Does DSD sound better than PCM?

[mansr]

All SDM DACs have noise from the modulator output... Then it is the actual conversion section that is supposed to remove it.

Yes, just as it's supposed to remove image frequencies. Let's look another DAC instead.

PCM 88.2 kHz:

DSD:

Which do you think is more likely to cause issues downstream?

 

[Miska]

Yes, just as it's supposed to remove image frequencies. Let's look another DAC instead.

DSD:

Which do you think is more likely to cause issues downstream?

Looks like something running at DSD64?

By the way, please show plots with 44.1k input, since something like 99% of content is RedBook.

 

[Miska]

From the old days, a CS4328 which one was one of the well known native 1-bit DAC chips running at 2.8 MHz, IOW a "DSD DAC".

No noise bumps here, but you can see a peak around 2.8 MHz switching rate and some other spurious tones. This was supposed to have a "built-in reconstruction filter". But example that even for DSD64 you can optimize DAC design not to have noise bumps. The analog filter design didn't have provision for "hires" content though, this chip was strictly 44.1/48 kHz PCM input. This could have benefited from a good external secondary analog filter though.

Kind of reminds me of my recent measurement of much more recent Focusrite Forte USB converter with 44.1k PCM input:

That hump at 2.8 MHz is actually quite dependent on the input, it is much narrower with 1 kHz input tone. You can also see "multibit modulator" noise hump reaching peak at around 200 kHz.

 

[mansr]

Looks like something running at DSD64?

By the way, please show plots with 44.1k input,

Ever heard of software resampling?

since something like 99% of content is RedBook.

By the same token, you should be using DSD64 for your "DSD goodness" plots since 99% of DSD content is that.

 

[Veri]

By the same token, you should be using DSD64 for your "DSD goodness" plots since 99% of DSD content is that.

I think the entire point of DSD256/DSD512 compatibility is to sample it yourself by software really. Next to no sources at those rates. Personally I wouldn't bother but there's a literal horde of people proclaiming it the next best thing since oversampling.

99% of DSD is DSD64, yes. A pity / failure of early DSD for sure

 

[Miska]

Ever heard of software resampling?


By the same token, you should be using DSD64 for your "DSD goodness" plots since 99% of DSD content is that.

Running DAC natively at DSD64 is almost like running a PCM NOS DAC at 44.1k. Both need oversampling. However, it is not so balanced because when you input 44.1k PCM to a DAC chip it typically then does it's own rate conversion. While when you input DSD64 it doesn't...

So either we use it for both or neither.

If we would be to compare DSD64, then we should also look for example ADI-2 when it's configured to also not use it's digital filters (NOS mode) when we input 44.1k PCM:

Now again, which one causes more problems upstream, PCM or DSD?

Or the iFi iDSD Micro for that matter:

 

[mansr]

I think the entire point of DSD256/DSD512 compatibility is to sample it yourself by software really. Next to no sources at those rates. Personally I wouldn't bother but there's a literal horde of people proclaiming it the next best thing since oversampling.

Oh sure, but then we should be comparing it to the best performing PCM rate. Miska has a habit of rigging his tests in favour of DSD. It couldn't possibly have anything to do with product sales, could it?

 

[Veri]

Oh sure, but then we should be comparing it to the best performing PCM rate. Miska has a habit of rigging his tests in favour of DSD. It couldn't possibly have anything to do with product sales, could it?

Lol I see your point I think so far in this thread his comparisons are quite fair though.
@Miska that RME ADI-2 sure beats the iFi in cleanliness/noise spikes!

 

[Miska]

I think the entire point of DSD256/DSD512 compatibility is to sample it yourself by software really. Next to no sources at those rates. Personally I wouldn't bother but there's a literal horde of people proclaiming it the next best thing since oversampling.

It is very handy for working around shortcomings in those resource constrained DAC chips that have pathetic digital filters and poor modulators. For SDM DAC, the modulator behavior and reconstruction filters are one of the defining factors how it sounds. Plus the conversion stage architecture.

But one cannot really expect that much from a DSP that sits on a $10 chip in close proximity to very sensitive analog parts and usually has other constraints such as that it cannot require heat sinks or active cooling.

I'd rather see DAC chips as they used to be in the R2R days. Dedicated chip that does only one thing as well as possible - convert digital samples to analog, as-is. Without DSP. The DSP can be then external, like it used to be in the old days. Like SAA7220 + TDA1541A. But the current trend is just price (BOM) driven.

If there would be standard to send multi-level SDM to a DAC, I would by all mean utilize that. Such just doesn't exist (yet).

99% of DSD is DSD64, yes. A pity / failure of early DSD for sure

It is as much failure as RedBook. You need to do upsampling for both to help the reconstruction.

 

[Miska]

Oh sure, but then we should be comparing it to the best performing PCM rate. Miska has a habit of rigging his tests in favour of DSD. It couldn't possibly have anything to do with product sales, could it?

That's why I posted the best performing PCM rate for ADI-2 as well. Do you really think I care if people use my software to upsample PCM to PCM or PCM to DSD?

Main point is that DAC chips fall short on digital filters sooner or later. With DSD output, I can run proper filters to 512x rate.

 

[Miska]

@Miska that RME ADI-2 sure beats the iFi in cleanliness/noise spikes!

It is extremely nice device, I have two of those. Also because of the 768k PCM and DSD256 recording capabilities.

And it allows control over the AKM DAC chip settings, such as DSD Direct mode and different PCM filters. TEAC NT-503 is same DAC chip, also runs in DSD Direct mode, but is not as clean and doesn't have as many configuration options. Only annoying part is that the internal headphone amp cannot be used with DSD Direct mode, because only volume control the device has is the digital volume on DAC chip. And that is bypassed on DSD Direct mode. So the main output works and the other headphone socket is muted in DSD Direct mode. The other headphone socket is active, but it is never in DSD Direct mode.

 

[Miska]

And about the claims regarding my tests. I test two cases:

1) Bit-perfect 44.1k PCM input to the DAC without any external processing see how it works when used in "traditional way"
2) Upsampled 44.1k PCM input to the DAC to find settings that make the DAC perform best (in my opinion), sometimes it is PCM (Chord, Devialet to name few) and sometimes it is DSD

I don't pay so much attention to hires content, PCM or DSD, because it is fairly small percentage of all the content. So I focus primarily on RedBook. Sure, in my opinion there's no question if DSD256 hires recordings are great played bit-perfect. If all recordings would be DSD256 there wouldn't be nearly as much use for my software.

But could we please focus on technical aspects instead of questioning someone's motives or making claims about persons in general? That would be great.

P.S. My software started many years before there was a single DSD DAC on the market. My first upsampling was running to pro-audio DACs of the day at 96/24 PCM.

 

[Roen]

2) Upsampled 44.1k PCM input to the DAC to find settings that make the DAC perform best (in my opinion)

How do you tell when something is "performing best"?

 

[Deacon Blues]

Miska - if I have a DAC that does 24/196 pcm - does your software upsample it? I guess I don't understand the settings. Thank you.

 

[Miska]

How do you tell when something is "performing best"?

By measuring both audio band and out of band behavior. Traditional measurements on audio band (20 kHz and 100 kHz). And checking that there are no correlated components or discrete spurious tones above audio band, such as images for example.

 

[Miska]

Miska - if I have a DAC that does 24/196 pcm - does your software upsample it? I guess I don't understand the settings. Thank you.

It can convert for example 44.1/16 to 192/24 in that case. Or if you happen to have 352.8/24 (DXD) content or some DSD, you can convert it down to that same 192/24. So essentially any PCM or DSD source to any PCM or DSD target.

 

[Deacon Blues]

It can convert for example 44.1/16 to 192/24 in that case. Or if you happen to have 352.8/24 (DXD) content or some DSD, you can convert it down to that same 192/24. So essentially any PCM or DSD source to any PCM or DSD target.

Can you tell me the best settings to do that? Its kind of a confusing program to me, no offense.

 

[Miska]

Can you tell me the best settings to do that? Its kind of a confusing program to me, no offense.

First in main window select "PCM" as output mode. Then for example poly-sinc-ext2 filter works fine, use TPDF dither and select 192k as output rate. You can check that your "DAC Bits" in settings is set to 24.

You can either use the library, or you can just drag-and-drop content on the HQPlayer window. Or you can play from Roon for example and there is also remote control app for Android (HQPDcontrol).

For every day use, there's also alternative GUI included in the player, you can open it from the toolbar (fullscreen / touch mode).

There's a PDF manual included, on Windows and Linux it can be found from Start-menu and on Mac it is straight in the DMG together with the application.

 

[Zek]

Where can this program be seen? Or I missed something.

 

[Deacon Blues]

Where can this program be seen? Or I missed something.

HQ Player, of course. It's rad, try it out.