Hmm, I suppose that Jan's analysis of the 21kHz sine wave was wrong (edit: or 'dumbed down') but the analysis of ringing in the square wave is correct?

The two analyses do seem to be contradictory, because if you use linear interpolation there would be no ringing in the square wave and impulse reproduction.

It seems to me that ringing is the inevitable result of having too low a sample rate. If you knew the input was a sine wave to start with you'd just keep the sampled points as they are and use linear interpolation in this part of the wav to reproduce the original signal as closely as possible. But since you don't KNOW this is a square wave and you this sampling rate usually can't reproduce frequencies above the Nyquist frequency correctly, you have to either choose to reproduce normal music correctly and square waves and sharp impulses incorrectly, or reproduce square waves and sharp impulses correctly and everything else incorrectly

Or use the windowing filter and lose HF.

Andre, are you saying that all that is needed to eliminate ringing is a rolloff from 15kHz on up? That you can't implement that with any digital or analog EQ, and that just doesn't seem right

Also, I suppose you and aos are saying that it's impossible to design a perfect analog brickwall filter but possible to design a pefect digital brickwall filter--and this is already done in today's DACs?

And you two seem to be saying that it is part of the JOB of a perfect brickwall filter to introduce oscillation BEFORE an impulse?

Hmm, is it POSSIBLE for a system that cannot detect frequencies above 21kHz to nevertheless detect ringing caused by removal of all frequencies above 21kHz?

From the old upsamling DACs thread:
Quote:



Excuse me--accurate upconversion requires interpolation, and interpolation works using sinc filtering, which is a frequency domain process, isn't it?

Why can sinc interpolation only be approximated not actually implemented? Can the correct interpolated value using sinc be calculated or can't it?

What, in your opinion (um, everyone that is), is the best filter and DAC chip out there?

Jan,

I do love the Analoguer but I find I use it either full-on or full-off. It's off most of the time but on some things I find I need it as much as it can. At that point it's a matter of choosing the lesser of two evils-deal with the 'artifacting' or lose some of the high-end. I can hear a distinct difference on each setting and the full-on setting still doesn't eliminate it so I keep it there when needed.

BTW, the bass-boost is perfect full-on when used with the K501s. HD600s only need the first setting when the crossfeed is on and even then, that's to taste.

On another note, I actually find *bad* recordings more pleasant to listen to.

Anders,

"It would be interesting to know a little about how RF noise and pollution affects the CD player."

The most common and insidious problem will be common-mode RF noise, and there are lots of solutions to prevent this noise from entering a DAC, or to control it within a CD player. Here's one example: a popular technique for mods and audiophile players use high-speed, wide-bandwidth opamps for various gain and buffering functions in the DAC. Almost every use of these opamps I've seen have very inadequate power-supply bypassing. And I don't mean they use crappy caps instead of audiophile caps. I mean that the power supply bypass is ineffective at the highest frequencies that the opamp can respond to. Why is response at 40 MHz important for 20 kHz baseband signals?

The reason is because of negative feedback. NF trades off gain and bandwidth for distortion reduction. Looking at any opamp data sheet, you will see that gain reduces as you near the opamp's open-loop bandwidth. For high-speed opamps, that's in the megahertz range, which is also in the RF noise range. If there is inadequate power-supply bypassing, RF noise can enter the opamp. If there is noise at the higher frequencies of an opamp's open-loop bandwidth, the opamp will have very little gain to reduce distortion, and if the RF noise is sufficiently strong, will cause the opamp to distort. The opamp can be operating non-linearly, and as a result intermodulate noise down to the audible range.

"There are two now highly praised DACs (e.g Stereophile) that are not extremely expensive: Musical Fidelity A3-24 and Perpetual Technologies. It's easy to ask, but you may know something?"

Sorry, I'm not familiar with these DACs.

--Andre

Kelly,

"I have two Panasonic portable players, the CT570 and the S320. I also have the Sony S7000 DVD player you mentioned explicitly.

In my experience, theres three players do exhibit the type of digital artifacting that haunts both Nezer and me."

Clearly something is going wrong, but I can't begin to guess what. I use the S7000 as a transport mostly, and have not listened to its analog outputs (and headphone output actually) very much. Perhaps if you could isolate an instance of music where you hear what you hear, we can at least have a good frame of reference.

My Panasonic PCDP is from 1994, and I don't know how they'd compare to the two you mention.

A possibility is a system effect where a particular combination of components somehow exacerbates the digititis problem.

--Andre

"Is Wadia really using a spline interpolator? That's nuts. (Doesn't that inject garbage as a side effect across the frequency spectrum?)"

Yes, and yes. They've improved their interpolater quite a bit since their first iteration, and it doesn't behave as badly as the original ones did. The intermodulation tests I saw looked very literally hairy.

--Andre

Joe,

"Hmm, I suppose that Jan's analysis of the 21kHz sine wave was wrong (edit: or 'dumbed down') but the analysis of ringing in the square wave is correct?"

It rings because of the way it was sampled, and because there is a discontinuity. The square wave cannot be a square wave post-sampling because you have to filter out frequencies above half the sample rate.

"The two analyses do seem to be contradictory, because if you use linear interpolation there would be no ringing in the square wave and impulse reproduction."

Yes, but you'd have far worse problems (as shown by Jan's wobble) if you linearly interpolate. No one has proven or disproven that ringing is important, BTW.

"Or use the windowing filter and lose HF."

This is actually an option in some Sony CD players where you can attenuate more high frequencies for a prettier-looking measurement. Jan's box does this. Wadia seems to be heading in this direction as well.

"Andre, are you saying that all that is needed to eliminate ringing is a rolloff from 15kHz on up? That you can't implement that with any digital or analog EQ, and that just doesn't seem right"

You can do this.

"Also, I suppose you and aos are saying that it's impossible to design a perfect analog brickwall filter but possible to design a pefect digital brickwall filter--and this is already done in today's DACs?"

No perfection here. It's possible to design a much better brickwall filter in digital than in analog. It's also cheaper, so that's why everyone does it.

"And you two seem to be saying that it is part of the JOB of a perfect brickwall filter to introduce oscillation BEFORE an impulse?"

No, but a consequence of the way a digital filter is implemented for brickwalls, you will have ringing. This is because its time-domain response is something less than infinitely long.

"Hmm, is it POSSIBLE for a system that cannot detect frequencies above 21kHz to nevertheless detect ringing caused by removal of all frequencies above 21kHz?"

I suppose, but what if I composed a piece of music that purposely has ringing in it?

"Excuse me--accurate upconversion requires interpolation, and interpolation works using sinc filtering, which is a frequency domain process, isn't it?"

They are two sides of the same coin. Frequency analysis is just a handy tool that happens to work on many real-word problems. Interpolation is filtering is interpolation. Sometimes it's easier to analyze a problem in frequency domain, and we do it that way. Sometimes it's easier in time domain, and we use that. BTW, it's sinc interpolation which is the equivalent of brickwall filtering.

"Why can sinc interpolation only be approximated not actually implemented? Can the correct interpolated value using sinc be calculated or can't it?"

Because you won't wait forever for your music to be played. It can be approximated very, very well, and in real life, we only have to approximate to an accuracy that falls below the noise floor.

"What, in your opinion (um, everyone that is), is the best filter and DAC chip out there?"

There are lots of good ones. The most important thing isn't what part something uses, instead it's how the whole thing is put together and implemented. Given the best chips, anyone can easily mess it up.

--Andre

If you have Pearl Jam's Binaural, all of Ed Vedder's vocals on 'Soon Forgot' after about the first phrase or two exhibit what I'm talking about.

I can extract a piece of this and show high-pass from about 12k up to make it *really* stand out. After hearing just the higher frequencies it's easier to tune-into that effect in the full-bandwidth stream.

I'll try (if I can make the time) to do this tonight. I'll provide both samples in Monkey's Audio. I'll use Cooledit Pro to implement the high-pass filter.

If you don;t hear any obvious problems, please step down to your onboard DAC and see if you can find it there. If not I want to know *every* piece of equipment you own, how much it costs and where to get it.

Ultimatly, all I care about is the sound and that means eliminating this digititus with brute-force means if necessarry.

Andre

So you've not actually listened to the analog out of the S7000?

Check the model number of your Panaonic portable player and I'll let you know how it compares. I'm pretty familiar with their history, now.

The "something else going wrong in the chain" argument doesn't hold up against switching components and leaving everything else the same, which I've done.

I'm sorry if I seem defensive. This is reminiscent of the "cables do so matter" thread going on elsewhere in the forum. I feel as if I'm trying to convince you to believe in ghosts. The digitis problem is not something I just invented from boredom.

Andre,

Thank you very much for the explanation of how common-mode RF noice degrades the performance of the CD player. Many implementations of techniques to prevent RF noice from entering the DAC seem far from successfull. If I understood you correctly, the RF noise comes from the power supply / AC line and results in distortion at audible frequencies through negative feedback and intermodulation.
If the CD player cannot handle this problem, couldn't it instead be solved by feeding the player with clean AC? Of course I here think of power conditioner and power cable solutions that really remove the RF noise (or a substantial part of it).

Jan said:

>>>With a true sinc interpolation there would be no wobbling. However, as any practical algorithm is just a approximation the wobbling can not completely be removed. It will of course be much less than shown in my article but please understand that I used linear interpolation for didactic reasons. Hardly any normal mortal knows what a convolution or a sync function is and what their relations to the FFT of a deltafunction are. To explain a problem one sometimes has to simplify.


Exactly as I've suspected. I thought Jan "dumbed down" the article for non-engineers to understand better, unfortunately it made it deviate from the truth and thus made it suspicious to people who do understand the process. As well as that there should be no wobbling in theory. Jan's filter definitely does something to sound, and helps improve the sound of real life equipment, though I still think that a good quality stuff shouldn't experience these effect to the point where processing by analoguer is necessary.

I agree with most things AndreYew said in his replies, about filters, ringing etc. so no need to repeat.

>What, in your opinion (um, everyone that is), is the best filter >and DAC chip out there?

The best digital filter is considered to be NPC5482. Few years ago, the best DAC chip was considered by some to be PCM63. However, newer chips have been released, so it is hard to say. I haven't heard CS43122 which should have superb digital filter as well as DAC, and new Burr Brown DACs are supposedly superb too.

>There are lots of good ones. The most important thing isn't what part something uses, instead it's how the whole thing is put together and implemented. Given the best chips, anyone can easily mess it up.

My portable DAC uses AD1866 which is a far cry from current cream of the crop, and its parameters are not impressive. But it sounds great because of use of ultralow noise power regulators, that regulate well into the MHz region, polymer capacitors that also behave well into the MHz region, and good analog stage (Jung design, basically the same as META42). You can use the best chip and then bypass the whole board with a couple of 5 cent capacitors a couple of inches away from chips and further use badly designed 2 pole opamp filter (classic Butterworth from a textbook, ignoring impedance matching, non-linearities, open loop gain etc.) and get bad results.

HEY! 500th post!! )

Quote:



How is it I have more than you?

Oh, because I'm a loud-mouth, opinionated, arrogant, know-it-all and you aren't.

If I found a DAC that cures this problem I'm hearing I'd buy it in a heartbeat (up to $3500, $5k w/ a transport).

So, what I'm hearing is that there is no inherient problem with 44.1k audio it's just problems with real-world implementations. This would mean that Wadia's solution might not have anything at all to do with the DSP and might be in design (though it doesn't sound this way from what I'm hearing).

Quote:



Strictly speaking, CD audio is not perfect. First, even though 44.1kHz sampling is sufficient to perfectly represent any <20kHz signal given an ideal DAC, there is the 16-bit quantization issue. For loud recordings, it doesn't tend to matter, but for wider range classical (e.g. Mahler's 1st symphony, or the 4th movement of Mahler's 2nd symphony), there is a perceptible difference in resolution as signals get quieter. Also, there is plenty of psychoacoustic evidence to suggest that harmonics above 20kHz do indeed affect our perception of sound, even though we can't hear them consciously. CDs are missing all of that.

Neither of these things would explain the artifacts you're hearing, though.

Personally, I believe that the job of DAC should be to convert the digital signal to analog as faithfully as possible. If listeners want to add extra frequency information that's not in the original recording (as the Wadia spline interpolation technique must do), listeners should add a separate effects processor after the DAC (or use a tube amp). Jan's Analoguer fits the bill nicely.

Although I for one would certainly appreciate being able to buy Wadia's processor outside of having to buy a $9000 DAC, I believe it is integrated for a reason. From reading these threads, I understand that the filtering may not be occurring in the digital domain as I initially thought, but that doesn't mean it's so easy to pull it out from the DAC and reproduce the same results.

I for one believe the Wadia DOES produce the analog signal more faithfully than other CD players. I would listen to one before assuming otherwise. I understand that their "technique" may not garner approval from purists but I'm forced to acknowledge results over methodology.

I'll definitely agree with you in that results trounce methodology, but there's no arguing with the mathematics. (And, as AndreYew points out, this is one case where the measurements do confirm what the theory predicts.) With their interpolation technique, there's no way the Wadia is reproducing the analog signal more faithfully than other CD players.

I'm not saying the Wadia doesn't sound better. Heck, it probably does. But that goodness isn't on the recording; it's being invented/added by the hardware.