Author Topic: Cooling voltage references?  (Read 2629 times)

0 Members and 1 Guest are viewing this topic.

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Cooling voltage references?
« on: May 20, 2018, 03:18:06 pm »
According to the datasheets I've read both thermal and 1/f noise increase with increasing temperature.  But I've not seen any mention of using a TEC to reduce the temperature of a reference.

Operating at low temperature would increase the thermal gradients, but I'd expect those effects to be stable and measurable.  In particular, reducing the 1/f noise seems very desirable.

I know TiN did a cryo run, but that's rather impractical to maintain unless you have to for a JJA.

Edit:  It got posted when I was just trying to add a blank line.
« Last Edit: May 20, 2018, 03:28:25 pm by rhb »
 

Offline lukier

  • Supporter
  • ****
  • Posts: 634
  • Country: pl
    • Homepage
Re: Cooling voltage references?
« Reply #1 on: May 20, 2018, 03:26:58 pm »
If I remember correctly DigilentMinds had the idea of TEC cooled LTZ1000, but his posts are now gone so I cannot double check. I doubt if this approach was actually tested in practice.
 

Offline kj7e

  • Frequent Contributor
  • **
  • Posts: 911
  • Country: us
  • Damon Stewart
Re: Cooling voltage references?
« Reply #2 on: May 20, 2018, 04:06:21 pm »
The whole point behind a voltage "reference" is precision and stability.  There is no practical way to maintain the high level of thermal stability the internal oven references can provide.  Its far easier to create and maintain heat energy than to remove it.  It may be best to first have a chilled environment then to heat the reference above that point to maintain stability.  But what do you gain with all this effort?
 

Offline pelule

  • Frequent Contributor
  • **
  • Posts: 513
  • Country: de
  • What is business? It’s other people’s money
Re: Cooling voltage references?
« Reply #3 on: May 20, 2018, 04:55:47 pm »
It may possible, I have the wrong in mind - so please correct if so.
Main function of the "heater" of a refernce is to ensure, the temp of the ref-diode is stable (not short term drift).
The selected  higher temp (f.e. 55°C) is to ensure heating regulation function at higher ambient temp (f.e. at 35°C).
Higher temp increases noise and aging (long term drift - see LTZ at 3458A with reduced heater temp).
Thus in my understanding one could build an encloser and cool it down to f.e. @ 5°C+/-1°C
at the same time use the LTZ1000 internal heater to keep the temp of the reference at f. e. 10°C.
The heater regulation is quick and low tolerance - thus reference temp is stabel - but temp is lower thus noise is lower and aging is reduced.
Am I wrong?
/PeLuLe
« Last Edit: May 20, 2018, 04:57:44 pm by pelule »
You will learn something new every single day
 

Offline Conrad Hoffman

  • Super Contributor
  • ***
  • Posts: 1928
  • Country: us
    • The Messy Basement
Re: Cooling voltage references?
« Reply #4 on: May 20, 2018, 05:19:10 pm »
You have to cool a lot to reduce noise significantly. No problem for peltier coolers. There is a huge problem that's hard to address- condensation. Get some condensation and you have current leakage. The device has to be in either vacuum or dry air/gas. It can be done- look at cooled astro cameras, but it's a big jump in terms of mechanical design. If you evaculate the housing, you can back-fill it with argon (common from any welding shop), which has poor thermal conductivity, making it easier and more efficient to control the device temperature. The right desiccant pack is also needed. I think GR filled some of their reference capacitors with dry nitrogen, so it's nothing new.
« Last Edit: May 20, 2018, 05:20:51 pm by Conrad Hoffman »
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14080
  • Country: de
Re: Cooling voltage references?
« Reply #5 on: May 20, 2018, 07:04:27 pm »
Cold temperature increases relative humidity. Ideally one want to stay at less than about 50% RH, as this reduces surface leakage with many common materials. So a temperature slightly above ambient is an advantage for this reason too.

Cooling is usually way more complicated than heating. TEC coolers have quite some thermal conductivity and thus are not easy to reach very stable temperature. Also power consumption tends to be much higher - so that battery power is not that practical anymore.

When it comes to lower noise, just using a 2nd LTZ1000 is not that complicated compared to a TEC cooler and still lower power and lower noise.

Aging is slower at lower temperature, but this also applies to the initial settling - so to keep initial settling in a manageable time frame a very low temperature is not really desirable.  If a possibly relative high temperature environment is given one might consider some cooling in case of high temperature, but still have the reference at some 40 C.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 16548
  • Country: us
  • DavidH
Re: Cooling voltage references?
« Reply #6 on: May 20, 2018, 07:14:08 pm »
I am just going to wing it here.

Most of the reference noise is flicker and shot noise which are independent of temperature and the big advantage of using a zener diode instead of a bandgap.  So the only reason to lower the temperature is to run the zener at a higher current lowering the shot noise.  (1) Considering the complexity and disadvantages, it would be better to use multiple zener references; to take advantage of lower shot noise, you will have to operate at higher power anyway whether one cooled reference is used or several non-cooled references.

Using cooling might make sense for added long term stability but I would not take it for granted.

(1) Does flicker noise depend on current or temperature?  It does not in MOSFETs.
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Cooling voltage references?
« Reply #7 on: May 20, 2018, 11:46:44 pm »
The LM399 datasheet shows ~35% less flicker noise at 25 C than at 90 C at 10 Hz.  So 50% reduction by cooling is a reasonable estimate.  Use 4 references and average and you've got another factor of 2 reduction so it would be 1/4 the flicker noise uncertainty.

Rather obviously if one is cooling it must be in a sealed enclosure which has been heated, evacuated and backfilled with dry gas. So it certainly complicates mechanical construction.  But it's not that difficult to do.  A rather substantial proportion of the EEVblog community have the facilities and skills.

Even with the heater off and used in temperature sensing mode, the reference is going to be the major heat source, so a good thermal conduit from the case to the cooler that does not involve the rest of the circuit would be important for eliminating thermal EMF.

So far as I know, the justification for heating is it's cheap and easy to implement. I rather doubt that cooling would improve the aging.  I think it would simply increase the time constant so that it took far longer to reach the asymptote.   That would be the case if aging is a visco-plastic phenomenum.
 

Offline kj7e

  • Frequent Contributor
  • **
  • Posts: 911
  • Country: us
  • Damon Stewart
Re: Cooling voltage references?
« Reply #8 on: May 21, 2018, 12:49:29 am »
The whole point behind a voltage "reference" is precision and stability.  There is no practical way to maintain the high level of thermal stability the internal oven references can provide.  Its far easier to create and maintain heat energy than to remove it.  It may be best to first have a chilled environment then to heat the reference above that point to maintain stability.  But what do you gain with all this effort?

if an LTZ1000 is implemented as an *unheated* version (at ca. 4-5ma or more even) the tempco
can be tuned to better than 50ppm/degC (sub 10ppm/degC, or even better using quadrature compensation)
as shown here:-

https://www.eevblog.com/forum/metrology/ltz1000-vs-ltflu-aka-fluke-vs-datron/msg1516690/#msg1516690

now the oven stability requirements are quite relaxed and the LTZ1000 is now running at ca. 23 degC
which essentially means higher stability in the long-term and maybe better noise performance (if using higher zener current)

(we do have some evidence of having a lower die temp. means better long-term drift specs.)

best regards.

-zia

Even if you where able to get 10ppm/C from a non-heated LTZ1000, that's still about 2000x worse than normal operation.   Expecting better long term drift at the expense of any short term stability which will be needed to measure the device in the first place.  Kind of like a GPSDO with a very long time constant steering a bare non-ovenized crystal oscillator.  Long term it may be bang on, but measuring it will be very difficult.
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14080
  • Country: de
Re: Cooling voltage references?
« Reply #9 on: May 21, 2018, 08:02:58 am »
The LM399 datasheet shows ~35% less flicker noise at 25 C than at 90 C at 10 Hz.  So 50% reduction by cooling is a reasonable estimate.  Use 4 references and average and you've got another factor of 2 reduction so it would be 1/4 the flicker noise uncertainty.
...
Going down from 90 C of the LM399 to unheated 25 C is a large step of 65 K down. It would be rather difficult and power hungry to get that much cooling. The 90 C of the LM399 is more like a not so good example for very good long term stability.

The LM399 datasheet mainly shows the frequency range (10 Hz and up) that is not rally interesting - the important range would be more like 0.1-100 mHz. It is not clear how the higher temperatures effect this range, as the so called 1/f noise is not exactly 1/f but in a good reference should go up slower than 1/f at least for the very low frequency part. There is even a chance the higher temperature might help to reduce the very low frequency noise as some noise processes would be shifted to higher frequencies.

So my estimate is more like cooling by maybe 10-30 K might give an noise improvement in the 10% maybe 20% range if one is lucky. This would be less than using 2 references in parallel, which would give a 30% reduction.
 

Offline Conrad Hoffman

  • Super Contributor
  • ***
  • Posts: 1928
  • Country: us
    • The Messy Basement
Re: Cooling voltage references?
« Reply #10 on: May 21, 2018, 12:56:26 pm »
It's all about ROI and in this case it seems there are better gains to be had, for a given amount of toil and trouble, then cooling.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf