• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Does DSD sound better than PCM?

Status
Not open for further replies.

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
This is what I would like to see too. Not that it should be on Miska to do it but it would be great if someone would make these measurements. I think both @Miska and @March Audio have good points, but without measurements it seems the conversation can just go around and around.
IM is a real problem. As I mentioned previously amplifiers that don't have appropriate input filtering are susceptible to issues created by out of band signals, not least of which is external RF.

As an example in the UK BBC Radio 4 has a 500kW transmitter broadcasting on 198kHz.
Serving the whole country and coastal waters. You do not want this or other AM transmitters getting into your amp. An amp that is wide open is a bad design. More than 100kHz is certainly asking for trouble.

Tests would be interesting
 
Last edited:

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
Beuna Vista, lots of brass, percussion so not to give an easy ride. HP filtered above 20 kHz. The readings arent FFT based but peaks in the time domain.

-36 to -40dB peak in that range. So nothing anywhere near 0dB. I previously talked about figures of around -60 in the transition band as a rough estimate which is quite believable with different music and the slightly higher frequency.

Beuna Vista.PNG


Copland beginning of El Salon Mexico

-50 to -53dB

copland.PNG


So any ultrasonic images are going to be lower in level than this. Any inter modulation created in an amp is going to be an order of magnitude lower again
 
Last edited:

Miska

Addicted to Fun and Learning
Audio Company
Joined
Feb 20, 2019
Messages
615
Likes
448
@Miska
Why not measure into the GHz range ?
In the many EMC tests I have done often there is a lot of life well above 5MHz.

I measure also higher, but there's usually nothing because the analog reconstruction filter cuts out. Typical analog filter is 2nd order with fc=100kHz so it has rolled off dirty output by 5 MHz already.

Taking one of the dirties devices I have (in that respect), Focusrite Forte. Almost all the the junk is in first 10 MHz band.
Forte-sweep-100M.png


Most amplifiers do start to roll-off above 100kHz though some do not. granted this is mostly 6dB or 12dB/octave as input filters are generally only 6dB/octave and the amp itself as well.

Same for DACs. I used 4th order (24 dB/octave) for my own design though.

Do you have any practical shots/analyses of certain (preferably often used/well known) amplifiers loaded with actual speakers/headphones that clearly show problematic behavior of amplifiers when fed signals in the audible range and say around 1MHz combined which seem to show that when the 1MHz signal (preferably not a multiple of the test tone) is included or not that the output of the amp differs within the audible range ?

Without starting to write my own story here now, here's one example from other people:
https://www.diyaudio.com/forums/class-d/321632-hypex-ncore-nc400-input-anti-alias-filter.html

Why not multi-tone? Images are correlated multi-tone because they reflect music spectra around multiples of sampling rate.

Preferably with the main frequency say -6dB and the higher frequency at -60dB or so.

Where does this -60 dB come from? It is not necessarily even that much down.

ADI2-sweep-44k1-sd_sharp-wide.png


I mean one can measure well outside of any audible threshold. What one can measure does not need to be audible. Some claim vice versa as well but I am not one of those people.

And I'm not one of those people who make claims about what other people can hear...

If it can be measured - it matters. If it cannot be measured, it still matters, but measurement resolution needs to be improved...
Is this your experience ?
Can you clearly correlate this ?
What amplifier / transducers are you using to listen for this (so I can avoid that one) as you clearly state they MAY create it I assume you own one to test.

One of the things you can simulate...

You seem to suggest that the often reported 'synergy' between DAC's and amps (almost infinite possibilities here) could (maybe) be related to this.

And loudspeakers. Metal dome tweeters combined leaky DACs is good example. Those tweeters have strong (typically at least +20 dB) ultrasonic resonance. Makes me feel like dentist drilling my teeth, or a bit like sound of the ultrasonic teeth cleaner they use (you may know what I'm talking about if you have experienced it).

This should be very possible to capture with an RME after an amp when loaded with a real load.

RME is not very voltage tolerant for such, unlike my measurement gear. :D

I don't like to make positive assumptions about amp behavior, rather negative worst-case scenarios. So I aim for clean DAC output that represents proper reconstruction instead of "half-assed" one.

Do you correlate listening tests with measurements ?
Are those listening tests 'blind'.
It's one thing to measure HF content but another thing to perceive it.

Yes, my own listening tests. Not blind. When I have two algorithms at hand, I have no reason to prefer A over B. I have my favorite algorithms based on listening. Some other people prefer other ones, that's completely OK for me. I offer options to choose from.

What's your current age (may I ask) and what is the max. frequency you can hear and what is the max. bandwidth you need to correlate recorded and actual sound quality ?

No you may not. :D

I don't know why what I hear matters to anybody else except me in first place. I usually don't talk much about my listening experiences unless specifically asked. Sometimes if people ask what are my favorite settings, then I tell my current favorite.

Do you have plots/graphs/examples indicating this within the audible band (preferably using nulling) ?

What I was talking about when this shit storm started, was @Blumlein 88 's test tracks and the oversampling filter in question there.

Original reference track, here you can clearly see the transient hitting Nyquist due to typical "modern" halfband decimation filter used:
Screenshot_2019-02-24_16-46-09.png

This creates aliasing band at the top, inside the transient. So this is the new form of TIM I was referring to.

Then the one that has gone through D/A - A/D loop. Since the DAC seems to have used ESS' "Hybrid" digital filter, it has largely removed that aliasing band and thus cleaned up the crap from the reference track:
Screenshot_2019-02-24_16-46-41.png

Difference in that about 2 kHz wide band is tens of dB's.

At other point in time, on other forum I was referring for example to original CD version of Pink Floyd DSOTM that looks like the lower plot here. While the latest "modern" remaster looks like the upper one.

As I said earlier, this is what I mean the DAC in question is fixing with the choice of digital filter. Some other ESS filter choices are halfband and thus leave that aliasing band intact in the output.

Now what I don't understand is why this particular thing has caused this kind of shit storm towards me. I was just sharing my point of view about this detail based on spending 20+ years designing digital filters.

My products are not involved in this case. They could be and you have choice of both types of digital filter there, just like with ES9038 if the DAC offers option to choose the filter (like for example Pro-Ject Pre Box S2 Digital does). One test these things on their own if they like.
 

svart-hvitt

Major Contributor
Joined
Aug 31, 2017
Messages
2,375
Likes
1,253
That's pretty low. The same argument can apply to miskas product but I wouldn't dream of suggesting it is the case.

It is a solution to a theoretical problem. If you believe it is a problem you can purchase the software. It is of no threat to my product sales. How could it be? You still need a dac to use the software.

I am challenging the propositions that miska is presenting. That's what this forum is about.

We have had this discussion before. You aim to market a product that is pretty good but has no uniqueness. I asked you about «uniqueness» on a previous occasion. So you cater to people’s taste through factors like look and feel. And you didn’t see the point of debating bringing sustainable products to the market.

@Miska ’s product is quite unique, isn’t it? It lets you do all sorts of conversions in software, plus much more. Updates are free of charge. Maybe his perfectionist approach is like calculating pi with 10,000 decimals, instead of using 3.14 as an approximation, or making audio products with an SNR in excess of 100 dB?

FWIW, I have the HQ Player and use it with Roon. So I have tried it. Do I hear a difference? Not so sure about that :)
 

Miska

Addicted to Fun and Learning
Audio Company
Joined
Feb 20, 2019
Messages
615
Likes
448
Beuna Vista, lots of brass, percussion so not to give an easy ride. HP filtered above 20 kHz.

-36 to -40dB peak in that range. So nothing anywhere near 0dB. I previously talked about figures of around -60 in the transition band as a rough estimate which is quite beleivable with different music.

Where are you pulling that 0 dB from?

Even with this small set of material you now checked you quickly found content with -36 dB. That is 16 times higher than your previous guesstimate.

Now you only need to complete scanning all the rest of recordings there are out there in the world...

For example out of my entire record collection, I've found I think two albums where -3 dBFS headroom is not enough to prevent inter-sample overs. So for example such do exist. And my record collection is very small compared to all the recordings that exist in the world.
 

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
Where are you pulling that 0 dB from?

Even with this small set of material you now checked you quickly found content with -36 dB. That is 16 times higher than your previous guesstimate.

Now you only need to complete scanning all the rest of recordings there are out there in the world...

For example out of my entire record collection, I've found I think two albums where -3 dBFS headroom is not enough to prevent inter-sample overs. So for example such do exist. And my record collection is very small compared to all the recordings that exist in the world.


From what you asked Blumlein to test.

-36 was a more severe example. Other tracks I have quickly scanned through are more like -50.

You are reaching now, oh I am sure you might be able to find a part of a track somewhere that can go higher than this, bu that would be entirely missing the point.

As I said above, any ultrasonic images are therefore going to be lower in level than this. Any inter modulation created in an amp from that is going to be an order of magnitude lower again.

I will record the same tracks out to 96kHz bandwidth, HP filter above the audio band and see what the level of the ultrasonic images are.
 

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
We have had this discussion before. You aim to market a product that is pretty good but has no uniqueness. I asked you about «uniqueness» on a previous occasion. So you cater to people’s taste through factors like look and feel. And you didn’t see the point of debating bringing sustainable products to the market.

@Miska ’s product is quite unique, isn’t it? It lets you do all sorts of conversions in software, plus much more. Updates are free of charge. Maybe his perfectionist approach is like calculating pi with 10,000 decimals, instead of using 3.14 as an approximation, or making audio products with an SNR in excess of 100 dB?

FWIW, I have the HQ Player and use it with Roon. So I have tried it. Do I hear a difference? Not so sure about that :)

As I have said, they are not the same product. They simply dont compete. I gain nothing from denigrating it. I am challenging the technical premise which I see as faulty.

So, if you hear no difference, or if being generous, it is not big enough to be readily discernible, what does that indicate to you about its beneficial effects? Has it solved a problem that needed solving? Well if it has you dont seem to be able to hear it.
 

Miska

Addicted to Fun and Learning
Audio Company
Joined
Feb 20, 2019
Messages
615
Likes
448
From what you asked Blumlein to test.

Absolute level doesn't matter, it is the relative level. If you test at 0 dB you don't need to normalize the results in your head. But I'm fine with test at -36 dB too. The relative level between test signal and alias is the same still.

I will record the same tracks out to 96kHz bandwidth, HP filter above the audio band and see what the level of the ultrasonic images are.

What are you trying to demonstrate with that? Now you switched to some other topic entirely. First we were talking about ADC decimation filter aliasing when running at 44.1k rate. Now you are talking about something else.

If you are talking about ultrasonic images, then 96 kHz bandwidth is not going to help you much... Only for checking your digital filter stop-band attenuation. Not attenuation above the frequency where the on-chip digital filter ends (352.8 kHz).
 

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
Absolute level doesn't matter, it is the relative level. If you test at 0 dB you don't need to normalize the results in your head. But I'm fine with test at -36 dB too. The relative level between test signal and alias is the same still.



What are you trying to demonstrate with that? Now you switched to some other topic entirely. First we were talking about ADC decimation filter aliasing when running at 44.1k rate. Now you are talking about something else.

If you are talking about ultrasonic images, then 96 kHz bandwidth is not going to help you much... Only for checking your digital filter stop-band attenuation. Not attenuation above the frequency where the on-chip digital filter ends (352.8 kHz).

This relative level you keep talking about is nonsense. You appear to be saying that the US image level wont reduce with reducing in band signal level.

Same topic. Same points as Archimago made. You have been talking about IM problems created in amps. lets see whats there but not using FFTs which you complain as being inaccurate.

An amplifier that is unfiltered to 352kHz is a bad design. Its actually called a radio. Again you are conflating this US output problem with bad amp design.
 

Miska

Addicted to Fun and Learning
Audio Company
Joined
Feb 20, 2019
Messages
615
Likes
448
As I have said, they are not the same product. They simply dont compete. I gain nothing from denigrating it. I am challenging the technical premise which I see as faulty.

So, if you hear no difference, or if being generous, it is not big enough to be readily discernible, what does that indicate to you about its beneficial effects? Has it solved a problem that needed solving? Well if it has you dont seem to be able to hear it.

Can you elaborate how technical premise of my product is faulty?

Are you saying it doesn't play music? Are you saying it doesn't enable you to play DSD music files to 192/24 DAC output? Are you saying it's multichannel speaker level/delay adjustment doesn't work? Are you saying the convolution engine for digital room correction, 3D headphone processing, etc somehow doesn't work?

Are you also saying that one cannot even hear effect of digital room correction or headphone cross-feed?

Or these are not beneficial effects? DAC plays music alone without a player?
 

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
Can you elaborate how technical premise of my product is faulty?

Are you saying it doesn't play music? Are you saying it doesn't enable you to play DSD music files to 192/24 DAC output? Are you saying it's multichannel speaker level/delay adjustment doesn't work? Are you saying the convolution engine for digital room correction, 3D headphone processing, etc somehow doesn't work?

Are you also saying that one cannot even hear effect of digital room correction or headphone cross-feed?

Or these are not beneficial effects? DAC plays music alone without a player?

I said your premise was faulty - regarding issues of US noise and that of alias.

You are presenting this as a problem that needs to be solved. That implies it has audible benefit. As Archimago has already shown, US image output is not at high levels with real music. Yes you can use test signals to make bad things happen, but that is not what is happening when people play music at home. Proper correlation with audibility (not audiophiles saying they can hear things) would be a good idea.
 
Last edited:

Miska

Addicted to Fun and Learning
Audio Company
Joined
Feb 20, 2019
Messages
615
Likes
448
This relative level you keep talking about is nonsense. You appear to be saying that the US image level wont reduce with reducing in band signal level.

First we were talking about aliasing and now you switched to talk about ultrasonic images? Can you please even stay on topic?

But yes, it reduces in exact same amount as you reduce the base band signal level. You turn down volume 10 dB and both base band and the images go down 10 dB. Their level difference stays the same.

An amplifier that is unfiltered to 352kHz is a bad design. Its actually called a radio.

Hmmh, show me an amp that doesn't have phase error at 20 kHz, and cuts input by >120 dB by 332.8 kHz (where audio band image starts).

Spectral amps go to +-1 dB to 1 MHz and -3 dB at 1.8 MHz.
 

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
First we were talking about aliasing and now you switched to talk about ultrasonic images? Can you please even stay on topic?

But yes, it reduces in exact same amount as you reduce the base band signal level. You turn down volume 10 dB and both base band and the images go down 10 dB. Their level difference stays the same.



Hmmh, show me an amp that doesn't have phase error at 20 kHz, and cuts input by >120 dB by 332.8 kHz (where audio band image starts).

Spectral amps go to +-1 dB to 1 MHz and -3 dB at 1.8 MHz.

Your comments have covered both as has this ongoing discussion.

Then I wouldnt touch a spectral amp with a barge pole. In the real world thats an idiotic design choice.

Show me your data demonstrating the audible effect of the phase error.
 

Miska

Addicted to Fun and Learning
Audio Company
Joined
Feb 20, 2019
Messages
615
Likes
448
I said your premise was faulty regarding issues of US noise.

OK, you switched topic from aliasing to ultrasonic images. Anyway...

That is one reason why all modern DACs are oversampling, digital filters are used to help removing those ultrasonic images. Some do it better, some not so well. But this is quite broadly accepted audio engineering concept. You are trying to tell investigating quality of this is not useful.

Yes, I know some people and audio engineers who believe NOS DACs are sufficient. I'll just put you on that short list I have. I'm not really interested to argue about this. I think we just need to agree that we disagree.

Show me your data demonstrating the audible effect of the phase error.

OK, so now phase response doesn't matter either. In fact, nothing seems to matter.

I'm actually now curious to hear how do you design your products. What are the things that matter and why? Why did you select that particular ESS filter for your DAC?
 

Miska

Addicted to Fun and Learning
Audio Company
Joined
Feb 20, 2019
Messages
615
Likes
448

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
Sorry to switch back to alias. So we see signals around -50dB in the transition band based the examples I showed above. @Blumlein 88 data showed a reduction of -85dB before you hit the audible band (20kHz).

So even if you look at the highest signal of -36dB, the alias are buried in the noise floor. Where is the problem?
 
Status
Not open for further replies.
Top Bottom