I think that we dont have IC dac with real resistors inside?
"R" component is made within in die transistors forming current circuits?
There are no passive resistors in IC DACs...
.
On the other hand in diskrete way, we have logic circuits, with complementary mosfets at output that driving R-2R pasive ladder network. Thes complementary devices has usually a bit different Rds on and other off state. So they adding a little mismatch in low tolerance passive resistors?
.
Actually some of IC dacs has thermomether encoding for 3-4 MSBs, so these thermomether added 7 to 15 bits are going to only "R" component without "2R" as AD1862 noted in Datasheets...
"R" component is made within in die transistors forming current circuits?
There are no passive resistors in IC DACs...
.
On the other hand in diskrete way, we have logic circuits, with complementary mosfets at output that driving R-2R pasive ladder network. Thes complementary devices has usually a bit different Rds on and other off state. So they adding a little mismatch in low tolerance passive resistors?
.
Actually some of IC dacs has thermomether encoding for 3-4 MSBs, so these thermomether added 7 to 15 bits are going to only "R" component without "2R" as AD1862 noted in Datasheets...
Last edited:
TDA1541A has the thermomètre iirc then the ladder. Plus bjt, maybe all of this explain its signature.
The only resistor in th ad1862 should be the one of the 3k feedback they outputted on the pin below the Iout one. I wondered if they made that to réduce the cost of the ownership for less passive external parts, or if it was to figth stray capacitance when adding the ne5532 of the datasheet on the pcb the closest as possible?
In the same fun TDA1541A has not the DEM timing cap embeded...and embeded in the little non-A batch. Some prefered that one vs the A withe same 470pF value...
To say than the layout MATTERS.
Not sure r2r discrete is a good idea but marketing. I do not know some Say Holo May sounds quite good as a partialy discrete with FGPA. But notice not à same gound sound between pcm or DSD as well.
The only resistor in th ad1862 should be the one of the 3k feedback they outputted on the pin below the Iout one. I wondered if they made that to réduce the cost of the ownership for less passive external parts, or if it was to figth stray capacitance when adding the ne5532 of the datasheet on the pcb the closest as possible?
In the same fun TDA1541A has not the DEM timing cap embeded...and embeded in the little non-A batch. Some prefered that one vs the A withe same 470pF value...
To say than the layout MATTERS.
Not sure r2r discrete is a good idea but marketing. I do not know some Say Holo May sounds quite good as a partialy discrete with FGPA. But notice not à same gound sound between pcm or DSD as well.
Last edited:
I don't understand what you mean, but it certainly is possible to integrate resistors.
The resistivity of silicon is measured in Ohms/square.
The actual resistance then becomes the length/width ratio, so for an R2R DAC, the length 2R is twice the R.
Dimensions in lithography can be accurately controlled, even if the resistivity itself cannot.
The actual resistance then becomes the length/width ratio, so for an R2R DAC, the length 2R is twice the R.
Dimensions in lithography can be accurately controlled, even if the resistivity itself cannot.
2R would rather be two unit resistors in series than a longer resistor. Matching is best when you use unit cells.
Yes it is posibile, but I dont know almost any dac that has integrated to many Rs inside?
There is https://www.analog.com/media/en/technical-documentation/obsolete-data-sheets/AD1865.pdf page 4: "The 18-bit D/A converters use a combination of segmented decoder and R-2R architecture to achieve consistent linearity and differential linearity. The resistors which form the ladder structure are fabricated with silicon chromium thin film. Laser trimming of these resistors further reduces linearity errors resulting in low output distortion."
Thanks. I will check if there is an information in other DACs data-sheets. So AD1865 is R2R ladder dac.
Pretty much all DAC chips were. It was the first DAC technology. When CD started off PCM56 was literally found in all brands except those that used Philips chipsets/mechs so TDA1540 + CDM0/1/2.
Such devices were fed by strange devices with copper wire and steel but that is another bedtime story. Good night!
Such devices were fed by strange devices with copper wire and steel but that is another bedtime story. Good night!
Last edited:
Let me introduce you then with the famous R-4R ladder which goes with the sign magnitude design (segmentation is optional but preferred). Both schematics are effectively identical and the bottom one has just two resistors in series and two resistors in parallel everywhere (except just one R at the bottom left and its precision is not important at all).
This has been a fun read!
I just wanted to make a couple comments - IC manufacturers (like TI) still make R-2R DACs, just not really specialized for audio. Most audio products also have significant digital content (digital filters, for example) and make more sense to be manufactured on a process node that has smaller digital capability. We make our R-2R DACs on precision analog process nodes that offer great matching. These enable very precise resistors, but not necessarily accurate resistors. We dont do a lot of laser trim anymore, but we have lots of e-trim capability. We use a mix of segmentation and R-2R ladders to acheive or highest resolution device. Our process nodes have improved so much, that we are now able to produce 16-bit string DACs as well.
Some of our latest R2R precision DACs, like the DAC82002 or the DAC11001B are great for audio, but that was not the intention when designed. Now the biggest issue when using an R-2R precision DAC is solving the digital headaches for I2S to SPI 🙂
PS I work for TI for our Precision DAC product line in Tucson.
I just wanted to make a couple comments - IC manufacturers (like TI) still make R-2R DACs, just not really specialized for audio. Most audio products also have significant digital content (digital filters, for example) and make more sense to be manufactured on a process node that has smaller digital capability. We make our R-2R DACs on precision analog process nodes that offer great matching. These enable very precise resistors, but not necessarily accurate resistors. We dont do a lot of laser trim anymore, but we have lots of e-trim capability. We use a mix of segmentation and R-2R ladders to acheive or highest resolution device. Our process nodes have improved so much, that we are now able to produce 16-bit string DACs as well.
Some of our latest R2R precision DACs, like the DAC82002 or the DAC11001B are great for audio, but that was not the intention when designed. Now the biggest issue when using an R-2R precision DAC is solving the digital headaches for I2S to SPI 🙂
PS I work for TI for our Precision DAC product line in Tucson.
Wow - interesting, Paul. I'm a valve guy and only know the basics of digital technology, but I'm curious to know whether TI is working on a new generation of chips that can be used in hi-end audio DACs? Is the Eval board for the DAC82002 relevant for audio, for example?
https://www.ti.com/tool/DAC82002EVM
https://www.ti.com/tool/DAC82002EVM
Indeed. My first job was at Philips Research Labs. Lots of interesting stuff going on like development of TFT LCD screens, can never forget seeing a postal stamp sized black and white TV screen with half the pixels working 😱
We are always working on the newest, latest, and greatest Precision DACs. My team does not focus on any specific audio DACs, but TI does create audio DACs and CODECs. We do have some new stuff in the precision DAC space, but I cannot disclose just yet.
For the DAC82002EVM, we added some flexibility for the output buffer to allow some mix-and-match testing. The EVM uses an asynchronous USB-to-SPI controller that would not work well for audio. Instead, we bring our digital signals out to a header so that you can connect it an FPGA or your controller of choice. The issue is that it is not trivial to convert I2S to SPI.
User "iitzex" is working on the FPGA to do this.
Is the tech from today is enough to use now PWM for digital to audio conversion with high resolution enough (21 bits?) as They did with the MOLA MOLA Tambaqui ? They did it but seems complex (4 FGPA, etc)
It seems there is an interrest at the view of this devive reviews?
It seems there is an interrest at the view of this devive reviews?
beyond my head... Is "plain old PWM" limited to 10 bits by design ?
Another way to ask if SD is the necessarily guest to make any dac today ? (or at least why many try to avoid it as the plague)
Another way to ask if SD is the necessarily guest to make any dac today ? (or at least why many try to avoid it as the plague)
PWM has problems by itself. You can get noise/distortion artifacts because its not randomized like sigma-delta modulation can more or less do.
However, sigma delta is not needed for some types of dacs. An R2R dac doesn't need it.
Regarding trying to avoid SD modulators, they can sound great if well designed. I guess it was mostly that ESS made a lot of dac chips that measured great but didn't sound so great, and they use SD modulation to do it. Doesn't mean the same problems will happen going forward. Marcel's RTZ dac with the PCM2DSD modulator can sound most excellent. And it can play using sigma delta PWM similar to what Mola dacs do. Not saying its cheap and easy to make it sound its best though.
However, sigma delta is not needed for some types of dacs. An R2R dac doesn't need it.
Regarding trying to avoid SD modulators, they can sound great if well designed. I guess it was mostly that ESS made a lot of dac chips that measured great but didn't sound so great, and they use SD modulation to do it. Doesn't mean the same problems will happen going forward. Marcel's RTZ dac with the PCM2DSD modulator can sound most excellent. And it can play using sigma delta PWM similar to what Mola dacs do. Not saying its cheap and easy to make it sound its best though.
Last edited: