LM399 under temperature controlled environment in oil

• LM399 under temperature controlled environment in oil
  Posted by IRFP on 20 Aug, 2018 11:18
• Hello,
  Here is my prototype LM399 based DC voltage standard, dipped in a dielectric oil oven maintained at temperature.
  The device supplies the following reference voltages: 10V / 5V / 2.5V / 2.0 V / 1.0 V
  The results seem rather conclusive but I do not have a voltmeter with a precision higher than 5.5 digits to refine the calibration.
  The oil bath has much improved temperature stability.
  Thank you for your feedback and advice.
  here are some photos

  ---

  

  

  
• #1 Reply
  Posted by IRFP on 20 Aug, 2018 11:58
• The container is a paint box modified to receive electrical bushings.
  The oil is dielectric mineral type ISOVOLTINE from TOTAL.
  Control is provided by an OMRON controller in PID mode, the controller controls a dive power resistor.
  Here are some more photos

  ---

  

  
• #2 Reply
  Posted by guenthert on 20 Aug, 2018 14:45
• Most impressive. 
  
  You don't need a long-scale DMM to monitor the drift of your voltage reference, just a sensitive Null meter (the 5.5 digit DMM might be sensitive enough) and a (at least as stable) 2nd reference...
• #3 Reply
  Posted by IRFP on 20 Aug, 2018 18:42
• Good evening,
  Thank you for your feedback.
  I have 2 identical 5.5 digits multimeters, and -Is it possible to have more details on the long-term monitoring of the reference?
  At the level of the diagram, the drivers for the references 10V and 5 V are made from AOP type OPA177 GP from BB.
  For 2.5V / 2V and 1V references they are divider bridges based on 15ppm resistors.
• #4 Reply
  Posted by Kleinstein on 20 Aug, 2018 20:06
• Is the LM399 operated with or without heater ?
  
  Normally the TC is about the best of the LM399 reference. Due to the high temperature the dirft is not that good. Also noise of the LM399 is not super low and can be a problem. Another good property of the LM399 is that there is little hysteresis - heating it to operating temperature usually erases direct memory of the past, except for long term aging.
  
  For a 5 / 10 V reference the divider is usually the more critical part than the LM399. So having the resistors in an oil bath may make sense, even with a normal used LM399.
• #5 Reply
  Posted by IRFP on 20 Aug, 2018 20:58
• Good evening,
  Yes the heating is used as in a standard assembly. The milliampemeter shows the heating current of the LM399.
  The LM399 is immersed in the oil, which is heated by a power resistor controlled by a process controller T ° in PID.
  A thermocouple probe measures the temperature of the oil to regulate it to T ° +/- 0.1 ° C.
• #6 Reply
  Posted by coppercone2 on 21 Aug, 2018 00:15
• How is the EMI performance of that PID controller?
  
  I actually put LCD disable switches on my precision stuff for that reason.
• #7 Reply
  Posted by guenthert on 21 Aug, 2018 06:53
• Quote from: IRFP on 20 Aug, 2018 18:42

  [..]
  I have 2 identical 5.5 digits multimeters, and -Is it possible to have more details on the long-term monitoring of the reference?
  [..]
  

  You'll need another very stable reference to measure the difference, I'm afraid.  Call your current "Mark I" and build now "Mark II", perhaps with an LTZ1000 or multiple LM399 (or LM329, since you're building the oven already).  Then you can compare those.
• #8 Reply
  Posted by IRFP on 21 Aug, 2018 08:34
• 
  Hello,
  OMRON regulator is type E5GC. It meets the standard industrial EMI specifications.
  I have not yet made noise measurements, but it is possible that the control system impacts the reference.
  If you have solutions for this type of regulator I am listening.
  OPA 177GP operational amplifiers and voltage dividers are in a shielded enclosure.
• #9 Reply
  Posted by IRFP on 21 Aug, 2018 09:12
• Hello
  I already realized some time ago a reference 10 V based on LM399, the reference is thermally protected in the cavity of the housing, its stability is suitable.
  And it possible to use it in comparison?
  here are some photos

  ---

  

  

  

  

  
• #10 Reply
  Posted by Echo88 on 21 Aug, 2018 09:15
• Are you using the linear current output of the Omron regulator to control the temperature?
  If not, please try to use it, since the other two modes (relay and SSR) will with high possibility show up in your measurements as short noise-peaks.
  You could also use a discrete temperature temperature regulator build with OPs, like in this thread: https://www.eevblog.com/forum/metrology/small-oven-controler-for-voltage-reference/
  Its already very good that you use a linear power supply instead of switchmode-supply, since those would also generate measurable noise.
  Like guenthert said: you need a better reference voltage to compare it to your LM399-reference (must have better long term drift than a LM399) with a good Nullmeter or you need a 8.5 digit DMM. I dont know if a standard 5.5 digit DMM is suitable as a nullmeter. Ebay sometimes has good nullmeters like the Keithley 155 / Fluke 845A / HP 419A. 
• #11 Reply
  Posted by IRFP on 21 Aug, 2018 09:29
• Hello,
  Yes indeed I use the current output of the regulator in attack of an optocoupler which drives a simple power transistor type 2N3055.
  For power, it is well linear type, multi-channel and filtered with quality capabilities. I have also installed a head filter at the network level.
  The transformer is of good quality, it comes from an old multimeter KEITLEY 192.
• #12 Reply
  Posted by Wolfgang on 21 Aug, 2018 11:48
• ... just asking,
  
  when you spent all this awesome effort, why did you not try an LTZ1000 for even better specs ?
• #13 Reply
  Posted by IRFP on 21 Aug, 2018 12:16
• Hello,
  The LTZ 1000 is the graal in terms of voltage reference, it is actually my future goal to work on this type of component.
  However, I prefer to do a soft learning, first with REF102 then the LM399.
  I am still novice in this area very interesting but terribly complex ...
  It is with the help of the many people of this forum that I advance in the exploration of this area ... And thank you ..
  
  To resume the debate on ovens, in your opinion, the use of oil is a good solution?
  There is a lot of constraint to the realization of the container, but I have the impression that the temperature stabilization is much better than in the air.
  Your opinions… ?
  
• #14 Reply
  Posted by Kleinstein on 21 Aug, 2018 13:57
• There are advantages with oil, but also quite some disadvantages: oil (or some parts dissolved in the oil) might diffuse into plastics. Repairs / changes can also be dirty stuff.
  
  Running the LM399 inside oil could cause special problems - there is considerably higher heat loss and thus more heater power needed. In my opinion the LM399 might be better used without heater if inside the oil. However it needs a really good regulation for the oil than.
  
  A reference more suitable for the oil bath might be 1N825A or the Chinese 2DW232 zener.
• #15 Reply
  Posted by Wolfgang on 21 Aug, 2018 14:04
• I am not sure that an oil bath is really needed when the reference has a stabilization feature in itself.
  
  How much better regulation did you get when running the LM399 by itself and in your oil bath ?
  
  I have also never seen an LTZ1000 bathed in oil, not even in 8 1/2 digit multimeters or any other equipment.
  Maybe there is a new frontier to be opened 
• #16 Reply
  Posted by guenthert on 21 Aug, 2018 14:29
• Quote from: IRFP on 21 Aug, 2018 12:16

  [..]
  To resume the debate on ovens, in your opinion, the use of oil is a good solution?
  There is a lot of constraint to the realization of the container, but I have the impression that the temperature stabilization is much better than in the air.
  Your opinions… ?
  

  I'm no expert in this matter, not by a far cray and I can't speak of the oil at all.  I got however the impression that temperature isn't all that critical. LM399 and LTZ1000 got already their oven and remaining TC can be measured and accounted for.  More frustrating is the long term drift, which cannot (by selection from a batch over a long time one can get better specimen, but that's about it afaik).
• #17 Reply
  Posted by macboy on 21 Aug, 2018 15:30
• Stability of your 5.5 digit meters depends on what they are. The Keithely 199 uses a basic "zero" tempco zener diode, and as a result, it drifts quite a bit (several digits at 5.5 digit resolution). In contrast to this, the Fluke 8800A uses the same fantastic LTFLU (aka SZA263) reference IC used in some of their metrology grade meters and voltage references.
• #18 Reply
  Posted by Kleinstein on 21 Aug, 2018 17:04
• For use as a null-meter, the reference in the 5 digit meter does not matter. What can matter is input bias, ground current and the resolution (preferably better than 1 µV). So the HP3478 with it's 30 mV range can be useful. If really needed a DIY null-meter is an option too - no need for many digits or a good reference, it's more about the input amplifier, which can be reasonable simple today with an AZ OP.
  
  For the LM399 or LTZ1000 reference itself an oil bath is likely not a good idea. It is more like the resistors for the 7 V to 10/5/2,5 V etc. step that could profit from a stable temperature and controlled / excluded humidity. As the power consumption is essentially constant - a air based oven might be good enough, if turbulent convection (and thus related noise) can be avoided. Just for handling I would prefer cotton pads over oil.  So it would be back to the old SZA263 based refs, with a zero TC zener and an classical oven to include the zener and the resistors.
• #19 Reply
  Posted by IRFP on 21 Aug, 2018 20:22
• Good evening,
  Thank you for these tips ...
  I am committed to this type of fluid because I find it easier to handle thermally than air.
  So in summary, it is better to stabilize in the oil the resistance sets of the divider bridges rather than the voltage reference.
• #20 Reply
  Posted by rhb on 23 Aug, 2018 01:51
• Any connection between dissimilar metals where there are two connections at different temperatures will give rise to a thermal EMF.
  
  It is certainly easier on a specific heat basis to maintain a uniform temperature with a liquid than a gas.
  
  I think your best result would come from putting the entire reference in an oil bath at the design temperature of the LM399 and monitoring the LM399 heater current to verify that the oil bath is maintaining the desired temperature.  A previous poster had experimented with using a reverse biased heater connection as a thermal sensor.  Unfortunately he got in a dispute with the moderators and was banned.  With the entire reference immersed in a bath and copper to copper connections in the bath and outside you should not have any significant thermal EMF errors as long as the connections don't oxidize.
  
  Long term, I think you would need some sort of conformal coating to make sure the oil did not alter the properties of the components.  Though that too, has it's own issues.
  
  Personally, at a certain point I prefer to measure the errors and correct them.
  
  I expect that TiN, Andreas et al will appear in due time.  They have vastly more experience at this than most.  In this game, my sole "qualification" is I know some mathematics that most people don't which is very useful for creating functional approximations that no one knew how to do 20 years ago.  Whether that is actually of any value awaits my getting a multiplexer constructed to monitor a set of references over a long period of time.  I have the parts, but am easily distracted. 
  
  I look forward to following your project.  The fabrication alone is very impressive.
• #21 Reply
  Posted by IRFP on 23 Aug, 2018 08:53
• Hello,
  Thank you for your advice.
  As I explained previously, it is the LM399 which is immersed in the oil.
  The ammeter on the device allows to visualize the internal heating current of the reference.
  The copper type connections are well immersed.
  The oil is of very high dielectric quality, generally used on high voltage installations.