LM3999

[GigaJoe]

Hello everyone,
I'm curious about LM3999, it a poor man LM399 in to-92 package, made by NS, discontinued. Is anyone had some experience to use it ?
i got 5 to try for around $1 it seems pretty good for this price, considering tempco and a simple heat shield to add on ; but have no clue about long term parameters; ...  but the source depleted already for this price.  Does it make sense to look at for more of this thing, or AD586 ( 587, 588) would be a better choice ( but i really love a heater) ...

or find a source with "group buy" if anyone interested in ... ( im thinking around $2 if lucky )

[Andreas]

Hello,

one problem might be the common heater ground pin.
So the zener output voltage is dependent on the current for the heater.
Assume 10 mA current variation times 10 mOhm for the bond wire (best guess)
= 100uV = 14 ppm output voltage change.

so why not use a LM399?

with best regards

Andreas

[GigaJoe]

Great,
completely miss this common connection effect ...
it consumed around 20ma "naked" and around 10-13 ma in 1in polystyrene cube
heater fluctuation around 1-2 ma ...it means ....  1-2 ppm fluctuation already taken by heater ....

if we can use, let say .... 20 , can averaging help, would it be better then 6 LM399 ( about the same $ spending ) , theoretically,  putting aside of resistors and schematics inperfection ?

BTW, i did fire up just one, it gives me 6.95... cold; and 7.171... hot ; I'm wonder if higher Vout tend to higher drifting ....

[Kleinstein]

The common pin for the heater current is a real problem. So if at all it makes sense to use it in a well isolated environment. If you really want several in parallel, I would even consider a common isolation around the whole circuit - so let the whole board run at maybe 50-60 C. This way heater current would be lower and the temperature of the rest of the circuit would also change less.

With an external heater, one could use LM329 - this is essentially like the LM3999 but without the heater. An external heater temperature could be a little lower and thus less prone to drift.

For the voltage: the TC of the unheated reference correlates with the voltage. So units with a certain voltage might be better - higher and lower voltage tend to have more TC (unheated). However with the rather good internal heater the TC is essentially not important anymore. Also the TC depends on temperature.

If it is for a cheap, low noise reference, one could consider the chinese 2DW23x zeners and add thermal stabilization around.

[wraper]

Hello,

one problem might be the common heater ground pin.
So the zener output voltage is dependent on the current for the heater.
Assume 10 mA current variation times 10 mOhm for the bond wire (best guess)
= 100uV = 14 ppm output voltage change.

so why not use a LM399?

with best regards

Andreas

I don't think there is only one bond wire, for that to matter. Most likely, separate bond wires for heater and zener are used. So lead and solder joint resistance is what matters.

[Andreas]

BTW, i did fire up just one, it gives me 6.95... cold; and 7.171... hot ; I'm wonder if higher Vout tend to higher drifting ....

yes, the sweet spot seems to be at 6875 mV (see also LM399 thread).
https://www.eevblog.com/forum/metrology/lm399-based-10-v-reference/msg854683/#msg854683

And I think best stability will be in hermetical metal can case,
with plastic you will have common cause effects (humidity, plastic creeping..).

with best regards

Andreas