Budget pecision resistor source

[Tj138waterboy]

I've been studying quite a bit on Precision resistors and ran across unusual characteristics that seemed to better than high-dollar Precision resistors with the exception of having fixed values in the ic. I am referring to the lt1997-3 op amp. Datasheet specs say "The resistors maintain their excellent matching over temperature; the matching temperature coefficient is
guaranteed less than 1ppm/°C. The resistors are extremely
linear with voltage, resulting in a gain nonlinearity of less
than 2ppm. Would this be a viable source for a transfer standard with tempco tracking being that good?

[SeanB]

The resistors are on the same die, and thus they do track really well temperature wise, and ratio wise, but your actual resistor value will not be constant over this range. As well this really only applies to equal value resistors, as they have to be the same physical size, and having big ratio's really does not track well.

[tszaboo]

You can buy these resistors separately, they are called LT5400. They are all right if used as dividers and such. Their absolute accuracy is not that great.

[Tj138waterboy]

So even if calibrated /trustworthy 7.5 to 8 5 digit meter or a cal lab could give me the exact resistor values at different temperature set points say from 20-25 degree c, I'm not understanding what else could affect the absolute values other than humidity and ageing.

[Zero999]

So even if calibrated /trustworthy 7.5 to 8 5 digit meter or a cal lab could give me the exact resistor values at different temperature set points say from 20-25 degree c, I'm not understanding what else could affect the absolute values other than humidity and ageing.

Temperature can affect the values, which will only be made to a certain tolerance. The LT997-3's resistors aren't particularly close tolerance. The data sheet specifies a common mode input impedance range of 19k to 26k, with 22k5 being nominal, which means the resistors have a tolerance of 15.6%; worse than even the cheapest resistors. What makes the gain tolerance so close is the ratio of the resistors is very tightly controlled, so if the feedback resistors are 20k, the input resistors will be extremely close to 20k, 6²⁄₃k and 2²⁄₉k, to give gains of 1, 3 and 9.

[RandallMcRee]

To emphasize the point, the  lt1997-3, in its datasheet is all about ratio versus temperature. The absolute TCR over temperature does not seem to be specified. Always a bad sign (if you think its important).

The technologically similar LT5400 specifies

Code: [Select]

(ΔR/R)/ΔT Resistor Matching [b]Ratio[/b] Temperature Drift (Note 5) l ±0.2 ±1 ppm/°C
but
ΔR/ΔT Absolute Resistor Value Temperature Drift (Note 5) l –10 8 [b]25 ppm/°C[/b]
see page 4.

You can often find S102K-type resistors on sale and/or surplus in various places. They are notorious for having great "typical" values but not-so-great actual max values, e.g. you can get a dog. If you *really* care you need to buy a bunch and choose. For example out of 15 AE 20K MQP (s102K from AE) https://www.newark.com/alpha-electronics/mqp20k000a/resistor-metal-foil-20k-smd/dp/96Y7872 I managed to put two together that had 10K +-0.1ppm absolute tcr. Thats the limit of my measuring equipment, though. For most purposes, I find the LT5400 to be cost-effective. Edit: for ratio purposes! It's best IMHO to avoid any circuit where absolute value is important!

Another option is to buy a high quality wirewound resistor or two from Edwin Pettis. Search the forum for his contact info. He can produce a good 3ppm resistor to a high specification although you would need to contact him for details.

[Tj138waterboy]

  thanks for the clarification I have contacted mr. Pettis in the past but I'm holding off for a more cheap alternative my requirements probably aren't as high as what he can produce as I only have a 5 1/2 digit multimeter that lost calibration I do have old fluke 931b differential voltmeter that im tearing down to repurpose and found fluke brand card wound resistors. Currently am l looking at ways to test the tempco of these amd debating of mailing them to someone on the forum if there is anyone willing to give me a measurement on them. Just didn't know if im wasting my time or if I should be looking at resistors that is just slightly better than specs of 3478a multimeter which I believe is aroumd 30ppm for 1 year. According to xdevs resistor test the card wound fluke are bad but not too bad comparing to vhp types from vishay.

[RandallMcRee]

Just FYI,
I took apart a Fluke differential voltmeter a few years ago and re-purposed many of the 1K WW resistors into a 13-bit DAC.

My measurements showed positive TCR from about 1ppm to 5ppm. The problem there is that the Fluke resistors are all from the same batch (as stated in the service manual). So its hard to find, for example a 1ppm resistor and a candidate -1ppm for TCR matching.

For my DAC I combined the Fluke WW with Vishay S102K obtained from ebay.

[Tj138waterboy]

I would think anything under 10ppm would be ideal and I will have to figure out the yearly drift on my own. Im pretty sure on some of the service manuals for high grade voltage standards the fluke differential voltmeter is called for in callibrating the ac ranges so I figure if they calibrate a calibrator it would meet pretty decent tolerences for a beginner. Eventually I will move up to a better meter such as 34401 or 3457a class but that would be as far as I go. Would the concensus be to test what I have or go for used ebay precision resistors? I hear good things about items from hifi seller on ebay.

[Tj138waterboy]

Also would it be ridiculous to make a 4wire standard enclosure with card type wirewounds or the s102 types? Keeping in mind this is for 5.5 digit and handheld meters.

[RandallMcRee]

Also would it be ridiculous to make a 4wire standard enclosure with card type wirewounds or the s102 types? Keeping in mind this is for 5.5 digit and handheld meters.

No, not ridiculous. The problem will be to get drift measurements using the equipment you have. Resistors are typically more stable than the DMMs you are using for the measurement, so you that is why you need a cal lab or a friend who is traceable etc.

something like this:
https://www.ebay.com/itm/ELECTRO-SCIENTIFIC-INDUSTRIES-STANDARD-RESISTOR-SR1-1-KILOHM-1/193259818310?hash=item2cff2ecd46:g:5SQAAOSwJmJe~O1f
should be a good investment if you can afford. The trouble with simply buying a quality resistor is that you then need someone to measure it for you--hence the market for something like a DMMCheck.
See: https://voltagestandard.com/

Maybe I will stop posting now. I feel like I'm leading you down the dark path of volt-nuttery.

[RandallMcRee]

Some volt-nuttery pics:

Measuring equipment, Keithley 2001, Keithley 2015, Advantest R6581T, HP3465A.

Pics of various types of precision resistors scrounged from old equipment and ebay.

Posting here because OP was curious.
The teal-colored resistors are also Fluke but they are much less fragile and less susceptible to humidity. The WW resistors on the mica cards are easily damaged.
The small black ones are Vishay Z-foil which is a different technology, not wire-wound, so ok for high frequency, etc.
Not included SMD types which are too small to show detail!