dCS Bartok

[GoldenSound]

If you look at the ring DAC block diagram from your post, there is absolutely nothing on the diagram at all which can be interpret as a delta-sigma DAC.

There is, this part:

This is the delta-sigma modulator. There isn't another type of modulation that would perform this task of converting 768khz 24 bit data to 3/6Mhz 5-bit data. (If you are aware of one please do say).
The only other option would be to literally do basic truncation or something and oversample the 5-bit result which would still drastically restrict accuracy of the dac

Its just like how a multibit DAC doesn’t have to be a delta-sigma DAC.

Multibit is a somewhat vague term that is used in different ways by different people and companies.
Technically, 'multibit' just means any converter that operates at a level above 1-bit (ie: multiple bits, multibit).
This could include 3-bit switched resistor or switched capacitor dacs, the ring DAC, R2R dacs, and could potentially make the argument to include PWM dacs too.
Some companies like schiit use it to refer to their native PCM converters (R2R, though IC based not discrete) which is technically correct but as a result some people have begun using the terms R2R and Multibit interchangeably which is not correct.

dCS also said on their forum that the 5-bit modulator is using “noise shaping technique” to convert from high-to-lower bits data

Noise shaping is a part of all delta sigma dacs. Without it they'd have abhorrent performance. The noise shaping is done to alleviate the issue of quantization error, which results from the modulation process.

I'd recommend giving these a read, they're quite thorough:
https://www.analog.com/en/technical-articles/behind-the-sigma-delta-adc-topology.html
https://www.ti.com/lit/an/slyt423a/slyt423a.pdf?ts=1641484292450&ref_url=https%3A%2F%2Fwww.google.com%2F#:~:text=The ∆Σ modulator is,outside the band of interest.

If you have a value of 0.7, but you can only describe it with a 1 or a 0, in this case you'd have a quantization error of 0.3.

Noise shaping is part of the oversampling process, and can be applied to both R2R dacs using oversampling, and Delta sigma dacs, but is required on delta sigma dacs to have any sort of decent performance.

This is because the quantization error on an R2R dac will be minimal (Because the difference between the true mathematical value of what an interpolated sample SHOULD be and what a 16 bit accurate R2R ladder can describe will be smaller than the true mathematical value and what a 5-bit converter can describe. As you reduce the linearity of the converter you will encounter more quantization error).

Noise shaping, dithering, truncation error and quantization error/noise are also not exclusive to DACs, it's a mathematical issue and so is also quite important in music production and processing, as when performing DSP, these problems can rear their head without proper dithering or noise shaping.

One of the main differences between a R-2R and a delta-sigma DAC is in its output section. R-2R contains resistors value R and 2R in a ladder-like network and its output the required voltages by turning the resistors on and off. Delta-sigma DAC output the conversion as bitstream which is basically in pulse-density modulation format (PDM). This is then connected to some kind of differential amplifier analog section and its output the required voltages by turning it self on-off really really fast.

Partially. Delta sigma dacs as mentioned previously can be done in all sorts of ways.

A true 1-bit converter will be pure pulse density modulation, but hardly any modern DACs are true 1-bit.

AKM modulates down to 6-bit (if I remember correctly) and uses a switched-resistor converter.

ESS uses a design practically identical to the ring dac but on an IC

Burr brown has a few designs but the most prevalent is likely their 'advanced segment' which uses a 6-bit R2R ladder, and converts the lower 12/18 bits with 1-bit delta sigma that has a top level value equal to 1 LSB of the R2R segment.

Again as mentioned previously, the key thing here is whether delta sigma modulation is in place. Whether you are reducing bit depth or not.
Even if you built an R2R ladder with say 14 bits, and used delta sigma to modulate from 16 bits to 14 bits. That's still technically a delta sigma dac.
If you did not use delta sigma then you would be limited to the linearity of the converter (14 bits).

The ring DAC shares a lot of similarities to R-2R ladder DAC and almost no similarities to a delta-sigma DAC. An R-2R ladder DAC contains resistors in a ladder-like network in which the ring DAC also has. The ring DAC also doesn’t looks like anything like a delta-sigma DAC but it does looks almost 100% like an R-2R ladder DAC.

It shares nothing in common with an R2R dac other than that it has resistors.
There are ways of doing a true 1-bit converter which to the naked eye looks like an R2R ladder, but that does not make it so. (links below)
R2R is a particular topology which is defined by the specific circuit arrangement, and its ability to convert PCM natively at the native bit depth.
A delta sigma dac cannot natively decode 16 bit PCM

I’ve also never ever see any delta-sigma DAC that contain resistors ladder network before and I do not understand why would a delta-sigma DAC need to use the resistor ladder network for either. So if you’ve seen one, please share and explain how the resistor network is being used in a delta-sigma digital-to-analog conversion. Please ensure to provide valid references should there be any further discussion about this.

There are a couple open source projects with circuit designs and gerbers for PCB printing available. Jussi Laako (who created HQPlayer, a tool for high performance oversampling and PCM/SDM conversion) created the DSC1 dac and this has been iterated on by others. It's quite an interesting one to check out.

https://puredsd.ru
https://www.signalyst.com/hardware.html

There are also some other proprietary designs which to the naked eye look like a resistor ladder. For example the Holo audio dacs have separate PCM and 1-bit DSD converters.

 

[SuperBurrito]

I’ve been lurking around here for years now… just didn’t feel the need to shared any thoughts until recently…

So…yes, hi… nice to meet you.

Welcome aboard!

 

[GoldenSound]

Also, on a note about the above discussion, delta sigma and R2R are not necessarily mutually exclusive. You can have for example a fast switching 6-bit R2R ladder that uses delta sigma modulation from the original 16 bit data. And you could also have a DAC that falls under neither the category of R2R or Delta Sigma.

For example you COULD make a 16-bit ring DAC. However it'd need 65535 elements minimum (or 16,777,215 elements for 24 bit) rather than the current 48 that dCS uses so it'd be completely impractical and the real-world technical challenges of building something like that would probably make it impossible. But still technically doable on paper.

It's not that there are only R2R dacs and delta sigma dacs. Just that R2R is for the most part the only practical way to do native 16 bit or higher PCM conversion with current technology. (Though there are also some clever variations to R2R designs too and I don't think any of the current R2R dacs I'm aware of are a 'pure' R-2R circuit. Much like almost none of the currently available delta sigma dacs are a 'pure' 1-bit PDM design)
Delta sigma modulation opens up a lot of possibilities as you can do a lot of stuff with a lower bit-depth converter (like the ring-dac or the pulse array) that would not be feasible at a higher bit-depth like 16 or 24 bits

 

[Ton AMG]

There is, this part:

This is the delta-sigma modulator. There isn't another type of modulation that would perform this task of converting 768khz 24 bit data to 3/6Mhz 5-bit data. (If you are aware of one please do say).
The only other option would be to literally do basic truncation or something and oversample the 5-bit result which would still drastically restrict accuracy of the dac

I don’t know what they are using in there as dCS themselves only saying that it is using “noise shaping techniques”. Actually, that was dCS replying to someone specifically asking is the 5-bit modulator using delta-sigma modulation or not?

So it could be that the person doesn’t really know what exactly it is using so just giving a generalised answer OR maybe they are using some kind of top-secret techniques that no one ever knew about and they want to keep it a secret? Or maybe there are literally flying pixies inside? Who knows?

The point here is that we can only guess at what they are really using in there.

Noise shaping is a part of all delta sigma dacs. Without it they'd have abhorrent performance. The noise shaping is done to alleviate the issue of quantization error, which results from the modulation process.

Agreed that noise shaping is a part of all delta-sigma dacs. However, not all noise shaping is a part of just delta-signa dac though. It is also a part of other type of dacs too.

Burr brown has a few designs but the most prevalent is likely their 'advanced segment' which uses a 6-bit R2R ladder, and converts the lower 12/18 bits with 1-bit delta sigma that has a top level value equal to 1 LSB of the R2R segment.

Again as mentioned previously, the key thing here is whether delta sigma modulation is in place. Whether you are reducing bit depth or not.

I’m not sure how this is an example of a delta-sigma dac… as it is literally said “Figure 65. Advanced Segments DAC”

You can even see how they got the naming from. Its from the “Advanced DWA” which is basically a mapper for which “Current Segment DAC” source to be used for the conversion. Hence the name, “Advanced Segments DAC”

This is similar to ring dac which uses “thermometer-coded” principles as a mapper for which “current segment” from their equal value resistors ladder should be used. Hence the name, “thermometer-coded” segments DAC.

When talking about what type of DAC it is… then we should be typing it by the main mechanism it is using to convert the data from digital to analog. We are talking about digital-to-analog conversion here, not what kind of preparation technique it is using to prepared the digital data for conversion.

A delta-sigma DAC uses delta-sigma modulation as the main mechanism to convert the data.

An advanced segment DAC uses advanced DWA and segment current source as the main conversion mechanism. Same as ring-DAC which uses thermometer-coded and segment current source for conversion. To say that both of these are “delta-sigma dac” then you are only describing part of the process they are using to prepared the digital data for conversion while totally ignoring everything else including the main conversion mechanism. It is very misleading.

If a DAC uses delta-modulation to prepared the digital data for pixies to convert it to analog, are you still going to call it a delta-sigma dac?… If yes, then you will be missing the very important part about the flying pixies inside…

Anyways, if anyone asked what kind of dac is it? for something like a ring-dac, I think the best answer is:

“Yes”

 

[GoldenSound]

I’m not sure how this is an example of a delta-sigma dac… as it is literally said “Figure 65. Advanced Segments DAC”

You can even see how they got the naming from. Its from the “Advanced DWA” which is basically a mapper for which “Current Segment DAC” source to be used for the conversion. Hence the name, “Advanced Segments DAC”

This is the name for their particular design approach. It's a product name/marketing term. Same as dCS 'ring DAC' or chord 'pulse array' or Holo 'linear compensation' or AKM 'velvetsound'.

The ford F150 is called the F150, but that does not mean 'It's not a truck cause it clearly says F150 on it and it has a different shape to other trucks'. It's just ford's particular way of making a truck.

A thermometer coded DAC is a particular design of DAC which uses equally weighted elements with their current output summed to represent a value upto the bit-depth of the converter, in this case a bit depth of 5, which can represent only 32 different values, and therefore requires delta sigma processing in order to achieve significant linearity.
You COULD make a native PCM thermometer coded DAC, but the physical size and challenge of even just connecting traces to and controlling 65536 elements would make it unfeasible. Technically possible, but likely not in practice (at least not with current tech)

To say that both of these are “delta-sigma dac” then you are only describing part of the process they are using to prepared the digital data for conversion while totally ignoring everything else including the main conversion mechanism. It is very misleading.

It is only part of the process yes, but it is the overarching category which defines whether a converter is a native PCM converter or not.
We can start describing other characteristics, but where do we draw the line as to what 'should' be said when mentioning a dac type?
Should we also include the oversampling filter design, noise shaping order, differential or not, whether a PLL is utilised, what rate the converter operates at, how the flip flops are operated, whether it is a current or voltage source, whether the output stage is opamp based or not, what manufacturers made the resistors and other components and what their values are?

When people say 'a car' they could be referring to a ferrari or a prius. Two VERY different machines, but people don't usually give the details about the engine, suspension, brakes, infotainment, ride height, fuel tank size, boot capacity etc of a car when mentioning it.
They say 'car' cause they're primarily just wanting to talk about a car not a boat

Besides, this is debating whether you should mention deeper down aspects of a design. Which even if we were to do so, does not exclude the ring dac from being categorised as delta sigma.
And if we only referred to products by their marketing material, we'd get even less useful information than if we just addressed them by general category. 'Velvetsound' or 'ring dac' gives you practically 0 information other than potentially what company makes it. Whereas 'delta sigma' does tell you about the general design approach of the product.

A delta-sigma DAC uses delta-sigma modulation as the main mechanism to convert the data.

An advanced segment DAC uses advanced DWA and segment current source as the main conversion mechanism.

No, delta sigma is not a specific circuit/converter design. It is a mathematical process, done in software, which is applied to various converter designs.

The two overarching categories of DAC are native PCM converters, and delta sigma converters.

You can either convert the full 16 bit sample at the native bit depth, or you cannot. It is a binary yes/no question, and the marketing label or product name a manufacturer gives to their design does not change this.
(You could potentially argue for a third category of 'PDM only' dacs which have no delta sigma processing, but can only accept DSD/1-bit information. Or require you to do the oversampling/modulation with an external tool like HQPlayer. An example of this being the DSC1 DAC or the Holo Cyan DSD)

The actual converter hardware/circuit itself often could fall under either category depending on design/implementation. For example as mentioned previously, you can make a delta-sigma ring dac, such as the one in the dCS dacs, or you COULD build a non-delta-sigma, native PCM converter ring DAC. But to do so would require 65536 elements minimum, not the 48 the ring dac currently uses.
You COULD also build an R2R ladder that operated as a delta sigma dac by creating a fast operating 6-bit R2R ladder with oversampling/delta sigma processing. And in fact both burr brown and AKM designs aren't too far off this approach.

you will be missing the very important part about the flying pixies inside…

what?

 

[edwardsean]

you will be missing the very important part about the flying pixies inside…‍♀

what?

@GoldenOne, please do try and give some consideration to the perspective I offer here. Post #1,260 of 1,265 It may be helpful to differentiate, for purpose of reviews, DACs that use DS modulation vs. a "Delta-Sigma DAC" (i.e., a more straightforward DS design)

When people say 'a car' they could be referring to a ferrari or a prius. Two VERY different machines, but people don't usually give the details about the engine, suspension, brakes, infotainment, ride height, fuel tank size, boot capacity etc of a car when mentioning it.
They say 'car' cause they're primarily just wanting to talk about a car not a boa

Right, so to use your analogy, it is technically justifiable to say that dCS DACs or Dave has aspects that are the best among delta-sigma DACs. But, it does have the same utility as saying a Ferrari puts up some of the best speeds among–cars.

That is they way your statements hit me in your review. Even after I understood the case for including the Bartok and Dave among DS DACS, I was still confused as to why you would make such a broad comparison among such carefully nuanced criticism.

When Motor Trend reviews a Ferrari's performance they are going to place it in its category among cars of similar engineering sophistication. Ferraris and Priuses (Priusai?) are "VERY different machines," so let's talk about them as such–or not. You're free, of course, to group DACs however you like for your analysis. I'm just relaying what might be helpful to readers of your reviews, that's all.

 

[GoldenSound]

@GoldenOne, please do try and give some consideration to the perspective I offer here. Post #1,260 of 1,265 It may be helpful to differentiate, for purpose of reviews, DACs that use DS modulation vs. a "Delta-Sigma DAC" (i.e., a more straightforward DS design)



Right, so to use your analogy, it is technically justifiable to say that dCS DACs or Dave has aspects that are the best among delta-sigma DACs. But, it does have the same utility as saying a Ferrari puts up some of the best speeds among–cars.

That is they way your statements hit me in your review. Even after I understood the case for including the Bartok and Dave among DS DACS, I was still confused as to why you would make such a broad comparison among such carefully nuanced criticism.

When Motor Trend reviews a Ferrari's performance they are going to place it in its category among cars of similar engineering sophistication. Ferraris and Priuses (Priusai?) are "VERY different machines," so let's talk about them as such–or not. You're free, of course, to group DACs however you like for your analysis. I'm just relaying what might be helpful to readers of your reviews, that's all.

I understand your opinion and am not disagreeing.

I don't wish to debate how best to group products together or not group things together, or how best to create useful contrast between product segments in reviews as its entirely subjective and there isn't really a right answer.

I chose to do so because in my experience, these two product groups are usually the most different from eachother subjectively (and differ from a design standpoint the most too). And therefore is the most apt divide to compare/contrast products between.
Others may disagree and that's fine. As said, it's subjective.

But regardless, the original debate arose due to the statement that 'the dCS bartok is NOT a delta sigma dac'. This statement is not true.
Whether its most ideal to call products by what design aspect/naming scheme is a separate debate to whether the dCS dac is/isn't a delta sigma dac (regardless of if other descriptions apply)

 

[edwardsean]

But regardless, the original debate arose due to the statement that 'the dCS bartok is NOT a delta sigma dac'. This statement is not true.

Funnily enough, if you go all the way back, you'll see that this whole debate originated with me responding to your review, questioning why you would categorize the Bartok as a DS DAC. The debate started for you in earnest when another poster questioned your response to me from a technical standpoint. So, yeah, the point I'm making is ancillary to your primary concern.

However, if this all precipitated from a technical paper, I would never have pushed in my pretty, but ignorant, little head. However, since part of your purpose is writing reviews (rather than designing DACs) I would hope that you would not so neatly cleave "ideal," helpful categorizations from technical ones. The two are not unrelated. Again, I know the Dave far better, but I don't know if dCS would be so sure as you that what they've designed fits as "a delta-Sigma DAC."

I'm not saying you can't make the case for it, obviously. I'm saying it–may–be misleading to those who read your reviews by way of–reductionism.

I don't wish to debate how best to group products together or not group things together, or how best to create useful contrast between product segments in reviews as its entirely subjective and there isn't really a right answer.

Not "entirely subjective" right? And also, why not? I respect your technical knowledge, but also your passion as a reviewer. And also,

I getcha. I'm tired of this for my part too, but I had some time before my dinner was ready. Apologies for buggin' ya.

 

[notofthisworld]

This thread makes me want to sell my Bartok. Time to head over to the Chorus threads!

 

[Ton AMG]

It is only part of the process yes, but it is the overarching category which defines whether a converter is a native PCM converter or not.
We can start describing other characteristics, but where do we draw the line as to what 'should' be said when mentioning a dac type?
Should we also include the oversampling filter design, noise shaping order, differential or not, whether a PLL is utilised, what rate the converter operates at, how the flip flops are operated, whether it is a current or voltage source, whether the output stage is opamp based or not, what manufacturers made the resistors and other components and what their values are?

I’ve already explained that we should be typing the DAC by the core mechanism it is using to convert the digital data to analog data.

When you see a block diagram for delta-sigma DAC you will see delta-sigma modulation very obvious, not hidden away as a sub-process.

It’s really not complicated.

When people say 'a car' they could be referring to a ferrari or a prius. Two VERY different machines, but people don't usually give the details about the engine, suspension, brakes, infotainment, ride height, fuel tank size, boot capacity etc of a car when mentioning it.
They say 'car' cause they're primarily just wanting to talk about a car not a boat

Depends on people, but if a car reviewer say a Ferrari with an engine derived from F1 turbo hybrid V6 ICE with Energy Recovery System, has the same kind of engine as a Toyota Prius… then I will be at the reviewer for conducting such a crime against humanity.

The two overarching categories of DAC are native PCM converters, and delta sigma converters.

99% of digital music available on the market is in PCM format. All DAC for digital music virtually converts PCM (or DSD) into analog audio signals.

There are no music encoded digitally in Delta-sigma modulation format. Why would anyone categorised the DAC in to two like that?

This make just about the same sense as my pixies DAC

**Edit**: I’ve just realised that DSD is actually stored as the direct digital output of delta-sigma modulation.

So if I play DSD files (which is the exact output of a delta-sigma modulation) through the Holo May DAC natively, is it now a delta-sigma DAC too? I don’t think so.

 

[cosmosland]

**Edit**: I’ve just realised that DSD is actually stored as the direct digital output of delta-sigma modulation.

So if I play DSD files (which is the exact output of a delta-sigma modulation) through the Holo May DAC natively, is it now a delta-sigma DAC too? I don’t think so.

In Golden's review of the May, he did say that it is both an R2R and an delta-sigma dac. Link is here:

The statement that it's both is at 11:54 and the explanation why starts at 10:59.

 

[Ton AMG]

In Golden's review of the May, he did say that it is both an R2R and an delta-sigma dac. Link is here:

The statement that it's both is at 11:54 and the explanation why starts at 10:59.

The problem here is that with native DSD playback the “delta-sigma modulation” is already done during the mastering process in the studio, not on the DAC itself during play back.

So, no I still don't think it is correct to call it a “delta-sigma DAC” when there is no “delta-sigma modulation” involved on the DAC side at all.

I think that’s why when playing back native DSD even with a “delta-sigma DAC” its kind of by pass everything and just go straight through to the output analogue section of the DAC.

(Although I’m not sure exactly how the Holo May is processing the native DSD files so feel free to correct me if you have the info)

 

[cosmosland]

The problem here is that with native DSD playback the “delta-sigma modulation” is already done during the mastering process in the studio, not on the DAC itself during play back.

So, no I still don't think it is correct to call it a “delta-sigma DAC” when there is no “delta-sigma modulation” involved on the DAC side at all.

Delta-sigma modulation isn't only done when converting from digital to analogue, its also done when going from analogue to digital. So yes DSD has already gone through delta-sigma modulation but thats when converting from analogue to digital. The DAC needs to convert from digital to analogue again (using delta-sigma modulation).

I think that’s why when playing back native DSD even with a “delta-sigma DAC” its kind of by pass everything and just go straight through to the output analogue section of the DAC.

This would mean DSD is analogue not digital which it definitely isn't. (Direct Stream Digital)

 

[Ton AMG]

So yes DSD has already gone through delta-sigma modulation but thats when converting from analogue to digital.

No, they don’t use DSD format in the studio. You can’t do anything with it once its in DSD. Virtually all studio will be working using PCM during mixing and mastering process.( There are some weird format like DXD too but I don’t think anyone is really using it)

If they decided to release in DSD, then the final mix will get convert from PCM to DSD by using delta-sigma modulation. It is a digital-to-digital conversion (DDC) and not a digital-to-analog conversion (DAC)

The DAC needs to convert from digital to analogue again (using delta-sigma modulation).

In a delta-sigma DAC, the usual input is in PCM digital format. The delta-sigma modulation basically convert PCM into bitstream which is the same as DSD format. Then it will be sent to the output amplification of the DAC to convert the bitstream into analog audio signals.

That’s why when playing DSD natively on a delta-sigma DAC, it will go straight to the output section of the DAC. Its already been through delta-sigma modulation from the studio. The DAC still convert the DSD from digital-to-analog but without using “delta-sigma modulation” on the DAC.

On the non delta-sigma DAC, my guess is that most will have to convert the DSD into PCM first or have a separate DAC that can accept bitstream as input for conversion directly.

I guess you can also use delta-sigma DAC too but then you are just using the “DAC” part of it only and not the “delta-sigma” part.

For Holo May, I’m not sure how that is done as it seems to be using different set of resistors ladder for DSD conversion. It seems there needs to be something to process the bitstream so that it can be feed into the dedicated resistors ladder. I don’t know how it is done.

This would mean DSD is analogue not digital which it definitely isn't. (Direct Stream Digital)

Yes, DSD is digital. It is the output of delta-sigma modulation process. It is not analog.

 

[TDinCali]

This conversation is so over my head. I try to follow along but I doubt I’m truly getting it. I do appreciate you guys sharing your knowledge and debating your points in such a civil manner.