LM399 heat loss within a vacuum

[maat]

I don't want to rain on your parade, but there are a couple of issues with this design:
1. Why spend $$$ and time on building a vacuum chamber for a mediocre LM399? The LM399 will typically return to a value within specs even after a longer time off. Just don't cool it down to very cold temperatures. Rather build a tiny 'oven' around it to keep it above lets say 15 °C (Kleinstein already suggested this).
2. The vaccum feedthrough will likely be kovar in a glass seal. Since the vacuum chamber obviously needs to be bigger than the diode and is likely hand-made, the temperature gradient in the feedthrough will most likely be bigger than from the diode substrate to the the PCB. So the thermal EMF will be a lot worse in this configuration.

If you want to build something around the LM399, I would recommend you buy a batch of LM399 (like 5-10) and bring about 6 months of time during which you monitor those diodes in a burn-in circuit (or the final circuit if you just want 7V out as the BOM cost ist relatively low). Burn them in for like 1000-2000 h and toss out any stinkers, that do crazy stuff like excessive popcorn noise or sudden 1-2 ppm jumps. Then subject the rest (probably half of them will get to this point), to a test, where you power cycle them to see how they behave. Again, toss those, that show unpredictable behaviour.

This is probably the best you can get with the LM399, maybe low single digit ppm uncertainty over a year. By the time you done with that, the market value of that diode would likely be around $300, due to the equipment and man-hours spent. At that point, you may consider going for the LTZ1000, ADR1000 or even wait for the ADR1001, which is about the same price range (dunno about the ADR1001, though).

[iMo]

Even the ADR1399 (also in smd, no kovar) would be a better choice, imho. Not sure on 1399's popcorning, but you can hardly find an LM399 without popcorn.

[Kleinstein]

Even when kept hot the performance of the LM399 is limited. A practical problem are the typical popcorn type noise jumps or some 0.5 ppm that essentially all lm399 show. Even those where they are less obvious may have them or start to show them after some time.  The LM399 is mainly an easy to use, moderate cost reference good for something like a 6 digit meter. With the internal heater loop and high temperature it has its limitations and is a bit unflexible.

For a transfer standard one could consider other alternative references instead of the extra effort of a vaccum and keep hot circuit. An LTZ1000 even without a keep hot circuit could be less effort and likely better performance. If in doubt have more than just 1 reference, possible different types.

The LM1399 is rather new - so the longer term performance is less proven. It's lower noise, but not necessary lower drift.

[iMo]

ADI offers "small portable standards" for hobbyists based on the ADR1399 classic, ADR1399 smd, ADR1000 and ADR1001. A small pcb powered by 5V from an USB, the reference is floating (via isolating switcher)..

[robert.rozee]

I don't want to rain on your parade, but there are a couple of issues with this design:
1. Why spend $$$ and time on building a vacuum chamber for a mediocre LM399  [...]  I would recommend you buy a batch of LM399 (like 5-10) and bring about 6 months of time during which you monitor those  [...]  By the time you done with that, the market value of that diode would likely be around $300, due to the equipment and man-hours spent. At that point, you may consider going for the LTZ1000, ADR1000 or even wait for the ADR1001, which is about the same price range (dunno about the ADR1001, though).

i am afraid you are missing the point, as are several other posters.

the LM399 is a good choice to experiment with because it is CHEAP and OBTAINABLE. i really couldn't care less about the zener diode properties: snap, crackle, pop and drift. all i am interested in is an ovenized TO-46, and exploring how it can be transported 'live' for a usefully extended period of time.

the ultimate solution - if indeed the vacuum idea does work - would likely end up being no more complicated or bulky than this:



... a technology dating from the WW2 era, that is relatively cheap and easy to achieve. if it works, then someone else may care to invest in an LTZ1000, ADR1000 or ADR1399 and carry the experiment further. i'm just interested in the thermals and energy budget, nothing more.


cheers,
rob   :-)

[iMo]

Added a simple naive prop reg into the LM399.
Assumptions - min heater current 5mA, max heater current 130mA, sensor temp 10C lower than the heater, prop delay between sensor and heater such it heats up to 90C in aprox 25secs, heater voltage 15V, heater max 300mW, vacuum 4700C/W, Volts=Celsius.. Has been made for fun only.. 

[maat]

i'm just interested in the thermals and energy budget, nothing more.

Ah ic. Jolly good fun them   

[robert.rozee]

Added a simple naive prop reg into the LM399 [...]

am i right in saying that your model suggests a decrease in power consumption from 250mW at 1atm (from LT's LM399 datasheet, pg3: 8.5mA @ 30v) down to less than 30mW when operating in a vacuum? ie, close to a 10:1 reduction in power consumption.

if so, this is very impressive and good grounds to experiment some more   


cheers,
rob   :-)

[iMo]

Nope, that simplistic model shows the system in vacuum, at the beginning at 25C, then you switch on the LM399 heater, there is a short spike 300mW as the system is cold, and the the heat propagates from chip/package/legs into the pcb. The 4700C/W (??) is the thermal resistance of vacuum I took from a previous post here. You may see the temperature of the hole system/pcb grows slowly up to some 80C in vacuum. With something like 200C/W (??) for the air the pcb temperature will go to some 40C.
The heater power goes down, but still there is an issue with the regulator model as the regulator does not include changes of the current (??), thus it is just a joke, imho. In reality the power loss of the heater at 25C air in equilibrium is something like 300mW (15V*20mA), this model shows rather BS in the air. Mind that is a pretty primitive model (everything there is just a guess), I made it while watching news in TV..   It is intended for somebody here as a motivation to improve it..

[Kleinstein]

Usually the thermal resistance from the PCB to the environment is way lower than 4700 K/W. There are at least the leads to the supply that have more heat loss.

Another point for the dynamic part is that the thermal capacitance is allways to ground - so no series connection of the heat capacitance part.

According to the datasheet 300 mW is the normal power needed in steady state air. For transients there can be more power.

[DavidKo]

I must agree with Kleinstein. Forget about vacuum and use foam.

Vacuum is hard to handle. Some remarks:

• every material is porous - the glass was usually used for vacuum chambers in the past and now it is used stainless steel
• ideal chamber will be smooth to have small amount of adsorbed air on the walls as possible (you can help with heat to speed up evacuation, but most of the time you pump the air adsorbed on the chamber itself)
• smaller chamber works against you - quickly fills with gas
• creation of vacuum tight chamber
• vacuum tight feedthrough is even bigger challenge than the chamber - as already recommended use Kovar and correct glass
• vacuum tight valve - how do you want to connect and disconnect your chamber from the pump?
• all apparatuses need to be constantly pumped - thus getter in tubes, mostly the gas get inside around the legs

In the picture that you have posted the tube itself may not be evacuated, resp. it can be that the vacuum is not essential (I do not see the Getter), it can be filled with inert atmosphere or small overpressure if needed. It was common way to use tube socket and tube housing for exchangeable parts. It can be easily filled with inert atmosphere or small overpressure if needed. It also has the good isolation properties etc. Except the fact that the tubes are fragile.

If you don't have the access to the tube or bulb production line (ideally big double filament lamp like P21/5W, W21/5W or similar), than it will be hard for you to evacuate and air tightly seal the chamber after.

In home conditions I think that the glass chamber is easiest solution. Take the W21/5W, cut the head off, install the reference instead of the filaments, weld the glass prolongation (you must be careful not damage the reference with the heat), connect to vacuum and collapse the tube with torch. Example how it can be done in "home" conditions when you have better access to correct materials is for example here.

[iMo]

This is with thermal capacities towards ground (ground is virtual ENV ambient), and in the "air" at 25C. There is still the issue with the heater power, we need a better model, indeed..
The current model is

P[W]=V(Vcc)*1.5*(20mA+130mA*(100C-V(chip_T)))

P[W]=V(Vcc)*(5mA+130mA*(100-V(chip_T))/90)

where Vcc is 15V and the chip_T is the temperature measured by the on-chip sensor (delayed by some prop delay, targeted 90C with 100C at the heater), 20mA 5mA is min current (obviously wrong) and 130mA is max current (the max could be ok, but there is something missing here), and 100C is max heater temperature (perhaps could be higher, like 150C ??).

PS: I've found a problem in the model - there should be a ratio deltaT/T90 in the model, as well as the scaling of the Power graph was wrong.
Now a little bit better model, imho, below..

[robert.rozee]

I must agree with Kleinstein. Forget about vacuum and use foam.

Vacuum is hard to handle. Some remarks:

every material is porous
[...]
smaller chamber works against you - quickly fills with gas
[...]
vacuum tight feedthrough is even bigger challenge than the chamber - as already recommended use Kovar and correct glass
[...]
all apparatuses need to be constantly pumped - thus getter in tubes, mostly the gas get inside around the legs
[...]
In the picture that you have posted the tube itself may not be evacuated, resp. [...]

as you have pointed out, stainless-steel and 'the right' glass are not porous. copper is also an option. and a non-porous small chamber will not fill with gas as quickly or slowly as a large non-porous chamber: that is, both will not.

feed-throughs are available, and there is an epoxy called 'Torr Seal' (see: https://www.lesker.com/newweb/fluids/sealants-epoxy/torrseal/) that i have used in the past for temporary (<12 month) fixtures.

the below tube is evacuated, and has likely remained so for at least 40 years without "need to be constantly pumped". the getter you refer to in a vaccum tube functions only ONCE when heated/activated with an external microwave heater after the tube has been initially manufactured. the purpose is to burn off any traces of contaminant not removed when the tube was initially evacuated. thereafter, the getter does nothing more for the remaining life of the tube.



alas, everyone seems quite intent on making the problem far bigger than it actually is. at this stage it is just the THERMALS and ENERGY BUDGET being considered. IF these fall into line, and a copper vacuum chamber verifies that there are useful real-world results, then there is a fellow in europe who makes nixie tubes who would no doubt be more than happy to pop LM399's, LTZ1000's, etc into evacuated glass tubes for anyone interested.


cheers,
rob   :-)

[Kleinstein]

A getter helps to keep the vacuum for a longer time and reduce the need on the pumping time. Especially the feed through part can be expensive or still a bit leaky with epoxy. The epoxy solution can be OK for a system that is constantly pumped, but I would not trust it for a level good enough to make a significant improvement on the heat loss (e.g. < 0.1 mbar ).
I would consider foam isolation and than also reducing the heat loss through the wires the better option. Faom is not that much worse than a vacuum im the 1 mbar range.
The loss directly to air is also only one part (maybe 2/3) of the heat loss.

Another option would be to use reduced heating when on transit. The LM399 does not show much hysteresis for cooling to normal room temperature. So it may be enough to keep it at above 15 C on transit with a different heater. For this it may be enough to keep a few mA of zener current all the time.

[Gyro]

...
the below tube is evacuated, and has likely remained so for at least 40 years without "need to be constantly pumped". the getter you refer to in a vaccum tube functions only ONCE when heated/activated with an external microwave heater after the tube has been initially manufactured. the purpose is to burn off any traces of contaminant not removed when the tube was initially evacuated. thereafter, the getter does nothing more for the remaining life of the tube.
...

Just a couple of corrections...

The ring shaped getter is actually activated, or flashed, by a relatively low frequency RF loop inducion heater that is passed over the tube, not Microwave energy. This causes it to burst open and evaporate the Barium metal flash onto the inside of the glass. The getter structure itself is merely a 'sealed container' to get the barium compound into the tube without exposure to atmospheric air during construction. The RF coil can also be used to do a final outgas of the anode by heating to a high temperature.

Regarding your definition of "functions only once": The Barium flash continues to work as a getter throughout the life of the tube, reacting with stray gasses that make it past the pin seals. The exception being Helium which of course is inert and can slowly build up. Indication of its continued operation is the slow fading around the edges of the flash in weak tubes, and of course going white in the case of a seal failure.

I regularly use former Soviet tubes that have been stored in cold damp warehouses for 40 - 60 years. The poor storage means that they (large volume power tubes in particular) sometimes need warming in a low oven for a while to speed up the chemical reaction rate of the getter flash and mop up any accumulated gasses before application of high anode voltages. Luckily the Mil spec tubes have several getters and lots of active flash.

[robert.rozee]

Just a couple of corrections...

i stand corrected      it was 30 years ago when i last worked with vacuum tubes/valves. at the time, i recall being told by my supervisor (university) that the external heater we were using operated at microwave frequencies, but my memory (it seems) was faulty.

for anyone interested in a refresher, see:
http://www.thevalvepage.com/valvetek/getter/getter.htm


in the present (LM399/LTZ1000/ADR1399) application, i'm only suggesting a vacuum of 0.001atm, at which level i believe a getter would contribute little. the pressure just needs to be lowered enough to (meaningfully) reduce convention heat transfer.


cheers,
rob   :-)

[DavidKo]

You can try Dalibor Farny (he did the video which I have posted and it is the guy who is producing the nixie tubes). Hard to say how tough will be connect wire to the legs. Probably some spacer will be needed to prevent shorting. To speed up the pump process I recommend to remove the termal isolation of LM399. Probably a lot of the gas will be trapped inside of the TO-46 case which will later leak and increase the pressure, but Getter should help there.

[robert.rozee]

You can try Dalibor Farny (he did the video which I have posted and it is the guy who is producing the nixie tubes). Hard to say how tough will be connect wire to the legs. Probably some spacer will be needed to prevent shorting. To speed up the pump process I recommend to remove the termal isolation of LM399. Probably a lot of the gas will be trapped inside of the TO-46 case which will later leak and increase the pressure, but Getter should help there.

Dalibor is likely the fellow i am thinking of, i saw a youtube video showing nixies being built around a year ago. another outfit doing this is Millclock (https://millclock.com/), or perhaps even the chemistry department of a local university may be able to fabricate something.

here is an interesting video showing feed-throughs being formed, suggesting this part of the process is not too difficult for someone skilled in glassblowing:
https://youtu.be/uHv1f4_tDv4?t=205

the LM399 (et al) wires can likely just be spot-welded, although there may be some thermoelectric issues to be resolved. i don't see supporting the weight of the part being a problem. the plastic thermal shield would definitely have to go as it would likely outgas for an extended period of time - and wouldn't do anything much useful in a vacuum anyway.

the problem of gasses trapped within the TO-46 metal case is an interesting one, although these gasses would only be at 15psi and hopefully the hermetic sealing of the metal case would provide adequate containment. and even if not, the internal volume of the TO-46 is relatively small. an extreme solution would be to carefully drill a tiny hole in the side of the metal case! a Getter is definitely NOT an option - firing it would likely destroy the LM399 (et al). but (hopefully) we don't need anything beyond 0.001atm anyway.


the above all said, at the moment i feel the questions of (1) thermals and (2) energy budget need to be addressed FIRST. this can most easily be achieved by placing an LM399 into a small copper chamber (1" diameter copper pipe, 6" length, with [solder] sealed ends) and pumping down to a modest 0.01atm (10mbar). then the behavior when powered up can be checked, to see if we actually get a drop in the heater current, and if the on-die temperature is stable enough for the zener/reference to perform within specs.


cheers,
rob   :-)

[Kleinstein]

A pressure in the 0.001 atmophere range would not help much for the thermal conduction. The curve shown earlier is for a very short distance of 10 µm. With a length scale of 10 mm instead one would habe mbar instead of bar. So to get a significant reduction one would need to reach the 0.01 mbar range.
At that level one would not use a metal getter (barium or titanium) like with vacuum tubes, but more like a some very porous (e.g. silica gel, perlite) material as an absobant that is heated for some time in vaccum. This effectively acts like a larger buffer so absorb at least water, but to some degrees also oxigen and nitrogen. For a single unit it woudl not be so bad to pump for something like 1 hour before sealing.

The hermetic seal of the TO46 can should be gas tight. However the vacuum would be at least some mechanical stress to the chip. With some parts in a metal case one can see an effect of atomspheric pressure.

Anway: vaccum is way more effort than foam insulation. The accuracy of an LM399 is also limited and it is not al all clear that cooling to some 0 C in between has a significant effect at all. At least just a normal power one - power off - power on cycle has not much effect (e.g. < 0.2 ppm , which is about what can be measured with the usual noise level). It is more that an extended power one time can lead to some drift.

[DavidKo]

...
, but more like a some very porous (e.g. silica gel, perlite) material as an absobant that is heated for some time in vaccum. This effectively acts like a larger buffer so absorb at least water, but to some degrees also oxigen and nitrogen. For a single unit it woudl not be so bad to pump for something like 1 hour before sealing.
...

Material that we have used for cryosorption was zeolite. I suspect that it is the similar stuff as it is used for the cats. The one that we have used on university have had gray color. If I have not mixed that up, to use it with liquid nitrogen, you should heat it above 300°C to open air (mainly to get water out of it) for few hours. For complete out gassing to open air >600°C should be used. For sure if are able to out gas it into vacuum it is even better, but what I remember from lessons, the heating was essential due to its big surface (pores are hard to pump, especially water). Remark DO NOT store zeolite in glasses with Ground glass joint you will not be able to open them later without destroying the bottle (after out gassing, but it is also risky to do it with material which have been previously on air).

Vacuum is pain in the a.. and if you do not have access to it at work or university, than it will cost you a lot of money. If you have no experience with vacuum than it is long way to get familiar with it. I liked to work with vacuum (nothing better to see than liquid nitrogen stops boiling and freezes), but I do not want to pay for it even second hand.

[iMo]

It depends on what you want to achieve with the "vacuum". My first job was in a lab with ultra vacuum and sure for that level you would need several stages - like mechanical pump, turbo-molecular one and that zeolite one. Btw, a couple weeks back I had short discussion with DaliborF on the exactly same topic and basically it is doable, but there are some issues when one wants to get a pcb with electronics put into the tube - like pretty high temperatures during production, those wires coming out of the tube are not easy to solder, and the pcb and the components on it should not produce any gases, of course.

[Conrad Hoffman]

Lots of DIY vacuum info here- http://www.belljar.net/articles.html

[MegaVolt]

Replace vacuum with airgel.

[iMo]

  Now it gets interesting - the targeted temperature 90C with heater at 92C (on-chip temp difference), and all in vacuum and starting at -10C ambient. The heater power is dropping down to 100mW and goes even less while the pcb with components goes towards 80C after 500secs .. So the simple model starts to show some signs of life.. 

[MK]

To get the epoxy to stop outgassing later you would need several weeks of pumpdown before sealing. You will need to draw all the moisture and volatiles out of the pcb or the vacuum will quite quickly get worse. I prefer the mickle approach of putting it in a dewar flask to help it stay at constant warm.