[volt-nuts] (no subject) -> LM399 Reference

[Andreas Jahn]

Hello Jan,

The LM399 is being far from bad as stated in the datasheet.
The 10ppm/khr (8-20ppm) is a figure which is very conservative.
(perhaps directly after soldering process).

Most of my LM399 are within 1-2 ppm/khr after a run in phase of 1-2 khrs.
And the typically drift of instruments with LM399 as reference
have about 1-2ppm/year drift after some ageing.

A large part of power supply sensitivity on output voltage comes
from heater temperature setpoint dependancy on heater voltage/current.
The effect is larger with lower supply voltages. (especially in the 
9-10V range).
You should always use a well stabilized (some applications specify 0.1%) 
power supply >= 15V with low tempco for the heater.

With best regards

Andreas




Am 29.03.2014 10:47, schrieb Jan Fredriksson:
> I was looking at a few different alternatives for a transportable,
> non-permanently powered on, DIY, voltage reference and I picked the LM399
> as one of the candidates.
>
> The LM399 is a buried zener reference with built-in temperature regulation
> and an integrated thermal insulation cover. It is very simple to
> implerment, only one resistor needed.
>
> The downsides:
> - it has a nominal drift in the order of 10ppm/sqrt1000h (as good as any
> except the LTZ1000)
> - Its a bit sensitive to input volgage, due to a 1ohm input impedance.
> - Its a bit noisy, about 10uVptp.
> - Output voltage tolerance is poor and at around 6.9V
>
> I set one of these up and, after a day of stabilizing, found the noise
> about as stated in the datasheet. However, noise in the datasheet is
> specified for 10Hz and up. The real problem to was at lower
> frequency. Overnight, the 10s averaged values where slowly drifting around
> about 10uV ptp.
>
> Now I made twelve LM399 parallelled on a simple breadboard, running at
> 1.8mA each, plus heating, a total of around 200mA.
>
> Like the first, I use simple 2K7 1% metal film resistors for current
> limiting at 12V supply. The resistors do not need to be very high-spec
> as errors are attenuated by a factor 1:2700 (1R/2K7). There is another 2K7
> resistor per zener for output averaging.
>
> This board, after a few hours stabilizing, measures 0.2uVptp, 10s averaged,
> over 6 hours.
>
> My 7.5digit NI PXI-4071 DMM has 0.1uV resolution on the 10V range. So I
> need a better method... But anyhow, id say it looks promising. It seems
> like the low frequency noise is cancelling well and does not come from the
> power supply.
>
> I will now leave this board for a half-year burn-in.
>
> Back to the original idea; I am assuming that much of the long term
> drift is due to the high, fixed to 90C, internal heating temperature in the
> LM399. To find out; after the half-year burn-in, I will turn off half of
> the board (six zeners) and turn them on for something like a few hours /
> times per week only, while leaving the other half on all the time. Over
> another half year, I should be able to see if there is any significant long
> term difference between the two sides.
>
> If whole idea turns out sour, it's still be interesting to me as a volt-nut
> wannabe. The construction is simple and the price of a dozen LM399 is less
> than one LTZ1000 with required precision resistors...
>
> At some point I will probably make a proper board with built-in power /
> voltage regulator, output buffer etc.
>
> In the mean time, I will build a few other boards with other ICs, including
> the LTZ1000.
>
> I will be posting progress now and then.
> _______________________________________________
> volt-nuts mailing list -- volt-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts
> and follow the instructions there.