Translate method from Waters 2690 to u3000

[karrun]

I am new to LC work and am not sure where to begin. I need to translate a method from an old Waters 2690 with DAD detector to a newer u3000 LC with DAD. I used the same method for the gradient analysis on both instruments and same column, but am getting about double the absorbance units on the newer instrument for the same sample. If the cell path length is the same, shouldn't the AUs be the same? What else can I check to see what could be causing the difference? Bandwidth, data collection rate? The Waters is quite old and the lamp hasn't been changed in almost four years (I have no idea how many hours its been on), could the lamp intensity cause the difference?

[tom jupille]

First of all, it really doesn't matter unless you're down near the LOQ or LLOD (and since the new instrument gives more absorbance, that's probably not something to complain about!). Remember that a calibration curve is only valid on a single instrument.

That said, absorbance is calculated from the ratio of transmitted light through the flow cell to light going to a "reference" photodetector. The intensity of the lamp should not affect that, since it cancels when you take the ratio (an old lamp will have more noise, but that's a separate issue).

If you truly mean "absorbance" (i.e., peak height not peak area), then sampling rate / bunching should not have an effect unless it's horribly slow (such that you are measuring less than 20 data points across the peak), and even then it's hard do imagine it having a 2x effect.

Spectral band width (or slit width) can definitely affect amplitude.

If you are using a high-efficiency column (one that gives narrow peaks) it's possible that the older system may have contributed enough extra-column band broadening to degrade the peak width (and therefore, the height). You should be able to check that by comparing the peak width on the two systems.

[karrun]

Yes, I am interested in peak height, not area counts. This method looks strictly at peaks at a certain time span and anything over 5 mAus in height is considered off-spec. No calibration is used.

If I understand you correctly, a column that is less efficient might give a broader peak for the component, resulting in a lower peak height. But I have compared the peak width of the two instruments for the component of interest, and the newer instrument actually has a broader peak width.

So looks like I need to compare bandwidth of the two instruments as a possible factor? Is a lower bandwidth related to higher amplitude, or vice versa (did I say I was new to LC)?

[tom jupille]

Actually, this is more a question of spectrophotometry than it is of chromatography.

The effect of bandpass on response depends on where you are relative to the absorbance spectrum. In the (very crude) sketch below:
- in the upper spectrum, you are measuring at the apex of a fairly sharp absorbance band. Increasing the bandwidth takes in a larger proportion of the lower absorbance part of the spectrum, so the "average" (measured) absorbance decreases.
- in the middle spectrum, you are measuring on a fairly steep slope. Increasing the bandwidth takes in both lower (left) and higher (right) absorbance parts of the signal, so the "average" stays about the same.
- in the bottom spectrum, you are measuring near the bottom of a trough. Increasing the bandwidth takes in a larger proportion of higher absorbance parts of the spectrum, so the "average" increases.


If this were my problem, I would start by setting all the detector parameters (wavelength, bandpass, sampling rate, etc.) as similar as possible, then prepare a solution of caffeine diluted down to give you a reasonable absorbance (10 mAU ?) at 275 nm. Disconnect the detector from the column and pull the caffeine solution into the flow cell with a syringe. If I *don't* get approximately the same absorbance on the two detectors, then it's time to start digging into the cause (e.g., are you sure the 2690 actually has a 10mm path? is there a chance the sample has degraded between the two runs? )

[karrun]

Wow, thanks for the very detailed answer. Your illustration really helps me understand the effects of bandwidth.
I would love to be able to take your advice and try the experiment you suggest. Unfortunately, the Waters instrument is not working right now, a big part of why I need to "qualify" the new instrument for this method.
I think it is quite possible the sample has degraded. I could only compare chromatographs of a sample that had been run back in November of last year when the Waters was running to a recent injection on the u3000. The thought had occurred to me about degradation, but I thought it likely it hadn't changed greatly. Possibly I'm wrong.
I did see that the previous analyst ran some single component standards. I am going to try to make up fresh standards and compare the response of both instruments.

Thanks once again, you've been very helpful.