Audio Asylum Thread Printer
Get a view of an entire thread on one page
|For Sale Ads|
Makes sense to me, gotta give Benchmark credit for sharing this:
Thanks that was VERY informative. As an old analog person that at FIRST was skeptical of digital, but learned to like it over my analog, I appreciate the explanation of hiss and distortion.
Is it ok if I post this over at the vintage forum?
As has already been pointd out earlier, this video has been around for quite some time now. It is very informative, in it's hands-on demonstrations utilizing continuous signals. What it leaves out is an effective demonstration of transient signal handling.
The sampling theorem is predicated not only on signals being band limited, but also on them being infinitely continuous - not only continuing past the end of the universe but, just like God, having always existed. Any other kind of sampled and reconstructed signal is mathematically imperfect. The degree of imperfection is a function of the signal's duration. The shorter the duration, the geater the imperfection. Conversely, as the signal duration lengthens, the more rapidly the imperfection shrinks to insignificance.
Signal duration caused imperfection manifests as a stretching/spreading/smearing of an transient's duration, and ALL real signals are transient to some degree or another (meaning, they begin and, eventually, end). The imperfection is negligible for many DSP applications, but not for all. This is what Meridian is primarily focused on with MQA. They believe that transient spreading is not an negligible effect with digital music application. I don't know whether that's corrct or not, but the reality of the transient duration smearing mechanism is easy enough to understand.
The trouble with what is shown is that it does not deal with anything but full level signals that are fully filtered. However, if you look at a 1K signal at -90db then you see 3 clear stair steps at 16 bits (Stereophile has been showing this on their measurements for years and years). If you raise the bit level then the stair steps at -90 start to disappear. If you looked at a 16K sign wave at -90 I don't think it would exist. Not enough bits.....this is why 24 bits is necessary....to get the low level signals and especially at higher frequencies.
Even more interesting is that if you use no analog filter and only a 8 times digital filter then you will see stairsteps even at 1K. He is showing a fully digital and analog filtered signal. If you use no digital or analog filter then it really looks horrible.
So...his information is useless except for nerdism. This is why digital has taken us so long to get right....guys like this think it is all ones and zeros........sorry, wrong. Every single thing you do to a digital signal anywhere along its path messes with the SOUND...this is the truth!
There is no myth....there is mystery. And you will only find the magic if you are adventurous. Just now put a piece of constrained layer damping material on the coax digital input jack inside my modified Gustard DAC.....OMG! what a revelation....way more real sound! Let me see you measure that! I can hear it....that is what matters! You can hear it too!
"If you raise the bit level then the stair steps at -90 start to disappear. If you looked at a 16K sign wave at -90 I don't think it would exist. Not enough bits.....this is why 24 bits is necessary....to get the low level signals and especially at higher frequencies."
Just out of interest, how do you reconcile that statement with DSD? Particularly those using 1 bit ADCs?
"Beauty is Truth, Truth Beauty.." Keats
" looked at a 16K sign wave at -90 I don't think it would exist. Not enough bits.....this is why 24 bits is necessary."
I have sympathy for your general argument but a 16 bit system would allow a -90dB 16K sine wave signal to be reproduced with a little to spare (it has a dynamic range of 96dB), at least it would still exist, this being your point. However that is a text book kind of argument and assumes a perfect 16 bit device. In practice as you have realised it isn't quite that straightforward. However 24 bit systems are way over the top for music reproduction and, as the engineers at Decca pointed out way back in the 1980s, a 20 bit system is perfectly adequate for all of our needs.
BTW, a processor that can deal with a 24 bit word length is not the same as an audio system that can utilise the consequences i.e. to fully exploit 24 bits means a system with a S/N ratio of 144dB. And of course listening in a room with ambient noise at -144dB. Even if you had this the changes to sound at lower levels would still be inaudible as they would be masked by the sound of your own blood flow through your ears.
Digital audio has brought us another "specmanship" race. In the 1970s it was based on ever diminishing distortion figures ( Your amp has 0.0001% distortion, ours is better as it has 0.00001%), now it's word length and sample rates. When I read of 32bit/768Ks/s for audio I laugh. No, cry.
I completely disagree with your thinking that 20 bits is enough. It is the low level signals that are improved the most by more bits. If you look at a -110db 10K signal at 20 bits then it looks no where as good as with 24 bits.
The signal to noise ratio of a system is just its maximum dynamic range.....what we need is pure signal up to the maximum dynamic range of the device. If a device has pure signal at -120db it will sound way better than a device with -130db that is not pure. Pure is not about numbers.....this is where you must listen. Even a single "bad sounding" resistor will wreck the sound of a great signal and yet it has no measurements. This is why some DACs with no filtering "sound better" than those with lots of filtering and great signal to noise ratio measurements........purity is not a number.....but you ear can hear it! Of course, having great measurements and super pure sound are the ultimate....someday we will get there.
The funny thing is this:
"If you use no digital or analog filter then it really looks horrible."
Audio Note is often considered to be the finest digital replay on the market and their DACs are no times oversampling, no digital or analog filters. Martin Colloms who has been reviewing CD state of the Art for decades hailed this as by far the best sounding CD player out there. No filters, no error correction.
The engineering discussions while mildly interesting to me - are really relevant After the audition. If something sounds utterly wonderful and the next one sounds utterly terrible - the engineer to pay attention to is the former not the latter.
Seems every post you make has some AN shill piece in it.
If you have another example of a CD player that uses no digital or analog filters I can probably use that as my example.
This would seem a relevant company to use as an example of the antithesis to the OP's provided youtube video. A CD player that is a virtual opposite in design and yet yields some of the highest regarded (sounding) players on the market.
Not recommending it in this thread - just pointing out a technology that runs completely counter to the video and yields very highly regarded listening results.
I recommend what I would would and do buy. My plan is to buy the $10,000 Audio Note DAC 3.1Xii (bad measurements and all).
For years now, ever since you were Awestruck (sold AN religion ?) by the Twits at that hole in the wall 'Audiohound' shoppe.
You've been Pushing AN 'stuff'.
Which is imo ok..no more/no less, but nowhere near Your estimations
EXPAND your experiences /horizons ?
Enough with the AN shilling.
It discredits your self appointed 'reviews'
No need to get testy - it's just audio. Do what I advise others to do - don't read my posts since chances are I will be recommending what I put my money on.
Typing this as I sit in Front of Line Magnetic (X 3 components), Marantz, Sony and KEF.
Post a Followup:
Post a Message!
This post is made possible by the generous support of people like you and our sponsors: