Ted explains DSD

DSD is explained by DirectStream lead designer Ted Smith

[CyrilleBoucanogh]

Ted is genius. Always a pleasure listening to him and learning from his deep understanding. God Bless you, Ted !

[ILO]

I like Teds presentation. He explains very well how PCM is different (DAC) and why it may sound different than DSD. I did something like DSD several decades ago for digital voltmeter, which was counting pulses in sampling interval to produce numbers. Clocks and precision of analogue components including op amp are very important. So less components there are is better. If you can avoid PCM part altogether it will increase accuracy of reproduction. But thing is in reality music recording and editing are not done this way.

[HollywoodRecordingStudio]

This is the clearest explanation of DSD that I have ever heard.

[mineralsound]

I'm a proponent of DSD as far as it advances demand for higher-quality recordings and playback systems. But since you can't edit or mix DSD without converting it to PCM, its realistic use is largely limited to single-take stereo pair recordings.

[alexandernazarenko]

amazing explanation! Ted, thank you very much!

[Synth]

Great explanation, thank you very much Ted! That's why many dsd enable DACs sound better in pcm mode.
They cut corners in the implementation of the filtering, then put the sticker with the DSD logo, while giving you a half baked version of it. Shame.

[ItsTheMastering]

What people should be talking about is...are the differences between PCM and DSD audible in such a way that would lead a reasonable person to prefer one over the other? My experience suggests they most definitely are not. If you approach it that way then the differences are irrelevant. They just are not as important as someone with a product or technology to sell you would like you to believe. FUD reigns supreme in the world of audiophilia, and it is that scintilla of doubt that you are not getting "the best" that drives a lot of sales in this space.

[judenihal]

DSD is still digital. Digital audio will never replace the audio coming out of an analog mixer. If someone can invent an analog recorder which captures the entire sound spectrum which outputs exactly how the mixing deck outputs the signal this will be the ultimate replacement for recorded audio!

[napalmhardcore]

I'm no expert (as will soon become apparent, as I only discovered DSD in the last day) but there are certain fundamental points that I don't understand.

With analogue I am aware of the advantages probably better than I can explain them, but basically it is a pretty accurate way of capturing things due to the lack of approximations processing entails. It is very directly influenced by the physical source being captured (sound in this case).

The argument against digital has long been that if you look at digital representations of a wave, it is comprised of steps that have to be averaged out, therefore it is not a true representation of what went in. However that technology has been being refined so that now those steps are incredibly small and being sampled at a much higher rate. Are we not to a point where the digital representation is more accurate than speakers can actually reproduce?

Something that perplexed me even more was reading a review somebody did of a DSD DAC. In this review they talked about converting PCM files so that they may be played back. I thought the point was, the further away from the original source you get, the less real and accurate it is. So if in this scenario the original source was analogue and it has already been converted once to PCM, what is the benefit of converting this to DSD? Surely you are then just reprocessing something that you know to be less accurate than the format you are converting to is capable of. I understand that there are efforts ongoing to have the master tapes from various bands converted directly to DSD. This I can understand, but I don't understand how converting PCM to DSD is beneficial.

Maybe I've just misunderstood the concept.

[Yozhk]

Could somebody please explain this. You need to "noise shape" in order to produce hi-res DSD. You dont need to "noise shape" if you produce hi-res PCM. So what is the point of DSD? In an environment when space for a 192/24, 6ch recordig is no longer an issue - why not just stick to hi-res PCM?

[mcintoshkid]

und where do i get this machine who gives me native pure dsd to my speakers??

[christopherpatefield]

Too complex for me. Does the computer record from the original analogue tapes and make it digital before stamping?

[laika]

So what's the easy practical way to listen to DSD? (As easy and practical as buying a CD/popping such CD in a player and hitting "play").

[okay]

This implementation of DSD reminds me of the spiel of proprietary schemes like MQA (s oil) I could not understand a word of what he said. Everytime someone says something cannot be measured, I am a bit uncomfortable with such assertions.

As we increase sampling rate, high sample rate PCM @ high bit depth, will approximate DSD, and any differences will be negligible (if such differences truly exist).

I had written the paragraph above - simply out of my own laymans reasoning, and lo and behold, I checked the Wikipedia page for DSD, and it has this paragraph included :

"Because of the nature of sigma-delta converters, one cannot make a direct comparison between DSD and PCM. An approximation is possible, though, and would place DSD in some aspects comparable to a PCM format that has a bit depth of 24 bits and a sampling frequency of 88200 Hz.".

Then in another youtube video from PS Audio, their Gus Skinner - Audio/Music Engineer/Producer says he would not wish to have more than about 30 multitracks in a DSD Mixing session, as it would end up adding too much noise. I could imagine a similar issue with high sampling rate PCM projects, where there would be a similar accumulation of any noise above the audible range, i.e noise which we humans would not hear, but which would at the very least compromise total noise.

But this is always going to be an issue with any multitrack system - multiple microphones, and multiple tracks, will increase the noise, for example if one records an orchestra with a single pair of microphones, or records an orchestra with a final mix of many microphones, the multi-track version will have accumulated noise and less dynamic range, from the total number of summed up noise in each track. Not a new problem and somewhat a common issue, so DSD mixing and PCM Mixing are somewhat having similar issues in that regard.

Ultimately these are issues that we will always have and the best solution is not to do all the fancy filtering, but rather to capture as is and play back as is. To improve quality - lets do the obvious 1st, double sampling rate in the Case of DSD such as DSD 128 or in PCM use 24 bit 88 khz sampling as the minimum, and achieve similar results.

MOst people do not have the kind of gear and listening environment that would make any appreciable difference in discerning a difference, so IMHO, just stick to what works for most people - PCM, which is so much easier to work with at the mixing and mastering stages. I doubt if DSD will ever become significant, in pop music, therefore DSD will remain the preserver of esoterics, the snobby ones, most of whom are old enough not to be able to hear any difference anyway, cos by the time you have made enough money to by the esoteric playback speakers, amps, listening room, and DSD recordings that only a few people can appreciate, you are most likely old enough to have lost the ability to hear anything above 14Khz anyway - So why all the bother.

Shakespear said it most elegantly - DSD is simply much ado about nothing. Higher bit rate and higher bit depth PCM is more than good enough - 24 bit @ 88.2 or higher is more than good enough PCM, and so much easier to mix and master than DSD, needing no specialised gear.

[danielraju]

The sheer intuition this gentleman possesses when he talks about how PCM uses dynamic element matching with randomised errors to compensate for any error during the translation process sums up his genius.

So, even though I get the time-invariant waveform in terms of energy to be perfect by patching up the errors, that doesn't necessarily mean it translates well into the energy associated with frequency bands in the s-domain and therefore, the phase of the signal, which is why PCM sounds artificial and sterile comparing to analog or DSD.

[rezathenightfly]

I wish this video was recorded using the "DSD" format. ;)

[stafonvoncamron]

(OK) MP3 = 92k, 128k+ (GOOD) CD / WAV = 16bit 144K (BETTER) DSD / SVCD = 16bit 196k (BEST) PCM / DAT= 16 bit or 24bit 196k or higher.

[luca]

1:09 First, it doesn't matter that the graph looks like a wave form. It isn't something that i have to understand, the computer has to, so it can appear in any way. It's not a plus.
Second, that likeness you're talking about is not even real: you drew the red line, it is not rappresented like that in the stream at all. 1 means the amplitude has increased compared to the previoys cycle, 0 that it has decreased or has remained the same, so there are lots of spikes and therefore lots and lots of noise.
In conclusion, the example is pretty but completely wrong and misleading

[cardtrix]

I think DSD is just another component to get some to spend yet more money on audio. I do 't believe most people can hear a real difference between high sampled digital & DSD. How many formats do we need?

[issadad]

What's the story with CDs that are marked DSD on the back but not marked SACD?