Thank you all for the suggestions. Unfortunately, due to a filament issue cropping up overnight (box and trap are suddenly jumping around by +/- 0.04 mA), I'll have to investigate this further later.
Travis - We are using Isodat. There is 20 seconds between peaks, and the backgrounds rises about 1 mV between the first and second peak. This may be relevant, as if I insert a reference peak where the sample normally appears and thus ~110 seconds passes before the next ref peak comes out, their difference is smaller (0.15-0.25‰).
Ben - I'll keep an eye on the actuator in the future. That's one of those things I wouldn't normally think to consider.
Andy - I have tried both with and without the EA triggered, and the effect is the same. Inserting a third peak did reduce disparity between the final and its prior peak, down to a difference of about 0.2-0.3‰. I have 10 peaks in my on-offs. The first peak is always the heaviest, though not by an extreme amount.
Stewart - We're using a DeltaPLUS XP. Hopefully it doesn't come to that!
Good Day Rich,
I trust you have a EA/Delta V Plus set up. If not read no further but might be useful for others. We had very similar issues in 2015 with our dual measurements CO2 ref values thinking it was due to jump calibrations or magnet calibrations errors or…. We ended up doing individual N and C only measurements.
The solution was to have the Delta’s electro magnet removed, cleaned and plates inserted to shield it from the Aluminum housing. Apparently it accumulates sticky static particles. This definitely did the trick after long troubleshooting days.
Hope this helps
I'm having an issue with CO2 reference peak deviation during an Elemental Analyzer N and C run. Two CO2 reference peaks are produced following the C sample peak. Originally, the first peak was consistently ~1.1‰ lighter than the second peak. After creating a new N2 to CO2 jump calibration, the first reference peak became consistently heavier than the second by about 0.6‰. We have sufficient distance between the sample and reference peaks, and the same issue results even if no samples are loaded. CO2 on-offs are fine, with a standard deviation of about 0.04‰. Does anyone know what may be causing our disparity?