As alluded to in the powder experiment log, the necessity for a new sensor (Transducer Voltage) has arisen from peculiarities in the pressure data.
The roller coaster pressure graph above is nothing out of the ordinary, as ambient temperature cycles with the presence of solar heating. However the dramatic change in the downward slopes is something new to us, and it really has us scratching our heads.
These 0.3-0.6 psi drops would seem an artifact of some temperature change or change in power but they don't couple to any of our temperature measurements nor phase change in the bulk. They've happened over 3 different AC frequencies and have a sort of periodicity to them - showing up about every 11 hours with the exception of the last drop (yesterday evening) shown above. I'm pretty pleased to see the instantaneous jump in all three lines at the right edge. The powder shifted phases to increase its impedance and thus its Ohmic heating and pressure went up as well. But why aren't these cliffs showing up anywhere else?
A closer look at yesterday evening's event shows that the loss occurs within a 6-8 minute period over a gradual curve instead of the sheer drop shown in the 1 minute average figures above. These data don't appear to be faults in the ADC, nor do they agree with temperature fluctuations in any of our 4 thermocouples or changes in resistance. The body of the sensor is well isolated. There is no threshold immediately present in any pressure, power, or what have you that would cause the transducer to go on vacation for a couple minutes.
Here's another close look. The wavy temperature line is due to the lab's air conditioning duty cycle since the supplied HVAC vent was blocked at that point in time. About the same slope as before and same drop.
We can only think of two possible remaining causes at the moment: either a change in excitation voltage (hence the transducer voltage channel) or interfering RF noise picking up on the various leads to the powder bulk. The supplied power for this transducer is coming from an external 9V plugged into a very busy power circuit. If the Transducer Voltage drops then also should the reported pressure. I have my gut feeling, but noise in the supplied power should show up elsewhere and so forth. We'll just have to see. If it's down to RF noise then a handheld radio should really take the data for a ride (Update tomorrow).
Are there any other possible explanations? With no correlation in any other sensor we're having exceptional trouble believing this to be a real physical effect, but clearly we're struggling to dismiss it just the same!
UPDATE #1 - Plan of Action
Back from a slight vacation, and R3 is up to its usual tricks - but with greater frequency. With all options well considered, there are two seemingly possible outcomes: either we have increasingly faulty data from our transducer, or increasingly interesting phenomena happening in the bulk. Take a wild guess which one seems more likely.
As suggested by Paul, the next best option for us would be a second transducer's opinion. We'll have to put it in the cell on the outside of the bellows valve and add another quarter turn valve on the following side, yielding a hydrogen line of: test cell--transducer in question--bellows valve--new transducer--quarter turn valve--reef hydrogen supply. I want to make this little addition as non-invasive as possible; we'll slightly increase the cell volume as well as keep the same hydrogen atmosphere with very little fuss so as to not disturb any potential lattice phase-changes. I would never forgive myself if we plugged in the new transducer and they both agreed perfectly, but with no more pressure anomalies. Fingers crossed.
To eliminate additional variables, the new transducer will also be put on the same HUGnetLab board and ADC1 as is the current transducer. The transducer voltage data channel will be discontinued for the project.
A little confusing? Maybe. All will be clearest when I return with pictures and a more detailed update. Stay up!
PS. Not 5 minutes after posting this blog update did I realize that the individual circuit reading the pressure transducer may have a hardware bug, since it's obviously not the whole ADC1. After all, this is not an uncommon problem for small individual pieces of the lab boards to fail spontaneously: the reef power supply control board was replaced a while ago for that exact reason. I can reconfigure the board to switch the pressure and P_Voltage pins. If there's a finicky electrical component, our Voltage driver is sure to see it. Why we didn't think of this before is unknown, but a simple fix indeed. I guess we just have our noses too close to the merchandise.
In summation: I will probe for another electrical issue BEFORE adding a new transducer (tomorrow or Friday).
UPDATE #2 - Culprit Found?
Results of last night's sensor switch are in, and the data look incriminating to my eyes. Just before 0900 the P_Voltage increased a little and then dropped in a fashion similar to what we've been watching in pressure. At 0844 we saw 0.080V, while at 0846 it was 0.079V - yes, we are looking at a miliVolt drop. It occurred at around the same time during a down curve (just like the pressure drops), so we have that going for us. . .
Additionally, there was no coupling from another sensor, which makes the situation fishier. It is odd though that T_Shell stepped out of line at the same time. If this really was a physical effect I'd expect to see ripples in either of the internal thermocouples, but T_Powder didn't follow suit. The little spike here is only an 80mK blip, that's really stretching the resolution of the sensor.
Puzzling stuff. It just doesn't quite look the same as the pressure anomalies. If there is a repeat offense by tomorrow I think we'll know where the problem lies.
For now, the jury is still out, if only a little bit.
Comments
Additionally it might not hold up well in a hydrogen atmosphere. Hydrogen diffusion through the silicon sensor membrane should be pretty poor, shouldn't it?
Of course, sensor artifact should be excluded first...
RSS feed for comments to this post