passive component tolerances for op amps used in front end of 18-bit ADC

Nicholas Kinar · 2009-08-04T20:39:37+00:00

Hello--I'm currently selecting resistors for the feedback loops of a few op- amps and in-amps used in the analog front-end of an 18-bit SAR ADC. The ADC reference is 5V, so the smallest theoretical voltage (with 0V to 5V input range) that can be resolved is 5V/(2^18) = 1.9E-5 volts.Page 719 of the the "Data Conversion Handbook" edited by Walt Kester gives an excellent overview of how resistor temperature coefficients can affect ADC errors. According to an analysis given in this book, resistors with a TC of 25 ppm/degC can be unsuitable for a 16-bit system.Now it would be possible to eliminate some error by using a resistor network with closely matched resistors. I would like my circuit to maintain 18-bit accuracy over a temperature range from -40 deg C to 10 deg C, which is a 50 degree range. Since the resistors are matched, the values will change in a similar fashion over temperature.I think that I might have to use extremely high accuracy parts (which are also extremely expensive), but I am hoping that there would be some other way to maintain accuracy.Could anyone recommend a manufacturer or source of resistors and capacitors which could be used in an 18-bit system? Vishay is the first company that comes to mind, but which series of components should be used?

Aug 04, 2009 24 Replies

Rate this thread:

Glen Walpert 16 years ago

Those Vishay Bulk Metal resistors:

formatting link

are made of a proprietary variant of 75Ni-20Cr-3Al, which is roughly an order of magnitude better than Shunt Manganin, and is the lowest resistance tempco material known in the vicinity of room temperature. One commercially available variant is known as Evanohm IIRC, available as foil and wire in case you want to make your own resistors :-).

When comparing tempcos of different materials it is important also to compare the specified temperature range or better yet look at the resistance vs temp curve.

Low resistance TC metal alloys get their low TC the same way Invar gets its low thermal expansion. Invar, at one specific temperature, exactly cancels thermal expansion with a gradual phase change to a lower volume phase. Low TC resistance alloys like Manganin and NiCrAl exactly cancel tempco at one specific temperature with a gradual phase change to a lower resistance phase. The further you get from the temperature which is exactly compensated the worse the tempco in either case. Specify TC very close to the compensated temp and you can make it arbitrarily low :-).

Regards, Glen

Vote

adriano.lopez 16 years ago

Not sure it can scale to the precision required (maybe far too gross in a 16-bit application), but sometimes amplifiers can be built ratiometric, meaning that the temperature coefficient of the gain- setting resistors does not matter as long as all resistors drift with the same coefficient.

Common inverting and non-inverting op-amp setups are of this kind, the gain being determined by the ratio of two resistors. So multiplying both for the same coefficient cancels out thermal drifts.

Once I applied a quick fix to a drifting amplifier replacing gain resistors with a multi-turn trimpot. My assumption was that being actually a single resistor halved by a sliding contact, both ends shared the same thermal behaviour. It worked in a 8-bit system.

Vote

JosephKK 16 years ago

this=20

Thanks. I appreciate it much better with the explanation of how it works.

Vote

Jim Thompson 16 years ago

How about OpAmp offset voltage? That'll do you in royally, unless you have an auto-zero periodically.

...Jim Thompson

Vote

JosephKK 16 years ago

for this=20

The few times i have seen using resistors like this, they were long since into fully ratiometric systems. With standard resistors in temperature controlled baths and all that.

Vote

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required