K.

[Jud]

Something that interests me is what appeared to be a mild disagreement between Miska and Fokus in comments in this thread and the 192/48kHz one (though I could easily be misinterpreting).

 

Miska I think is looking at things from a time domain point of view and saying filter ringing should be audible as smearing of transients. Fokus I think is looking at things from a frequency domain point of view and saying for people like me who can hear to perhaps 15-16kHz, ringing should not be audible.

 

Am I accurately restating what each of you has said? Any further thoughts or comments?