Influence of ambient parameters

[branadic]

Hi,

almost all investigations made here are focussing on temperature coefficient. This is okay as this is the parameter with the biggest influence on the stability of our references. Once T.C. is almost canceled out of the reference most of the work is done. But chasing into ppm and sub-ppm level other influence need to be taken into account.
Beside some rare publications there is pratical no investigation on influence of humidity, ambient pressure or radiation as they are hard to investigate.
I'm pretty sure a 3458A is also sensitive to ambient pressure in some way.

I do measure ambient temperature, humidity and pressure (SHT25, MS5611) but not radiation. On the other hand I haven't put one of my references nor the 3458A into one of our climate chambers.
For ambient pressure a pressure container and a pressure controller are needed. We do have a calibrated pressure controller but I'm lacking the pressure container yet.

Has anyone had the possibility to investigate on things like that in some way (reference or multimeter)?

-branadic-

[chuckb]

I have a pressure test chamber with two LTZ1000A voltage references PCB inside. I tested them for voltage shift with pressure and vacuum 6 months ago. Since then they have just been aging. I will recheck them for voltage shift under vacuum tomorrow.

[branadic]

@chuckb

That would be great, thank you. 

-branadic-

[Echo88]

The only mention ive ever seen regarding ambient pressure dependency:

"A Fluke 7004T
“Transref” transportable voltage
standard has been extensively
evaluated for pressure coef-
ficients[3] over the range 835
mb to 1200 mb and exhibited
a linear pressure coefficient of
-1.5 nV/mb. This is extremely
low and is negligible in rela-
tion to the pressure differences
likely to be encountered during
normal use as a travelling
standard."

http://www.elcal.ch/files/11749-eng-01-a.pdf

[Tomorokoshi]

EMC and hazardous location test reports typically include both pressure and humidity.

[Andreas]

Beside some rare publications there is pratical no investigation on influence of humidity, ambient pressure or radiation as they are hard to investigate.

I do measure ambient temperature, humidity and pressure (SHT25, MS5611) but not radiation.

Hello Branadic,

what do you mean with "radiation"
There is a wide range from "low frequency radio waves" over "light" to "radioactivity".

As I did a new measurement setup for ageing measurements on ADC25 + ADC26 where I have the laptop far away from my desk, I decided to use WLAN and doing the setup per "remote access" to sample the data.
But this was not so a good idea:

At minute 230 and minute 390 during "remote access" you can see that the minute averaged values drop down by 2-3 uV. (so 0.5-1 ppm).

At the same time the ADC maximum peak/peak noise increases up to 90uVpp.
At minute 410 I started to move the ADCs and the LTZ#4 as far as possible from the laptop.
From about 10 cm to now 50 cm on opposite sides of the (small) table.
Noise immediately decreased to below 10uVpp (which I regard as a normal value).

But I fear in the long term I will need much more (metal) cookies boxes.

with best regards

Andreas

[chuckb]

I have a small die cast metal box for pressure testing parts. I can apply a good vacuum or 15 psig to the parts inside the box. Hermetic EMI filters pass the voltage signals through the wall of the case.
The vacuum pump is a Welsh 1400 series that can achieve a 0.0001 Torr vacuum. I’m certain my plumbing leaks or outgases preventing it from reaching that level.
I currently have an LTZ1000A voltage reference PCB inside the chamber. It is the normal LTC schematic running at 50 deg C. The Zener has been running for the last 6 months. 
I used a HP34420A Nano voltmeter (NPLC 200) to monitor the voltage difference between the Zener under test and 7.14 V provided by a F720A (KVD) that is supplied by a F732A (10V Standard). 

I do not have a fine control over the pressure in the chamber. I just have a ¼ turn valve and a 3” mechanical pressure/vacuum gauge. So typically vacuum is applied quickly then the valve is shut and the pressure equalizes (leaks) to ambient over the next 10-15 minutes.
When full vacuum was applied the Zener voltage increased 7 uV (1ppm). There may be a few reasons for this like –
Stress on the LTZ1000A case and die, unlikely
Loss of convection cooling and the resultant temperature rise of all the internal components, more likely

NOTE: The chip heater was always in regulation during the testing

The voltage returned to normal after the vacuum was shut off and the pressure was about half way back to ambient. So the vacuum has a threshold effect. For this setup that threshold is 400 hPa or about 7200 meters of altitude. So it will be stable until you get to the top of Mt Everest. Anybody want to verify this in the field?

From 400 hPa to an ambient of 950 hPa (at this location) I could not detect a voltage shift from the baseline. I certainly believe the 7004T specification that the pressure sensitivity is negligible (-1.5nV/mb).

[branadic]

Thanks chuck for your measurements. What about the opposite pressure direction? Does this make any difference? Sad that the measurements are not correlated, so that it would be possible to plot reference voltage vs. pressure.

-branadic-

[MisterDiodes]

Be careful to distinguish between TC and true pressure effects especially on LTZ Vrefs - remember when you put the LTZ bare board in a vacuum you're drastically changing the board thermal flow.  An LTZ is basically a power in vs power out device with a stable Vref as by-product.  The vacuum environment on a bare board can mean you've over-insulated the LTZ or whatever thermal-stabilized Vref you're testing.

Watch your heater current also during vacuum test and if you see that going very low or to zero you know you're out of regulation, and Vref is not valid - and neither is the test.

In our testing the pressure effect is near nothing IF you maintain the same thermal flow off the board in sea-level atmosphere vs. vacuum (i.e. test bare board under oil).  These tests are typically required for aerospace applications, and when enclosure is purged - usually dry N2, CO2 or He.

On LTZ you shouldn't see much pressure effect during vacuum or below sea-level pressure - if you do you want to look for the real problem, which is usually a thermal flow issue, which can be a bigger contributor.

[chuckb]

Thanks chuck for your measurements. What about the opposite pressure direction? Does this make any difference? Sad that the measurements are not correlated, so that it would be possible to plot reference voltage vs. pressure.

-branadic-

Six months ago I checked it with pressure also. I did not detect any changes with pressure. I checked it with 5 psi above ambient. I think the LTZ1000A has a very nice, predictable variation with reasonable pressures. There is zero correlation so we don't need to worry about it. That's good.

I have seen different pressure sensitivities numbers published by NIST and BIPM for the Fluke 732A and the Fluke 732B. The F732B can have several different pressure sensitivities depending on which Zener chip is used. Fluke lists three different pressure sensitivities in the attached document.

From Witt -
"The F732B units with -2nV /hPa used an obsolete type M Zener." I assume this is the Motorola SZA263 that is also used in the F732A.
"The F732B units with +18.3nV/hPa use a newer type L Zener." I assume this is the LTFLU Zener.

The use of F732B L type Zeners should not be a issue unless the user and the calibration labs are at greatly different altitudes. The calibration error can be up to 0.06 ppm / 1000 ft. Reference page 10 of the attached Fluke document.


reference
T. J. Witt "Pressure coefficients of some Zener diode-based electronic voltage standards,” IEEE Trans. Instrum. Meas., vol. 48, no. 2, pp.
329–332, Apr. 1999

Complete Characterization of Zener Standards at 10 V for Measurement Assurance Program (MAP)
Yi-hua Tang and June E. Sims
IEEE Trans. Instrum. Meas., vol. 50, no. 2, pp. 263–266, Apr. 2001. This does not have a copywrite so it is attached here.

NOTE: This paper was researched and written in 1999/2000 time frame. NIST does not say but they probably tested the Fluke 732B and the Datron 7000 Zener references. I believe ZenerA is the Fluke 732B and ZenerB is the Datron 7000 Zener reference. The main clue was ZenerA has a Thermistor. The 732B has a thermistor and the Datron 7000 does not.

[chuckb]

This is fun!

Fluke notes (in 1996) there are three different pressure sensitivities for the 732B. They are
0, -0.03 and -0.06 ppm / 1000 ft.

One of the national labs also noted three different pressure sensitivities.
-2nV / hPa, 10nV / hPa, and 20nV / hPa

Then the guys in the LTFLU discussion noted there were two different Part numbers for the LTFLU chips...


Re: The LTFLU (aka SZA263) reference zener diode circuit
« Reply #221 on: September 29, 2017, 05:20:39 PM »

I'm pretty sure the difference is the die attach. As you can see in the former opened LTFLU that the die attach is deviated while the latest pictures shows a well defined area for the die attach. They will not have changed the die itself.

-branadic-

Maybe it goes like this
0 ppm / 1000 ft = original Motorola SZA263
0.03 ppm / 1000 ft = LTFLU-1ACH (the improved Zener with a different die attach for less coupling of package stress)
0.06 ppm / 1000 ft = LTFLU-1CH

[branadic]

We did have a change of 66hPa during one year (min 980.4hPa to max. 1046.6hPa). Assuming -1.5nV/hPa this is a change of 100nV or 0.01ppm on 10V output.

-branadic-

[Edwin G. Pettis]

In order to make valid readings on any given parameter, all other parameters should be held as constant as possible.  There is no way to measure a given parameter when you have a bunch of variables interacting at the same time, this always leads to inaccurate results and mistaken conclusions.  If you want to find out what effect humidity has a on a given component, then all other major parameters that may have a significant effect on the component must be controlled.  I have not seen this with most of the measurements that have been posted, this leads to incorrect conclusions about what a given parameter has on a given part.

As noted here, temperature is a major parameter, therefore it is one of the parameters that must be controlled.  You should preferably know which parameters affect the type of component you are measuring, for example Manganin (and its derivatives) are known to be a bit sensitive to barometric pressure, Evanohm not so much.  Manganin is slightly affected by moisture, Evanohm isn't.  The type of construction may also determine a sensitivity to the environment, most molded packaged film/foil resistors are sensitive to humidity, epoxy resistors vary depending on who made them.

Usually the two most significant parameters that resistors respond to are temperature and stress, stress being a component of heat and winding techniques or die attachment as in film/foil.  These two parameters are difficult to separate in many cases as they both cause a shift in resistor value.  Generally speaking, internal stress tends to produce a non-linear curve in the resistance, unfortunately many resistors do not have an inherently linear TCR so again, it can be difficult to separate the two from each other.

Don't forget, there are also measurement uncertainties which can vary depending on the circumstances, instrument drift and a pile of other small errors that can sneak into your measurements without you noticing them.  Measurements at PPM and sub-ppm levels are not trivial to accomplish and just because an instrument may display a measurement to .01PPM resolution doesn't mean it is also accurate, uncertainty wipes out most anything below 0.1- or 0.2- PPM at the very least, there is also source noise, thermals and EMI to consider.  Too often the output of a digital instrument is taken as absolute, it isn't and don't forget those inaccuracies are included and amplified in math calculations so be careful what you believe.

[MK]

if you rig up a pressure regulator with a hard vacuum on the bonnet/spring you then have a true psia pressure source, check it with ambient on the bonnet first, then turn the vacuum pump on and you can get 5psia/10psia etc, then you don not need to worry about pulling a hard vacuum onto the ltz.

[TiN]

Best would be design dedicated pressure test venicle for this tests, with LTZ chip placed on own PCB that can be sealed in the pressure chamber, and interconnected to the actual reference circuit outside of the temperature chamber box. Or perhaps easier could be to use just hermetic vacuum seal with feedthru terminals, and LTZ chip mounted to it in dead-bug style? This way all temperature/flow/stress variations of reference circuit would be eliminated.

And as result measurements pressure/voltage could be related to LTZ chip itself, which might be the desired data (or not, as even if no pressure correlation to LTZ exists, it still affect remaining parts on the board of final reference).

Anyhow, thanks for the efforts. Even if the data is not represents what some might want, it represents overall behaviour of that specific design, including other effects like temperature, humidity, etc. I'd wish to see data with slower ramps of pressure, to get better detail on what is going on. I noticed 5-15 minutes delay from temperature ramps when actual PCB temperature reaches the setpoint change of the chamber. I'd expect simiar exist with pressure variations too, as it takes time for the board to "stabilize" on new conditions.