Hello zlymex,
it is really a pleasure to see, read and learn from your great experience and knowledge!
Your DIY 10k seems to be extremly stable.
I wonder which combination of resistors (parallel, series or comb.) gives the very best resulting TC and stability.
Also I am interest to know how to meassure and use the alpha and beta values.
Thanks a lot for sharing
Bye
quarks
Thanks quarks,
I can open a new thread for DIY resistors - I made some before and this one is probably my best.
This particular one, made from selected 4 Vishay VHP101 hermetic resistors from a batch of 50, 9.999k, 0.01%.
I first tested those resistors for values and tempco for sometimes and type in the data in Excel, selected 4 best matched base on alpha and beta, then its not necessary to use compensation late, though I do have a spread-sheet way of compensation for tempco and value, which have been used in other of my DIYs.
Connect 2 in series first, and then parallel. This formed something like a bridge, those orange wires are for 'null detector' output.
As far as the performance is concerned, parallel first or series first is exactly the same, provided they are all equally weighted(same power). I believe that 1k to 10k range is the best for Vishay foils resistor as the value is concerned. I may have ordered the value at 2.499k but that will eliminate the option for single-resistor use at 10k.
I didn't design, select or control the stability, it mainly depends on the resistors themselves, just lucky to be this stable.
I did some small range temperature cycle before hand of -18 deg C in a fridge and about 80 deg C hot water bath, just for mental comfort.
Made sure not to introduce any unstable elements into it by some considerations like long-lead, less stress. When soldering use thick tweezers with heat conducting paste to prevent it reaching the body.
Well, the leads is copper, with about +4200ppm/K tempco, had been taken into account during selection and measurement.
No heatsink, thermal equalizer or grease were used because I was thinking those may do more harm than good for stability, and because it only dissipates 10mW, it can be handled well with four elements.
I do use Teflon board and dual case, the inner case is electrically isolated and also is the guard output.
Those small resistor along side were German brand, Czeck made Vitrohm of 0.1%, and also very small in value, in series with VHP101 to facilitate the adjustment to final value. I ordered those VHP101 at slightly less than 10k for that purpose. I do sensitive analysis for each components to make sure they don't affect the end results in an unexpected way.
The binding posts was taken out of my Dataproof scanner(164A), very low thermal EMF according to them.
This particular standard have been traveled in China for about two years among my Ohm-nuts friends and return to me recently.
For tempco measurement of alpha and beta, put them in a car-fridge and run a temperature cycle, record value and temperature. The measurement is always done by my own scanner/switch, where I connect my two SR104 as a reference, data was feed to 3458A direct or indirect(by a Warshawsky bridge), and against one of my SR104, to eliminate the drift of 3458A.
Then, plot the temperature-value scatter diagram in Excel like this:
![](https://www.eevblog.com/forum/metrology/teardown-standard-resistors/?action=dlattach;attach=207525;image)
This is an example, but is the actual plot of a BZ3 1k standard resistor, also see the attached spread sheet for how to plot and calculate.
Then make a trend line of second-order polynomial(red line, by right click the data), this is the least-squares regression of the best fit of the scattered data.
Also, there is this 'display equation' and 'R square' option of this trend line in the form of
y = c + bx + ax^2
where y is the resistor value and x is the temperature
the R^2 is the measurement of the fit quality and should be close to 1. It is also the indication for hysteresis.
beta is immediately available: beta = a/c
actually alpha0=b/c but this alpha is at zero degree C, we normally use alpha at 23 degree C, so a little conversion has to be done in spread sheet(enclosed) to transform the equation to the form of
y = c1 + b1*(x-23) + a*(x-23)^2
and finally alpha23=b1/c1
In order to make the good result of alpha and beta, use slow temperature cycle to reduce hysteresis, use a good temperature sensor close mounted with the resistor. I use DS18B20 for sensor(part of the data collection system) and checked it by ice-water and mercury thermal meter of the Chinese 2nd grade(to 0.1 degree C).
SR104 is regularly calibrated. Recent is done at Hong Kong HKAS two months ago. I also monitor the thermometer and made small corrections if necessary. As for the measurement of alpha and beta, absolute value of the resistor reference is not critical at all.
I measured alpha/beta of one of my SR104 this way and the procedure/result were on 38hot.net
http://bbs.38hot.net/thread-1352-1-1.html(Sorry its in Chinese, may be you can use google translate)
To use the alpha and beta, just measure the temperature and use the formula y = c1 + b1*(x-23) + a*(x-23)^2 to obtain the exact resistance at that temperature. Or better make a temperature chart(like attached) to get the resistance quickly.
The tempr chart in the inner case of SR104 is not good because it does not show the deviation explicitly and is error prone when add it up.
![](https://www.eevblog.com/forum/metrology/teardown-standard-resistors/?action=dlattach;attach=208431;image)