Does DSD sound better than PCM?

[Miska]

So you listen at home at 110dB? Really? Well you won't be hearing 20kHz for very much longer.

That is about exposure times. Professional musicians tend to have challenge about that though.

http://www.nehearingandspeech.org/audio-exposure-times-how-loud-and-how-long/
http://www.sengpielaudio.com/PermissibleExposureTime.htm

I'm fine with 1 or two seconds of exposure about once per year.

It's not about doing research its about understanding relevant information to what your doing. You know, things like the actual real world ability of people to hear above 20kHz

It-is-not-as-simple-as-that.

But anyway, if you have something to talk about converter design that's fine. But i don't have more to say perception research since I don't have any new information to offer in that area.

If you think my work doesn't matter I'm fine with that too. For me it is enough that it matters to me.

 

[March Audio]

That is about exposure times. Professional musicians tend to have challenge about that though.

http://www.nehearingandspeech.org/audio-exposure-times-how-loud-and-how-long/
http://www.sengpielaudio.com/PermissibleExposureTime.htm

I'm fine with 1 or two seconds of exposure about once per year.



It-is-not-as-simple-as-that.

But anyway, if you have something to talk about converter design that's fine. But i don't have more to say perception research since I don't have any new information to offer in that area.

If you think my work doesn't matter I'm fine with that too. For me it is enough that it matters to me.

I'm a qualified HSE noise officer so yes I know about exposure times. I have performed many noise surveys and modelling of complex noise environments, mitigations etc.

You have completely missed the rather obvious point, or you have dodged it, not sure. So you need 110dB to hear your castenets HF output above 20kHz. Are you now going to listen to your Flamenco album at 110dB? That will be more than 2 seconds.

Of course you won't. So how are you going to hear the >20kHz sound?

If it's not that simple please explain.

 

[Miska]

I had to rethink my definition of ADC/DAC since to the audio world e.g. a DAC is the DAC and all the support circuitry around it.

That is fairly recent development, mainly cost-driven to put all DSP along with the converter itself on a single chip. ESS takes this even further by also integrating things like S/PDIF receiver.

If you remember earlier R2R days, converters and digital filters were separate chips... But the days of laser-trimmed R2R chips are long gone.

Now there's again interesting trend of more discrete designs appear, and SDM DACs are actually very suitable for discrete design. Now there's Holo Audio, Denafrips, Playback Designs, EMMLabs/Meitner, dCS and Chord doing such DACs, to name a few. And if I'd have to name a closest competitor to me, it would be Mola-Mola. That is closest in similarity to what I call a "software defined DAC" (a little bit like there has been longer a "software defined radio" you may know about).

However, denigrating what the vast majority of the designers do to create commercial products seems in poor form. There are aspects and facets of board design every bit as challenging as transistor physics, transistor-level design, algorithm design, etc. I continually try to stop myself from thinking my part of the whole is so greatly harder and more important than another part.

Sure, no doubt about that, but that is just outside of my scope. My scope is more similar to those who design chips at ESS or AKM for example.

By the way, those board-level changes and many other things you are talking about have impact much lower in level and higher in frequency than what is being claimed to be audible limit here. And would be such much more hidden by masking too.

There is just common mix-up for meaning of "DAC". For some "DAC" means the COTS chip or the discrete process. For others "DAC" means the end-user product.

 

[Miska]

If it's not that simple please explain.

Since you claim to know human hearing so well, I'd rather ask you to explain the heard differences from perceptual point of view.

I'm trying to explain possible correlations between heard differences and measured results that I've found to match quite well.

We can of course fall back to the age old mantra that measurements just simply don't tell it all. Without even trying to explain any correlation.

Then of course there is the effect of masking.

For those about masking, here's an example I created earlier for similar discussion. File has -6 dBFS white noise with 250 Hz sine mixed into at -26 dBFS level (IIRC). No problem hearing the sine though:
https://www.sonarnerd.net/noise/noise250.flac
With some practice no problem hearing -36 dB or even further into the noise. This is not so far from some of the stuff I was teaching passive sonar operators to do at the Navy.

 

[hetzer]

Excuse me. Am I in a right forum? I thought that this forum is about DSD and PCM. I can't even write any opinion because I need to read several pages of debate to catch the flow.

 

[Roen]

Excuse me. Am I in a right forum? I thought that this forum is about DSD and PCM. I can't even write any opinion because I need to read several pages of debate to catch the flow.

Yeah, you're in the right place. Grab some popcorn.

 

[solderdude]

I'm also primarily a headphone listener. With some 90% of listening through Sennheiser HD800.

I meant mobile phones.. but since were on the subject I too do most of my listening on (EQ'ed) HD800's.

Now recent increase in high-quality headphone offerings and portable DACs is very good and welcome indeed development in this area.

As these are complete devices (DAC and amp in one) do you see any objective improvements here ?
I'll elaborate.. most of these stand alone devices have popular DAC chips and play 16/44 files (most even MP3 which is filtered below 20kHz so there won't be much to complain about aliasing in the 20-24kHz band)
As these devices mostly use rather wideband opamps as an output do you believe that HF garbage from the DAC chips can 'upset' the internal opamp output devices ?

Instead of waving hands about perception research which is still changing all the time, how about just listening? ...You can then decide for yourself if it matters to you and whether it is worth the investment. You cannot condense all aspects of hearing into couple of parameters.

Nope that last bit was what I meant when you replied to me that it could be simulated. Hearing is too complex to simulate.
And yes, I am not a measurement guy (at least not DAC's) and use my ears combined with measurements.
Have fallen way too often for 'sighted' tests so when I really need to know I do it blind.
But I do use my (oldish) ears and decide for myself.
Which is the only correct way IMO


While we are on topic... does DSD sound better to you than PCM ?

 

[RayDunzl]

For those about masking, here's an example I created earlier for similar discussion. File has -6 dBFS white noise with 250 Hz sine mixed into at -26 dBFS level (IIRC). No problem hearing the sine though:
https://www.sonarnerd.net/noise/noise250.flac

That's sort of a parlour trick... Not that I don't like it.

I suppose you can hear the 250Hz tone because the level of the surrounding "tones" making up the -6dBfs white noise are 20dB (maybe) below the 250Hz sine:

 

[Miska]

Nope that last bit was what I meant when you replied to me that it could be simulated. Hearing is too complex to simulate.

Generating simulated test case signals and listening those is not about simulating hearing.

But I do use my (oldish) ears and decide for myself.
Which is the only correct way IMO

Yes, certainly.

While we are on topic... does DSD sound better to you than PCM ?

In many cases yes. But that is a bit broad question. With what source, what oversampling filter (if applicable) and modulator, to what destination? How about PCM, what source, what decimation and oversampling filters (if applicable) to what destination?

Otherwise it is like asking which one is better, white wine or red wine.

 

[Miska]

I suppose you can hear the 250Hz tone because the level of the surrounding "tones" making up the -6dBfs white noise are 20dB (maybe) below the 250Hz sine:

Yes, except they are not any specific number lower, it depends on bandwidth you look at a random process like white noise. You can keep making the FFT longer and the noise keeps going down while the discrete tone stays at same level. If you sum up all those FFT bins you again end up at -6 dB level.

Hearing has similar kind of effect with different kind of split into sub-bands, where the bands also partially overlap and so on.

I could make this harder to look at computer by splitting the sine into very short bursts. Then it is even easier to hear instead of stationary tone. But for FFT it becomes more difficult because it begins to disappear into the noise because it's length is notably shorter than length of the FFT needed to dig it out.

 

[Sal1950]

Otherwise it is like asking YOU which one is better, white wine or red wine.

A pretty basic question, no?

 

[Miska]

A pretty basic question, no?

It's not! I like both, but for my whitefish meal I'd rather have dry Riesling than Pinot Noir.

 

[Sal1950]

It's not! I like both, but for my whitefish meal I'd rather have dry Riesling than Pinot Noir.

Fancy dancy,
For me just give me a good glass of home-made Italian red and it's a done deal.

 

[Gomjab]

That's sort of a parlour trick... Not that I don't like it.

I suppose you can hear the 250Hz tone because the level of the surrounding "tones" making up the -6dBfs white noise are 20dB (maybe) below the 250Hz sine:

You can find old ham radio operators that can pull very weak CW signals out from the noise as well.

The question is would it be easier with DSD or PCM?

 

[Sal1950]

CW signals

The original Digital stream
Data rate was up to the sender and receiver.

 

[solderdude]

Generating simulated test case signals and listening those is not about simulating hearing.

Your answer was too short for me to extract this information.

In many cases yes.

This is what the thread is basically about.
I know it is not that cut and dry though.

 

[March Audio]

Since you claim to know human hearing so well, I'd rather ask you to explain the heard differences from perceptual point of view.

I'm trying to explain possible correlations between heard differences and measured results that I've found to match quite well.

We can of course fall back to the age old mantra that measurements just simply don't tell it all. Without even trying to explain any correlation.



For those about masking, here's an example I created earlier for similar discussion. File has -6 dBFS white noise with 250 Hz sine mixed into at -26 dBFS level (IIRC). No problem hearing the sine though:
https://www.sonarnerd.net/noise/noise250.flac
With some practice no problem hearing -36 dB or even further into the noise. This is not so far from some of the stuff I was teaching passive sonar operators to do at the Navy.

Well that comes across as a bit of a dodge

It's really not about what I know, it's about the decades of well established, repeatable and scientifically accepted research that contradicts what you are saying.

"you've found" doesn't really cut it. BTW I'm very, very much a measurist. I still live in the real world however and accept that there are limits to hearing, physically and perceptually

Erm.... try masking with music. So Mp3 doesn't work? Are you really saying masking has no effect? If so that's another contradiction of well established and accepted research, and with mp3 a real world example of it in action.

https://en.m.wikipedia.org/wiki/Auditory_masking

So, are you listening to flamenco albums at 110dB?

 

[March Audio]

That's sort of a parlour trick... Not that I don't like it.

I suppose you can hear the 250Hz tone because the level of the surrounding "tones" making up the -6dBfs white noise are 20dB (maybe) below the 250Hz sine:

View attachment 22674

Precisely. another contrived situation.

 

[Rock Rabbit]

We were now talking about frequencies between 20.0 kHz and 24.1 kHz where you don't need as much level. Then you can factor in potential 20 dB boost from metal dome tweeter resonance there too. (I personally cannot stand sound of metal domes, so I don't use such speakers)

Speaking of SPL, I was yesterday at Steven Wilson's concert, not very close to the stage 3/4ths to the back from main speakers and I measured SPL around 110 dB(A). And that is certainly not among loudest concerts. And yes, I was wearing ear plugs...

But in real world, castanets can produce SPL around 105 dB at player's distance (just measured) and given 20 dB difference to said frequency band that makes around 85 dB SPL. Now playing at realistic level and factor in 20 dB resonance boost are you are again at 105 dB.

Some overall SPL levels:
Claves: 100 dB SPL
Wood block: 100 dB SPL
Maracas: 90 dB SPL

With impulse SPL level mode I measured about 3 dB more. Maximum I got for castanets was 108 dB and for claves 107 dB. Reported maximum short term peak 130.8.



I would consider potential music listeners to be younger. I was audiophile already around age of 12.

• The same ASA study says that the threshold rises rapidly over 15 or 20 kHz, in any case you don’t need less at 20 kHz than at 28 kHz. The cited text doesn’t talk about a threshold curve over 20 kHz, cause obviously there’s no curve with that small sample. You could not generalize a fluke of three ears, much less without a statistical significant study.
• The ear isn’t a low pass filter at high frequency, over the high frequency limit the response is zero! And the limit is different per ear. In reality we are talking here of “detection” not real hearing because of the very high threshold.
• 110 dBA live!, that’s the correct way “A” weighted to model the AVERAGE ear curve with -10 dB @ 20 kHz and not more high (obviously the electret could go on).

Castanets, and many percussion instruments are a real nightmare of transient SPL. But we’re are not the player or recording engineer. I suppose we’re here for the pleasure and interest in sound reproduction at a “healthy” level, we are in command of the volume knob, so the original SPL is not an issue.

In fact, any CD demo with percussive sound samples comes with clear warnings for your speakers and ears.
Certainly a golden ear could potentially find some problems near Nyquist filter, but at least a well engineered CD with percussion has -100 dB at 20 kHz (I tested a castanets sample).
The very real problem with that kind of sound is the loudspeaker crossover in the kHz range, that’s the big problem not the statistically improbable issue over 20 kHz, and then only if the percussion is recorded in near field and mixed without attenuation.
As you see audio problems have a hierarchical order, that hierarchy is managed by your absolutely non linear ear that evolved to interact with nature, not to be a very linear high frequency transducer.
So please be realistic and put the problems of sound reproduction in relation to the average ear, very sensitive in 1-4 kHz and probably with capacity to reach 16-20 kHz (detect a tone).
We all gonna learn something in that way

 

[finneybear]

And if I'd have to name a closest competitor to me, it would be Mola-Mola. That is closest in similarity to what I call a "software defined DAC" (a little bit like there has been longer a "software defined radio" you may know about).

No, no, Mola-Mola is basically a 128 node FIR design. The DSP part is not as sophisticated and flexible as HQPlayers.
DSP is the central part of modern delta-sigma based DACs, moving the DSP to external hardware/computer is the right direction.
Whether it's online or offline, that's another matter.

I am with you, Miska.