Improving RSD?

[Ytterbium]

Hi,

I had been trying to build a calibration curve but was unsucessful, so I checked the RSD of my system and it was as follows:

275 - 235 - 220 - 215 - 285 (nm)
2.5 - 1.1 - 4.0 - 6.4 - 3.4 (%RSD)

This was calculated from the peak area of a single peak (no other peaks in chromatogram) across 25 samples, where there was 5 injections of 5 samples all taken from the same stock solution, total loop fill of 1ul. wavelenghts were selected from spectrum as 275 maxma, 235 minma, 220-285 flat reigons.

Long story short

I have now got the following:

275 - 235 - 220 - 215 - 285
0.6 - 0.6 - 0.5 - 0.6 - 0.6

Again across 25 samples in the same protocol.

Should I be happy with these RSD's or should I continue to try and improve them? If I should improve them what areas should look at improve the RSD?

I have a Agilent 1100 system but with a Gilson sampler and rheyodyne loop. I have already replace the the flow cell, lamps, rheyodyne and seals of injection port needle, needle.

I have lots more infomation if needed

[Mark Tracy]

First, check the specifications for your autosampler. As long as you are comfortably within spec, you should be good to go. (For context, the spec on mine is 0.8 %RSD; the last qualification came in at 0.2 %RSD.)

Given that your first set showed different %RSD at different wavelengths makes me think that you might have had either contamination or peak integration problems. The second set, being consistent, suggests that you have eliminated that problem.

[Ytterbium]

Gilson specify:

0.5% RSD for a 200ul total loop fill across 10 injections by HPLC

<0.9% for a 20ul total loop across 30 injections (using fluoresence as detection).

I suppose 0.5% for 1ul by HPLC is OK?

[Alfred88]

Normal sampling loop is 20-50 uL. You should inject a higher volume to have a lower RSD (how about 20uL?). In our tests, we usually inject 5uL or more. (3uL is very rare)
If injection volume is 1uL, RSD may exceed 2%.
Alfred.

[Ytterbium]

I don't do a partial loop fill, I have a 1ul loop so do total loop fill.

I could change up to a 5ul loop, again total fill an dilute my sample more.

Thanks for the advice

[Peter Apps]

Hi Ytterbium

How good does the rsd need to be for your analyses to provide useful data ? As long as it meets that target you do not need to make it any better.

Peter

[DR]

<1% is pretty much universally acceptable for any assay using an 1100 and any decent autosampler. Sure, you can do better, but nobody I know would worry about rerunning a set w/ <1% RSD.

The variance across wavelengths is actually indicative of a detector problem - wavelength accuracy/consistency - were not up to par. Evidently the new lamp & flowcell fixed that.

I hope you changed one component at a time so that you don't spend more $ than is needed next time you have these issues...

[Ytterbium]

Thanks for the advice everyone, I'll try to do a calibration curve and check the result for that.