Ayre wants $1.5K for DSD'ed QB-9

[Charles Hansen]

From the 6Moons Auralic Vega review:

 

"However converting such 1-bit data to multi-bit doesn't equate to DSD/PCM conversion. I did check the final output of Sabre's chip on DSD. That signal exhibited text-book ultrasonic noise typical of DSD which meant that it hadn't been converted to PCM."

 

Trusting a reviewer to have the knowledge of an engineer is a mistake. It would be like asking your junior high-school biology teacher to do brain surgery on you.

 

This guy clearly has no idea what he is talking about.

 

"DSD" by Sony's own definition (they are the ones who made up the marketing term) means that it is a 1-bit signal. A one-bit signal can have no signal processing done to it whatsoever without changing it to multi-bit (PCM) -- except a time delay.

 

When a DSD signal is sent to a DSD DAW (eg, a Sonoma), it is converted to 8-bit wide PCM. To turn it back into DSD, it must be sent to a modulator to be remodulated. This is basically like sending an analog signal to an analog-to digital converter. When this PCM signal is sent to the 1-bit modulator, it must be noise shaped as the S/N ratio of a 1-bit modulator is next to nothing unless it is noise-shaped. So even if some (or all) of the out-of-band noise had been filtered out during its journey through the PCM stages, more out-of-band noise will be added back in.

 

So it is no surprise that there was high-frequency noise present at the output of the ESS DAC.

 

Stanley Lipshitz did a mathematical analysis of it in an AES Journal article around a decade ago and if a DSD signal is converted to PCM and back to DSD more than a certain number of times, you end up with more noise than signal -- in other words a negative S/N ratio! I don't recall off the top of my head how many trips were required, but it was quite low. Somewhere between four and seven, as I recall.

[57gold]

Mr. Hansen,

 

Thanks for your comments/insights and congrats on your company's success and reputation for high quality and musical products.

 

I have heard several of your products, and would have liked to purchase your QA-9 DAC but it is missing an important function for a computer audio only system, a volume control/attenuator.

 

Prior to embracing computer audio, I simplified my music system to rely on one source, CDs. Sold the expensive tube preamp as I used it solely to pad the output of my player as I relegated my vinyl, cassette and FM tuner gear to the obsolescence waste bin. No need for all the switching, inputs and outputs and the fancy circuitry of a preamp to attenuate a strong, ultra high quality signal designed to drive cables and an amp input. Oh, and the direct approach sounded more like live music.

 

With the convenience of having 1,600+ disc rips and downloads that i can actually find in seconds versus rummaging through the "library" of hundreds of plastic cases, the significant improvement in audio/musical quality of memory players like A+ and PM over spinning laser discs, and the advent of hi rez downloads, I have to believe that many music lovers have/will go computer audio, soon there will be no CDs, only downloads...yes, I am a master of the obvious.

 

But that takes me back to the question of why your computer audio DAC is missing a key and desirable feature to be more broadly used? A remote control attenuator.

 

Thanks again for your expert and generous participation on this forum.

[beetlemania]

. . .the Fourth National Scandinavian Nose Whistler's Orchestra in super-fidelity . . .

 

Actually, some of that Swedish sh*t is pretty good!

http://www.computeraudiophile.com/f14-music-analysis-objective-and-subjective/trondheimsolistene-marianne-thorsen-mozart-violin-concertos-14973/

 

You can get a free sample download here: High Resolution Music DOWNLOAD services .:. FLAC in free TEST BENCH

 

Make sure to try that Mozart Violin Concerto. I'm going to sample the DSD version after I get my QB-9 DSDed.

[wisnon]

Trusting a reviewer to have the knowledge of an engineer is a mistake. It would be like asking your junior high-school biology teacher to do brain surgery on you.

 

This guy clearly has no idea what he is talking about.

 

"DSD" by Sony's own definition (they are the ones who made up the marketing term) means that it is a 1-bit signal. A one-bit signal can have no signal processing done to it whatsoever without changing it to multi-bit (PCM) -- except a time delay.

 

When a DSD signal is sent to a DSD DAW (eg, a Sonoma), it is converted to 8-bit wide PCM. To turn it back into DSD, it must be sent to a modulator to be remodulated. This is basically like sending an analog signal to an analog-to digital converter. When this PCM signal is sent to the 1-bit modulator, it must be noise shaped as the S/N ratio of a 1-bit modulator is next to nothing unless it is noise-shaped. So even if some (or all) of the out-of-band noise had been filtered out during its journey through the PCM stages, more out-of-band noise will be added back in.

 

So it is no surprise that there was high-frequency noise present at the output of the ESS DAC.

 

Stanley Lipshitz did a mathematical analysis of it in an AES Journal article around a decade ago and if a DSD signal is converted to PCM and back to DSD more than a certain number of times, you end up with more noise than signal -- in other words a negative S/N ratio! I don't recall off the top of my head how many trips were required, but it was quite low. Somewhere between four and seven, as I recall.

 

Charles,

 

Miska and many other engineer types have said the same thing long ago. I also read similar on whatsbest forum:

---

...Are we slave to the PCM world?

In the time domain, DSD looks like little clouds of noise, with an average cloud-size about 714 nanoseconds. Signal modulation is carried out by varying the size of these clouds; on any given group of samples, it's going to be very difficult to discern how this clump of noise is any different than another clump of noise.

 

PCM, by contrast, is convertible to an exact number for every single sample. The magnitude of that sample, unlike DSD, does not need to look to its neighbors to figure out how big it is; it's already precisely specified in the PCM code itself.

 

DSD is somewhat akin to Frequency Modulation, or its close cousin, Phase Modulation. FM, of course, is used for FM radio, and the sound carrier for analog television. FM is also notorious for not tolerating signal manipulation; ripples in the frequency response, or more importantly, variations in group delay vs frequency, translate into outright distortion when demodulated back to analog. SECAM color television, which uses an FM carrier for the chroma information, is very difficult to work with in the studio; my understanding is that SECAM territories ended up using PAL in the studio, and converting to SECAM at the transmitter.

 

Sony has plenty of experience working with NTSC, PAL, and SECAM broadcast regions, since they made a big push into professional broadcasting gear as far back as the early Seventies, when they challenged Ampex for domination in the VTR world ... and beat Ampex at their own game. One side effect of their broadcast work is they made some of the earliest time-base correctors, which allows a helical VTR with an inherently jittery time-base to be accurately phase-locked to network timing standards.

 

A TBC is basically a FIFO buffer that uses precise external sync (from the network) to clock out the contents of the buffer and an over/under sensor to speed up or slow down the motors of the VTR. The digital converter for the TBC is also interesting: NTSC analog video is converted to 8-bit digital at 14.32Mhz (4x the 3.58MHz chroma frequency used in the Americas and Japan). Hey ... that sounds like DSD-wide! Yup. Pretty much like digital video, just not as fast.

 

Which gets us back to DSD-narrow. It's an OK way to capture analog in the digital domain, but actually doing anything with the digital signal (aside from storing it) is close to impossible. Convert to a form of PCM, which DSD-wide is, and you can do what you want. Although reducing from 8-bits back to 1-bit (discarding all of the smaller bits) is intellectually unattractive, nothing has been lost compared to the DSD-narrow 1-bit original.

 

You do have to watch for buildup of ultrasonic noise, though. Remember, for DSD-narrow the ultrasonic noise is the carrier. Every conversion in and out of the 1-bit domain carries a noise penalty; that's probably the reason that Scarlet Book specifies no more than 50% modulation for the analog signal going into DSD-narrow, in order to prevent clipping from added noise over several conversion steps.

[freddie40]

Im back were I was before I started tweaking the... never mind. Listening to other types of music I’m now at:

- 12

- 2 000 000

- 1.0

- 0.30

 

Setting 0 for pre-ringing and steepness is too fuzzy. I realize that now...

 

Thank you

 

Dave

[SoNic67]

There are over 3,000 pure DSD SACDs and about 1,500 analog transfers.

No, there are not. Pure DSD SACD's would mean that where recorded LIVE, in ONE TAKE without any post-processing. Who cares for that kind of music?

 

As for the "analog transfers" - you know that any modern-day LP cutter has a delay line that is based on digital PCM part of the “margin control” system? All that 'pristine' analog is actually PCM-derived.

 

@Charles Hansen: My respects! Good business decision to sell them 'DSD-compatible' upgrades, I understand now that you really know what are you doing.

[MikeJazz]

No, there are not. Pure DSD SACD's would mean that where recorded LIVE, in ONE TAKE without any post-processing. Who cares for that kind of music?

 

I care also for like that kind of music.

(and I simply don't mind if it comes from DSD, or PCM, as long as I can enjoy!)

 

One of the most memorable presentation I had was a recording made with a chorus on a church.

 

The recorded sound was made with not much more that a mic and a A/D converter (I guess it was a lyngdorf millenium ADC, with the resulting digital audio directly recorded on hard disk, no edits).

When later reproduced through the system, and after listening my comment was that it sounded simply like analog unamplified music (wich the source was of course).

The system simply got out of the way! The voice richness was natural. The spacial clues were there. The easiness in listening was there. The natural acoustics of the space were there and I was transported there!

 

Even considering that it was a very "simple" audio event...no small feat!

[wisnon]

No, there are not. Pure DSD SACD's would mean that where recorded LIVE, in ONE TAKE without any post-processing. Who cares for that kind of music?

 

As for the "analog transfers" - you know that any modern-day LP cutter has a delay line that is based on digital PCM part of the “margin control” system? All that 'pristine' analog is actually PCM-derived.

 

@Charles Hansen: My respects! Good business decision to sell them 'DSD-compatible' upgrades, I understand now that you really know what are you doing.

 

Sonic, no disrespect...I dont take these things too seriously...but have you listened to DSD files played thru a dedicated DSD Dac (not SACD, but DSD comp audio)? If so, what do YOU think about the SQ to you? If not, why not get a demo and make up your own mind independent of specs and processes? It may just be your cup of tea.

 

As to Charles, I have no problem with him having his opinion, not his peeve with Sony (likely very justified). I am sure there was a lot of marketing hyperbole in the past, but jjust like the humble CD that was sold as perfect years ago, progress in Dac IMPLEMENTATION has virtually made RBCD a HiRes format!

 

 

On my lampi Dac, its easy to confuse a welll recorded RBCD song with hires! I await the Lampi DSD Dac just being prototyped now, as the designer was an agnostic, but now has changed opinion 360degrees...

 

Repeat after me, the magic is in the IMPLEMENTATION!

[Charles Hansen]

not SACD, but DSD comp audio

 

Wisnon,

 

What is the difference to you whether the bits come from a silver disc made of polycarbonate or a silver disc made of aluminum with a magnetic coating?

 

Thanks,

Charles Hansen

Ayre Acoustics, Inc.

[Charles Hansen]

Wisnon,

 

Also please let me know when you want me to explain the lies in the Sony graphs showing the impulse response that you keep posting.

 

Thanks,

Charles Hansen

Ayre Acoustics, Inc.

[Onkle Je]

Hello Charles,

 

Thank you again for taking time to answer us. Do you have an idea of when the QB-9 DSD and the upgrade will be available worldwide (for instance, where I live, in France) ?

[MikeJazz]

Wisnon,

 

Also please let me know when you want me to explain the lies in the Sony graphs showing the impulse response that you keep posting.

 

Thanks,

Charles Hansen

Ayre Acoustics, Inc.

 

I am interested!

Thanks!

[SoNic67]

but have you listened to DSD files played thru a dedicated DSD Dac (not SACD, but DSD comp audio)? If so, what do YOU think about the SQ to you? If not, why not get a demo and make up your own mind independent of specs and processes? It may just be your cup of tea!

This is a match for the other thread.

[beetlemania]

I am interested!

Thanks!

 

See this post:

http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/ayre-wants-%241-5k-dsded-qb-9-a-15650/index4.html#post225870

[beetlemania]

To which a highly regarded mastering engineer who owns a QA-9 noted, "Having worked with the Ayre QA9 as the capture ADC in the mastering chain here, I agree completely.** There is something about the QA9 that appeals to the ear with a sound that feels less processed and more natural.* If there exist subconscious reactions to digitally recorded music* that signal “digititus” to the listener they are absent here.* This may well be down to the zero feedback no coupling cap analog stage for all I know, but whatever is in the mix, the QA9 stands apart, IMO."

 

Hey Charles,

 

I downloaded "Jul" earlier this week.

http://www.computeraudiophile.com/f14-music-analysis-objective-and-subjective/jul-john-martorella-recorded-john-marks-16120/

 

I'm curious if the "highly regarded mastering engineer" in your quote is John Marks. Perhaps I might look forward to some recordings from another engineer?

[wisnon]

Wisnon,

 

What is the difference to you whether the bits come from a silver disc made of polycarbonate or a silver disc made of aluminum with a magnetic coating?

 

Thanks,

Charles Hansen

Ayre Acoustics, Inc.

Implementation!

[wisnon]

Wisnon,

 

Also please let me know when you want me to explain the lies in the Sony graphs showing the impulse response that you keep posting.

 

Thanks,

Charles Hansen

Ayre Acoustics, Inc.

 

I think you did a fine job last time, but if you have more, let's have it. I am no Sony fan (clearly you are not either) and I have not kept posting anything. The graphs were all posted at virtually the same time...all before your first response. Please fire away with more for the erudition of all.

[Axiom05]

Hey Charles,

 

I downloaded "Jul" earlier this week.

http://www.computeraudiophile.com/f14-music-analysis-objective-and-subjective/jul-john-martorella-recorded-john-marks-16120/

 

I'm curious if the "highly regarded mastering engineer" in your quote is John Marks. Perhaps I might look forward to some recordings from another engineer?

 

The recording details can be found here: John Marks Records - Jul Downloads

 

Additional details can be found here: The Fifth Element #78 | Stereophile.com

 

I downloaded the 24/192 file yesterday but have not yet had a chance to listen to it.

[Charles Hansen]

I'm curious if the "highly regarded mastering engineer" in your quote is John Marks. Perhaps I might look forward to some recordings from another engineer?

 

No, this was another guy based in NYC. He does mastering work full-time. The two projects I've heard had already been recorded when they were brought to him, so they weren't recorded solely with the QA-9 (unlike John Marks' organ recording). His comments have been made after using the QA-9 for a wide variety of tasks for a wide variety of purposes.

 

Best regards,

Charlie Hansen

Ayre Acoustics, Inc.

[Charles Hansen]

Implementation!

 

Wisnon,

 

If you believe that a DAC playing back DSD-encoded bits from a computer hard drive can sound far superior to a DAC playing DSD encoded bits from a polycarbonate disc due to (differences in?) "implementation", would you mind telling us exactly which parts of the implementation are different and exactly how they would impact the sound (either for better or worse)?

 

Any speaking hypothetically, do you suppose that it might be possible that the differences you hear between PCM and DSD might also be due to "implementation" also? Or do you think that the differences are more like the commandments that Moses brought down from the mountain -- engraved in stone, and that DSD is inherently superior to PCM, and always will be?

 

And if the latter is the case, what exactly do you believe that it would be that accounts for DSD's "inherent superiority"?

 

Thanks,

Charles Hansen

Ayre Acoustic, Inc.

[firedog]

I have listen to those Police SACD and they are no better than my original Police CD's. I listen them trough my Denon universal player (SACD-R capable) and PC (Foobar + E-MU 1820m). Looks like the old PCM files where plain converted to DSD, probably the analog tapes are either damaged or lower quality now than original digital transfers from 10 years ago.

DSD makes sense only if the native recording is in that format or something higher (DXD format) like they do it here: High Resolution Music DOWNLOAD services .:. FLAC in free TEST BENCH

 

I don't know what or how you listened, but I have the Police albums in vinyl, CD, (also inclucing some MFSL versions), and DSD. They're all good. But the DSD versions are clearly the best. Especially Synchronicity - it just sounds amazing.

 

Again, in spite of what's being written here, there are lots of Analogue Tape direct to DSD conversions. (The ABKCO Rolling Stones catalogue, for instance). Many of them are amazingly good sounding - IMO often better than the original LP.

[Charles Hansen]

So Charles, how do you explain this?

 

[ATTACH=CONFIG]5567[/ATTACH]

 

Wisnon, in your original post you had two similar graphs. I am only going to refer to one graph in the interest of simplicity. There is a lot of confusion surrounding filters, and since that is what this graph is supposed to illustrate, I will explain:

 

1) The first thing that you should understand is that there is no inherent difference between an analog filter and a digital filter. Once you wrap your head around this, it will clear up a lot of confusion. Now it turns out that some things are easier to do with an analog filter or a digital filter, but basically whatever can be done with one can be done with the other. And in accordance with the "NO FREE LUNCH LAW", when you have two filters that give the same response (transient, frequency, or whatever) they will act in exactly the same way. It does not matter whether they are analog or digital.

 

2) The second thing to understand is that the steeper the filter and the sharper the corner of that filter, the more it will ring when a transient comes along. The only filter that has no ringing, undershoot, or overshoot is a first-order filter. A first order filter has only one element to change the frequency response and the it will change the response at 6 dB/octave = 20 dB/decade. In general, the steeper the filter the sharper the corner. This is not always true. When a filter rolls off the high frequencies, that element is referred to as "pole" and when it increases the high frequencies, that is referred to as a "zero".

 

If you make a second order filter, generally the two poles will happen at roughly the same frequency. If the happen at exactly the same frequency, the corner will be as sharp as possible. If one pole is removed by several octaves from the other pole, there will essentially be two separate corners, but this is not a general filter configuration. Instead the poles (or zeroes) are moved slightly closer or further apart and this gives a different shape to the corner. These different shapes generally have names, after the mathematicians who first described them -- Butterworth, Bessel, Chebychev, Linkwitz-Riley, and so forth. All that these different filters mean is that the distance between the poles is slightly different and so the corner will be either more or less sharp, and the sharper the corner the more it will ring when a transient hit it.

 

3) The third thing to understand is that every electrical system has a limited bandwidth. If were not true, I could shine a light on the input wire and a brighter light would shine from the output wire of an amplifier. But no electrical system made today has anywhere close to that kine of bandwidth. In fact audio is considered to be a very low bandwidth system. The minimum bandwidth of a telephone system is around 3 kHz (yet we can reliably identify familiar voices with ease, despite this limitation), the commonly accepted range of normal human hearing is 20 kHz. And there are many engineers who believe that the response of the system should be at least 10x what is required in order to achieve proper sound fidelity. (Clearly the CD does not meet this last criterion, but very few source components do, nor do the loudspeakers used that we listen to.)

 

A wide bandwidth means that the system can respond to a very short event. The higher the frequency, the more quickly the signal changes. At extremely high frequencies, the signal will change its amplitude in a very short period of time.

 

~~~~~~~~~~

 

So no we have enough information to make sense of the graph shown above.

 

At the far left is the imaginary input signal, to be reproduced by various systems. It is a positive-going pulse. They give the amplitude as "-6 dB" which is a bad beginning as a decibel is a ratio. Without knowing what the reference level is, we don't know what the actual level is. -6 dB is half the voltage of the reference level. It is not given in this case. With analog tape, it is usually to a specific magnetic flux density on the tape itself. In a digital system, it is usually given with reference to the maximum level that the system is capable of, which is why you often see "dBFS", which is short for "decibels relative to Full Scale".

 

However, they do say that this pulse lasts for 3 µsec. For a "back of the envelope" calculation, we can take the reciprocal of the period to give us the frequency. So if a signal had a period of 1 µsec, it would have a frequency of 1 MHz. With this pulse of 3 µsec, we need a system with a bandwidth of (at least) roughly 333 kHz to pass it.

 

And indeed, if we sent this signal through a well-designed analog system with a bandwidth of at least 333 kHz, when it came out the other side, it would look more or less the same as when it went in. But of course this forum is about computer audio and that, by necessity, means digital audio, and due to these guys named Nyquist and Shannon, we know that we have to be very careful about what frequencies we send in to the digital system or we will not get the expected output. Instead we will get all sorts of nasty noises that have almost no relationship to the input signal.

 

So much of digital audio is concerned with filters and filtering. Which is why this graph you posted is important, and why it is important to see the lies that are contained in it.

 

Let's start off by sending this signal into a system that is not on your graph. What happens to this signal if we send it through a well-designed analog system that only has a bandwidth of 100 kHz, rather than the 333 kHz required to reproduce it properly?

 

That's an interesting question, isn't it? After all, 100 kHz should be more than enough bandwidth to allow us to hear the signal perfectly, yes? I mean, we can only hear up to 20 kHz, so if the system goes up to 100 kHz, we should be in pretty good shape, yes?

 

Well, what happens when we send the signal into this bandwidth-limited (but otherwise perfect) analog system is that it cannot respond quickly enough to track the input signal. It cannot rise as quickly as the input and it cannot fall as quickly as the input. So the "sides" of the pulse are less steep, and the signal does not reach the full height of the original signal -- there isn't enough time for the signal to climb all the way up to the top before the input starts to fall.

 

It doesn't take a lot if imagination to see that the more limited the bandwidth that the wider and shorter the output pulse will be, yes?

 

~~~~~~~~~~

 

Now with a well-designed analog system, there is normally only one pole (bottleneck) that limits the bandwidth of system. This is done for a very good reason. 99.999% of all audio circuits made in the last 50 years use negative feedback to lower the distortion and increase the bandwidth. But if the gain is greater than 1 (unity) when the frequency response is falling off by more than -6 dB per octave (one pole), then the system will have enough phase shift that the negative feedback can become positive feedback at some frequencies and the system will oscillate. This means that there will be an output signal even when there is no input signal, which of course is not a very high-fidelity system....

 

~~~~~~~~~~

 

Back to the misleading graph. The nest pulse shows the response of "professional" quality analog. As you can see, it precisely duplicates the input. Therefore the bandwidth of the professional quality system must exceed 333 kHz. Well, while this may be true for some professional equipment, it most certainly is NOT in general. Almost every professional recording studio will use transformer-coupled microphone preamplifiers. It is very difficult to get response above 80 to 100 kHz with an audio transformer. It is virtually impossible to get response to 333 kHz with a microphone transformer. It is absolutely impossible to get response to 333 kHz with any analog tape recorder designed for audio.

 

So already with the first graph they show, they are lying. The only professional quality audio equipment that can give a pulse response as show will be a line-level preamplifier with no transformers in the signal path. Virtually NO other piece of equipment, from power amps to speakers to microphones to microphone preamps (including transformers) to even metering consoles will display the transient response that they claim. It is a flat-out lie, designed to deceive. After all BILLIONS of dollars were at stake, so why not play fast and loose with the facts? It fooled you, and you know more than 99.99% of all consumer electronics customers!!!

 

The second graph shows "consumer" quality playback. According to this graph the frequency response of "consumer" quality systems will be down perhaps -1 dB (~11%) at 333 kHz, BUT there will be ringing after the transient. Remember what causes ringing is that the frequency response falls off more rapidly than -6 dB per octave. Well, this is quit clearly bullshit. It is just some made-up stuff to make you feel that "professional" equipment is better than "consumer" equipment. But if any consumer equipment had a bandwidth that was down -1 dB at 333 kHz and rang that much, it would be so prone to oscillation that it would self-destruct in short order.

 

Second graph, and the second lie...

 

The next two graphs show the response of 44 kHz sampling and 96 kHz sampling. Guess what? They are finally telling the truth! Well more or less...

 

As we noted before, when the bandwidth of the system is limited, the input pulse will come out wider and shorter than when it came in. The input pulse in this case is 333 kHz while the bandwidth of the CD is only 20 kHz, so it is no surprise at all that the pulse is wider and shorter than the original. When the sampling rate is increased to 96 kHz, the bandwidth is increased to about 44 kHz, so the pulse is wider and shorter than the original, but not to the degree of the system limited to 20 kHz.

 

Again, this is completely normal and completely expected and is true whether the signal is analog or digital, PCM or PDM or whatever system you choose.

 

However you will also notice that there is ringing at both the beginning and the end of the pulse. But we already know that ANY system with a filter steeper than -6 dB per octave will ring. The steeper the filter, the more it will ring. CD is really in a bad situation here because people have been brainwashed not to buy something unless it is "flat to 20 kHz". But Nyquist and Shannon have told us that for the system to work properly we have to remove all of the energy (by filtering) below half of the sampling rate, or 22.05 kHz. So that means that in about 1/6 of an octave, we need to reduce the level of the signal by 96 dB (the dynamic range afforded by 16 bits). That works out to a 96 pole filter.

 

A 96 pole filter would be horribly expensive, whether done in the analog domain (as in the original CD players) or the digital domain, as in all current CD players. So they fudge it and let some aliasing happen and cheat a bit here and cheat a bit there. But they still end up with a very steep filter that rings like a bell at the cut-off frequency (20 kHz).

 

Well I'm going to let you in on a secret.

 

If you took the world's finest stereo system and played only pure analog master tapes, 1/2" half-tracks running at 30 ips and then you stuck a brickwall filter into the system somewhere, it would sound like dogshit.

 

The filter would kill the music. It would ring like a bell, it would introduce time smear, and it would cause so many problems that you would be better off listening to an AM car radio.

 

~~~~~~~~~~

 

Notice in the third graph that all of the ringing is at the end of the transient, while in the next three (digital) graphs, the ringing is equally distributed at both the beginning and the end of the transient.

 

The first is typical of an analog filter and is called a "minimum phase" filter, as it has the minimum phase shift possible for a given frequency response. The second type of filter is called a "linear phase" filter and is typical of FIR (Finite Impulse Response) digital filters. The phase shift is constant with frequency, which means that there is a time delay, but it is the same for all frequencies. When digital audio was being developed in the late '70s and early '80s, there was a big trend to try to build "linear phase" loudspeakers. This belief spilled over into the digital realm. Plus it turns out that (in general) it is much cheaper to build a linear phase filter than a minimum phase filter.

 

The only problem is that they don't sound very good.

 

In the real world, no effect occurs before the cause. It's impossible (at least according to the laws of physics as we know them today). In contrast, all kinds of things happen after the cause. Every single sound you ever hear will have echoes after the sound. Even outside there will be sonic reflection off the ground or the trees or some other physical object. But you will never hear an echo before the original sound -- except with digital audio.

 

In general IIR (Infinite Impulse Response) digital filters and analog filters will both be minimum phase filters, while FIR's will be linear phase filters with "pre-echo". But it is possible to make any kind of filter work either way. So in the Ayre equipment, all of our filters are implemented as minimum phase FIR's. There are a lot of reasons for doing it this way, but let's just say for now that we believe it sounds the best of any of the available choices.

 

~~~~~~~~~~

 

Back to the lying graph. The next graph shows something a bit different than the other graph you posted. The graph labeled "DSD" here looks very much like the graph labeled "192 kHz" in the other green graph that is supposed to look like an oscilloscope trrace.

 

The main difference is that this graph shows some ringing that the other graph doesn't.

 

Well, they are both full of lies. For any system to play that 3 µsec pulse back without widening it, the system would need a bandwidth of 333 kHz. Well I think that everybody here already knows that no commercially available digital system has a response up to 333 kHz. So there is your next lie.

 

The last graph in this image shows a pulse-density modulation system sampling at 20 MHz. That would be roughly DSD-512. First of all, it doesn't exist. Second of all, every time that they show these graphs, they only show the RECORD side.

 

The Scarlet Book standard for SACD requires a third order lowpass filter at 50 kHz. This is because the noise of DSD climbs very sharply above 20 kHz, reaching about -20 or -30 dBFS at 100 kHz. By adding the third-order (typically analog) filter in the player, there is another -18 of filtering at 100 kHz so that the noise in the system is "only" about -38 or -48 dBFS at 100 kHz. Now it is true that this won't blow up any tweeters. But I have heard some otherwise lovely tube-type preamps go berserk when presented with this out-of-band noise signal. I believe that the problem was an interaction with some coupling transformers that had peaks in the 100 to 150 kHz range. When excited by the out-of-band-noise from DSD, you could hear all kinds of beautiful chirping bird noises coming from your speakers. The only problem is that the birds were not present at the original recording.

 

So the BIGGEST lie of all on this graph is that the bandwidth of the full record + PLAYBACK of a DSD system is no wider than that of a 96/24 system, let alone a 192/24 system. Furthermore, the presence of the third order filter in the player will cause visible ringing when a pulse is input into the system. As the filters are generally analog, this ringing will all be post-ringing.

 

The RECORD side of a DSD system generally has no filtering. So there is nothing to kill the music. The playback side has a bandwidth no better than a 96/24 system, although there will be less ringing than a TYPICAL PCM system. But if Sony hadn't LIED about the true transient response of the SACD system, it would look virtually identical to the 96/24 system, but with no pre-ringing -- only post-ringing.

 

~~~~~~~~~~

 

In a separate post you talked about IMPLEMENTATION.

 

Well, it is certainly possible to implement PCM in a far better way than has been done before. In our QA-9 A/D converter we use a delta-sigma modulator running at 256 Fs that puts out 6 bits. We run this through a low-pass filter to turn it into bog-standard PCM that can be used by any bog-standard DAW. BUT we don't use a brickwall filter. Instead we use a filter that give a frequency response very much like an analog tape machine running at 30 ips.

 

This filter has NO pre-ringing, NO post-ringing, NO ringing at all. So it has even BETTER transient response than DSD.

 

And since we are running at 4x the rate of DSD and with 6 bits (64 times as many levels as DSD), we have an information density that is 256x greater than DSD. This means that we can end up with a signal that not only has better transient response than DSD, but also has essentially ZERO out-of-band noise. And since it is bog-standard PCM, it can be used with any tools (hardware or software) that can handle quad-rate signal for any type of signal processing that is desired.

 

Following is a graph that shows the response of our filter to a single pulse at 256 Fs. This pulse is less than 0.1 µsec wide. It goes through three filters. The first time it comes out with a rectangular shape, the second time with a triangular shape, and the third time with a Gaussian shape. Please note - AT NO TIME IS THERE ANY OVERSHOOT OR RINGING. NONE.

 

Compare the truth against Sony's LIES and you will see why it is not that hard to make a PCM system that outperforms (both by measurement and by listening) the flawed format that is "DSD".

 

Best regards,

Charles Hansen

Ayre Acoustics Inc.

[Charles Hansen]

I don't know what or how you listened, but I have the Police albums in vinyl, CD, (also inclucing some MFSL versions), and DSD. They're all good. But the DSD versions are clearly the best. Especially Synchronicity - it just sounds amazing.

 

Again, in spite of what's being written here, there are lots of Analogue Tape direct to DSD conversions. (The ABKCO Rolling Stones catalogue, for instance). Many of them are amazingly good sounding - IMO often better than the original LP.

 

Hello Firedog,

 

I had one of the Rolling Stones SACD's, "Beggar's Bangquet", I also had Rosy Music's "Avalon", CCR's "Green River". I didn't feel like any of those sounded better than the original vinyl. On the other hand a friend has the SACD of Allison Krauss and Union Station (I'm spacing on the name right now) that sound pretty danged amazing.

 

But maybe it's just that my system is getting better. I recently bought a bunch of the Warner's 192/24 downloads from HD Tracks and was thoroughly mind boggled. At least half of them sounded better than I remember the original vinyl sounding, and not by a little bit. I would highly recommend both "Crosby, Stills and Nash" and "CS&N" as beating out the vinyl -- at least my memory of it. I haven't heard the vinyl in years, but I was hearing all kinds of things I've never heard on those recordings before -- even some of the vinyl 45 rpm re-issues by Classic Records.

 

Best,

Charles Hansen

Ayre Acoustics, Inc.

[Hiro]

From the 6Moons Auralic Vega review:

 

"..However converting such 1-bit data to multi-bit doesn't equate to DSD/PCM conversion.

 

Agreed, whether the converter is using 1bit SDM or 4bit SDM, it's still SDM.

[wisnon]

Wisnon,

 

If you believe that a DAC playing back DSD-encoded bits from a computer hard drive can sound far superior to a DAC playing DSD encoded bits from a polycarbonate disc due to (differences in?) "implementation", would you mind telling us exactly which parts of the implementation are different and exactly how they would impact the sound (either for better or worse)?

 

Any speaking hypothetically, do you suppose that it might be possible that the differences you hear between PCM and DSD might also be due to "implementation" also? Or do you think that the differences are more like the commandments that Moses brought down from the mountain -- engraved in stone, and that DSD is inherently superior to PCM, and always will be?

 

And if the latter is the case, what exactly do you believe that it would be that accounts for DSD's "inherent superiority"?

 

Thanks,

Charles Hansen

Ayre Acoustic, Inc.

Charles,

 

I did not say anything about inherent superiority! I said another great flavour of ice cream and who doesnt like ice cream? LoL

 

Implementation is something you should agree with as I suppose you consider your Dac better implemented than average, no?

 

A Universal player or even a midrange dedicated SACD player with typical opamps will not fare well against my Lampizator Tube Dac, that I can tell you. (I expect your QB should be able to dispatch those solutions likewise). The Lampi man is also approaching DSD playback with active tube filters and though I have not heard that yet, I like that approach intuitively.

 

In summary, IMPLEMENTATION is the biggest determinant, but we also have to agree that there are differences in PCM and DSD. PCM is digital, DSD is not, it is analog.

 

PS thanks for the long explanatory post. I will print read and try to digest. Very good of you to take the time to explain your PoV in detail. Much appreciated.