Hi Going back to the original math:
The objective is a device that is stable to about 3 ppm and has a noise level of << 3 ppm (if possible). Ideally one would want to get below 1 ppm on both numbers. A TL431 has a noise level of roughly 5 to 15 uV (depending on your sample) running at 2.5V. That puts it’s noise level at 2 to 6 ppm. If you take a look at the noise plots on the TI data sheet it’s pretty apparent that the bulk of the noise is at very low frequencies. Noise is climbing by the time you hit 100Hz and it’s at 220 nV / sqrt(Hz) by the time you get to 10 Hz. They don’t plot it below there, but the curve is moving up at a pretty good rate at 10 Hz. The 0.1 to 10 Hz plot in the same sheet shows a pretty active process going on. There are other parts out there with the same output voltage that have a *lot* less noise: http://www.linear.com/product/LTC6655 Yes, they do cost more money. Bob > On Sep 6, 2015, at 2:38 AM, Perry Sandeen via time-nuts <[email protected]> > wrote: > > Hi, > > Wrote:No, the temperature is not stable, not even close. > > Wrote:Given that the TL431 is quite good at measuring temperature, that > makes itunsuitable in this case. > > Wrote:There is a LOT more to making a good temperature-controlled > >environmentthan hacking together a DIY oven. Thatmay be fine for stability > of tenths of a percent (10e-3), but when stability of > >> 10e-13is at issue it doesn't pay to start with 50ppm parts and put them in a >> DIYoven. You need to start with low-tempcoparts. > > Waita second good people! We have a bit of confusion about what I wrote. > > Anamplification of what I wrote: > > TheTL 431 is listed as a low noise shunt regulator. It doesn’t measure > temperature at all. Theonly noise spec I could find is about 7 uV. A LM723 > data sheet specified 2.5 uV noise. The REF 012A specifies 5 uVnoise. The > first two are around 50 cents from Mouser. The REF102A is about 10X more. > > I’mnot sure the original reference diode had that low noise but that’s a > guess onmy part. > > Thetemperature circuit I referenced uses a 220 ohm load resistor in series > withthe positive output of the regulator and Vcc. The temperature is > regulated by a resistor from Vcc to the programminglead. Then there is a > thermistor going from the programming lead to ground. > > Usingthe formula given in the data sheet. One can then calculate the desired > voltageratio and then with a series control pot set the temp to the desired > value. > > Thisis a simple circuit and with 1% 100ppm vishey metal film resistors it > would begood for .1 degree stability. That should suffice for Paul’s > application > > Charlesis correct that if one is getting into the TN class of .01 or .001 > degreestability a far more elaborate circuit is required. > > Goingback to my original post, this was only a suggestion for an > inexpensivesolution for the problem. Since I don’t have a 5065A, I can’t > prove or disproveif it would give the desired results. > > WhenI get them set up, my demon to solve is how to get my 5370B’s to be cool > andquiet. There has to be a better solution to the PS heat. > > Regards, > > Perrier > > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
