[time-nuts] Voltage standards

[Bruce Griffiths]

WarrenS wrote:
> A Time-nut group may not be the best place top discuss Voltage references. 
> Where is the Volt-Nuts site?
>
> First thing to be aware of if you are planning to build your own reference is:
> Good voltage reference parts are "SELECTED" which is just another word 
> for cherry-picked. My better modified AD587LN Geller Lab voltage 
> reference's are good to 0.1 PPM transfer accuracy over a 5 degree F 
> temperature change and a 1 Hr time range, They also stay with in about 1PPM long term. 
> I  think that  will 'qualify as a voltage standard'  to some.
>
> The TC correction circuit I use is just two standard parts. 
> A100K pot is placed between an added transistor's collector 
> and ground with the pot's wiper going to the transitor's base.
> What that does when current is applied to the transistor is to make  
> the collector voltage vary with temperature by an adjustable amount. 
> This temperature sensitive voltage is then added with the trim voltage 
> to the reference's trim input to cancel the 1st order temperature drift.
> The results I got was a settable 4 degF wide ZERO TC, 
> turnover point at any desired room temperature.
> The added parts fit on the small Geller's PCB.
>
>   
In other words an adjustable VBE multiplier??
> The other suggestions that I have heard so far on this post take more skill 
> and additional precision parts that quite frankly the average home builder 
> is not going to have.
>
> I do not mind posting my mod and results on line, if soneone will tell me a place to do it.
>
> Some notes from my own  'make a state of the art voltage reference' efforts.
> I.ve tested everything I could get my hands on, looking for simple solutions.
>
> 1) All AD587 are NOT created equal, but the 'noise' of the better ones is in 
> the same general range as the 732B &/or the 3458A.
> The best AD587 parts show that calibration and measurement is settable 
> short term to better than 1uv (0.1 PPM).  
> The poorer parts  show approx 4 uv step pulses every so often, 
> but even then, If using a strip chart recorder, it is no problem to see 
> their non stepped value to 1uv on most of the ones I tested.
>
> 2) The existing best references need to be kept ON to insure their calibration, 
> A better low cost solution for long term stability seems to be to keep it OFF most 
> of the time (using the right kind of selected zero warm up no hysteresis part).
>
>   
With a thermal conditioning cycle, before use, the LTZ1000 stability has
good stability (according to Fluke) even when powered off.
Implementing such a conditioning cycle would probably require using a
microprocessor to implement the conditioning cycle algorithm.
> 3) My conclusion on 399's after months of long term test is,. they are great for having  a 
> wide temp range low TC, but not so good for noise and long turn stability at the sub PPM levels
> Holding a part at 90 degC is how you burn them in to age them 64 times faster, 
> not how to hold good long term stability.
>
> 4) I personal didn't like LTZ1000's because their accuracy is very dependent on 
> the stability of the external 2 terminal 100 ohm resistor that sets its zener current. 
> If I remember right, it has only 10 to1 rejection so you still need a 1 PPM 
> resistor to get sub PPM performance, and that's an absolute value on 
> two terminals, not a tracking resistance.
>
>   
The LTZ1000 datasheet states that 100ppm change in this resistor
produces a 1ppm shift in the LTZ1000 reference voltage.
It is possible to use a 4 terminal resistor for this part to improve
accuracy somewhat as the base current of th compensation resistor is
around 1uA and its emitter current is 100uA.
This isn't ideal but the emitter current is about 40x lower than the
zener current.
If the LTZ1000 output is scaled to 10V any humidity sensitivity of the
scaling resistors will probably contribute more error than the 120 ohm
resistor.
> BTW, I have no connection with Geller Labs other than buying some of this parts.
> I must add that he was very helpful in my past quest for a good stable voltage source. 
>
> WarrenS
> **************
>
>   
>

Bruce