It seems, that a fluctuating column bleeding is responsible for the problem.

we have the same problem in this period with the same column using a Waters ELSD. How could you fix this problem?

Several main concerns come to mind reviewing some of the HPLC conditions you report. (1) Your system back-pressure of only 28 bars is far to low to be reliable (You wrote: "System pressure with detector around 28 bar (400 psi)"). Most HPLC pumps can not provide highly uniform flow, composition and baseline stability at such low pressures (ignore what the screen shows and measure true S/N of peaks). Most pumps are optimized with system backpressures > 50 bars. If the back-pressure is not well controlled, instability may result and the nebulizer will not atomize the effluent well resulting in fluctuating droplet size. *Your column shows a max pressure limit of ~ 67 bars. (2) "House Air" is often far too impure for use with ELSD or CAD modules. Just about everything in the air will be detected by those detectors so the standard recommendation is to use ultra pure nitrogen similar to what we use with LC-MS systems. 'Tank' gas is also easier to regulate than house air, which often varies, esp during the course of 24 hours. Particulate levels and oil are two key specs that must be minimized in the gas used. Install extra filters. (3) For sugars, one of the most important factors which effects reproducibility is proper column washing procedures after analysis. This is extremely important when using only an isocratic method as the samples quickly build-up and foul the column (esp running only water, as you are). You must develop and use proper washes after each analysis. Some of the columns benefit from the addition of some alcohol in the mobile phase (for washing, not analysis). Refer to the column supplier/manufacturer for advice regarding which chemicals are compatible for use with YOUR column. (4) Are you aware of the fact that those ELSD (and CAD) detectors fill up with sample over time? It gets baked onto the inside, then flakes off over time. They require regular disassembly and internal cleaning to remove baked on contamination. The manufacturers tend to not bring this up (bad for sales), but it is needed. (5) Detector Optimization: Each sample and method must be fully optimized to the detector (each one). Finding the best gas flow rate, drift tube temperature (your seems too low, esp for pure water), atomization, flow rate(s) for each method must be optimized one-at-a-time, using peak S/N values (not the peak area counts, which are misleading). This is a time consuming process, but is required when using these finicky detectors. Poor selection of detector settings results in very poor %RSD.

• "Evaporative HPLC Detectors; CAD (Charged Aerosol Detector) and ELSD (Evaporative Light-Scattering Detector)"; https://hplctips.blogspot.com/2017/12/evaporative-hplc-detectors-cad-charged.html

In addition to what Bill stated above, I would add that your sample matrix (enzyme digest) is interfering. How are you killing the enzyme before injection? Are any enzymes eating the HPLC column (thus bleeding)?

Thank you very much for all the helpful comments! I will update this conversation when i have improved my measurement!

Here is one more article that may be of some help to ELSD or CAD users. The article referenced below was written for a different model of ELSD, BUT the same concepts of optimizing for gas temperature, gas flow and so on... all apply to any brand of ELSD or CAD system. *Optimization, using peak S/N ratios, not area or height counts, is required to optimize the settings of these finicky detectors with EACH sample type.

" Technical Report Evaporative Light-Scattering Detector (ELSD) Optimization Pr...

Have you checked the carry-over possibility? Inject reagent blanks between injections and test the reproducibility again. We should be sure about the properly working pump and sampler first to think about ELSD. Ligand exchange columns are polymeric and highly stable ones and your LC flow, column temperature looks fine. The pressure at these conditions is expected any we could say that it won't be any problem if the fluctuation is not occurring. You said that RT reproducibility is ok then we can exclude this for now. I would suggest adding diluted acid (say H2SO4) to your mobile phase. Afterward, I would try to test the whole system with nitrogen-supplied ELSD rather than air supply. ELSD linearity can be problematic in some cases. Inject multiple injections from the same vial and varying concentrations to catch the altering point and to find out whether it is conc. dependent or system dependent.