How many bits, how fast, just how much resolution is enough?

[semente]

With this image I am triying to describe graphically the advantage of a 24 bit file in regards to resolution as it was explained to me by Barry Diament:

 

- the effective resolution of a 16 bit file at -60dB is 6 bit

 

- the effective resolution of a 24 bit file at -60dB is 14 bit

 

Whether this is audible would depend I guess of quiet your room is, of how loud you are playing, of how "transparent" your speakers are.

 

R

 

 

 

[semente]

The article appears to address that as a misconception?

 

[ATTACH=CONFIG]31053[/ATTACH]

 

"Above: Sampled signals are often depicted as a rough stairstep (red) that seems a poor approximation of the original signal. However, the representation is mathematically exact and the signal recovers the exact smooth shape of the original (blue) when converted back to analog"

If I'm not mistaken the points are joined using some form of interpolation and the more points you have/take the more accurate curve will be reconstructed.

 

 

R

[semente]

See here:

 

https://en.m.wikipedia.org/wiki/Quantization_(signal_processing)

[semente]

You are mistaken. See Nyquist.

 

The bit depth matters, of course. Sample rate does not once it is sufficient for the input signal.

Where did I mention sample rate?

[semente]

Your remark about interpolation using more points implied it.

In the post you quoted the image on the right has the same sample rate but more "amplitude" values.

The increased resolution is obvious.

 

R

[semente]

That image makes little sense alongside your words.

Care to explain?

[semente]

Usually when someone is explaining digital to you and use the staircase waveform, STOP. Stop listening to anything else they have to say.

 

The higher sample rate allows more bandwidth. You can as the video demonstrates with purely analog gear at each end fully reconstruct a waveform up to the limited bandwidth with only two samples per wave. The "extra" samples of a higher sample rate aren't constructing a more resolved or more accurate waveform. Or looked at another way, a high sample rate 1 khz tone is not more accurately reconstructed than a redbook 1 khz tone even though it has more sample points per wave. You couldn't tell the difference between the two. The staircase idea would mislead you into thinking the high rate 1 khz tone is more accurately reconstructed.

 

I didn't use a staircase waveform, there's no waveform in this example only, as far as I know, quantisation error:

 

 

And I don't know why people started talking about sample rate when I was referring to bit depth...

 

R

[semente]

Sample rate and bit-depth is linked values. If bit-depth is constant, but sample rate increased 2 times, then noise floor decreased 2 times.

 

Since I never spoke of sample rate, let's consider 16/44.1 vs. 24/44.1.

 

Doesn't higher bit depth lower the noise-floor?

And if the noise-floor is lower, don't we get increased resolution?

 

I still don't understand why everyone keeps talking about sample rate...

 

R

[semente]

Well in my case because I took my own advice, and stopped paying attention when I saw the stair case looking graph. Also because you mentioned the more points you have taken the more accurate. That describes a higher vs lower sample rate.

 

OOPS, mansr answered in between and gave a much shorter answer.

 

I wrote points when I should have written levels, otherwise there's no mistake in what I wrote (before that).

PCM digitisation has an amplitude and a time scale; more bit levels are also points.

And, yes, dithering masks quantisation errors but 24 bit still has more resolution.

 

Can you hear that?

Like I said, perhaps. If your room is dead quiet, if you play loud enough, with transparent equipment and the right recording it's impossible not but I find it very difficult...

See post #19.

 

R

[semente]

Uh, that's the reason the word cheating is in quotes. Yes, it is introducing low level noise to linearize low level signals to keep from a situation where only one or two of the LSBs is being manipulated by changing the PCM to virtual PWM. But it's a bandaid when compared to actually adding more bits.

 

That's it.

Masking through dither, even when used below the audible threshold, is still a crutch.

 

Alternatively one could always turn down the volume or open the windows to increase the noise in the room.

 

R

[semente]

There is physically no way to correctly compare 16 and 24 bit from technical point of view.

What if we reduce the gain of a recording that maxes at 0dB by say 36dB on both Redbook and 24/44.1?

Won't this give us an idea of 16 bit limitations?

 

Perhaps we could do this with Mario's PlayClassics files...

 

R

[semente]

The way I understand it, theoretically 16-bit / 44.1 kHz may be good enough for domestic playback but in practice:

 

• 16-bit — the resolution is down by 6-bit at -36dBFS and with wide DR musical programme that has not been mastered to top somewhere close to 0dBFS this may be audible

 

• 44.1 kHz — there are audible top end problems due to filtering at 22.05kHz being to close to the audible band and if I am not mistaken this is true whether or not you over- or upsample

 

The first and so far only time I was able to determine differences between 16 and 24 bit and 44.1 and 96kHz was with Mario's files.

I am happy with the sound I get from Redbook and have yet to buy HR files or discs.

 

R

 

 

 

 

 

P.S.: an example (Berlioz: Lélio, ou Le retour à la vie Op. 14b / Radio-Sinfonie-Orchester Frankfurt, Eliahu Inbal / Denon)

 

[semente]

Rob Watts has always said that there is a lot more that 16-bit audio can give.

When you listen to the Hugo, you know exactly what he means. 16-bit files played thorough my Hugo sound better than higher bit-rate files played through my other DACs.

Which leads me to believe that 16-bit audio is 'good enough', and that the variable in all this is the type of decoding.

 

That doesn't make sense...

The question is not whether Redbook with your current DAC sounds better than HR with your previous DACs but how do Redbook and HR compare with your current DAC.

 

R

[semente]

Not so sure about that... the question is whether 16/44 is good enough as source data to properly deliver the content within the hearing range of humans (frequency & dynamic range) on DACs than are designed well enough to let that happen.

 

"How many bits, how fast, just how much resolution is enough?"

 

... if one of his DACs tells him redbook is enough, then his post is right on topic in my book. If his other DACs don't, that's not necessarily redbook's fault.

 

 

---on a another note---

 

If we (CA) agree that freq range 20hz-20khz is being captured and reproduced, then dynamic range shortfall of 16 bit is next to examine, which is fine on paper but not much music contains huge dynamic ranges like that... DR Database for example makes for some depressing reading.

 

 

;-)

 

I understand your comments but I think that if he wishes to know if Redbook is "enough" with his current DAC he should compare it to HR converted by that same DAC and not by another one (which I assume didn't perform as well with either format).

This is what I would do if I wanted my findings to be meaningful.

 

I recommend using Mario/PlayClassics test files.

 

My comments from that thread regarding Redbook vs. HighRes comparisons:

 

Yesterday morning I spent some time listening to the Format Test Files kindly provided by Mario Martinez.

Even though I generally find listening evaluations a humongously tedious affair, I must have been in the right mood and the neighbourhood was very quiet which was also helpful.

 

The proceedings of my listening session were as follows:

 

16/44.1 track

24/44.1 track

16/44.1 track

-

16/44.1 track

16/96 track

16/44.1 track

-

24/96 track

16/44.1 track

24/96 track

 

First of all, it was a very interesting experience to listen to different performers and instruments playing with such a tonal and acoustical seamlessness, almost as if I were listening to a single musical event.

 

Then something happened which really surprised me: I had listened to two of the recordings included in the set a reasonably amount of times before in 24/96 and it didn't take too long to identify the shortcomings of the 16/44.1 version.

 

And once I nailed the characteristics of those shortcomings it was hard to miss them.

I can still enjoy 16/44.1 but I am convinced of it's lower fidelity.

R

[semente]

You also should do the same comparison using HQP and/or AuI to upconvert to the native format of your DAC from these source files.

 

This will test whether the differences you hear are related to the filtering/sampling done by the DAC (vs capabilities of CPU/software)

Indeed.

There are so many variables: upsample to PCM or DSD, to a multiple frequency or maximum DAC sample rate, minimum phase or linear phase filters, SOX vs izotope vs HQPlayer vs Aul vs DAC ASRC, in line or off line, etc.

 

This is what I dislike about computer audio, it's as fussy as vinyl...or more.

 

R

[semente]

@elsdude is correct, if the system is only responsive to 20khz then 44 khz is enough to sample the relevant transients. Conversely, if the system is sensitive to faster transients then it is *by definition* responsive to > 20 khz.

 

I'm not sure the 20khz limit is set in stone. That could be the limit for detecting isolated tones but since the (human auditory) system is highly nonlinear, that doesn't mean the entire system's response is limited to 20khz.

 

Consider the retina: although not sensitive to infra-red, the skin is :-)

 

It is possible that 30 khz frequencies are interpreted as "irritating" or other nonlinearities such as the response of amplifiers/electronics to "ultrasonics". Consider why we need to filter out > 20khz from DAC outputs: the system itself is sensitive to ultrasonics. These ultrasonic effects on electronics (that manifest themselves in the audible range) are examples of nonlinearities in the electronics -- similarly the auditory system itself may have (other) nonlinearities.

 

As they say, when you assume ...

 

In any case we really aren't doing science here so folks should judge for themselves. I find that DSD256 goes well with scotch.

 

Let's not forget the oil can resonance of many hard-domed tweeters which once triggered may very well have repercussions further down in the audible range:

 

 

Focal Aria 936, cumulative spectral-decay plot on tweeter axis at 50" (0.15ms risetime).

Read more at Focal Aria 936 loudspeaker Measurements | Stereophile.com

 

R

[semente]

Certainly it is not. And for 96 kHz PCM you end up doing over 10x decimation factors from the actual AD conversion stage. And again over 10x interpolation factors for the DA conversion stage. Just wasted effort.

 

 

 

Processing 384k, even tens of channels in realtime is completely non-issue these days. Really.

 

 

 

Going down >100 dB in 30 kHz to 48 kHz band is practically brickwall. It is more than 100 dB/oct, that is not "gentle".

 

Would it be possible and what would be the disadvantages of having two "knees" in the filter, say 6dB/Octave at around 20kHz and then 24dB/Octave higher up?

 

R

[semente]

In my view, the problem with using audibility as a gauge is that there are aspects of equipment performance that can be affected by out of audible range noise, such as the tweeter example that I mentioned previously and if I am not mistaken the fact that some amplifiers are negatively affected by high in level supersonic noise.

 

Besides, the cumulative result of a few tiny inaudible improvements may end up being audible.

 

R

[semente]

We can listen and make a subjective determination whether something sounds more real or more appealing, but we can't know if what we are hearing is "accurate" to the recorded event. Yet, people make these accuracy claims all of the time.

 

In my opinion there is no such thing as accuracy to the recorded event; but there can be (a reasonably) accurate reproduction of the recorded signal, also called high fidelity reproduction of recorded music or "garbage in, garbage out".

 

Speakers (and cartridges) are the worst offenders when it comes to (in)accuracy...

 

R

[semente]

Here's what Nelson Pass says:

 

 

Source: https://passlabs.com/press/phase-coherent-crossover-networks

 

Until a few years ago I owned a pair of Pass XVR-1 active analogue crossovers:

 

[ATTACH=CONFIG]31280[/ATTACH]

 

I spent many months playing around with all types of crossovers. But eventually, the only ones I could live with were 1st order.

 

Mani.

 

In a perfect world we would (/should?) all prefer phase and time coherent speakers but the reality is that the shallow slope of a first order filter will affect frequency and power response and is ineffective at tackling break-up resonances (although you may get away with it by narrowing the operating band of the driver):

 

 

Here's an interesting essay on the subject:

 

Phase, Time and Distortion in Loudspeakers

 

R

SaveSave