Why do many "high-end" DACs have low specs?

[curiouspeter]

Just for fun, I was browsing the specs of a few "high-end" Roon Ready DACs. It turns out many of them are limited at PCM 24/192 and/or DSD64 over network. Why is that?

@NYfan2 the OP was about over network . Most DAC's support a local USB connection of very high bitrates . so thats not the question

Then it must clearly be the streamer implementation inside these DAC's .

* Limitations of the streamer software and DAC mfg is bad at software and do not supply this themselves it is some app they like and they bundle that .

* Limitations in Linux ?

* Limitations in RaspberryPi hardware which is the base of many streamers , they all have USB problems for example.

* A wellt hougth of tradeoff ! hogging the network with to large streams may cause reliability issuses for many customers that do not have the latest in network tech, so it is sane imho to let the server for example roon downsample on the fly . As there is no downside your server would hardly notice and soundqality will not change . Flaky network behavior can doom a product a mfg will not go there.

Work around build a roon player in small computer like an intel NUC connect the DAC via USB .

So if you really want to keep the original very fast rate its a two part solution a roon endpiont buildt from computer hardware with good USB spec and then a modern DAC with good USB implementation.

I was earlier suggesting that this may be wild goose chase and really a no problem I was just questioning OP assessment that there is a problem to be solved . If it easy go for it , but to much hassle and money into it for not much return ? it was no ill intent.

Probably it's the streamer software then. My $70 Pi 4B has no issues with DSD256 (not that I normally use that rate). On the contrary, I had a few instances of stuttering with 16/44.1 when I was using the USB 3.0 port of the Pi.

The customer support angle is also a valid one, although networking is increasing a non-issue as more people stream 4K Netflix, which uses 15-25 Mbps.

Using a NUC as a Roon Player is an overkill. A NUC is more than enough even as a Roon Core. Also a NUC dumps too much heat.

 

[charleski]

1-bit DSD cannot be edited, but multi-bit PDM can be used to some extent.

Even Roon can do DSP on DSD using multi-bit PDM.

A note on semantics:
DSD is PDM (pulse density modulation), since it uses only one bit and varies the density of the bits within its overall envelope.
Any representation that uses multiple bits per sample is a form of PCM (pulse code modulation), since the multiple bits are a ‘code’ that represents the amplitude of the signal.
There’s no such thing as ‘multi-bit PDM’, as PDM is single-bit by definition.

Although classical Redbook PCM is 16bits/44.1kHz, it’s perfectly possible to use different forms which have a lower bit length and correspondingly higher sampling rate.

[Edit] I suspect the confusion may arise from the fact that most high-quality DACs these days are higher-order 'multi-bit delta-sigma' designs, and some may associate delta-sigma with DSD. But 'delta-sigma' is strictly a technology for converting between the analog and digital domains and is agnostic concerning the coding format used. A block diagram for a multi-bit delta-sigma ADC looks like this:

(taken from here, and if you scroll down to figure 15 it shows how the Sonoma and Pyramix products manage to 'mix DSD' by converting to an intermediate format.)

 

[Matsuo Mayu]

Pricey Integrated Amps generally have relatively low-spec DACs of 24/192 bulit-in, some go to 32/384. But does a Amp with a higher spec DAC which is up to 32/784 & costed at a lower price , sounds better than the pricey one? Especially Class-D amp vs Class-AB amp？ This issue confuses me more~

 

[Mnyb]

Probably it's the streamer software then. My $70 Pi 4B has no issues with DSD256 (not that I normally use that rate). On the contrary, I had a few instances of stuttering with 16/44.1 when I was using the USB 3.0 port of the Pi.

The customer support angle is also a valid one, although networking is increasing a non-issue as more people stream 4K Netflix, which uses 15-25 Mbps.

Using a NUC as a Roon Player is an overkill. A NUC is more than enough even as a Roon Core. Also a NUC dumps too much heat.

Maybe they have to move to PI 4B my old rasberrys are all generation 3 (or 2 ?)they work fine for my purposes .

The original Pi had a hardware design where the Ehternet port is really an Ethernet to USB port build into the USB implementation . So it can craok if stuff moves in at very high rate at the ethernet and then out at the USB I'm not technical enough to really explain .

Suppose 4B is much better.

Would not Roopie work for you then in a Pi 4B with USB 3 ? keep the Pi 4B use USB to a DAC ?

 

[curiouspeter]

Maybe they have to move to PI 4B my old rasberrys are all generation 3 (or 2 ?)they work fine for my purposes .

The original Pi had a hardware design where the Ehternet port is really an Ethernet to USB port build into the USB implementation . So it can craok if stuff moves in at very high rate at the ethernet and then out at the USB I'm not technical enough to really explain .

Suppose 4B is much better.

Would not Roopie work for you then in a Pi 4B with USB 3 ? keep the Pi 4B use USB to a DAC ?

Yes, 4B is much better. Sharing a bus between USB and Ethernet can be an issue for audio.

I wonder if people are still paying more money for the Allo USBridge.

I use RoPieee in a 4B now. My Gustard X16 is happily connected to one of the USB 2.0 ports.

 

[Mnyb]

Yes, 4B is much better. Sharing a bus between USB and Ethernet can be an issue for audio.

I wonder if people are still paying more money for the Allo USBridge.

I use RoPieee in a 4B now. My Gustard X16 is happily connected to one of the USB 2.0 ports.

Seems like a better solution to me than having the dac mfg integrate tech they dont understand as well as making DAC's .

And it will quickly date the product ? Your X16 will happily chug away on its USB2 port as long there is backward compatibility and it's probably still transparent to human listeners 10 years from now and we are at Pi 10B or something

 

[Wombat]

Re the OP,

The average buyer is not buying on specs. Feel-good is more likely, the things that marketers target rather than objective idealogue preferences.

 

[mansr]

DSD is PDM (pulse density modulation), since it uses only one bit and varies the density of the bits within its overall envelope.

Proper PDM has continuously variable pulse density/rate. DSD is at best a pseudo-PDM with quantised pulse positions. For all practical purposes, it is much easier to treat it as 1-bit PCM. If you wanted, you could also regard it as a degenerate PWM, not that that would do you any good.

 

[charleski]

Proper PDM has continuously variable pulse density/rate. DSD is at best a pseudo-PDM with quantised pulse positions. For all practical purposes, it is much easier to treat it as 1-bit PCM. If you wanted, you could also regard it as a degenerate PWM, not that that would do you any good.

Obviously a continuously variable pulse position would imply an infinite sampling rate, so they had to make some sort of compromise .

 

[mansr]

Obviously a continuously variable pulse position would imply an infinite sampling rate, so they had to make some sort of compromise .

It's not a compromise. It's simply something else.

 

[charleski]

It's not a compromise. It's simply something else.

I think the discussion here revolves around digital audio, rather than a hybridised form of analog. Laserdisc, for example, used a PWM scheme to modulate the video signal onto an FM carrier and thus encode an analog signal on a digital medium without encoding it. But it was still clearly an analog video format.

 

[danadam]

Although classical Redbook PCM is 16bits/44.1kHz, it’s perfectly possible to use different forms which have a lower bit length and correspondingly higher sampling rate.

Yes, I've read that PCM 1bit/2.8224MHz is popular among some people

 

[josh358]

DSD was a solution in search of a problem that didn't exist in consumer audio in order to sell a licensing scheme. Since people began figuring that out, it is and has been on the way out.

So now you know why DSD support in DACs is no longer popular.

I was working with Sony at the time and you have to remember that the OCM equipment of the day was of limited quality, owing to factors like jitter that weren't well understood. Sony even found that CD pressings even sounded different because of differences in pit geometry -- something that should not have made a difference, but did with the circuits then in use.

DSD was developed because, at the time, it sounded better when making archival recordings. Then it was extended to new recordings and the consumer market, because again, it sounded better with the technology of the time. But that meant mixing, and so they developed DXD.

Today, it makes little if any sense. As I understand it they've successfully ABX'd high sampling rates, but I doubt very much that you can hear the difference between 88 kHz and anything above, maybe even 48 kHz.

24 bits (more like 21 bits given limitations on converter linearity) is of more use in the studio than at home. In theory, it's necessary for uncompressed music with the highest dynamic range played at natural levels in the quietest room. In practice, I don't know of any commercial recordings that qualify. But some headroom is required when making a live recording, so bits are wasted, and when a recording is compressed, low-level signals increase in amplitude and dither could potentially become audible. It would, I think, require a very quiet room, music of wide dynamic range, and playback at natural levels for that to become audible.

 

[raistlin65]

I think that OP wanted to know why expensive DACs don't support higher resolution formats when cheap DACs do. He did not invite everyone to bash him for wanting to listen DSD. It is becoming like a dogma with PCM and redbook.
Why do some people have strong feelings about DSD?

I think you are confusing the aims of these types of conversations at ASR where a primary goal of this community is to educate people about audio.

When someone has misconceptions about audio technology, these conversations can become more like an intervention and deprogramming. And of course, the more misconceptions someone has, the more contentious they can become because the person who holds onto the beliefs can be uncomfortable with the conversation.

In fact, could it be that some of the conversation is making you uncomfortable because it's asking you to confront some misconceptions you have about DSD? Perhaps you are confronting "your strong feelings about DSD"??? That could explain why you perceive it as "bashing."

 

[Thomas savage]

DSD gets some closer to the original revenue stream.

 

[curiouspeter]

Look, sometimes I pick a format for a subset of my library simply because it is odd. DSD is an interesting format. I hate analog.

I appreciate this community. I can sense that most people are trying to help.

But this is a "product management" kind of discussion...

 

[curiouspeter]

I was working with Sony at the time and you have to remember that the OCM equipment of the day was of limited quality, owing to factors like jitter that weren't well understood. Sony even found that CD pressings even sounded different because of differences in pit geometry -- something that should not have made a difference, but did with the circuits then in use.

DSD was developed because, at the time, it sounded better when making archival recordings. Then it was extended to new recordings and the consumer market, because again, it sounded better with the technology of the time. But that meant mixing, and so they developed DXD.

Today, it makes little if any sense. As I understand it they've successfully ABX'd high sampling rates, but I doubt very much that you can hear the difference between 88 kHz and anything above, maybe even 48 kHz.

24 bits (more like 21 bits given limitations on converter linearity) is of more use in the studio than at home. In theory, it's necessary for uncompressed music with the highest dynamic range played at natural levels in the quietest room. In practice, I don't know of any commercial recordings that qualify. But some headroom is required when making a live recording, so bits are wasted, and when a recording is compressed, low-level signals increase in amplitude and dither could potentially become audible. It would, I think, require a very quiet room, music of wide dynamic range, and playback at natural levels for that to become audible.

More bits will still be useful for digital volume control, right?

After attenuating 30+ db, there will be a lot fewer than the original 24 bits.

 

[mansr]

More bits will still be useful for digital volume control, right?

After attenuating 30+ db, there will be a lot fewer than the original 24 bits.

Now imagine how few bits will remain if there was only 1 to begin with.

 

[curiouspeter]

Now imagine how few bits will remain if there was only 1 to begin with.

Not if it gets converted to multi-bit DSD and re-modulated.

 

[josh358]

More bits will still be useful for digital volume control, right?

After attenuating 30+ db, there will be a lot fewer than the original 24 bits.

In theory, it shouldn't, not at 24 (really 21) bits. The reason is that the noise+distortion of the DAC should be inaudible at any practical listening level, so it doesn't matter if the DAC is going full out, as it is when you use digital volume control. (Often, higher precision, typically 32 bits, is used in signal processing before the DAC, with appropriate dither added when the signal is converted to 16 or 24 bits.)