Why NOT to use these DACs for audio ?

[percy]

So I ran into this list of R2R DACs from TI that have very good INL and DNL specs. They are not typical 'audio' dacs but they are still afterall a classic r2r dac. I'd imagine they would be the NOS DAC builder's dream dacs?! Or even os dacs following an oversampling filter or asrc, for the matter of fact.
But I have never seen any mention of these types of dacs around here. Am I missing anything ? Do you see any reasons why these dacs cannot be used for audio purposes ?

Some of them are -

DAC8811/2
DAC8830
DAC8871

[hifimaker]

I have not looked at these specific DAC's, but in general non-audio DAC's will not work. Audio DAC's have input sections for decoding standard audio data streams like I2S. Sometimes there are also issues with linearity and resolution.

-David

[percy]

All those dacs have standard serial interfaces. Are you aware of any spdif receivers that do not have a serial interface output ?
And what "issues with linearity and resolution" are you talking about exactly when a 16bit dac has ±0.5LSB DNL and ±1LSB INL and 16bit monotonicity ?

[tritosine]

doubt these are classic r2r dac-s , once we suppose a classic r2r dac has glitch . Looks like you only look at the numbers. These arent even I out or i cant find specs?

[anli]

Originally posted by tritosine
These arent even I out or i cant find specs?

DAC8871, for example, has output impedance specified (having voltage rather current output).

[tritosine]

yeah and with 2nV of glitch its anything but classic

[zilym]

I looked at those R2R DAC's a long time ago and decided against using them because of how much they cost (>$13 each). Since they only provide one channel each, you have to buy two of them to get stereo audio. That suddenly prices these parts beyond high end audio DAC chips, while limiting you to only 16-bit audio. Then the question becomes, why would you want to use these DAC chips for audio when audio DACs can do more for less?

hifimaker has a point that these chips do not implement support for standard I2S audio streams. These use something more like an SPI interface, which could probably be hacked into accepting I2S audio by having some supporting logic chips or a microcontroller to arbitrate data signals. But at this point, why bother?

Instead, why not wire up a set of hand selected, matching resistors in R2R fashion against a cheap 16-bit microcontroller? Resistors are super cheap and you can buy a pair of PIC24FJ64GA004 16-bit microcontrollers for way less than a pair of those TI R2R DAC chips. Plus you've got some room for extra features, like re-clocking internally, etc.

Another idea I'd like to see someone try is using a microcontroller with a ton of PWM outputs like dsPIC33FJ64GP206 in a PWM/Resistor Hybrid DAC configuration. In theory, with enough PWM pins and properly selected resistors, this design could achieve greater bits of resolution while using fewer components. I did it using two PWM pins per channel to get to 16-bit audio, but never had time to try anything further.

I was using a clock source that was not an exact multiple of the audio playback frequency. I think that resulted in muddy high frequency sounds. I would definitely try to use a clock source that is a perfect multiple of the audio playback frequency if I were to try this stuff again.

[percy]

Originally posted by tritosine
yeah and with 2nV of glitch its anything but classic

but you realize that the glitch is only a 'carry over' glitch - i.e. only when the input code changes from 32767 to 32768 and vice versa, right ?

Originally posted by zilym
[B]I looked at those R2R DAC's a long time ago and decided against using them because of how much they cost (>$13 each). Since they only provide one channel each, you have to buy two....

thats hardly an argument against it when the purpose is solely personal/diy/hobby. lets ignore cost for now.

Originally posted by zilym
hifimaker has a point that these chips do not implement support for standard I2S audio streams. These use something more like an SPI interface, which could probably be hacked into accepting I2S audio by having some supporting logic chips or a microcontroller to arbitrate data signals. But at this point, why bother?

why complicate it ? unless you are talking about slaving the source via i2s, I see no point in using an i2s connection. I'd rather rely on my own clock than on the one coming over an i2s from 'xyz' source. And with jitter rejection abilities of an spdif receiver like that of the wm8804, spdif to spi becomes even more viable.

Originally posted by zilym
Instead, why not wire up a set of hand selected, matching resistors in R2R fashion against a cheap 16-bit microcontroller?

To be honest, I have considerd that.

[tritosine]

ultra low glitch is a feature and definately not in a "classic" way. Perhaps invest some time and read what they mean with "segmented" architecture.

[rfbrw]

Originally posted by percy

And with jitter rejection abilities of an spdif receiver like that of the wm8804, spdif to spi becomes even more viable.

Perhaps you could post a design.