Ti's ina1620 integrates a thin-film resistor. It is said that the accuracy of this resistor is very high. Is it more accurate than the resistance corrected by laser in the R2R chip? If the semiconductor process resistance is more accurate, does it mean that TSMC is theoretically capable of manufacturing R2R DAC chips with 32 bit accuracy? Is the cost lower than the previous manufacturing process?
The DS says the matching is 0.004% which is pretty good - it suggests about 15bit performance if used in a DAC.
Reading a bit of the fine print it turns out that this is a typical matching, the worst case is 0.02% which becomes about 12bits. Reading further it turns out that this figure only applies to the matching of the two resistors in a particular pair. So far I'm not seeing any spec for the matching between pairs, it could be as poor as 15%. Without a between pairs spec it wouldn't be a good idea to try building an R2R DAC from a few of these chips.
Reading a bit of the fine print it turns out that this is a typical matching, the worst case is 0.02% which becomes about 12bits. Reading further it turns out that this figure only applies to the matching of the two resistors in a particular pair. So far I'm not seeing any spec for the matching between pairs, it could be as poor as 15%. Without a between pairs spec it wouldn't be a good idea to try building an R2R DAC from a few of these chips.
Last edited: