For more than a decade, the ADI's AD5791 was the king of precision DAC until Ti took the crown over by introducing their rival DAC11001B in Dec'21. The INL, DNL & noise spec are just fascinating, any interesting project on-going?
https://www.ti.com/product/DAC11001B
The AD5791B and DAC11001B are about comparable from the specs. So nice to see come competition at the high end, though the 2 are not direct drop in replacements.
Looking at the AD5791 data some of us gathered
here and the typical INL chart shown on the DAC11001B datasheet, both DACs have about the same INL over all 20 bits (1.4uV/V).
The AD5791B has a 6-bit segmented region and a 14-bit R2R region. This is both a blessing and a curse. Curse; there is mismatch between the 2 sections, the INL goes from 0.6uV/V for just the 6-bit segmented region, to 1.4uV/V for all 20 bits. Blessing; if you just use the MSB 6-bit, you improve linearity a lot (if that is the key design requirement rather than bit depth).
TI sometimes hide lower level detail in their datasheets so we don't know if the same trick can happen until we see some real data.
A couple decades ago Agilent employed the 6 bit MSBs with trimmed uniform current sources, and the lower bits with R-2R ladder in their Griffin 16GSPS DAC in a SiGe Silicon BiCMOS Process.
This architecture seems to be popular with precision DACs!!
The Griffin DAC utilized Dynamic Element Matching which dynamically interchanged the 6 MSBs in a pseudorandom fashion with "noise" pushing to move the mismatch "noise" to out-of-band frequencies. Remarkable achievement and a high performance high frequency DAC that later became the center-point in their high frequency AWGs.
Anyone know if these precision AD or TI DACs employ DEM techniques, we didn't spot such in the data sheets.
Best,
Going a little off topic but hopefully still on trend for 'precision analog ICs'...
Anyone know if these precision AD or TI DACs employ DEM techniques, we didn't spot such in the data sheets.
Not a DAC, but best-in-class ADC AD4630 uses a DAC internally with DEM. See DOI: 10.1109/ISSCC42614.2022.9731652 for a decent level of detail of the internal working of this ADC. I have measured the INL of this ADC to match very well with the curve shown in the paper, in the region of 0.35 to 0.5 uV/V with a +/-15V range analog frontend. Fantastic for a minimal part count solution IMO, could easily make a 6.5d DMM using the same ADC for both NPLC and digitizer modes.
That's a nice ADC according to the data sheet, couldn't fine the paper via the link tho.
Recall Agilent used something like 80 individual precision uniform current sources for the 6 MSBs, then twiddled these to produce the desired result with DEM. The lower 8 bits were conventional R-2R ladder.
BTW think some of the newer 6 1/2 digit DMMs from the lower tier OEMs utilize a single chip ADC for the NPLC and higher speed digitizer functions as you've mentioned, altho KS and Keithley rely on a discrete ADC for the main ADC I believe.
Best,
Have the utmost appreciation for the enormous intellectual creativity & power involved in these chip designs with Jesper Steensgaard leading the efforts at AD!!
Best,
