Jump to content
IGNORED

DSD Offshoot Discussion From MQA Topic


Recommended Posts

2 minutes ago, jabbr said:

The concept is that an R2R ladder could be linearized beyond say 17bits using a different encoding (and faster switching) at least for the lower order bits.

 

I dont know if commercially available audio DACs do this, but it’s also hard to see how folks claim 24 bits with any discrete ladder, let alone any ladder.

There are DAC chips from TI and ADI, possibly others too, that work along those lines.

Link to comment
12 minutes ago, jabbr said:

No the data transmission rates are certainly different. The number of bits per second is the same : understand?

Yeah I meant that the frequencies were different — wrote the wrong thing — do you understsnd the overall point I am making?

 

To day another way, the difference between PCM and DSD is not that one is an integral or differential of the other. In both cases the quantization noise is removed with LPF. 

Custom room treatments for headphone users.

Link to comment
53 minutes ago, jabbr said:

No the data transmission rates are certainly different. The number of bits per second is the same : understand?

 

Nope, can't say I really understand where you are coming from. It's a context issue though, and I am physically pretty tired, so I will go away and think about it some more. :)

 

Anyone who considers protocol unimportant has never dealt with a cat DAC.

Robert A. Heinlein

Link to comment
38 minutes ago, jabbr said:

Yeah I meant that the frequencies were different — wrote the wrong thing — do you understsnd the overall point I am making?

If by frequency you mean sample rate, then yes.

 

38 minutes ago, jabbr said:

To day another way, the difference between PCM and DSD is not that one is an integral or differential of the other.

Right. Because they are the same thing.

 

38 minutes ago, jabbr said:

In both cases the quantization noise is removed with LPF. 

Not quite. With flat dither, all the noise below the low-pass cut-off remains. If the sample rate is high enough, the noise can be shaped to fall mostly above this frequency, and then it will be removed by the filter.

Link to comment
1 hour ago, jabbr said:

The concept is that an R2R ladder could be linearized beyond say 17bits using a different encoding (and faster switching) at least for the lower order bits.

 

Problem with R2R is always MSB accuracy.  As you go from LSB towards MSB your accuracy requirement doubles on each step. Combined error of all resistors, in any possible combination, above LSB must be less than +-0.5 LSB.

 

1 hour ago, jabbr said:

In any case the point Im trying to make is that PCM and SDM differ primarily in the resistor weightings one R2R and the other equal.

 

No, they are massively different because they use different encoding altogether. In addition to massive difference in the content itself.

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
58 minutes ago, mansr said:

It so happens that an equal-weighted summing network can be used both for converting a single unary coded value and as a FIR filter for 1-bit values. They are, however, conceptually performing vastly different functions.

 

...or anything between... As long as you operate at multiples of unary coded words. So let's say you'd have quite common five level SDM (2.3219... binary bits) for which you need minimum four elements to convert. Of course nothing says that you would be restricted to a DAC that has only four elements. You could as well have 32 or 64 elements. In fact, such is not uncommon at all.

 

Essentially SDM DACs always contain some form of filter as part of the conversion stage, because typically the analog filter after I/V is usually only 2nd or 3rd order.

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
3 minutes ago, Miska said:

...or anything between... As long as you operate at multiples of unary coded words. So let's say you'd have quite common five level SDM (2.3219... binary bits) for which you need minimum four elements to convert. Of course nothing says that you would be restricted to a DAC that has only four elements. You could as well have 32 or 64 elements. In fact, such is not uncommon at all.

Sure, you can string together unary coded words of any length and get a longer unary coded word representing the sum.

 

The point I'm trying to get across is that converting a multi-bit word (regardless of coding) to a voltage is a different thing than summing/averaging multiple sample values (whether 1-bit or wider), even if the circuit doing it looks the same.

Link to comment
10 minutes ago, mansr said:

Sure, you can string together unary coded words of any length and get a longer unary coded word representing the sum.

 

The point I'm trying to get across is that converting a multi-bit word (regardless of coding) to a voltage is a different thing than summing/averaging multiple sample values (whether 1-bit or wider), even if the circuit doing it looks the same.

 

Sure, one can decide how many unary bits he wants to slam in at once. Anything from 1 to 32 for example. And of course modulator doesn't need to be restricted to binary word lengths. Instead something like five levels is quite common for multi-level. That has zero-level, plus two levels above zero and two below.

 

Number of levels used it mostly just a small curiosity (until you want to make a power-DAC).

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

However one wants to view the mechanism, and after all the mathematics have been thoroughly digested - what actually counts as regards subjective SQ is the quality of the implementation of the particular method; one tiny poor decision, or a weakness or senstivity to interference effects in some area will be enough to bring down the whole shebang - the circuit never delivers the sound quality that theoretically it's capable of.

 

My attitude is that every technique can deliver "good enough sound" - just get the actual bits on the board to work as they should, assessed by measuring, and then listening; the real arguments should be about how to optimise particular converter approaches to ensure that they produce good sound.

Link to comment
3 hours ago, Miska said:

No, they are massively different because they use different encoding altogether. In addition to massive difference in the content itself.

 

This is a subjective "massive" difference. Conceptually they are different. They may be produced in a different way. I suspect that your "massive difference" is the difference in the algorithms used to filter and upsample.

 

From a purely functional POV, both the PCM and DSD DACs send bits through a resistor or set of resistors to convert to current, and then both through an LPF. To me if the output is similar and the circuits are very similar, then the inputs are also similar.

Custom room treatments for headphone users.

Link to comment
26 minutes ago, jabbr said:

This is a subjective "massive" difference. Conceptually they are different. They may be produced in a different way. I suspect that your "massive difference" is the difference in the algorithms used to filter and upsample.

 

From a purely functional POV, both the PCM and DSD DACs send bits through a resistor or set of resistors to convert to current, and then both through an LPF. To me if the output is similar and the circuits are very similar, then the inputs are also similar.

 

For example with SDM you can use any of the conversion elements to convert any of the bits. While that certainly doesn't work with R2R.

 

And many SDM DACs don't use resistors, but instead bits drive charge pulses from capacitors to another. For example Cirrus and AKM are of this type.

 

image.png.c6bcd990d33851f5db23117841db9056.png

 

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
3 hours ago, Miska said:

And many SDM DACs don't use resistors, but instead bits drive charge pulses from capacitors to another. For example Cirrus and AKM are of this type.

 

For extra credit I will leave the reader to implement a PCM version 😂

Custom room treatments for headphone users.

Link to comment
4 hours ago, jabbr said:

 

This is a subjective "massive" difference. Conceptually they are different. They may be produced in a different way. I suspect that your "massive difference" is the difference in the algorithms used to filter and upsample.

 

From a purely functional POV, both the PCM and DSD DACs send bits through a resistor or set of resistors to convert to current, and then both through an LPF. To me if the output is similar and the circuits are very similar, then the inputs are also similar.

 

That's kind of like saying that life is just a series of chemical reactions. It may be a bit over simplified. How you do something is often quite important, perhaps this is one of those cases. 

 

I will leave it as an exercise for you to figure out the ramifications of that. There are a few. :)

 

-Paul 

 

Anyone who considers protocol unimportant has never dealt with a cat DAC.

Robert A. Heinlein

Link to comment
3 hours ago, jabbr said:

For extra credit I will leave the reader to implement a PCM version 😂

 

That is also possible, but never heard of anybody implementing such for audio. For 24-bit, biggest capacitors would probably go beyond what is possible on a chip. Getting accurate capacitors is even more tricky than getting accurate resistors. Luckily for FIR type construction I use the accuracy of resistors is not critical because inaccuracies don't cause nonlinearity or distortion, they just slightly alter filter's roll-off.

 

Nice thing with switched capacitor DAC is that you can have both DAC and a LP-filter as part of the same design.

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
4 hours ago, mansr said:

It's already PCM. You mean binary coded PCM.

Man, you are holding on to this idea like it is a religious maxim or something. Ok - I won’t even argue with you that DSD is a special case, or that how you interpret the information in a signal defines what the signal is, at least for practical purposes. 

 

It it reminds me of a chap I know who demands everyone realize the truth about digital signals and data; that truth being that there just isn’t any. Digital signals or storage that is.

 

Everything, when you get right down to it, is purely analog. That digital signal is really an analog signal. And so on. And he is correct as far as that goes.

 

But, with his argument, the big kicker is pretty simple; how we interpret that analog signal defines whether it is a digital signal or not. Very basic, and also very true. Being digital is just adding one or more layers of agreed upon definitions to an absolutely analog signal.

 

So it is true that a digital signal, any digital signal is also an analog signal, though not the reverse of course. Amazingly enough, my friend is a digital signal designer, and he does really good work. Perhaps his “always on” awareness of the underlying analog reality of a digital transmission helps him be so good.

 

I suggest we may have a slightly similar situation here.

 

Anyone who considers protocol unimportant has never dealt with a cat DAC.

Robert A. Heinlein

Link to comment
5 minutes ago, Paul R said:

Man, you are holding on to this idea like it is a religious maxim or something. Ok - I won’t even argue with you that DSD is a special case, or that how you interpret the information in a signal defines what the signal is, at least for practical purposes.

There is a distinction between what a signal is and how it is encoded. A sampled signal is the same whether encoded as binary numbers, unary numbers, or carefully arranged tea leaves. That some encodings make certain tasks easier doesn't make them fundamentally different to others.

 

8 minutes ago, Paul R said:

Everything, when you get right down to it, is purely analog. That digital signal is really an analog signal. And so on. And he is correct as far as that goes.

That's a common justification for all manner of audiophile voodoo. When someone brings it up, it's typically a good idea to stop listening.

 

9 minutes ago, Paul R said:

So it is true that a digital signal, any digital signal is also an analog signal, though not the reverse of course. Amazingly enough, my friend is a digital signal designer, and he does really good work. Perhaps his “always on” awareness of the underlying analog reality of a digital transmission helps him be so good.

If you're designing the physical layer, you do need to care about the analogue aspects of the signal. Specs and datasheets don't pretend otherwise. Once you move past the circuit that interprets voltage wiggles as zeros and ones, you no longer need to care. Each 1 or 0 bit is as good any other. I'm sure your friend knows this.

 

In general, it helps to have an understanding of the layer below (and above) where you're working and an awareness of the one below (or above) that. This applies whether we're talking about communication protocols, programming languages, or something else.

Link to comment
1 minute ago, mansr said:

There is a distinction between what a signal is and how it is encoded. A sampled signal is the same whether encoded as binary numbers, unary numbers, or carefully arranged tea leaves. That some encodings make certain tasks easier doesn't make them fundamentally different to others.

 

 

Well, we agree there. Seems to all be a matter of perspective. 

Anyone who considers protocol unimportant has never dealt with a cat DAC.

Robert A. Heinlein

Link to comment
40 minutes ago, Miska said:

I guess we can conclude that everything in the world is PCM, even water flowing in the river and these words I'm typing in the browser. We don't need to waste bandwidth anymore by mentioning PCM. :D

 

 

Would you consider 2 bit PCM as equivalent to 2 bit SDM ... perhaps not. 1 bit PCM?

Custom room treatments for headphone users.

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...