http://www.pa4tim.nl/?p=3624. Prema 5017 7.5 digit benchmeter. Very solid build. The meter does not switch of totally uing the front powerbutton, the reference stays powered.
Teardowns not on this kind a stuff, only if i have to open it because of repairs. I do not like more dirt and dust inside just for pictures ;-)
I see the link to the modified 7081 allready is given by fellow Voltnut Fabio.
I have a Solartron 7061. I made some modifications too. I have not replaced the reference diode, if I remember well Fabio has done some experimenting in biasing it for better tempco. I just did a check up according the manual. And then I mounted a heatsinks on the outside with a fan on it. I made a temperature controller to drive the fan. The sensor is mounted inside de meter and the fan spins slow around 18 degrees roomtemp, speeding up if roomtemperature changes. It keeps the inside of the meter ( at the measure point) at 39 degrees.
The biggest problem with voltnutting is environment. My lab changes temp from about 17 to 33 degrees over a year and relative humidity from 20 to 40 %
If you want to keep everything tracking you need climate control. But that is not easy and i think costs a lot. Those of you in a all year long hot country probably have airco so maybe it then is possible.
I did not know the appnote number from the Williams reference , it is in my book Analog circuit design part 1, but i looked it up for you, according the book It should be number 86 a standards lab grade 20 bit DAC with 0.1 ppm/C drift. Look at the picture of the setup, a bunch of 3458, two KV dividers and his reference, that is a very expensive table.
My LM399 is made this way. I also used two KV's and the bufferamp. But I made a 10 turn potentiometer on the front that gives me about 800 uV. This for easy adjusting abd keep my KVs free for use with the 332. ( one for deviding, the second for checking) This pot is the weakest part. The tempco is probably horrible. And mist of the drigpft at this time probably comes from this. If i am going to make the oven for it, the portentiometer will be changed. I will make is so that 10 turns give me about 100 uV in that case tempco will not do much because one full turn (1K) will give 10 uV so if the tempco is 1000 ppm the resistance of the 10K pot will change 10 Ohm in total. So maybe this gives a 0.1uV fault / degree ( if i do the math correct, i have dyscalculus and no calculator at hand right now)
On the other hand if I ditch the pot I maybe do not need the oven. Resistors I use are 0.05 to 0.01 % from dead a Fluke 8500 benchmeter. There is guarding, i avoided mechanical stress ( the lm399 anf LTC1052 are mounted on sockets but without the ( leaking and dielectric) plastic. So no thermal stress to the components. After this everything is cleaned with ipa.
Only problem i have to solve is EMC ( if i power down everything except the prema or solartron and the LM or 332 or whatever and use a KV i get a nice linear behavour. So 1uV is 1uV on the meter and 10V is 10V. But if I leave the lights on ( halogene on a rail) everything goed wrong. There is about 150 uV deviation over 10V in that case. And the strange thing the readings go down. It dies not matter if I use the 332, LM or whatever. The meter straight on the 332 and using its own divider gives no problem so it must be the cables from source to KV and from KV to meter ( but shielded, twisted, coaxial, triax ect all give the same problem.
Welcom to the darkside of voltnutting ;-)
Home
Help
Search
Login
Register
