On 10/18/2012 02:05 PM, Andreas Jahn wrote: > Hello Bob, > > whats the matter with you. Infected by precision virus like me? > You wanted to have a standard with about 10ppm and now you blame a 3-4ppm > drift.
:-) - afraid so Andreas! >> >> The LM199A is hermetically sealed. >> > > The PCB, the 8K Resistor and the voltage Regulator are not. > >> Although a small sample, the two references appear to be similarly affected >> by >> whatever caused the drift - similar range of drift, similar time constant. >> > On the first view I would blame it on the meter. > It is very unusual that the drift of 2 different references has nearly exact > the > same amount of ppm and direction. > But on the other side you state that there are several HP3458A which recorded > the drift. > It is not probable that all came freshly from calibration of a other location. > >> Any ideas about what could cause the drift we are seeing? > >> From time constant it could be the humidity change. This is my best guess. > My 2 LT1027CCN8-5 references which ara mounted only with 1 Pin > to the PCB have time constants in the range of 4-7 days. > The epoxy material of a PCB should lie in the same ball park area. > The LT1027 are influenced by around 0.5 ppm per percent humidity change. > > For the hermetically sealed brand new references LT1236AILS they state in > their > new product catalog > a humidity change of less than 10ppm for 25% humidity change. (page 36) > http://cds.linear.com/docs/Product%20Info/NPC.pdf This is most interesting - so even "hermetically sealed" units are influenced by humidity! > I asked them whether this is from mechanical stress from the PCB and they > confirmed to me > that with a dead bug mounting the influence of humidity will be virtually > unmeasurable. > So they will delete the parameter from the data sheet. Mine are not "dead bug" mounted. The 'PCB' is in fact a Radio Shack perf board - certainly not the best substrate to mount them on - I don't think it is FR-4 material. > So for the LM399 it might be mechanical stress introduced by the PCB. > > > When looking at your cirquit there are several points to mention: > One common failure source will be the LM78L15. A output voltage change will > influence the supply of MAX6350 and the reference current of LM399. > PSRR of MAX6350 is about 2-5 ppm/V above 10V supply. (without self heating > effects). LM78L15 spec is 1mV/C. This would result in .02 ppm/C on the LM199A (operated at 1 mA with 8K resistor providing the current). For the MAX6350 1mV/C and 5 ppm/V => .005 ppm/C > The LM399 resistor will give a current change of about 10% per Volt (100uA) > resulting with 0.5 Ohm impedance in about 50uV/V or 7ppm/V see note above. > Other weak points of the cirquit are: > The LM399 heater voltage is not stabilized. this will give about 0.5ppm/V It is stabilized by the power supply - a 24 volt, .02%/C unit => 5mV/C => .0025 ppm/C. > And finally: was the LM399 always in the same orientation during measurements? > (will be difficult with a cylindrical housing). > My LM399 drift 3-4 ppm by tilting orientation. Not sure - but from the consistent results (consistent drift and apparent settling) it likely was. The above still leaves humidity induced PCB changes causing mechanical stress as likely. I am going to look for some better board material for when I construct TS2. Thanks Andreas! -- "As we act, let us not become the evil that we deplore."
<<attachment: smither.vcf>>
_______________________________________________ volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts and follow the instructions there.