LM399 Reference - Heater Current

[drazziweht]

I am trying to build a 399 Voltage Standard mostly as a reference for a DIY SMU.

Initial results seem to be worse than expected !.
I have just a LM399 connected to a lab 15volt PSU with heater across the supply and zenner across the supply via a 3K resistor.

All in a polysytene box to minimise drafts etc.

Results of voltage and heater current below.
Voltage in the graph is after subtracting the constant 7.1345 and shows about 30 microvolt variance over 24 hours, I had hoped for better !!
The odd thing is it seems to be highly correlated with heater current, is this nurmal or have I done something 'wrong'

The Room temp varies by about 5 degrees over the measured period.

Measurement is via a HP 34401A for the voltage and the built in current measurement of the 6632B PSU for the current.

Note I actually have 2 LM399 hence the 31ma. They are connected in parellel with the ouputs summed by 2 3K resistors to the DVM, this to try and reduce noise but the results were almost the same with a single unit.

[branadic]

Why 3k and not 7.5k? The 3k is valid when having the reference in a 10V output bootstrap configuration.

-branadic-

[razvan784]

Can you explain, or provide a drawing, showing how components are connected (including how the wires are connected) and where exactly you are measuring the output voltage?
It is normal for the heater current to vary with ambient temperature, but the voltage drop caused by it in the ground wires should not sum with the reference voltage. That is why ideally you'd want to measure the output directly at the LM399 pins. This technique is sometimes called a "Kelvin connection" in similar contexts.

[drazziweht]

This is how it is connected.

I will check to see if the heater ground current path is on the right side of the meter connection/move it to the actual 399 pins directly to remove any doubt.

The variation in heater current is 1ma with a voltage variation of 30uv so an implied ground resistance of 0.03 ohms, which might be the problem !!

The 3K resistor is because I started with a supply voltage of 10V intending to move to a bootstap config..... From reading it does not seem to be a very critical value ?

Will rejig the DVM connection to see if heater ground current is what is causing the problem.

[drazziweht]

Revised the DVM Connections to look like this, will see if it makes a difference

[Kleinstein]

The LM399 still reacts a little to a change in the current. A change in the supply voltage can effect both the heater current and ref. current and this way also the ref voltage.  The direction of the seen correlation fits: lower supply voltage gives more heater current (about +2% in the curve), if from a drop in the votage this would be 2% drop of some 30 mV. With 3 K this would be 10 µA  - so this would be some 2 Ohms of differetial impedance, which is a bit on the higher side.

With a new reference there can be some change for the first 10 hours. Only later the dirt rate goes down. 

[miro123]

General speaking LM399 is quite easy to easy device.
Few thiks must to be taken into account

1. Avoid current loops with kelvin connections
2. Zener line must be as symmetrical as possible. LM399 Kovar pins creates termocouople with the PCB = 35uV/C
3. Proposed higher value bias resistor has many positive effects. Lowe sensitivity to power supply is one of them.

BTW 30..32mA@15V seems to be ok for two lm399

[drazziweht]

Thanks for the hints.
It turned out layout was the problem heater current via ground was causing most of the error.
Not quite got my head around how careful you have to be if you want microvolt accuracy :-)

I seem to have one 'bad' LM399 and 2 good ones out of 3 purchased from RS.
The bad one has random 40 microvolt noise spikes downwards around once per half hour.
I guess even the bad one is is well within the data sheet spec, just noticable worse than the other 2.

New DVM (fluke 7510) arriving tomorrow so will be able to see the data better.

[oz2cpu]

suggestion, have a look at this zener diode implementation

at 24:00

https://youtu.be/pspuqOsWG44

the idea is the zener voltage, is used to make an amplified voltage, that drive the zener via a resistor
this means the current is now constant and stable, add this idea to LM399 to make it even more fantastic ??

[Kleinstein]

The LM399 internally regulated the current to the actual zener inside. The current is also derived from the zener voltage, at least in parts.
The current to the ref side of the LM399 is usually generated from an amplified (e.g. 1.5 x) the ref voltage. So this kind of bootstrpping is already standard practice.