 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
71.202.60.151
I noticed a post on MSB's site regarding the Sabre DAC not being 32-bit, but actually 6-bit.
Here's a short snippet from their post:
"So the hot new 32 bit DAC is actually a 6 bit DAC! Right from their own white paper. It is undoubtably a good DAC for the mass consumer market it was designed for. It is certainly nothing of interest to the high-end community, especially as the the DAC, digital filter and sample rate converter that cannot be dissabled are all bundled in a single chip so no opportunity exists to improve its performance."
I understand that more bits don't necessarily translate into better sound, but just from a technical perspective is what MSB stating on their site accurate?
Follow Ups:
Hi,Well, I think we need to define what we mean with "32 Bit DAC" if we want to see if the claims are true or not.
We can interpret it in many ways:
1) A 32-Bit DAC is one that shows 32-Bits equivalent analogue resolution, that is 183dB dynamic range, measured the traditional analogue way. It goes without saying, not only can we not measure such a DAC should it actually exist, but of course no such thing exists nor is it possible, outside a laboratory system suspended in liquid nitrogen and maybe not even then.
2) A 32-Bit DAC is one that uses 32 individual binary weighted bit switches, so it is theoretically capable of producing 2^32 discrete steps, though it's analogue dynamic range is less than the postulated 183dB (or 192dB as some may say). No such thing exists either (yet), but it is at least theoretically possible to make such a device.
3) A 32-Bit DAC is a DAC that accepts a 32-Bit wide data word and outputs whatever real resolution it is capable of, in other words it is marketing number without any appreciable meaning.
It should be added that recent specifications for computer based audio (Intel/M$ HD Audio) call for systems that are able to handle 32 Bit words, simply because this is how computers like to work, they want 8/16/32 Bit words to work with, not 24 Bit. This is the reason for the 32-Bit DAC's now becoming more common.
This has no meaning other than a DAC should accept a 32 Bit word, for compatibility, not that it that actually does anything meaningful with the whole 32 Bit's. In fact, several "32-Bit" DAC's simply take the 32-Bit Data and dither it down to 24-Bit which is then applied to manufacturers 24-bit DAC Core.
However, as it is very easy for marketing departments to point out that "32-Bits are better than 24-Bits (even though under my definition in 1) there is no such thing as a 24 Bit DAC either) just as 24-Bits are better than 16-Bits" and so on, the 32-Bit part has become a major marketing macguffin for DAC (Chip) manufacturers and their customers.
Okay, on to the Sabre (and for that in principle almost all modern DAC's).
In the year of our lord 2010 all but one DAC targeted at audio use a concept that used to be called "hybrid DAC". This means these DAC's combine several bits worth of multibit core with a delta sigma modulator (aka one bit DAC).
The combination is used to achieve the total resolution by using a process called noise shaping from the "real" resolution of the DAC (that is the number of levels that the DAC can directly represent in the analog domain and the additional resolution attained using noise shaping. It is a little difficult to understand and even for those who understand to explain where this extra resolution comes from, but we do not worry ourselves here about details.
The bottom line is that the DAC will have a number analog levels that can be represented directly by 1-Bit and Multibit conversion. The rest has to be produced using noise shaping. The views on noiseshaping vary, my own is negative, where sonics are concerned, compared to having enough real resolution.
Back to the Sabre.
For the "6-Bit DAC" for the Sabre, this is both correct and incorrect.
The Sabre has 2^6 or 64 so called "unitary weighted" or "thermometer code" bit switches. These are able to represent directly 2^6 or 64 individual levels.
Further, the Sabre uses asynchronous sample rate conversion on ALL Input data and converts into a clock rate of 40MHz. If we assume for ease of calculation a 50KHz data sample rate (close enough to to the 44.1KHz used on CD) we can represent as many as 40MHz/50KHz or 800 individual levels using classic pulse width or pulse density modulation.
For ease of calulation I will round up to 1024 levels, which is equivalent to 10 Bit resolution.
This means that the raw resolution build into the ESS DAC is around 16 Bit for single speed (44.1/48KHz) Data, 15 Bit for double speed (88.2/96KHz) and 14 Bit for quad speed (176.4/192KHz) data.
I have to say that this is appreciably more real, raw resolution than most DAC's in the market offer. The ESS Sabre DAC's can actually represent CD Data in the analogue domain with no or very little noiseshaping, which may explain the fact that many find it superior to many other DAC's.
For reference, a highly regarded (by some anyway) 32-Bit DAC by another manufacturer uses a 32 Level (5 Bit) multibit section and 128 Times oversampling at all datarates (7 Bits), thus meaning the actual core of the DAC is able to provide only 12 Bit real resolution without noiseshaping.
Certain others are even more miserly on real resolution, because real resolution costs real money and why bother if you can fake it in the measured performance by agressive use of noise shaping?
It means the ESS DAC relies appreciably less on noise shaping to represent the full needed resolution than most (or at this time perhaps all?) others using the same principle, though it is less than what is attainable using a true multibit DAC. In fact, it is barely able to match the mid 1980's TDA1541 in terms of real (non-noiseshaped) resolution.
For reference, if we combine analogue resolution (24 Binary weighted bits) and the possibility to run at 8 Times oversampling (3 Bits) the Burr Brown PCM1704 (the last true multibit Audio DAC in production) allows us in effect 27 Bit of analogue levels. Sadly this chip is hampered by a SNR/Dynamic range of much less than 120dB, so much of that possible resolution resides below the noisefloor and is of no use. We woudl have to parallel humungous numbers of PCM1704 DAC's to push the noisefloor low enough to make use of the extra bits.
Of course, non of what is written above has any direct relation or mapping onto perceived sound quality. However, a personal observation is that I seem to like DAC's sonics in about the inverse of the amount of noise shaping used. Equally I also know (of) experienced listeners who have a reasonable track record judjing sound quality whose reaction is the opposite (the more noiseshaped the system the better they like it - SACD/DSD being one extreme example).
So the bottom line is - listen for yourself and select what sounds best to you.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
Edits: 07/31/10 07/31/10
T,
I feel your definitions of DACs are driven predominantly by your personal aversion of noise shaping, which is not very scientific.
I suggest redefining class 2) :
a 32 bit DAC is a DAC whose architecture is aimed at attaining 2^32 discrete output states in the bandwidth of interest (even when this is obvisouly not possible). This allows the complete spectrum of amplitude-time trading, so from PCM1704 to DSD so to speak ;-)
You can always add to it that, personally, you only value the leftfield.
As for class 3) it allows a type of device that isn't entirely devoid of meaning. Consider the traditional signal flow in a consumer replay environment:
-24 bit medium (pray)
-oversampling filter internally expands to 48+ bit
-filter output reduced to whatever the DAC itself takes
-conversion
Now add a stage of signal processing between source and conversion. This conversion is new but all to real. Think of that people are doing with computer-based replay, DAWs, or even what appears (or could appear) internally in multi-channel processors.
-24 bit medium
-user-side processing expands to 32-64 bit
-reduce to 24b to send to DAC chip
-oversampling filter internally expands again to 48+ bit
-filter output reduced to whatever the DAC itself takes
-conversion
So there is one wordlength reduction that could be removed if only the DAC accepted longer words (and of course used then directly in the initial filter stage).
I think the latter is intended by at least some of the '32 bit' DAC manufacturers, although they did a bad job of making it clear, and a very unfortunate new numbers war might be the result ...
BTW, I can testify that the management levels of semiconductor companies active in the (very cruel) consumer electronics world tend to be quite cynical indeed. As for the design engineers themselves, they are like mushrooms: fed shit and kept in the dark.
BTW2: there is no way a PCM1704 has real 24 bit performance. Such is impossible in the process used. It is not just thermal noise that intrudes. Beyond 20 bits or so device matching is just a dream, and the process goes ape. And there is only so much laser or flash trimming can achieve, especially on the long term.
bring bac k dynamic range
Dear Werner,
> I feel your definitions of DACs are driven predominantly by
> your personal aversion of noise shaping, which is not very
> scientific.
It is actually scientific in the most basic sense of the word. I have arrived at my dislike of noise shaping as a result of trying to find indications why my empirical research into what sounded good returned certain results.
> I suggest redefining class 2) :
>
> a 32 bit DAC is a DAC whose architecture is aimed at
> attaining 2^32 discrete output states in the bandwidth
> of interest (even when this is obvisouly not possible).
This suffices to exclude any system that uses large amounts of noiseshaping, as even with multiple limitations applied (which classic multibit technology does not so require BTW) there is enough of an inbuild error mechanism, that at any reasonable level of precision 2^(n) where (n) is the number of bits levels cannot be attained.
> This allows the complete spectrum of amplitude-time trading,
> so from PCM1704 to DSD so to speak ;-)
DSD is so heavily noiseshaped, it is (compared to 24 Bit 192KHz pCM which requires less bandwidth to transmit) that it stands no chance in hell to do this even with an upper limit of 20KHz, never mine of 80KHz.
That of course is the point of my original definition.
> As for class 3) it allows a type of device that isn't entirely devoid
> of meaning.
Of course not, for it the class of device nowadays routinely as (n) Bit Audio DAC where the term (n) seems to undergo inflation faster than the Zimbawean currency.
> BTW, I can testify that the management levels of semiconductor
> companies active in the (very cruel) consumer electronics world
> tend to be quite cynical indeed. As for the design engineers
> themselves, they are like mushrooms: fed shit and kept in the dark.
Indeed. And if you know this you also know how target specifications for new Audio DAC's (and ADC's, ASRC's et al) are arrived at, namely by instructing the designer to design something at halve the budget (silicon area) that offers slightly better measured performance than the best measured performance the competitors offer.
> BTW2: there is no way a PCM1704 has real 24 bit performance.
Agreed. I'd call it a 19 Bit DAC, on a good day, with tailwind.
But that leaves the question, what do we call the others?
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
I think the D/A chain needing noise shaping at all is an indication of a flawed design, because the noise shaping suggests *lossy* conversion is taking place, and hence the need for the "band aid" application of noise shaping to mitigate the losses.
A good D/A should process the data at full resolution, in which all internal conversions are "lossless". In such case, noise shaping wouldn't be necessary.
Todd,
I personally think that any AD and DA chain that needs noise shaping is flawed, yet that is what the industry gives us.
And SACD/DSD is nothing but noiseshaping but many like it (I don't like it as much as CD done well or hi-rez PCM).
I am looking very hard at alternatives, but it seems not easy.
So we have choices between various unpalatable alternatives regarding ADC's and we are almost at the same point regarding DAC's now.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
Thorsten, what about the MSB Diamond DAC. Hype or some truth to their claim?
Hi,
Your link is to a complete assembled Digital/Analog Converter device (eg including box, supplies etc.). The original discussion was primarily about the core part (usually a chip) for a complete DAC.
From what I can make out, the DAC used is a Multibit DAC, which would fall into my "Category 2" in the above list.
That is it has the ability to represent 26 (or 27, the information seems a little confused) Bit's worth of levels through it' bit-switches, however noise levels are such that the lowest levels are actually below the noise level.
I seem to remember around 130dB A weighted Dynamic range (A-weighting tends to add around 10dB to the numbers), so the real dynamic range is somewhere between 20 and 22 Bit.
Based on Gordon Rankin's measurements of the ESS Sabre the Sabre seems to offfer a little more dynamic range, but it uses noise shaping to attain it.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
Here is a schematic of the core DAC itself. I think the main difference between the Platinum, Signature, and Diamond core DAC is the use of higher tolerance discrete resistors in the core DAC as you move up the line.
Hi Thornsten, I see your point. The reason I pointed out the entire DAC was to show that it seems they have gone in great lengths do decrease the noise floor of their power supply, among other things, for the Diamond DAC. There noise floor seems to average between 160 and 170 dB. I'm not sure many complete DAC package with those kind of noise figure for the power supply. Even if the effective dynamic range is "only" 130 dB for the Diamond DAC, at least power supply noise won't drown it out.
"The Diamond DAC contains EIGHT 24 bit DACs per module for a true 26 bit resolution. "
26?
Hype ...
bring bac k dynamic range
May be but could it be at least 24 bit? It is afterall a multibit DAC, not delta sigma. They seemed to have gone to extraordinary length to recover the last couple of bits from noise (electrical, magnetic, and thermal).
Provided what you stated here is true, this could be the single most-informative post on digital audio I've ever seen on AA....... And although I cannot determine whether what you stated is true, it does seem very believable, based solely on my personal listening experiences. (I don't see anything technically questionable either.)
But if what you stated is true, it's also a slamming indictment on modern DAC technology. Even though it also gives validity and need for designing a 32-bit DAC with delta-sigma architecture such as the SABRE. As it turns out, this would *not* be overkill, as I first claimed, if some of the "resolution" is indeed processed with noise shaping. (Although such approach in itself I find very questionable in the first place.)
It would also explain why the best designs were 1990s technology. Which I personally thought had been the case based on listening, but never expected a possible technical basis for this. (I always gave benefit of doubt that these processors had true resolution, due to "technical advancement". I only questioned how this technology sounded. The only technical negative I knew about was how ASRC processes jitter.)
What is also wacky is these explanations suggest that high-rez playback has been rendered as something somewhat less than "high-rez". No more resolute than CD playback.
This might sound some alarm bells in the industry. Maybe for the betterment of digital audio design in the future. Thank you much.
Dear Todd,
What I have presented is based on the best of my arguably meagre and limited knowledge. And maybe it is somewhat carried by prejudice (such as Werner has already taken me to task for).
You may wish to look at Dustin's responses (Dustin designed the Sabre DAC) for an alternative take.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
"The Sabre has 2^6 or 64 so called "unitary weighted" or "thermometer code" bit switches. These are able to represent directly 2^6 or 64 individual levels."
This is true on a per pin basis. Each pin has 64 uniary weighted DAC's conneted to it. So in the stereo config many people use, thats 8 pins so thats 8*64 = 512 DAC per channel.
"Further, the Sabre uses asynchronous sample rate conversion on ALL Input data and converts into a clock rate of 40MHz."
Again, true, but not he whole story, you "CAN" use the ASRC if you like, or not use it by simply clocking the XIN pin syncronusly (at an interger multiple of) to the BCLK and then the ASRC drops it self out in this case, revering to a more conventional method as the other DAC's Im aware of do.
"If we assume for ease of calculation a 50KHz data sample rate (close enough to to the 44.1KHz used on CD) we can represent as many as 40MHz/50KHz or 800 individual levels using classic pulse width or pulse density modulation.For ease of calulation I will round up to 1024 levels, which is equivalent to 10 Bit resolution.
This means that the raw resolution build into the ESS DAC is around 16 Bit for single speed (44.1/48KHz) Data, 15 Bit for double speed (88.2/96KHz) and 14 Bit for quad speed (176.4/192KHz) data.
"
Using the ratio of oversmapling of the FS to the Xin clock (the 40MHz in this case) and deriving the amount of extra bits from that assumes a moving average type filter. Meaning it just takes the average over a certain period of time. This is somewhat aprroximate to 1st order shaping. So the assumption that the 1024x upsampling can only give you 10 more bits of resoultion is true only if you use 1st order shaping to your quantization noise. (1st order is 6dB/octave, there are 10 octaves in 1024x, so that means 60dB, ie the 10 bits.) Now imagine you shape with a filter that does 12dB/octave (2nd order) , thats 120dB now, or 20 bits for a 1024x oversampling. The Sabre DAC has a 5th order modulator meaning it does 30dB/octave, however to be fair, the 5 poles in the Noise transfer function don't all kick in at the same point, so it wont do the 30dB/octave * 10 octaves = 300dB. It does however have a digital noise floor of -200dB up to about 200kHz when clocked at 40MHz. This noise is impossible to reach in the analog domian since that is the amount of noise that a 1 ohm resistor will genreate at room temp. (aproximatly).
More details for those who are interested.
Noise shaping took me a long time to wrap my head around, but I think I have it figured out. At first I thought, ok, If I have 1024 pulses that I can set either to a 1 or a 0, then surely, I can get no better then 1/1024 resolution out of this system. This is only true if you use the type of filtering mentioned above, Ie a 1024 tap long rectangular shaped filter. (This makes a SINC filter than has a 1st order rolloff to the peaks of the lobes so this is why I say its "aproximately" 1st order rolloff above)
But the part that is left out of any text book I have read on DS (or SD) modulators is that the post filter after the DAC actually weight's the pulses into it in time by its own impluse responce. This is the "magic" trick to how higher orders of modulation can actaully give you better then 1/1024 resolution with only 1024 pulses. However, the better than 1/1024 is only valid for a certain period of time, thats the tradeoff. This time is set to be 5us or 200kHz and I think we can all agree that sufficient for audio purposes. It is up to the fitler after the DAC in a analog domain to supress the "out of band noise" so it cannot affect things down the line like a preamp or a amplifier.
Someone pointed out the implementation of the filter after the DAC is instrumental to its percieved (not measured) audio performance and this explanation above shows why.
I could go on forever about noise shapiong and how cool it is, and how it shows up in placed we would never suspect, but I will leave it at that for now.
Dustin
...article? I would think you'd be a little irritated and would take them to task.
Hi Dustin,
Nice to see you here.
(For those that do not know, Dustin designed the Sabre DAC).
Thanks for the additional insights. I was trying to keep things simple and avoid lengthy and difficult calculations.
One question, if used in stereo mode, is the whole internal logic of the DAC reconfigured to treat the DAC as 9 Bit DAC or is it simply eight parallel 6 Bit DAC's?
As for noise-shaping, it is a tool like all others available.
However I am sure that you will agree that the more real resolution a DAC is capable of from the hardware side (that is more bits in the DEM Multibit section and higher oversampling ratios) and the less noise shaping is used to attain the final resolution, the better? Or simply, a little can be very good, too much is a bad thing.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
"One question, if used in stereo mode, is the whole internal logic of the DAC reconfigured to treat the DAC as 9 Bit DAC or is it simply eight parallel 6 Bit DAC's?"
This is programmable. It can be both. For the end user to decided what is best for their implmentation. :)
"However I am sure that you will agree that the more real resolution a DAC is capable of from the hardware side (that is more bits in the DEM Multibit section and higher oversampling ratios) and the less noise shaping is used to attain the final resolution, the better? Or simply, a little can be very good, too much is a bad thing."
This is something that people all have different opinions on. The more "true" bits the better is a general overstatement in my opnion. On one side of the fence are the extreme bit resolution NOS guys, this can be made to work quite well from what I hear. On the other side is the noise shaping camp. Neither are "better" than the other, its all a matter of implementation. There is nothing incorrect with noise shaping the data, if this was true, then filtering the data in any way (think of a crossover in the speaker) would be no good. They do the same thing. (i.e. use the impulse responce of some band limiting filter to weight the data in time) One company has pointed out that 6 bits cannot be made to match the resolution of a higher bit DAC. Then if you look at their website you will see that their own ADC uses noise shaping. So for a DAC noise shaping is no good, but for an ADC its fine. Anyways I digress. Back to the question, is more bits of "true" resolution better? My opinion (FWIW) is not in all cases. If one can make 6 bits with some noise shaping create more resolution than a 24 bit with noise shaping then its better, if one cannot its not. There is no right answer here, It more a matter of prefference.
BTW Thorsten, who do you work for?
I have heard NOS and OS DAC's both sound amazing, and both terrible. One cannot generalize on a certain topology. (Or atleast they shouln't if they have an open mind)
Dustin
Dustin, Thorsten is the designer of all the AMR electronics:
http://www.amr-audio.co.uk/
The most common metric for effective bits in DAC performance is monotonicity, the equivalent of no missing codes in a A/D. The other issue is linearity which is a % deviation from the ideal which itself has to be specified as 'best straight line' or 'end-points' depending upon whether the gain is important to the application. Then there are secondary things involving glitches and such.
After reading the article by MSB bitching about the competition I looked at their Specs. Well I tried to but they don't seem to have any. Pot and kettle.
Rick
Hi,
I did not particularly comment on MSB, but they seem to use a multibit system, semi-discrete. And in their case they claim a 24 Bit DAC and they have 24 binary weighted bit-switches.
So in the sense of my definition #2 the MSB DAC is a true 24 Bit DAC. It is not of course a true 24 Bit DAC in the sense of my definition #1.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
"So in the sense of my definition #2
'2) A 32-Bit DAC is one that uses 32 individual binary weighted bit switches, so it is theoretically capable of producing 2^32 discrete steps, though it's analogue dynamic range is less than the postulated 183dB (or 192dB as some may say). No such thing exists either (yet), but it is at least theoretically possible to make such a device.'
the MSB DAC is a true 24 Bit DAC."
Maybe. I'm being a little fussier than you are in that if each step isn't at least a step in the right direction (monotonic) then I don't think it counts.
"It is not of course a true 24 Bit DAC in the sense of my definition #1
'1) A 32-Bit DAC is one that shows 32-Bits equivalent analogue resolution, that is 183dB dynamic range, measured the traditional analogue way. It goes without saying, not only can we not measure such a DAC should it actually exist, but of course no such thing exists nor is it possible, outside a laboratory system suspended in liquid nitrogen and maybe not even then.'"
That appears to be be true. As near as I can make out it doesn't achieve 24 effective bits.
I bring MSB up because apparently their advertising department is the spark for this thread. Off hand it's not clear to me who has the best DC performance, let along overall audio performance. But to me their offensive against their competitor is, well, offensive. At this point they have exactly the same credibility as the old Shell commercial I used to watch on TV demonstrating how adding 'platformate' (gasoline) to the tank increased the distance one could drive. It would be good for companies to realize that if their marketing is bullshit, and that that's often all we have to judge the rest of their company by then their reputation becomes so also.
On the other side of the coin I'd love to hear their devices, I bet they both sound good.
Rick
Is this what you're looking for?
I read the white paper, the snippet is used out of context to 'prove' so totally arbitrary (faked) point. Duh.
We're missing the most important question though. How's it sound?
My new Peachtree NOVA with the 9006 sabre dac sounds great. Beat up my non oversampling dac.
"If people don't want to come, nothing will stop them" - Sol Hurok
Of course, the non-oversampling dacs will always get beat up in equal side-by-side comparisons. The detail is not there.
But to those who need non-oversampling dacs for reasons of smoothness, digital fatigue, I see no reason why sabre would do anything in this regard.
Is the Nova more fatiguing and harsh over several hours than a non-oversampling dac?
Haven't noticed any fatigue factor. I did an a/b at the dealer's with my dac and the sabre dac. No contest.
"If people don't want to come, nothing will stop them" - Sol Hurok
The "6 bits" are the transmission the DAC utilizes at very high frequencies. It supposedly decodes this six-bit high-frequency data to 32 bit data at the 192 kHz rate. Or decodes to SACD data. How this occurs is a mystery, and is probably proprietary to the chip manufacturer.The only way to verify the resolution is provided the input can handle 32-bit data (in which the DAC would have to run in non-oversampling mode to handle it), take a recording where the LSB is being toggled, and check the output on a scope. But such evaluation would likely be inconclusive since the amplitude gradation would be at -192 dB, and would likely be unnoticeable in the presence of ambient noise a good 100 dB above such signals.
There has been so much emotional and high-strung opinion regarding what DACs actually do....... It has become a religion to some people. There has been insufficient information with how these devices work down to the lowest levels, it has often come down to one individual’s speculation versus another’s..... The only comments we can really make pertain to those in reality.
First of all, a 24-bit DAC, running at 96 or 192 kHz, if utilizing true 24-bit data (as opposed to data upsampled from a lower resolution), has resolution down to -144 dB. Since most recording venues at best have signal-to noise ratios of 90 to 100 dB (which is optimistic), such resolution is vast overkill in the first place. The six to eight least-significant bits in such recording are handling almost all noise. What these LSBs are in essence doing is switching on and off to random noise, which if anything produces negative side effects (additional RFI) rather than further capturing relevant musical information from the performance. So adding yet 8 more bits gains absolutely nothing, in my humble opinion. Even if used as an interpolation filter on a 24-bit recording, it would be acting on noise elements at these levels, rather than the signal.
(I can see the importance for a lot of bits in the A/D process, where it creates margin for error in recording levels where limiting or compression is not an option. But it serves no purpose in playback.)
Since I do think the added switching of bits is destructive in the form of additional RFI generation, from both a listener and technical perspective, this is why I’ve never been a proponent for high-resolution digital audio playback and high bitrate digital filtering. I think beyond the 18th bit, the drawbacks in extra bits outweigh the benefits. (If I were to decide on an ideal digitized resolution for audio, I’d say 18 bits at a 60 kHz sample rate. The minimum resolution where I think digital interpolation filtering wouldn’t be necessary.)
Aside from the extra bits, my main gripe with the SABRE DAC is it does use asynchronous sample-rate conversion for CD playback. (It is purportedly defeatable in the SABRE chip, but I don't know if the products using the chip provide such option.) I did listen to the chip once at the LA Can Jam a couple years ago. It sounded decent for an ASRC chip, but I was unable to listen long enough to find out whether this particular chip produced the same sonic ills that other ASRC products produce.
As far as the MSB article is concerned, the explanation could be deemed as "deceiving", but the best way to disprove the article is to somehow demonstrate the resolution. But 32 bits resolution is almost impossible to prove or even disprove. As stated above, the resolution at the analog output is buried so deep in ambient noise, such resolutions IMO are pointless in the first place.
Edits: 07/28/10
Hi Todd,
> There has been insufficient information with how these devices work
> down to the lowest levels, it has often come down to one individual’s
> speculation versus another’s...
That is patently untrue. There are plenty of AES Papers and JAES Articles that cover the low level working in great detail.
A good basic and 'popular science' level overview of the subject at a state in 2005 is in an akademic paper by Ivar Løkken titled "High-Resolution Audio DACs - Final Report - A Review of the Digital Audio
Conversion Process". It has extensive references and is a simple and easy introduction to the subject (which is a bit more complex than this simple and short report intimates).
The only significant change is the advent of the Sabre who has it's own extensive papers on it's operation, some of which are not available without signing drakonian NDA's, but who's operating principles are not dramatically different from what is covered in Ivar's paper.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
When I saw Ivar Lokken's name, the alarm bells went off.......This section contains the same information in Lokken's ASRC white paper that conjured up the notion that jitter can somehow be eliminated with asynchronous sample-rate conversion. The very paper where I shot down such notion by explaining how such application transforms jitter into noise. (See link below.)
Since this guy wrote such a technically-corrupt explanation of the "virtues" of ASRC, I wouldn't be so confident about the other portions of the article. I think it's just another man's "speculation". (The best sources for information on how these chips work at the lowest level are the manufacturer data sheets themselves.)
Edits: 07/31/10
Hi Todd,
I repeat, check the references section for the various AES Papers etc. in which the manufacturers staff describe how they designed the DAC's, if necessary sign up at the AES as associate or whatever they call it so you can acess their library without paying 5 bucks each time.
You must forgive Mr. Lokken his tardiness on many of the subjects that he covers, he endeavors to provide a wide ranging summary useful to pre-graduate students and laymen, which he does rather well I feel. Those more deeply immersed in the matter would be expected to pursue the matter in more depth and for those Mr. Lokken has provided his sources.
I'd wager that his ASRC paper just like this one is merely a summation and amalgamation of reference materiel, as is so common in modern akademick work, and not a work providing unique insights or independent research, such silly tomfoolery being quite discouraged in most of akademia these days.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
The following link is good.
"Since most recording venues at best have signal-to noise ratios of 90 to 100 dB (which is optimistic), such resolution is vast overkill in the first place. The six to eight least-significant bits in such recording are handling almost all noise."Yes there are random bits. But there may not be as many of them as you think, particularly where instruments are closely miked. And the random bits from the microphone preamplifier are "good" analog random bits, not "evil" digital random bits (which are probably not really random, just "pseudo random" and maybe not even random enough in some implementations).
It would make no sense to use SABRE's asynchronous resampling feature in a DAC which (1) has two crystals to support the 44.1x and 48x family of sample rates and (2) is running off an internal clock and using it to control the transport data rate (as with Gordon's USB designs). However, for any given design one might want to investigate as there are undoubtedly "nonsensical" designs out there, although one would wonder how many that use high end chips.
There is little extra processing involved with going to 32 bits, as the internal word lengths needed for good quality upsampling are much greater than 32 bits. One could argue that in some applications there is actually less processing required to use 32 bits, as there will be no need to do any dithering.
If your comments about RFI and hi-res were universal and not peculiar to some playback systems, then I would expect that 44/16 material would sound better than 88/24 material, etc. That's not been the case at my place. The higher sampling rate material usually sounds better when the same recording is available in multiple formats. I have never found a case where it sounds worse. In any event, the high frequency noise output from the SABRE chip is very low because of its design, nothing like what one gets with SACDs.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Edits: 07/29/10
“If your comments about RFI and hi-res were universal and not peculiar to some playback systems, then I would expect that 44/16 material would sound better than 88/24 material, etc. That's not been the case at my place. The higher sampling rate material usually sounds better when the same recording is available in multiple formats.”My experiences have been just the opposite- I have yet to hear one example of 24/96, 24/192, or SACD that I thought was enjoyable for greater than a ten minute duration. The playback always leaves me a feeling of “distress” which ends up overriding any enjoyment I might have for the music. It’s like “dentist drills in the ears” that is foreign to how I think music should be experienced. (The ill feeling stays with me after the music stops, and takes roughly an hour to subside. I think this could be demonstrated with EEG testing.)
I also think my comments about RFI and high-rez playback *not* being universal might be the crux of this whole digital playback issue. It is in my opinion a serious issue that curiously gone under-addressed in recent time. (Although the opinion of the ills of ASRC has become more universal. In spite of new products that still use it.) I think half the problem is this fatiguing phenomenon still exists in a great majority of CD playback systems, and under such circumstances, high-rez is indeed superior.
I may have been spoiled by CD playback relatively void of this fatiguing artifact. Or at least where the artifact doesn't become objectionable. (Although it took me a long time finding such source. Aside from the Prism DA DAC products and Don Allen-mod Philips changers, I don't know much else that I could live with.) I think the difference is while a small portion of CD playback has this fatiguing artifact put in check, I have yet to encounter one such example from high-rez playback. (If I ever do, I'd sing its praises.) What's difficult in attaining listenable playback with CD might be impossible with high-rez formats. And I think the higher bit density/switching per unit time is to blame.
Edits: 07/29/10
I assume that you are not bothered by live concerts, even if sitting in the front rows.
If I were you, I would be very disturbed that my evaluations are diametrically opposite to most others who hear differences between the various digital formats. As such, it would be my highest priority to understand what was causing this. Presumably, it reflects something either in your hearing (or other portions of your body) or in the equipment that you use. Coming to understand this distinction would appear to be the first step.
Do you still express the same preference while listening to other people's systems? Do other people listen to hi-res on your system and agree with your conclusions? Working in this way you might be able to determine whether it was a hearing issue or a system issue (possibly even a room issue).
Another test on those lines would be to listen to a live microphone feed using high resolution equipment. This will be all analog and there should be nothing but microphones, microphone preamplifiers, amplifiers and speakers involved. Do you have problems listening to this?
At this point you could introduce digital ADC DAC loops and see what happens.
In case the problem is caused by radiated signals rather than acoustic, you could introduce nearby computer equipment, including ADC DAC loops that aren't doing anything but are powered up, etc...
It should be possible to get to the bottom of your unusual situation. By the way, I wouldn't consider this to be a problem. I would consider it to be an opportunity.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
“I assume that you are not bothered by live concerts, even if sitting in the front rows.“
Depends on the concert..... I once attended a concert where the “instruments” were mostly laptop computers, and I had that “ill feeling” to a degree I thought I’d keel over. (I was at Hollywood's HOB when this occurred, about midway amongst the audience.) Needless to say, I left early. But I never had such reaction to normal rock, jazz, or especially unamplified/classical concerts. Even if the sound was otherwise horrid for different reasons.
“If I were you, I would be very disturbed that my evaluations are diametrically opposite to most others who hear differences between the various digital formats.”
Being in the small minority has never bothered me, and never will.
Although I do question that it’s “diametrically opposite to most others” in the high-end realm. I do think this has affected more people than we’d like to believe, it’s that most of us don’t realize what’s taking place. I do think audio digitization alters the perception of music, and I do think it has changed the course of how music has evolved since digital audio’s inception roughly 30 years ago.
If I were to add one thing, the few times I’ve stated this ill effect while listening to a high-rez source with others present, I almost never get met with dispute. Or even kind disagreement. It’s as if the advocates themselves realize they’re experiencing the phenomenon too. (Unlike some other issues with audio, where I have had heated disputes. On numerous occasions.)
“As such, it would be my highest priority to understand what was causing this.”
It has been a super-high priority with me. For years. That’s why I bring it up all the time.
“Presumably, it reflects something either in your hearing (or other portions of your body) or in the equipment that you use. Coming to understand this distinction would appear to be the first step.”
I’ve heard the ills in every system I’ve listened to, mine or others. And as I stated above, whenever I mention the ills, people listening with me react as if they’re sensing those ills too.
“Do you still express the same preference while listening to other people's systems?”
Most definitely. Even compared to CD sources that I thought were fatiguing. Although I might eventually encounter a system where I think the CD source is more-fatiguing than the high-rez source. There are a lot of awful CD sources out there.
“Do other people listen to hi-res on your system and agree with your conclusions? Working in this way you might be able to determine whether it was a hearing issue or a system issue (possibly even a room issue).”
The few times I had high-rez piped into my system, those listening with me never disputed what I was hearing, but often thought the problem was with my system. But the last time this was done (which I say was a good six years ago), the guy demo’ing the high-rez source was *dumbfounded* over how pleasant the CD source sounded by comparison, through the very same system. (After being made aware of the problem, he told me he was noticing the ill effects on his own system. I think he ultimately dumped high-rez himself.)
“Another test on those lines would be to listen to a live microphone feed using high resolution equipment. This will be all analog and there should be nothing but microphones, microphone preamplifiers, amplifiers and speakers involved. Do you have problems listening to this?“
Not at all. But I never tried it. This would be a good blind test for venues that use sound reinforcement.
“At this point you could introduce digital ADC DAC loops and see what happens.”
I’ve experienced that. Digital room correction using a vinyl source. On numerous occasions. The in-room response was improved in some (but not all) cases, but the music always sounded "digitized" by comparison. This is why I’ve been a detractor of digital room correction. (Unless integrated into a single D/A conversion with digital sources.)
Heck, I’ve even noticed the ill effect with vinyl sources that use switching power supplies. I remember someone upgrading the power supply of his Linn turntable. From AC synchronous to a Lingo switching power supply. He initially raved about the upgrade, but when I heard it, I thought something had gone big-time wrong. The sound was great analytically, but the music sounded "broken". I think someone else voicing similar opinion made him try the synchronous PS again, and then when I listened again, I told him I thought it was a lot better this way. After reading up on how the Lingo PS worked (I used one personally when I had a Linn, but never suspected it was causing problems at the time), I concluded that RFI can even cause such ills with pure analog playback. Which is why whenever someone does an analog vs. digital comparison, I suggest the digital source be powered down and removed from the system while the analog source is played.
“In case the problem is caused by radiated signals rather than acoustic, you could introduce nearby computer equipment, including ADC DAC loops that aren't doing anything but are powered up, etc...”
I will say I’ve never thought PC-based playback was listenable in cases where the DAC was not galvanically isolated from the computer's CPU. My PC rig has been using outboard DACs with galvanic isolation. (I do prefer Toslink transmission over isolated USB, by the way.)
“It should be possible to get to the bottom of your unusual situation. By the way, I wouldn't consider this to be a problem. I would consider it to be an opportunity.”
I think I have done so...... I am 99 percent certain that excessive RFI is what causes the ills of digitized audio playback. If not that, I wouldn't have a clue regarding what might be causing the ill effects. I don't think jitter alone would cause it.
I do think if the industry were to buy into the notion that RFI is a major issue with digitized audio playback, and addressed this problem, I really, really think it could potentially take major strides in making digitized audio more listenable, to where even high-rez might become an attractive format for a far greater number of people. For I do think the reason why high-rez hasn't gotten popular is that people are actually hearing these ill effects. Without consciously realizing it. For if it were not for that, these formats would be and should be a big-time slam-dunk for ultra-fidelity audio playback.
This ill effect, is it like a sour taste in the back corners of the mouth? A metallic feel at the back of the throat and ears?
"This ill effect, is it like a sour taste in the back corners of the mouth? A metallic feel at the back of the throat and ears?"
The second item is close..... My ears do have a feeling of excessive HF stimuli- "Dentist drills". This strain often extends to the head, neck, and upper chest.
Have you figured out whether the effect is coupled acoustically or through the presence of RFI directly affecting your body?
If coupled acoustically, is it just the presence of the interference by itself or does it involve some kind of interaction between the interference and the music?
You might be able to answer some of these questions with appropriate tests. For example if the effect doesn't involve the interaction of the music with the interference it should be detected when the equipment is turned on and playing a silent track. At that point you could disconnect the speakers and see if the effect persists. If the effect involves the interaction of music with the interference it will be more difficult to conduct tests, e.g. it might require two systems in the same room and running (and powering up and down) various combinations of gear.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
"You might be able to answer some of these questions with appropriate tests. For example if the effect doesn't involve the interaction of the music with the interference it should be detected when the equipment is turned on and playing a silent track."
If I were really masochistic, I'd try this with the volume maxed out.............
I've noticed similar effects with ultrasonic bug repellers, but unlike with listening to music, the part I notice most is elevated tinnitus. It's not that I listen for it as much as the tinnitus getting my attention.
"the part I notice most is elevated tinnitus"
Me too. And also computer monitors, especially CRT ones.
The scary thing is that after a few weeks a new monitor no longer causes problems (if the refresh rate is constant) and I think it's because it's wiped out my hearing sensors for those frequencies. We may not 'hear' ultrasonics and obviously by definition we don't, but that doesn't mean that they aren't dangerous.
Rick
I do not want to start a war here but would just like to say that having a minority opinion does not necessarily mean that you are wrong. The majority can often be incorrect in there opinions. To say that Todd should be disturbed is simply downright insulting and is based on a I(we) am right and you are wrong. Just my opinion. It would be nice if would could just talk about our experiences without the insults but that doesn't seem to be the case
Alan
"that having a minority opinion does not necessarily mean that you are wrong."
I do not think that Todd is wrong. I did not post that Todd was wrong. Please read my post carefully. Todd reports what he hears and there is no way anyone can prove him wrong. I am not like some of the people (generally found on Prophead and a few other places) that deny that other people hear things they have reported and call them "deluded" or "fools". These people typically cite questionable "research" as "proof" of their assertions. Some have been banned from the asylum, but there are others who skirt the edges of polite demeanor and are still around.
"To say that Todd should be disturbed is simply downright insulting..."
I did not state that Todd was disturbed nor that he should be disturbed. I would not say such a thing. If you read my post you will see that I said that if I experienced what Todd has been experiencing then I would be disturbed. Since it is my own feelings that I was describing, I can explain what I meant.
When I am "disturbed" I mean that my mind is constantly nagged and bothered. From experience, I tend to be patient and ignore minor annoyances if they do not last long. However, should they persist then I interpret the constant nagging as an indication that there is something to be resolved, which might be physically changed or mentally changed, e.g. it might mean getting to the bottom of the problem and learning exactly what it is, so that matters could be put right one way or another. (This might be fixing something, learning to make the best of an imperfect situation or figuring out how to avoid a situation.)
In my post I used the word "opportunity". The opportunity here is if Todd is one of a few people who can hear a particular sonic defect then he potentially has the capability to improve the state of the art, not just for himself but for all of us. This could be done in a number of ways, such as potentially teaching other audiophiles how to hear what he does, helping psycho-acoustic researchers improve their understanding of human hearing, or helping designers and manufacturers produce better sounding gear. (Of course there might be other personal interests and responsibilities that might preclude Todd doing this and that would not be for me to judge. I was speaking to my own priorities, being semi-retired.)
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Here's what I recently posted at another forum after I asked Wyred 4 Sound about whether there is asynch upsampling in their Sabre32 DAC's:
--------------------------------------------
I was curious enough to ask W4S about the somewhat unclear upsampling/oversampling situation with the ESS DAC chip. I was pleasantly surprised that I got answers quickly and to my satisfaction:
"Yes, our dac uses the internal oversampling. Basically, the unit will oversamples the signal to 386 times the rate of the data."
I then asked the following:
"Thanks for the reply. Are you saying if I feed it 44.1kHz redbook data, it will oversample it 386x to 17022.6 kHz?
What if I'm feeding it 96kHz signal, does it oversample it 386x to 37056 kHz?"
The answer from W4S:
"Yes indeed... 44.1kHz will go to 17,022.6kHz, and you are correct about the other rate as well. It is actually oversampling, instead of up-sampling."
SABRE has a number of sections and can do both synchronous upsampling and asynchronous sample rate conversion. These are optional features that can be enabled or disabled.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Sheppard,
Furthering what Tony said and he is totally correct.
It would be like saying hey the top of the line XXX dac can't be 24 bits as it says it streams sigma delta at 1 bit.
Everything in the end has to go out some way. All dacs uses either ladder logic like the MSB or sigma delta (1 bit) or some combination (6 bits). Which is better really depends on the design and output structure.
Heck I have heard (and made) some 16 bit dacs that better most the 24 bit units I have heard.
Anyways, believe me it is a 32 bit dac.
Thanks
Gordon
J. Gordon Rankin
Gordon,
Would you mind explaining how the math works out to get the 32-bits for the Sabre? I'm interested in just understanding the math, even if as you mentioned how an entire DAC sounds depends more on the design and output structure.
Is it that there is an unspoken standard that when someone says a 32-bit DAC chip, they mean that it's 32-bits at 44.1 kHz, or something along those lines as Tony alluded to?
The function behind the bits is signal to noise ratio, and this is only meaningful in the context of a given bandwidth. Basically, for a given channel as the bandwidth is increased the noise also increases. This is true for white Gaussian noise in an analog system, as well as "shaped" noise created by a digital converter. For a meaningful comparison between two systems that have different bandwidths one needs to provide some weighting of the noise. Conventionally this has been done by measuring the noise in the range 0 - 20 kHz. Better is to provide a complete noise spectrum as this will allow for a range of comparison metrics.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Sheppard,
I just explained this over on Tube Asylum so I will link it here.
http://www.audioasylum.com/forums/tubediy/messages/18/186791.html
Thanks
Gordon
J. Gordon Rankin
The information is blatant misinformation by a web site selling a competing product. Either the people who wrote it are technically incompetent (at best) or dishonest (at worst). I would dismiss the entire site as bogus. "False in One, False in All."
Those 6 bits are clocked at 40 MHz. This is not like clocking them at 44.1 kHz. This should be compared to 1 bit being clocked at 2.8 MHz which provides excellent sound on SACD recordings. To quote the number of bits without discussing the clock rate is deceptive. Furthermore, when run in 2 channel mode there are actually 8 bits, not 6.
There are various portions of the chip that can be independently configured, so the statement about not being unable to disable features is false.
Finally, the SABRE chips are presently being used in high end DACs, such as the Denominator Module sold by Wavelength Audio.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
"So the hot new 32 bit DAC is actually a 6 bit DAC! Right from their own white paper.
Post a Followup:
| FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
||||||||||||||
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
