Page 1 of 1
%RSD for injection repetability
Posted: Thu Jul 26, 2012 3:21 pm
by raman_srividya
%RSD for 6 injections of API(in 100% acn) is about 1.7% if injected from the same vial , but is less than 0.5% for 6 injections from different vials. the autosampler temp is 5C so there is very little chance of evaporation. Is there an explanation for this senario?
Re: %RSD for injection repetability
Posted: Thu Jul 26, 2012 5:40 pm
by michaelbarnes42
is this just from one set, it is easily within the realm of possibilty to get those results given a small enough data size.
Re: %RSD for injection repetability
Posted: Thu Jul 26, 2012 6:03 pm
by remesquaddie
I would try that using self sealing septa.
Re: %RSD for injection repetability
Posted: Thu Jul 26, 2012 7:07 pm
by JGK
What injection volume are you removing?
Could you be creating a vacuum effect in the vial which could affect the ability of the autosampler syringe to draw samples.
A vacuum could be created if you overtighten cap and septa at room temperature then cool to 5C which could affect injection performance.
Re: %RSD for injection repetability
Posted: Tue Jul 31, 2012 9:04 am
by aceto_81
Can you see an increasing or decreasing trend with your injections from 1 vial?
Or is it just random?
Ace
Re: %RSD for injection repetability
Posted: Thu Aug 02, 2012 4:31 pm
by lmh
If the difference is reproducible time-after-time in successive experiments, ignore part 1 of this post!
Part 1:
coincidentally, see Veronika Meyer in LC-GC this month (August 2012) for a rather long investigation of the errors in experimental measuerments of errors. If you've measured the standard deviation of the mean in one experiment of n replicates, you could go on to carry out m experiments, and get m different standard deviations. Then you could calculate the standard deviation of the standard deviation. Meyer's point is that it's often quite big relative to the standard deviation of the mean. If you found your difference in only a couple of experiments, it's quite possible it's down to chance.
To be honest I can't imagine many scenarios where even as much as a 2-fold random error in the standard deviation of the mean of analytical replicates really matters.
Part 2:
I agree with aceto_81 that ongoing trends are more worrying than obviously random fluctuations; JGK's right about vacuums. Another thing I've seen happen in a cooled autosampler is that a drop of condensation can appear in the dimple left at the top of a vial after a needle's gone through the septum. If you're unlucky, it then dilutes the next injection, and so on...