Yes - there's appreciable noise modulation in the higher frequencies. In the bass its awesome sounding to my ears, but the treble sucks. Makes a real mess of violin tone for example in this recording - http://www.amazon.com/Vivaldi-Quatt...DYU/ref=sr_1_7?s=music&ie=UTF8&qid=1432946013
Actually it was a very interesting paper, and I found it interesting regarding the problems with DSD and if it was not implemented almost it seems perfectly.
I personally don't find digital replay grating or problematic as many do, but do suspect that some equipment may be prone to high frequency noise, that could be one of the causes of 8the difference in noise spectrum as an all analogue playback system will not have the higher frequency noise created by digital circuitry...
I personally don't find digital replay grating or problematic as many do, but do suspect that some equipment may be prone to high frequency noise, that could be one of the causes of 8the difference in noise spectrum as an all analogue playback system will not have the higher frequency noise created by digital circuitry...
Benchmark DAC2 HGC D/A processor/headphone amplifier Measurements | Stereophile.com . OK doesn't go above 100KHz but below that looks nice and clean.
Schiit Audio Bifrost D/A processor Measurements | Stereophile.com. Not as good, but the highest spike-o-grunge(tm) is still below the vinyl noise floor.
dsd (ignore text see graph) shows the noise shaped floor of DSD. Yeah does look nasty if you are a dog or a bat.
Now as DS converters fed PCM do not have this noise issue it would appear the problem is not DS DACs, but rather DSD. As 99.9% of my digital music is red book or 24/96 I can live with that. I can play SACD in my blu ray player but dont have a surround decoder to take any advantage of it.
skimming the lipshitz paper he points out as well that DSD is the evil thing and that 'Audio grade' DS converters are 5 or 6 bits in general.
So I guess we can agree that feeding your average DS DAC yummy multibit PCM does not give us all these digital nasties? I am happy with the view that SACD/DSD is evil. I Have held that view since it was first announced.
Schiit Audio Bifrost D/A processor Measurements | Stereophile.com. Not as good, but the highest spike-o-grunge(tm) is still below the vinyl noise floor.
dsd (ignore text see graph) shows the noise shaped floor of DSD. Yeah does look nasty if you are a dog or a bat.
Now as DS converters fed PCM do not have this noise issue it would appear the problem is not DS DACs, but rather DSD. As 99.9% of my digital music is red book or 24/96 I can live with that. I can play SACD in my blu ray player but dont have a surround decoder to take any advantage of it.
skimming the lipshitz paper he points out as well that DSD is the evil thing and that 'Audio grade' DS converters are 5 or 6 bits in general.
So I guess we can agree that feeding your average DS DAC yummy multibit PCM does not give us all these digital nasties? I am happy with the view that SACD/DSD is evil. I Have held that view since it was first announced.
Which noise issue are you referring to specifically with 'this' ?
I suggest more than just a 'skim' so you might notice SL's significant oversight.
As a handwaving "explanation", what poor DS appears to do is lose resolution, the effective number of bits is well down on what it theoretically should be and, it largely appears to be a dynamic behaviour - it varies in its effect depending upon what the signal is that's passing through - so, "modulation noise" as abraxalito calls it.
The result is that what many people call micro-detail goes missing in action, it literally just vanishes - I've been gobsmacked at times listening to recordings that I know by heart on other, sometimes pretentious systems, and whole chunks of what's going on is no longer there, it's been scrubbed out! The less significant bits of the waveform have been discarded in action, I'm no longer listening to 16 bits of information, it's more like 8! And the noise of the discarded information is not hiss, but a beautifully dithered greyness of tone.
The result is that what many people call micro-detail goes missing in action, it literally just vanishes - I've been gobsmacked at times listening to recordings that I know by heart on other, sometimes pretentious systems, and whole chunks of what's going on is no longer there, it's been scrubbed out! The less significant bits of the waveform have been discarded in action, I'm no longer listening to 16 bits of information, it's more like 8! And the noise of the discarded information is not hiss, but a beautifully dithered greyness of tone.
The ultrasonic hash caused by the noise shaping.
Really not worried if SL has missed something. I will only get music in DSD if that is the only way. I'm not even convinced that I need 24/96. The measurements available to me suggest that current DS DACs fed PCM are more than adequate and free me to invest my funds elsewhere in the chain where I have real issues that need addressing. But one day I may find time to dedicate to it.
Really not worried if SL has missed something. I will only get music in DSD if that is the only way. I'm not even convinced that I need 24/96. The measurements available to me suggest that current DS DACs fed PCM are more than adequate and free me to invest my funds elsewhere in the chain where I have real issues that need addressing. But one day I may find time to dedicate to it.
So then your original sentence, expanded with the latest explanation reads now -
'Now as DS converters fed PCM do not have the ultrasonic hash caused by the noise shaping....'
which is clearly wrong as ultrasonic hash is inherent in such noise shaping. Otherwise there's nowhere to push the in-band noise to.
Incidentally what SL missed others in the field seem to have missed too which impacts multibit noise shaping, not just 1bit. If you're still interested I suggest seeking out the RMAF presentation (I think in 2012, not certain) by Martin Mallinson from ESS on issues with noise shaping.
'Now as DS converters fed PCM do not have the ultrasonic hash caused by the noise shaping....'
which is clearly wrong as ultrasonic hash is inherent in such noise shaping. Otherwise there's nowhere to push the in-band noise to.
Incidentally what SL missed others in the field seem to have missed too which impacts multibit noise shaping, not just 1bit. If you're still interested I suggest seeking out the RMAF presentation (I think in 2012, not certain) by Martin Mallinson from ESS on issues with noise shaping.
Last edited:
potato, potarto. Up to 100KHz the links I added show no hash. If it's above 100KHz or less than -80dBFS its not something that keeps me awake at night. For me the current crop of commercial audio DS converters which are not single bit internally are good enough for me.
Maybe I am just easily pleased.
Maybe I am just easily pleased.
So you don't see 100kHz+ output noise from a DAC as any sort of issue at all? How about 400kHz-ish output noise from a classD amp when fed into some electronics (an AP audio analyser for example) - see any problem with that?
AP see a problem with it (the analyser stage is not too far different from a DAC's analog output stage, lots of opamps) so don't recommend classD amps are fed into their analyser without an add-on LC filter between the two. Now which commercial D-S DACs have you seen where an LC filter is used between DAC and subsequent opamps?
Incidentally I see billshurv linked to the propaganda about DSD on PF. Be very skeptical there as the graph is patently mis-labelled with supposed 24/96 PCM noise floor which stems from a mis-interpretation of the FFT.
Incidentally I see billshurv linked to the propaganda about DSD on PF. Be very skeptical there as the graph is patently mis-labelled with supposed 24/96 PCM noise floor which stems from a mis-interpretation of the FFT.
Not for listening to music at home no, as I don't have a classD amp or an AP analyser. I don't get your point. The SL paper was about 1-bit DS as implemented in DSD. He accepts that 5 or 6 bit DS gets around the quantisation problem. Well implemented DACs using said DS converters have excellent noise floor up to 100KHz above which I am not bothered for domestic listening. When not spinning vinyl I listen almost exclusively to PCM. In my mind I don't have a problem. What am I missing?
(been listening to swan lake off the server tonight. Not like being there, but most satifying).
Do you want to get it? You've not been getting me all along this thread but I don't see that as any kind of issue myself.
Did you notice his oversight? Why does he accept this?
Why do you assume there's a 'noise floor' with such converters? Because the FFT looks flat?
What are you missing subjectively in your listening experience or what are you missing in terms of the current discussion? If the latter then you missed an earlier question of mine which I'm curious about having an answer to.
I think you will find the latest AP stuff handles classD just fine without additional filters. Have a look at the APx range specs. Older APs that were designed before classD came along does need filters.
The supposed 'propaganda' link was all I could easily find to show the DSD noise floor. PCM noise floor was on the links to stereophile measurements. If you have a better comparative graph I would be interested to view it.
Thanks, I hadn't noticed that, I shall definitely go check it out. Any ideas on what they've changed?
The propaganda nature of it is clearly visible, not merely 'supposed'. Its patently false in its marking of the 24/96 noise floor.
I told you I'm not going looking for his oversight. The noise floor I reference for commercial DACs are the ones measured and published. How much of that is analog noise, how much shaping noise and how much the converter noise floor I do not know.
Your last comment is cryptic and I do not understand. I find both vinyl and digital satisfying. Which question did I miss?
Your last comment is cryptic and I do not understand. I find both vinyl and digital satisfying. Which question did I miss?
No Idea how they do it, just that I had the debate with Tomchr as he is planning to test a hypex. Ignoring the 24/96 noise floor, is the DSD noise floor as plotted correct?
I don't know what 'correct' would be given there are no parameters given for the FFT - its clear that Lynn misinterprets what's shown on the FFT in his article. So at best its only partially correct because the bandwidth isn't provided - just as tomchr doesn't provide a bandwidth when he references noise from his FFTs.
I've had the pleasure of spending the last month with some classically trained and extremely talented musicians, pianist, cello, harp, and vocalist. What really impressed me every day was how wonderful they all were at sharing their wide ranging expertise on classical music. Yes they enjoyed the fact I have a couple of pretty good sounding systems but never once did anyone mention how much they preferred the turntable , versus CD, versus whatever the upgraded iPod was putting out. The main thing was that any of them could deliver enough of the music experience to be thoroughly enjoyed.
- Status
- Not open for further replies.