[SlimDevices: Audiophiles] Question about Stereophile's "16 bit undithered sine at -90dB"

[darrenyeats]

I have a question about this.
http://www.stereophile.com/content/quality-lies-details-page-6#DQEqFEA63RHRhQcK.97

I thought such a waveform would be impossible with 16 bit data when fed
into an upsampling, 24 bit DAC. I thought the samples would be
interpolated at higher rate and greater bit depth, leading to a smooth
sine even at -90dB. But looking at any number of reviews on Stereophile,
the stair steps (with Gibbs ringing) are seen.

I infer that the "16 bit undithered sine at -90dB" is a deliberately
staircased waveform delivered in a hi-rez container. My question: is
this inference correct? If not, DAC upsampling doesn't quite do what I
thought it did!



Check it, add to it! http://www.dr.loudness-war.info/

SB Touch
------------------------------------------------------------------------
darrenyeats's Profile: http://forums.slimdevices.com/member.php?userid=10799
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Julf]

Post by darrenyeats
I have a question about this.
http://www.stereophile.com/content/quality-lies-details-page-6#DQEqFEA63RHRhQcK.97
I thought such a waveform would be impossible with 16 bit data when fed
into an upsampling, 24 bit DAC. I thought the samples would be
interpolated at higher rate and greater bit depth, leading to a smooth
sine even at -90dB. But looking at any number of reviews on Stereophile,
the stair steps (with Gibbs ringing) are seen.
I infer that the "16 bit undithered sine at -90dB" is a deliberately
staircased waveform delivered in a hi-rez container. My question: is
this inference correct? If not, DAC upsampling doesn't quite do what I
thought it did!

Note the explanation. "Fig.8 shows the ideal waveshape (assuming an
infinite bandwidth);", so the so-called "signal" is actually just a
simulated graphic (as a real system wouldn't have infinite bandwidth).
It continues with "a nearly perfect reproduction of this signal is shown
in fig.9, which overlays the fig.8 waveform with small amounts of Gibbs
Phenomenon ringing at the bit transitions owing to the CD system's
limited, 22kHz bandwidth.", leading me to think that that waveform, too,
only exists in a plotting program, not as a real signal.



"To try to judge the real from the false will always be hard. In this
fast-growing art of 'high fidelity' the quackery will bear a solid gilt
edge that will fool many people" - Paul W Klipsch, 1953
------------------------------------------------------------------------
Julf's Profile: http://forums.slimdevices.com/member.php?userid=42050
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[darrenyeats]

It's actually figure 9 I'm questioning. Fig.9 is actually the
measurement made on every digital source reviewed by Stereophile. The
result varies in each test, and is a measurement of the analogue
output.

Note this was discussed a while ago here:
http://forums.slimdevices.com/showthread.php?94054-Has-anyone-listened-to-the-NAD-C-390DD/page3
and John Atkinson even replied, though I didn't understand how his reply
addressed my concern.



Check it, add to it! http://www.dr.loudness-war.info/

SB Touch
------------------------------------------------------------------------
darrenyeats's Profile: http://forums.slimdevices.com/member.php?userid=10799
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[arnyk]

It's actually figure 9 I'm questioning. Fig.9 is an example of the
measurement made on every digital source reviewed by Stereophile. The
result varies in each test, and is a measurement at the analogue output.
Have a look at almost any review there e.g.
http://www.stereophile.com/content/chord-electronics-hugo-tt-da-headphone-amplifier-measurements#Vur2dhfuZZR2KXxg.97
figure 5.

I think that when dealing with sources that present themselves as having
superior knowledge, one important thing is to keep one's eyes on the big
picture and not get caught up in what may turn out to be minor or even
irrelevant details that will only waste time or distract if taken too
seriously.

The big picture of digital audio applies to this situation as follows:
An undithered digital signal is well known to cause problems, and the
standard recipe for recording digital audio signals includes applying
adequate dither.

If someone chooses to evaluate audio gear with undithered digital
signals, well its a free world and they can do what they want to do, but
they are basically breaking applicable rules for making digital
recordings.

Out another way, a recording of music for general listening that lacks
dither is mystery meat. There is no guarantee that it will fully exploit
the media and equipment that you use with it. Furthermore, it can be
shown that virtually every commercial recording of music is dithered,
even if not intentionally dithered, as a natural consequence of being
recorded in the real world as we know it.

In short, this business of testing digital audio gear with undithered
recordings is a technical rabbit hole.

It is hard for me to imagine why someone who is well-informed about
digital audio would in good faith, do such a thing.


------------------------------------------------------------------------
arnyk's Profile: http://forums.slimdevices.com/member.php?userid=64365
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[darrenyeats]

Thanks arnyk. I agree with what you say, I think.

Nevertheless, I am curious as to how this 3-level waveform ever makes it
out of a 24 bit, upsampling DAC - unless, as I infer, the "16 data" is
actually nothing of the sort, but instead a 16-bit-like staircase
rendered in a higher-rate, higher bit-depth file.



Check it, add to it! http://www.dr.loudness-war.info/

SB Touch
------------------------------------------------------------------------
darrenyeats's Profile: http://forums.slimdevices.com/member.php?userid=10799
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Julf]

I mentally glossed over it being a 1kHz cycle. So there are 44 samples
on each level of figure 9. The correct over-sampling behaviour is indeed
to sustain three plateaus, with Gibbs ringing in between. (If each level
were one sample, as I had in my mind, then a smooth sine would be
interpolated.)

Exactly.



"To try to judge the real from the false will always be hard. In this
fast-growing art of 'high fidelity' the quackery will bear a solid gilt
edge that will fool many people" - Paul W Klipsch, 1953
------------------------------------------------------------------------
Julf's Profile: http://forums.slimdevices.com/member.php?userid=42050
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Julf]

It's actually figure 9 I'm questioning. Fig.9 is an example of the
measurement made on every digital source reviewed by Stereophile. The
result varies in each test, and is a measurement at the analogue output.
Have a look at almost any review there e.g.
http://www.stereophile.com/content/chord-electronics-hugo-tt-da-headphone-amplifier-measurements#Vur2dhfuZZR2KXxg.97
figure 5.

So yes, what they are doing is having an undithered 16-bit signal that
has been padded to 24 bits (so that the DAC can't interpolate - it will
try to reproduce the artificial signal unchanged).

http://forums.slimdevices.com/showthread.php?94054-Has-anyone-listened-to-the-NAD-C-390DD/page3.
Best to start at post #22. John Atkinson even replied in post #34,
though I didn't understand how his reply addressed my concern.

He is basically verifying that it is a completely artificial test signal
designed to illustrate the symmetry of the DAC (and it is only relevant
on old-fashioned non-oversampling DACs anyway).



"To try to judge the real from the false will always be hard. In this
fast-growing art of 'high fidelity' the quackery will bear a solid gilt
edge that will fool many people" - Paul W Klipsch, 1953
------------------------------------------------------------------------
Julf's Profile: http://forums.slimdevices.com/member.php?userid=42050
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[arnyk]

Post by Julf
He is basically verifying that it is a completely artificial test signal
designed to illustrate the symmetry of the DAC (and it is only relevant
on old-fashioned non-oversampling DACs anyway)

With the vast number of DACs now sigma-delta types, this test of
obsolete, But I still include it because it also shows you at a glance
how high above the DAC's noisefloor the LSB transition lies

Exactly. Also note that he correctly said that delta-sigma DACs
eliminate the problem of bit-for-bit linearity. Any distortion in a
sigma-delta DAC is in its analog components which number very few.

Oversampling does not address the problem of DAC amplitude linearity.
The first generation Philips players were oversampled, but used R-2R
ladder DACs that were therefore highly dependent of parts quality for
amplitude linearity.


------------------------------------------------------------------------
arnyk's Profile: http://forums.slimdevices.com/member.php?userid=64365
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Julf]

Post by arnyk
Oversampling does not address the problem of DAC amplitude linearity.
The first generation Philips players were oversampled, but used R-2R
ladder DACs that were therefore highly dependent of parts quality for
amplitude linearity.

True. Fortunately most of us have moved on from ladder DACs (I do have
fond memories of them - my first one was a 8-bit one made out of
discrete resistors on the output of a parallel port, but that was 40
years ago....).



"To try to judge the real from the false will always be hard. In this
fast-growing art of 'high fidelity' the quackery will bear a solid gilt
edge that will fool many people" - Paul W Klipsch, 1953
------------------------------------------------------------------------
Julf's Profile: http://forums.slimdevices.com/member.php?userid=42050
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Mnyb]

Unfortunately the measurment migth stilll be needed ? Are there not new
r2r ladder dacs popular in some camps of "fantasy audio" or "novelty
electronics" :)



--------------------------------------------------------------------
Main hifi: Touch + CIA PS +MeridianG68J MeridianHD621 MeridianG98DH 2 x
MeridianDSP5200 MeridianDSP5200HC 2 xMeridianDSP3100 +Rel Stadium 3
sub.
Bedroom/Office: Boom
Kitchen: Touch + powered Fostex PM0.4
Misc use: Radio (with battery)
iPad1 with iPengHD & SqueezePad
(spares Touch, SB3, reciever ,controller )
server HP proliant micro server N36L with ClearOS Linux

http://people.xiph.org/~xiphmont/demo/neil-young.html
------------------------------------------------------------------------
Mnyb's Profile: http://forums.slimdevices.com/member.php?userid=4143
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[arnyk]

Post by Mnyb
Unfortunately the measurment migth stilll be needed ? Are there not new
r2r ladder dacs popular in some camps of "fantasy audio" or "novelty
electronics" :)

These are typically made out of NOS chips that should have been tested
20-30 years ago when they were the best we had.

Any golden ear reviewers who tests them will have to come up with some
clever euphemisms to candy-coat their designed-in poor frequency
response. They generally don't do measurements, anyway - right?


------------------------------------------------------------------------
arnyk's Profile: http://forums.slimdevices.com/member.php?userid=64365
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Julf]

Post by Mnyb
Unfortunately the measurment migth stilll be needed ? Are there not new
r2r ladder dacs popular in some camps of "fantasy audio" or "novelty
electronics" :)

Would the fans of such devices care about any measurements anyway?



"To try to judge the real from the false will always be hard. In this
fast-growing art of 'high fidelity' the quackery will bear a solid gilt
edge that will fool many people" - Paul W Klipsch, 1953
------------------------------------------------------------------------
Julf's Profile: http://forums.slimdevices.com/member.php?userid=42050
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Mnyb]

Post by Julf
Would the fans of such devices care about any measurements anyway?

Probably not :) I can't imagine taking the ladder DAC concept to 24bit
it , but that actually exists ?

The delta sigma ,one bit DAC and similar designs is just such an elegant
way of solving the linearity problem .

I have seen new ladder DAC's not only the old Phillips chips , a company
called MSB does them with discrete hand built modules, the price is
what you'd expect and that pretty much explains why it's not widely used
.



--------------------------------------------------------------------
Main hifi: Touch + CIA PS +MeridianG68J MeridianHD621 MeridianG98DH 2 x
MeridianDSP5200 MeridianDSP5200HC 2 xMeridianDSP3100 +Rel Stadium 3
sub.
Bedroom/Office: Boom
Kitchen: Touch + powered Fostex PM0.4
Misc use: Radio (with battery)
iPad1 with iPengHD & SqueezePad
(spares Touch, SB3, reciever ,controller )
server HP proliant micro server N36L with ClearOS Linux

http://people.xiph.org/~xiphmont/demo/neil-young.html
------------------------------------------------------------------------
Mnyb's Profile: http://forums.slimdevices.com/member.php?userid=4143
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Wombat]

There is a 20bit ladder Analog Devices AD5791



Transporter (modded) -> RG142 -> Avantgarde Acoustic based 500VA
monoblocks -> Sommer SPK240 -> self-made speakers
------------------------------------------------------------------------
Wombat's Profile: http://forums.slimdevices.com/member.php?userid=4113
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[arnyk]

Post by Wombat
There is a 20bit ladder Analog Devices AD5791

For a price - about $60 - 80, or more than 10 times a Sigma Delta DAC
with similar amplitude resolution performance. It also does faster
conversions.


------------------------------------------------------------------------
arnyk's Profile: http://forums.slimdevices.com/member.php?userid=64365
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Spin]

The Soekris DIY R-2R DAC
(http://soekris.com/products/audio-products/dam1021.html) will
apparently resolve 24 bits and is around $200 (you'll need a case and
transformer to complete it though!)



Bridged Silent Server -> Transporter -> Lyngdorf TDAI 2200 Amp with Room
Perfect -> Art Emotion Speakers
------------------------------------------------------------------------
Spin's Profile: http://forums.slimdevices.com/member.php?userid=15298
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[arnyk]

Post by Spin
The Soekris DIY R-2R DAC
(http://soekris.com/products/audio-products/dam1021.html) will
apparently resolve 24 bits and is around $200 (you'll need a case and
transformer to complete it though!)

Are there any independent technical tests that confirm these exceptional
claims?

For example, transformers are susceptible to creating quite a bit of
noise, distortion and frequency response errors by the standards of even
mid-fi audio gear. Any equipment that needs them to operate is therefore
automatically suspect.

For example R-2R DACs generally have linearity errors that are large
compared to 24 bits.

If built using discrete parts there are problems with temperature
tracking which is a leading reason why most modern examples such
components are not built out of discrete parts.

That's 3 strikes...


------------------------------------------------------------------------
arnyk's Profile: http://forums.slimdevices.com/member.php?userid=64365
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Spin]

Hi Arny,

When I said that a transformer is required I meant that it needs a mains
transformer to power the on-board power supply, not that transformers
are used in the audio path!

As for the technical spec's, i have no idea whether the claims are met
or not. The spec's are published (http://soekris.dk/dam1021.html) and I
guess you either believe the company or not (although I guess that a
reputable company is taking a risk if they are lying!).

I'm not a fan of R-2R DAC's but I think that what this guy is doing is
great interesting.



Bridged Silent Server -> Transporter -> Lyngdorf TDAI 2200 Amp with Room
Perfect -> Art Emotion Speakers
------------------------------------------------------------------------
Spin's Profile: http://forums.slimdevices.com/member.php?userid=15298
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Julf]

Post by Spin
As for the technical spec's, i have no idea whether the claims are met
or not. The spec's are published (http://soekris.dk/dam1021.html) and I
guess you either believe the company or not (although I guess that a
reputable company is taking a risk if they are lying!).

I don't see any claims about being able to resolve 24 bits. 127 dB of
SNR corresponds to roughly 21 bits. 0.006% THD corresponds to
approximately 14 bits.



"To try to judge the real from the false will always be hard. In this
fast-growing art of 'high fidelity' the quackery will bear a solid gilt
edge that will fool many people" - Paul W Klipsch, 1953
------------------------------------------------------------------------
Julf's Profile: http://forums.slimdevices.com/member.php?userid=42050
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Spin]

Post by Julf
I don't see any claims about being able to resolve 24 bits. 127 dB of
SNR corresponds to roughly 21 bits. 0.006% THD corresponds to
approximately 14 bits.

Sorry for my poor use of words. Clearly no DAC can resolve 24 bits - the
laws of physics tends to intervene ! I meant to say that it had 24 bit
resolution (although I'd have been wrong about that too, it's a 28 bit
DAC not 24 !)



Bridged Silent Server -> Transporter -> Lyngdorf TDAI 2200 Amp with Room
Perfect -> Art Emotion Speakers
------------------------------------------------------------------------
Spin's Profile: http://forums.slimdevices.com/member.php?userid=15298
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Julf]

Post by Spin
Sorry for my poor use of words. Clearly no DAC can resolve 24 bits - the
laws of physics tends to intervene!

indeed.

Post by Spin
I meant to say that it had 24 bit resolution (although I'd have been
wrong about that too, it's a 28 bit DAC not 24 !)

Well, yes, but what is the point of the extra bits if neither linearity
nor SNR are anywhere near that resolution?



"To try to judge the real from the false will always be hard. In this
fast-growing art of 'high fidelity' the quackery will bear a solid gilt
edge that will fool many people" - Paul W Klipsch, 1953
------------------------------------------------------------------------
Julf's Profile: http://forums.slimdevices.com/member.php?userid=42050
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Spin]

Indeed.
Well, yes, but what is the point of the extra bits if neither linearity
nor SNR are anywhere near that resolution?

If I remember rightly I believe that the DAC uses a digital volume
control that works by throwing bits away and the extra 4 bits of
resolution allows you to apply quite a lot of attenuation before audio
quality is affected.



Bridged Silent Server -> Transporter -> Lyngdorf TDAI 2200 Amp with Room
Perfect -> Art Emotion Speakers
------------------------------------------------------------------------
Spin's Profile: http://forums.slimdevices.com/member.php?userid=15298
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Julf]

Post by Spin
If I remember rightly I believe that the DAC uses a digital volume
control that works by throwing bits away and the extra 4 bits of
resolution allows you to apply quite a lot of attenuation before audio
quality is affected.

That makes sense for internal processing (and many DACs actually use 32
bits or more for that), but not for the actual ladder output.



"To try to judge the real from the false will always be hard. In this
fast-growing art of 'high fidelity' the quackery will bear a solid gilt
edge that will fool many people" - Paul W Klipsch, 1953
------------------------------------------------------------------------
Julf's Profile: http://forums.slimdevices.com/member.php?userid=42050
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Spin]

Post by Julf
That makes sense for internal processing (and many DACs actually use 32
bits or more for that), but not for the actual ladder output.

Just searched through the thread at DIY Audio and the designers actual
words are 'I decided on 28 bit, to have headroom for a perfect digital
volume control. At -72 db volume you still have 16 bit resolution with
perfect linearity thanks to the sign magnitude architecture'. Whether
that makes sense I guess that you would know better than I.



Bridged Silent Server -> Transporter -> Lyngdorf TDAI 2200 Amp with Room
Perfect -> Art Emotion Speakers
------------------------------------------------------------------------
Spin's Profile: http://forums.slimdevices.com/member.php?userid=15298
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[Julf]

Post by Spin
Just searched through the thread at DIY Audio and the designers actual
words are 'I decided on 28 bit, to have headroom for a perfect digital
volume control. At -72 db volume you still have 16 bit resolution with
perfect linearity thanks to the sign magnitude architecture'. Whether
that makes sense I guess that you would know better than I.

Unfortunately it does not really make sense. Yes, 72 dB is 12 bits, so
"giving up" 12 bits to the volume control still leaves you with 16 bits
- but what is the point, if you are only 4 bits above the noise floor?



"To try to judge the real from the false will always be hard. In this
fast-growing art of 'high fidelity' the quackery will bear a solid gilt
edge that will fool many people" - Paul W Klipsch, 1953
------------------------------------------------------------------------
Julf's Profile: http://forums.slimdevices.com/member.php?userid=42050
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[arnyk]

Post by Spin
Hi Arny,
When I said that a transformer is required I meant that it needs a mains
transformer to power the on-board power supply, not that transformers
are used in the audio path!

You need to be clear when you write. Its not unusal for DACs like this
to require an output transformer. Yhe product is incomplete as sold,
lacking part of its power supply, input connectors, output connectors, a
proper case, etc. The product write up seems to direct audiophiles to
attach any kind of headphone to its outputs, but its given output
impedance is actually about 100 times or more to high to handle that
kind of usage.

Post by Spin
As for the technical spec's, i have no idea whether the claims are met
or not. The spec's are published (http://soekris.dk/dam1021.html) and I
guess you either believe the company or not (although I guess that a
reputable company is taking a risk if they are lying!).

The spec sheet seems to be incomplete. For example no +/- dB tolerance
seems to be given for its frequency response spec. No corresponding FR
plot is shown. What is shown is mediocre performance given the price of
the incomplete product that they offer.

Post by Spin
I'm not a fan of R-2R DAC's but I think that what this guy is doing is
great interesting.

That may be true if one finds watching someone try to sell the audio
equivalent of mud pies for $100s of dollars.


------------------------------------------------------------------------
arnyk's Profile: http://forums.slimdevices.com/member.php?userid=64365
View this thread: http://forums.slimdevices.com/showthread.php?t=105769

[arnyk]

Post by Spin
The Soekris DIY R-2R DAC
(http://soekris.com/products/audio-products/dam1021.html) will
apparently resolve 24 bits and is around $200 (you'll need a case and
transformer to complete it though!)

This graphic related to the Soekris DIY R-2R DAC
(http://forums.slimdevices.com/attachment.php?attachmentid=20661&d=1466593657)
shows a DAC that barely has 16 bit resolution, if that:

20661


+-------------------------------------------------------------------+
|Filename: fft_-1db_1.jpg |
|Download: http://forums.slimdevices.com/attachment.php?attachmentid=20661|
+-------------------------------------------------------------------+

------------------------------------------------------------------------
arnyk's Profile: http://forums.slimdevices.com/member.php?userid=64365
View this thread: http://forums.slimdevices.com/showthread.php?t=105769