DSD? Any point to it at all?

Audio Gear Talk
81

The processing delay is less than one sample at 44.1k. And PCM noise-shapers add much more delay.

For realtime playback HQPlayer has much more playback output buffer anyway, regardless of output format. All processing runs over one second ahead of the time in any case.

And if you are doing mixing and mastering, the lag from the record label to your HDD/SSD is much more.

P.S. Your SDM DAC will add much more delay with PCM input than this chain and direct SDM playback. Remember that it’s oversampling filters will have delay plus it’s modulator will have delay too. And it cannot have much ahead of time, because it is synchronous process, unlike this which is asynchronous running at GHz speeds.

82

I know I am definitely not in the “real-world”. :laughing:

83

But if you want to do DSP volume, don’t you need to do it on the endpoint to avoid the delay due to network buffering? Can a Pi do it? Also, why does the volume in HQP have such a big lag?

84

Exactly. If I upsample it myself I can do it a dozen times with different parameters. I suppose if I turn some knobs to 11 it might even produce a (barely) audible difference :slight_smile:

For, let’s say, PCM sampled at 48+ would there be any inaccuracy in the audible band? Sure, having perfect reconstruction up to 40KHz would be great, technically, but unless I am bitten by a radioactive bat it is of purely academic interest.

85

Can one access those measurements somewhere, to see the difference all that wizardry makes, with line level signals at least?
I have a really hard time digesting all the hyperbole and confidence in posts about audibly improved sound, especially when considering what can be measured acoustically after the amp-speaker-room chain, no matter how much money one throws at it!

EDIT:
I’d also love to see your reference system’s acoustic response in the frequency and time domains - you can see mine here and here.

86

There is no reason for DSD to just listen to music.
But that is also true for 16/44 FLAC or 320 MP3.
Just listen to your favorite (digital) radio station, they broadcast in 128 kbs mp3. What more do somebody need?

3 Likes
87

These aren’t reasons for listening, but rather for recording, and as we’ve already discussed DSD is only a small percentage of available recordings. This suggests PCM is far from irrelevant.

Incidentally, the thread is about DSD not HQP.

Oh, and vinyl is more about collecting and lifestyle, especially for a generation that grew up on Spotify. Having a tangible “record” of their musical journey is important; arguably it’s autobiographical, and this is something you can’t do that with streaming.

2 Likes
88

When I’m using HQP I do not expect better amp-speaker-room since I think HQP plays it’s part before DAC. (and the impact is much smaller and maybe better felt with headphones)

I love your speakers and the response (I wish one day for them) and share your curiosity on measurements from DAC outputs (I think jussi posts a lot of those in different threads)

89

Because it happens as part of all the other DSP and after that there’s the output FIFO and hardware buffers. Volume is also smoothly ramped, it doesn’t just jump from one value to another one.

NAA will never do any DSP, it is jus moving data around, nothing else, on purpose.

You can halve the FIFO lag by using the “short buffer” option. But that increases risk of drop-outs.

90

Yes, in many cases how the DAC is driven affects the performance throughout.

I sometimes post some of the results. But quite a bit less recently. These kind of threads already take too much valuable time.

My main reference / development system is T+A HA 200 + Solitaire P headphones, corrected with Oratory’s EQ. The other one is Holo Spring 3 DAC + Ferrum Oor + Hypsos + Sennheiser HD800 also corrected with Oratory’s EQ.

91

Why the hostility toward the one person that really knows what he talking about when it comes to DSD?? The vast majority of his posts in this thread are about DSD specifically. When does bring up HQPlayer, it is in reference to how DSD functions in the real world in a real world product talking to real world DACs.

4 Likes
92

There is zero hostility. I’m simply reminding everyone that the subject of the thread is DSD. I and others welcome @jussi_laako’s input, and I completely understand that he is answering some questions, but a detour into HQP capabilities isn’t entirely relevant to the topic.

95

But wouldn’t that be a problem with a specific DAC implementation, not PCM vs DSD specifically? Not like there weren’t tons of native DSD DACs that measure like crap either…

1 Like
97

Minor quibble. Octave Records does state it for all their albums

99

So, in scrabbling around the Web yesterday, I came across this essay on the Mojo Audio site by one Benjamin Zwickel, which seems to be consistent and informative. I’m not going to buy one of his $10K R2R DACs (see the conversation going on over at Stereophile), but the discussion he presents seems like a good one. I now better understand some of what Jussi is saying.

Does he get anything wrong?

100

Yes. I can compile a list if there’s interest.

1 Like
101

You have to be very careful of people who clearly want to sell you something. Especially multi-thousands-of-dollars “audiophile” gadgets. And he starts with getting some basic math wrong.

Not surprisingly, Mojo DACs (not to be confused with Chord Mojo ,which is ugly and overpriced but works quite well) measure like crap.

103

As we see from the Stereophile discussion.

104

Sure, have at it! Here I thought I was beginning to understand this…

105

Ok, here it goes…

If the SACD format was DSD128 instead of DSD64 and 5-bits instead of 1-bit it would have made a huge difference in performance.

Long before the DVD, SACD, or DSD formats were developed, the Bit Stream DAC chip was introduced to the consumer market as a lower-cost alternative to the significantly more expensive R-2R multi-bit DAC chip. Bit Stream DAC chips have built-in algorithms to convert PCM input to DSD, which is then converted to analog. Once again, the result was a huge cost saving at the expense of fidelity.

Just a couple of examples of unsubstantiated remarks about performance differences.

So a 16-bit 44.1KHz Red Book CD has 28,901,376 sampling points each second (44,100 x 65,536). And a 24-bit 192KHz recording has 32,212,254,000,000 sampling points each second (192,000 x 16,777,216). This means 24-bit 192KHz recordings have over 111,455 times the theoretical resolution of a 16-bit 44.1KHz recording. No small difference.

The number of “sampling points” is not relevant in any way. It’s audio, not video.

Of course when studios convert a 48KHz multiple format to a 44.1KHz multiple format or visa versa they introduce quantization errors.

That’s not how quantization errors are introduced. They are introduced whenever you reduce the bit depth, i.e. convert from 32- or 64-bit (fixed point, floating point or integer) to 24-bit integer or 16-bit integer, or from 24-bit integer to 16-bit integer. That happens during much simpler DSP than sample rate conversion, i.e. even during gain adjustments. Also, even if you convert from a multiple of a base frequency to another multiple of the same frequency, the new samples have to be quantized, thus introducing quantization errors.

I should add that converting from a multiple of 44.1kHz to a multiple of 48kHz and vice-versa does not add any more errors compared to converting between multiples of the same frequency, and the computational effort depends only on the frequency ratio. For example, converting from 44.1kHz to 96kHz is only about 9% more demanding than converting from 44.1kHz to 88.2kHz, all else being equal.

For example, when Sony decided to archive their analog master libraries to DSD64 back in 1995, they were wrong to believe that these masters would be future-proof and able to reproduce any consumer format. The fact is, these masters could only properly reproduce a format that was divisible by 44.1KHz. So any modern 96KHz or 192KHz recording created from DSD64 master files have quantization errors.

Another example that shows misunderstanding of both quantization errors and rate conversions.

But because naive consumers wrongly believe that the higher the sampling rate the higher the fidelity they demand 192Khz falsely believing it is better than 176.4KHz, so that is what record companies market.

Actually this is something I agree with :slight_smile: I would say consumers wrongly believe anything over 44.1kHz is better.

Since quantization noise is present around the sampling frequency of a PCM recording, a 44.1KHz recording has quantization noise one octave above the human hearing limit of 20KHz.

This one’s really off. On the one hand, 44.1kHz recordings have a bandwidth of less than 22.05kHz. Then, raw quantization noise (i.e. in the absence of dithering and noise shaping) is distributed throughout the entire frequency band. That is, for 44.1kHz recordings, quantization noise spans the whole 0-22.05kHz band. It continues like this:

Because the quantization noise is only one octave above audibility the filters used have a very steep slope so as to not filter out desirable high frequencies.

It’s actually worse than that: those filters have to go from unit attenuation to zero attenuation between 20kHz and 22.05kHz. That’s a lot less than one octave. And it’s not because quantization errors, it’s because Nyquist.

Sorry to be the one to burst your bubble, but despite what many audiophiles may believe, less than one person in a thousand can hear anything above 20KHz as a child and there is almost no one over the age of 40 who can hear much above 15KHz.

This one is spot on. It’s not all bad…

3 Likes