|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
83.168.10.231
In Reply to: RE: I have just read this and-thanks posted by knewton on March 17, 2014 at 18:15:13
leaves me no wiser on the posted text on the player, nor presumably anyone else.
Follow Ups:
Well, that would figure, since I don't know what's inside AMR's player. I only wrote what it sounded like AMR may have inside based on their marketing copy.No, I'm not going to take the time and effort necessary to teach you about NCOs, nor lead you to proof that they actually exist, can perform well, and are widely utilized. If, for some reason, you doubt those facts, then you can easily research them for yourself via Google.
_
Ken Newton
Edits: 03/18/14 03/18/14 03/18/14
actually it sounded like nothing. I am familiar with numerical methods, having used them extensively in computer modelling.
I feel that our exchange is on the verge of degenerating, so, you can have the last word.
_
Ken Newton
Edits: 03/18/14 03/18/14
Maybe you or someone else can point us to some specific technical papers on NCOs that describe various methods of implementation and include models and formulas describing and quantifying the jitter these generate. Without this basic knowledge one is in no position to evaluate manufacturer's specifications except by purchasing products and testing them, something that I'm in no position to do. Worse, without this knowledge one may not even be able to think of appropriate tests. (I do know some ways of implementing NCOs, but all the one's I've looked at wouldn't meet my standards of what is needed for high quality audio or are absurdly expensive, as with zillions of crystals.)
Note: I am not interested in manufacturer's spec sheets or technical white papers. I've looked at some of these and, like most spec sheets show specific information calculated to make the devices look good, e.g. numbers of "typical" devices (no guarantee as to the one you have) or RMS averages (no good description of how averaging was done or what the actual worst case behavior might be).
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
http://analogictips.com/using-ad9850-direct-digital-synthesis-dds-ic/
not very impressive since the assumptions in programming the chip are all important
These chips output both analog (sine and cosine) waveforms and digital clock pulses. The analog waveforms are used for software controlled radios, while the digital output is the one that we are concerned with for digital audio applications.Looking at the spikes, I would think the true harmonics wouldn't matter after conversion to digital since they will be synchronized with the fundamental frequency and hence would affect the transition times of all transitions equally. This statement does not apply to the bogus 4th harmonic in the plot. It looks like some kind of aliasing. This could be bad because this can cause periodic jitter and the pattern of this could vary as a function of the synthesized jitter. This could cause people to hear different sound as a function of transport clock frequency.
It seems like digital synthesizers can be made to work very well if the synthesizer DAC uses analog circuitry (including its own DAC) that works sufficiently better than the audio DAC that is actually clocking the audio. This seems like a serious fail to me, not only in cost, but if one can build a better DAC one should just use it for audio in the first place. If the two DACs are equal in performance then it appears like there will be more jitter noise in the audio output compared to just using a fixed reference clock and ditching all the DDS machinery. Here the problem is the noise floor in the analog signal that results in the digital clock. Any noise here appears as jitter in the output audio when the audio DAC uses the jittery clock.
Having built some software generated audio clocks (square waves) I am aware of the various artifacts these devices can produce. The ones that I built suffered from different tone quality due to varying aliases with different synthesized frequencies. This was a hard fail for me, as I was interested in human musical pitch perception at the time, i.e. how different tuning systems for keyboard instruments sound. I was generating square waves, as this was all I could do at the time (early 80's) using software in an Apple II. Eventually, I coded up a generator that produced synthesized square waves at a sampling rate of 1 MHz. At this point the aliases were sufficiently low level that performance was adequate for my purposes. (If you are wondering how one can generate samples and play them at the computer's clock rate when a single instruction uses several clock cycles, that's a different story.)
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Edits: 03/19/14
a look at the waveform in the link that I provided.
Edits: 03/19/14
I missed the link in your earlier post because it wasn't hot.Thanks for pointing out my omission.
From looking at the waveform (not very accurate) one can see that a clock edge triggered on the rising edge might be reasonably consistent, but one triggered on the falling edge would vary from cycle to cycle, probably producing significant deterministic jitter. It is likely that this jitter pattern will be very different when synthesizing different frequencies. Hence the sound of a DAC clocked by this DDS will be dependent on the average frequency of the input clock. This means that the same DAC might sound significantly different, even when being driven by two essentially perfect transports that differed only in their average clock frequency. From a sonic perspective a straight reclocker would be preferable with a big buffer so that one could play for a long time without overflowing the buffer, thereby avoiding all this dodgy unanalyzable mixed signal circuitry.
Having any of this non-linear interaction between the transport and the DAC creates a situation where audio performance is unlikely to be adequately predictable, as it will vary from unit to unit. None of this nonsense is necessary since there is no need for any kind of adaptive master clock in the DAC in the first place.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Howdy
Chase down the patents... I used to work with John Melanson before he went to Cirrus Logic. I think some of his patents might have the kind of detail you are interested in. I noticed that he has more than 100 of them, but looking for DDS and/or sigma delta might narrow them down.
When I tested them next to a high quality crystal they fared fairly well. I suspect that since their jitter error is fairly white that their errors aren't as audible as they might appear from the numbers. I'd use them in a lower cost DAC.
-Ted
I'll have to look some more details on DDS circuitry. The patent that I read was mostly about the complete system and didn't contain much information on the specific DDS circuit. What it said pretty much matched my model of how such a device might work, except for some additional detail of the DS DAC used to generate the quadrature synthesized waveforms. (Which I didn't understand, in that no one really understands DS modulators, or at least if they do they aren't talking... )
Thanks for your help. Incidentally, I would be curious to see how your DAC sounds when playing PCM files converted in the computer to DSD128 by various programs such as HQPlayer, vs. how the same files sound when the native PCM is converted by your FPGA. Perhaps some of the reviewers can conduct these listening tests.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Thanks. It looks like US 7,557,661 is probably the one. Printing it out now.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
that DDS hardware I have come across and used (2 clocks) are not any better than others.
Having said that, there is also a claim that relatively high phase noise fundamental SAW chips make better clocks!
I think these claims are misguided at best, and that the "improvement" is likely to be a preference for masking noise that hides some more objectionable distortion. However, there are systems where the addition of noise actually improves operation according to theory, measurement and listening. Details matter. Issues involving clocking can be quite difficult.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Were I a commercial DAC vendor, I would implement an NCO if my main jitter related concern were the rejection of transferred jitter, over having the absolutly lowest intrinsic phase noise local oscillator. This is simply because an NCO can be very closely 'tuned' to track the remote clock source's frequency over a very wide range. Sort of like an VCXO, but with an impossibly wide pulling range.Interestingly, Naim appears to have been seeking both the lowest intrinsic and the lowest transferred jitter when they designed their DAC. So, instead of a pullable VCXO, or a programmable NCO, Naim utilizes a series of individual XOs that are tuned to slightly different frequencies. One of these XOs will likely be close enough to the source clock's frequency, and so can be dynamically selected to serve as the local oscillator. Of course, there are practical limits to how many individual XOs are available. At some point, there won't be an close enough frequency match available, whereas an single NCO can be tuned to closely match a great range of remote source clock frequencies.
_
Ken Newton
Edits: 03/19/14 03/19/14 03/19/14 03/19/14 03/19/14 03/19/14 03/19/14
Yes, it was the Naim approach that I had in mind when I mentioned "an absurd number of crystals". With only a few crystals then one will have to have a huge buffer, with resulting limits on responsiveness to user play controls (and suitability for live recording applications) and length glitch free playback.
None of this complexity is necessary if the DAC just outputs a clock that the transport can accept. The problem is that none of the consumer standards except async USB address this issue, making the proper solution impossible to obtain in a standardized plug and play fashion except via USB. IMO USB is not a very good system to have anywhere close to the mixed signal circuitry in a DAC and so this adds substantial costs to a high quality DAC to achieve the needed isolation.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
It's not a question of noise but a question of the harmonics generated. Some chips are fundamental whilst others are 3rd harmonic. Rig up a FFT (I did this sometime ago with the Tent Clock) and one sees that odd order harmonics are quite strong and there does not seem to be a suppression scheme. It has been claimed that the Crystek low phase noise chip is not as good sounding as a SAW chip (hifiduino and Epson or Citizen).
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: