Flow Proportional Counter Backflow Gas Regulation

[dawncruth]

Hey Karsten,
In October 2018 you said:

In any case, probably another reason to put something on the exhaust. I tried filling our bubbler with diffusion pump oil (to avoid the somewhat nasty dibutylphthalate) but it was way too viscous at room T. So now I'm trialling Alcatel 200 rotary pump oil, which seems to have the right sort of viscosity (with bubbler close to half filled 26 nicely shaped bubbles per minute, with P10 pressure regulator set at 16 kPa, flow regulator to around 1.15 ml/minute). From the specs it is hopefully fairly clean and long-term stable ("double distilled hydrocarbon fluid, low backstreaming ... strong oxidation resistance, ... for corrosive applications..."), so I'll see how that goes. It is a double chamber bubbler so even in the case of a detector window failure it shouldn't suck the fluid all the way back into the detectors (in theory, at least...). If anyone has a better idea what to use let me know. I'd still be interested in the back pressure regulator setup even at low altitude such as in our case.

Cheers,
Karsten

What was the outcome of your experiment? I'm going through the same process.
Dawn

[Karsten Goemann]

Hi Dawn,

I still have the Alcatel oil in the bubbler and it seems very stable (no discolouring, no change in the level, constant bubble rates...). I don't have a dataset similar to what Brian has done to be able to verify if and how much it reduces drift in the bias settings. I'm hoping it does not only prevent air (with changing humidity) backstreaming into the detectors but also to have at least some backpressure regulating effect.

Cheers,
Karsten

[Brian Joy]

Below is a more or less final version of my plot of GFPC anode bias versus atmospheric pressure:



In contrast to previous versions of the plot, I’ve now contoured it (by hand/eye) for dew point temperature, and I’ve also color-coded the data.  The curves that I’ve drawn on the plot are given by the following equations, which work well for interpolation:

Channel 1:  bias [V] = 7 V/kPa * P_atm [kPa] + 0.52 V/°C * T_dew [°C] + 909.4 V
Channel 4:  bias [V] = 7 V/kPa * P_atm [kPa] + 0.76 V/°C * T_dew [°C] + 929.7 V

Some unexplained scatter is still present on the plot, and so additional variables are likely significant in addition to atmospheric pressure and dew point temperature.  For instance, I don’t really have a good handle on time required for equilibration, which obviously would be particularly important when atmospheric conditions are changing rapidly.

When determining the appropriate anode bias, I should note that I used the JEOL “base level” scan rather than the “high voltage” scan and then adjusted the bias in 2 V increments until I got a distribution centered at/near 4 V.  I made the final scan using a step of 0.1 V and dwell time of 1 s; generally I repeated the slow scan at least once in order to assure reproducibility.  This is a tedious process, but it produces better results than the “high voltage” scan, which tends to have a broad, gently sloping “peak.”



So I guess now I need to do something about this problem.  I’m reluctant to add liquid to the bubbler at the exhaust, as I worry about backstreaming in the event of counter window failure.

[DavidAdams]

Wow, Brian!! This is a fantastic data set! I'm really happy that you took the time to do such intensive testing and compile all this. This definitely shows VERY similar behaviour to what I have observed for a lot of years. As to a fix, I'm still scratching my head on that myself.

I don't think you should be too concerned about backstreaming. I've had fluid in all of my instruments and have also had window failures. I've never had any fluid pulled back into the system. The key seems to be to only fill the bubbler to just where the bubbles start and no more.

-dave

[Brian Joy]

I've attached a copy of the measurements in an Excel file in case anyone wants to examine them more carefully and/or plot them differently.

[sem-geologist]

I have no experience or knowledge with Jeol probe construction, but I am very familiar with Cameca probes. I want to point to some very important physical behaviors/issues which can influence the observed optimal bias drift and is more plausible to proposed water diffusion against gas flow.

At first, It is not enough to be sure that your room temperature is rock stable. Can gas bottle be heated by floor or wall? (Or more precisely how different room air temperature gets in contrast to wall and floor temperatures (in day cycles, in seasonal cycles)? Hopefully the bottle is not outside of lab - that would be the fatal flaw). Air humidity is very important in air <-> object temperature exchange rate. The higher humidity - the faster object's temperature equalizes to the temperature of the air. Temperature of the bottle influences the temperature of P10, and so that influences the density of the gas in the proportional counter which will shift optimal bias value. Same counts for the probe, just differently to static bottle, probe has lots of heat sources and many passive and active cooling solutions. So even if Air temperature in the room is kept extremely stable, the humidity variations will change the temperature of the objects. BTW, pressure also affects the cooling rates.

I want also to point out that between setting bias value (or scanning bias values with digital interface that is Your computer and software, or values which You are displayed on the screen) there is the complicated electronic system which in different parts is actively and passively cooled. So again, any change in humidity will swing lazily/carelessly designed analog signals and that will result in observable PHA shifts and optimal bias shifts. On some instruments there are such design atrocities as analog signal squeezed-down (down-voltaged) with primitive voltage divider (pair of resistors) which will toss the PHA left and right depending from humidity... And changing proportional counters with SDD's wont change a lot. The EDS systems (where electronics part has much better engineering) has exactly same issues (spectrum shift after humidity change), at least I am aware of these in our labs.

[Brian Joy]

Perhaps you could present some actual data to support your argument??

I have no experience or knowledge with Jeol probe construction, but I am very familiar with Cameca probes. I want to point to some very important physical behaviors/issues which can influence the observed optimal bias drift and is more plausible to proposed water diffusion against gas flow.

At first, It is not enough to be sure that your room temperature is rock stable. Can gas bottle be heated by floor or wall? (Or more precisely how different room air temperature gets in contrast to wall and floor temperatures (in day cycles, in seasonal cycles)? Hopefully the bottle is not outside of lab - that would be the fatal flaw). Air humidity is very important in air <-> object temperature exchange rate. The higher humidity - the faster object's temperature equalizes to the temperature of the air. Temperature of the bottle influences the temperature of P10, and so that influences the density of the gas in the proportional counter which will shift optimal bias value. Same counts for the probe, just differently to static bottle, probe has lots of heat sources and many passive and active cooling solutions. So even if Air temperature in the room is kept extremely stable, the humidity variations will change the temperature of the objects. BTW, pressure also affects the cooling rates.

I want also to point out that between setting bias value (or scanning bias values with digital interface that is Your computer and software, or values which You are displayed on the screen) there is the complicated electronic system which in different parts is actively and passively cooled. So again, any change in humidity will swing lazily/carelessly designed analog signals and that will result in observable PHA shifts and optimal bias shifts. On some instruments there are such design atrocities as analog signal squeezed-down (down-voltaged) with primitive voltage divider (pair of resistors) which will toss the PHA left and right depending from humidity... And changing proportional counters with SDD's wont change a lot. The EDS systems (where electronics part has much better engineering) has exactly same issues (spectrum shift after humidity change), at least I am aware of these in our labs.

[sem-geologist]

Perhaps you could present some actual data to support your argument??

I should apologize at first. I forgot to say thank You a lot for this extensive excellent measurements and data. I wish I could had this at hand 6 years ago and could show that under nose of our management, so they would not cheapskate on air conditioning (AC) system in our lab.

I am not questioning credibility of data, I am questioning only the proposed mechanism for observed bias drifts.
So lets look to my claims step by step.

I have no humidity logs in our lab, albeit there is plan for those. The hardware is sitting and waiting for configuration, but I had no time to configure it. I have at the moment only extensive temperature data (logged every 1 minute for a bit more than a year with some brakes) from different places such as outside wall near window (old site for AC control panel (and internal thermometer used for T control), inner build wall where P10 gas bottle is presently placed by (also a new site for AC control panel), P-10 bottle metal case, backside of EPMA (air at 20 cm from floor, 10 cm from microprobe), microprobe chasis at two points (column isolation valve screw (EP6), junction of second ion pump and column).



Initial idea for thermometer network came from some weird behavior of one of spectrometers, so I wanted to see if air streams from AC are not effecting more one of the spectrometers (the one with weird behavior) and so wanted to measure separate temperatures for every spectrometer (using DS18B20, digital 1-wire thermometers connected in network). Before I started long term logging I had re-modified the network to troubleshoot vacuum leak problems (I was suspecting either metal gasket of that valve or metal gasket of second ion pump, thus the weird positions given below). Also I had expanded that network with other points in lab to prove that AC control panel was initially badly placed, and P10 Gas bottle had to be moved from external wall to internal wall (I was getting critique from management for moving stuff, and needed hard proof). After P10 gas movement we saw a lot of improvement in calibration stability, but that was not enough. Unfortunately I started logging temperatures with that network only after AC and P10 Gas placement modifications, so I can't compare directly with old setup. You can see in above picture that at winter the temperature of outer wall gets much cooler, and gas bottle (when it was standing by outer wall) was also surely affected. Also AC logic within c/p in old place was often getting very biased temperature readings. I.e. in cold winters we were getting very warm room (i.e. when -20 C outside), while I have no data record from those times, it was very obvious for human senses.

I should mention that our AC panel is set to keep in between 23 (start cooling if above) and 22 (heat up if below) degrees, and integrated thermometer reading on AC c/p is showing 22 or 23 whole year. I also should mention that previously we had in room installed some USB temperature humidity logger which was also showing very stable temperatures (Multimetrix DL 53), basically lulling us into believing that there is no issues. That useless piece of USB logger was installed by Cameca with our new probe. The "PRO" thermometers/loggers have one significant drawback - they are encased in plastic packaging and measure highly averaged temperature - they measure something from alternative reality - or like going to listen for opera but putting the bucket on the head. That is why I made network of thermometer chips (DS18B20) which are bare-exposed to the stuff which it is supposed to measure (i.e. direct contact to gas bottle with thermal paste in between chip and bottle, but well isolated from direct air blow; or other example - directly exposed/hanged in the air for air temperature measurement - so that room humidity and air streams would affect the thermometer in exactly same manner as the machine). Before assembling the network of thermometers I had tested them all by simultaneously measuring temperatures while keeping them immersed in same liquid container to see that they get exactly same (+/- 0.0625 C) temperatures in range of 30-15 degrees. So the temperature differences showed in charts are not artificial but real. DS18B20 is able to differentiate temperatures with 0.0625 C steps, which is pretty good and it is able to capture not only daily and seasonal temperature fluctuations but also individual AC cooling/heating cycles. (In the end this precision helped me to locate and fix vacuum leak).

The temperatures are pretty stable during autumn, winter and spring. As an example 4 days during stable period:


However it is pretty terrible during hot summer days:


Here comes so waited autumn, which begins the stable cycle again (the figure below depicts logged temperatures with transition from hot summer weather changed with atmospheric front going through location and lowering outside temperatures):

 
This I think summarizes currently gathered all temperature measurements:


So why during summer our lab gets so bad variations? I think (I still need to install humidity measurements hardware to be 100% sure) that it is so at summer as lab is coolest place compared to the surrounding rooms and halls in the building and gathers/traps humidity, that is outside from lab (inside building) and outside of building temperatures gets significantly higher during hot day than inside the lab. The increased humidity significantly increases the rate of feedback between air conditioning injected cold air streams and AC c/p (where we can see very steep negative peaks recorded on all temperature readings). As control panel logic recognizes that AC is over-cooling it switches off AC very rapidly and waits for threshold to be crossed again. Thus there is much larger cycles than during stable period, where AC works in PID mode (like i.e. water chiller; it is integrating temperature measurement and controlling AC air stream strength and temperature not so violently - thus on chart we see smaller cycles which switches much more often). And so these are the data which I base my question about if being sure that room temperature is rock stable, and being sure about all heat fluxes.

Talking about P10 gas temperature and its influence to counting, I have rather "anecdotal" experience without real measurements (sorry, no data). When we change P10 gas, old calibrations gets instantly outdated; But then after some time the newly made calibrations gets outdated, and old calibrations gets again to be good. The effect is clearly correlative with temperature differences between lab room, that is: if weather is mild (spring or autumn) and temperature of transported bottle is similar to our lab, then there is no or very week calibration misalignment observed. But if gas bottle is transported and changed at cold winter or hot summer, we get 1-2 weeks of constantly out-dating calibrations until gas temperature stabilizes with room temperature. Then we can use again old calibrations. I know that gas bottles are stored in some not-heated or cooled shed by gas producer, and so its temperature equilibrates with environment/weather before being transported to our lab.

Now talking about electronics, most of datasheets of electronic components present in one or other form how the values of described component react to temperature changes and (optionally) humidity. Resistors are particularly sensitive to temperature and so rating can change (and so voltage values, i.e. which are fed into ADC, or other way around - the digital values which are converted from digital to analog and amplified (there are always some resistors in control loops of OPAMPs in most cases)). I don't think I can give any data for that.

[Brian Joy]

JEOL uses metal film resistors where necessary; these typically have tempco = 50 ppm/°C and also low voltage coefficient.  As far as op amps, even the humble 741 is specified by TI to have an average input offset voltage drift of 15 µV/°C; for variation of a few degrees C, this drift is at least an order of magnitude lower than the input offset voltage specified at 25°C (typical maximum = 1 mV, 5 mV guaranteed [for +/-15 V supplies]).  For a precision op amp, the corresponding values are at least an order of magnitude lower.

Can you present data to show that the centroid of the pulse amplitude distribution shifts position significantly (perhaps a few hundred millivolts) as a function of temperature (in any part of the lab or instrument)?  Of course, as you pointed out, the water content of the air will tend to rise as temperature of the air rises, and so the issue becomes more complicated.

I'm not trying to say that I have all the answers, but I can show conclusively that the PHA centroid varies systematically with variation in dew point temperature as well as atmospheric pressure.  Though I presented a suggestion to explain the effect, I can't prove a cause-and-effect relationship -- only a correlation.  I collected data over a period of more than a year and found the results to be largely reproducible, even when comparing data from different seasons.  Of course it's difficult to compare values from summer and winter here, as the winter dew point T in the lab rarely exceeds 5°C, and summer values typically fall between 10 and 15°C.  I never write in terms of relative humidity since a given value only applies at a specified temperature.

I also obsess over temperature in the lab (though maybe not quite as much as you), and I also have to deal with people who think that temperature variation is of no consequence.  I see effects on peak positions (for all diffracting crystals, not just PET) and count rates (including sealed Xe counters), and I attribute this to expansion and contraction of metal parts as temperature varies.  Obviously this will change the geometric relationship between sample surface, diffracting crystal, and counter window.  JEOL spectrometers are constructed almost entirely of brass, which has a relatively large coefficient of thermal expansion. 

Now talking about electronics, most of datasheets of electronic components present in one or other form how the values of described component react to temperature changes and (optionally) humidity. Resistors are particularly sensitive to temperature and so rating can change (and so voltage values, i.e. which are fed into ADC, or other way around - the digital values which are converted from digital to analog and amplified (there are always some resistors in control loops of OPAMPs in most cases)). I don't think I can give any data for that.

[Probeman]

I also obsess over temperature in the lab (though maybe not quite as much as you), and I also have to deal with people who think that temperature variation is of no consequence.  I see effects on peak positions (for all diffracting crystals, not just PET) and count rates (including sealed Xe counters), and I attribute this to expansion and contraction of metal parts as temperature varies.  Obviously this will change the geometric relationship between sample surface, diffracting crystal, and counter window.  JEOL spectrometers are constructed almost entirely of brass, which has a relatively large coefficient of thermal expansion. 

I also have obsessed about temperature stability in the lab, though now that we have that under control in our new facility it never enters my mind any longer.

I agree with Brian that PET crystals show the largest changes in intensity with respect to temperature, but I note this previous post from him on this question:

https://probesoftware.com/smf/index.php?topic=854.msg5422#msg5422

Brian: how much of the intensity change on PET crystals do you now think is due to temperature changes, and how much to exposure to the beam (electron trap issues)?  On Cameca instruments we don't need (or have) electron traps because all our crystals are behind "column separation" windows.  But I still saw significant intensity changes, especially on my PET crystals, correlated with temperature though I can't place my hands on the data at the moment.

[Brian Joy]

Brian: how much of the intensity change on PET crystals do you now think is due to temperature changes, and how much to exposure to the beam (electron trap issues)?  On Cameca instruments we don't need (or have) electron traps because all our crystals are behind "column separation" windows.  But I still saw significant intensity changes, especially on my PET crystals, correlated with temperature though I can't place my hands on the data at the moment.

Ah yes, fond memories…

The effects of failure of the static filter depended on Bragg angle.  I routinely monitor Ca Ka count rate, and, at this peak position, the decrease in count rate due to failure of the static filter was small (2 or 3 per-cent).  Although it pains me to do this, below I’ve presented a plot of Si Ka count rate on PETL at a point in time before failure of the filter and also after; both scans were collected under identical conditions.  The variation in count rate that I attribute to variation in lab temperature is in the range of 1%.  I haven’t investigated the effect systematically because I’m not willing to subject the lab to abnormal temperature variation intentionally.  The changes in peak position with varying temperature are obvious, especially on PET.  At any rate, I think the plot below will answer your question.  By the way, I now monitor Si Ka count rates on the PET crystals, and I also check the static filter periodically with an electron mirror.