LeavIng us to still laugh at a marketing phrase like multi-bit DSD to stop the "pure DSD" ocd-ers from having a heart attack at the unavoidability of PCM (unedited recordings exempt of course)
There's 2 reasons why DSD can be superior to PCM
1: It allows bypassing the resource constrained SRC/SDM facilities in SDM ASIC's
2: It's a superior medium if absolutely no editing is required. Such as direct copies of analog tape, or DSD direct recordings where tuning was done with room acoustics, and mic placement, rather than DSP.
# 1 is quite important as 95% of all of today's DAC's use SDM ASIC chips.
I don't think this math from early 1900s is completely accurate. Information might be quantum-coupled between different locations, so a system without memory can fail Markov property and exhibit sequential behavior, as long as it interacts with other system with memory. When superstition (opposing "key" multipoles) is used to "teleport" information, it's the key multipole what is remembered, and not the information itself.
This is the opposite of experimental method. This results in hierarchical brain preference control, the most superstitious ones follow in secret the preferences of the least ones, but continuous exercising of experimental method is required to keep more certainty. The opposing "key" multipoles used for quantum coupling of information are actually unhealthy biological conflicts, so if you get fed up of others conflicts just assume they intimally love any conflict they project over others, to the point their conflicts are moved to their heart neurons, and you do not love any of that unhealthy biological noise.
Concerning DSD, remember that signal is ruined by jitter, as information is encoded as skew. Has anyone ever seen the amount of jitter in the "eye pattern" of a CD in oscilloscope screen? Not to mention magnetic media.
These square waves from delta sigma are only accurate for IC to IC communication (like a TL3016 sending DSD to a IRS20957). But any "powerful logic" IC (like a DSP) is quite noisy and skew inducing inside, it's not a good place to do DSD to PCM conversion. So in the end this conversion is better performed in the audio codec IC.
These square waves from delta sigma are only accurate for IC to IC communication (like a TL3016 sending DSD to a IRS20957). But any "powerful logic" IC (like a DSP) is quite noisy and skew inducing inside, it's not a good place to do DSD to PCM conversion. So in the end this conversion is better performed in the audio codec IC.
If we use the Mola Mola DAC block diagram as a reference. In this topology, PCB 1 is the PC with powerful Intel processors, and software 64 bit floating point DSP engine. All audio is upsampled to 100Mhz. The 100Mhz data travels via fiber Ethernet to PCB#2 in the speaker. At this point it's clocked by the ultra low phase noise master clock, and low pass filtered to obtain the analog signal.
How do you digitally process unclocked data? This data can only be translated to analog, to PCM. More operations please?
And the golden rule for conflict: It either comes from a learning process or from a profit process. If neither learning nor profit is shared, the conflict can be safely rejected.
edit: It's PCM as long as it's clocked. A SHARC won't accept anything unclocked, and if transition timing functions are used to measure pulses then the signal is aliased to SHARC clock. You are aliasing (trying to decide past certainty).
And the golden rule for conflict: It either comes from a learning process or from a profit process. If neither learning nor profit is shared, the conflict can be safely rejected.
edit: It's PCM as long as it's clocked. A SHARC won't accept anything unclocked, and if transition timing functions are used to measure pulses then the signal is aliased to SHARC clock. You are aliasing (trying to decide past certainty).
Last edited:
And the assumption that the first premise is generally true is false unless it is backed up by some real evidence apart from your subjective opinions.
You also forgot to add that doing any kind of raw processing with DSD required for say an active speaker implementation is both pro-grammatically cumbersome and resource intensive. In other words a waste of clock cycles for little or no gain !
cheers
Last edited:
It's pretty much standard knowledge in the industry that ASIC DAC chips have sub-par SRC/SDM capabilities. This is why all of the best DAC's, even the ones that use ASIC chips, do their own SRC on FPGA or DSP chip, prior to the ASIC. ASIC chips are built to a price point. They are designed mainly for high volume mass market applications. The Sabre chip designer told me 98% of all Sabre chip sales go into DAC's around the $1000 price point like the Oppo's. Yes they could make them better, but they would cost way more, and they wouldn't be able to sell enough to cover the R&D costs. ASIC chips are only feasible for ultra high volume applications.
There's a reason they do the raw DSD processing in DAW workstations. And it's not just so they can get a worse result. The resource intensive end of things is only an issue if your solution uses DSP chips that don't have enough power for the task.
Last edited:
A couple of months ago Roon came out with built in SRC/SDM in their media player. I felt it was sub-par and told them this straight up on their forum. They were downsampling all DSD to standard PCM rates before applying the DSP. Fanboys (proponents of mediocrity) revolted thinking I was being too harsh, and it resulted in a pile of posts getting deleted. Anyways they ended up doing exactly what I requested and the end results can be found in this thread and others. People are floored by the difference in sound it made doing "DSD wide" processing compared to downsampling to standard PCM rates.
https://community.roonlabs.com/t/dsp-up-sampling-features-in-roon-1-3/19293/35
https://community.roonlabs.com/t/dsp-up-sampling-features-in-roon-1-3/19293/35
Last edited:
Well Oppo can afford to put 2 of the more expensive $75 9038's in a $1000 full function 4K bluray player with all sorts of functionality besides just being a DAC. Several Chinese DAC's using dual 9038's for under $1000 as well. These are the units 98% of all Sabre chip go into. Not the Ayre QX5's of the world (that happen to use an FPGA based SRC engine before the 9038, along with all other top end Sabre based DAC's)
Last edited:
And the FPGA costs how much ?
Your flawed reasoning is that because there is an FPGA in front of the DAC then it must automatically be superior and that justifies paying 10K for the DAC instead of 1K !!.
Did it ever occur to you that you shouldn't need to pay more than 1K for an extremely good quality DAC ?
And speaking of Ayre according to his website, doesn't seem to have too much good to say about DSD contrary to the way that you paint the story http://www.ayre.com/insights_dsdvspcm.htm
cheers
Last edited:
...For most, a thick aluminum case, and blue LED's, has far more impact on the pleasure neuron's than the sound coming out of the speakers
+1
Last edited:
The FPGA doesn't cost all that much. The knowledge to program it is far more expensive.
No they don't just put the FPGA in there for a decoration. It's there for a purpose.
I suppose if you look at it that way you shouldn't need a stand alone DAC at all. Your average smartphone should be good enough for anyone. Look what $9 gets you from Apple:
Apple Lightning Adapter Audio Quality Measurements
Really should be an endgame for all audiophiles.
And why even bother going active? A simple passive setup is all everyone should need.
Last edited:
But the baseline Sabre DAC is still not good enough for you unless it has an FPGA and a 10K price tag. Yep I get it now
LOL
Last edited: