[pottesjaak on den hoak]

hi;

I have been working on a problem at my work.

we have 2 HPLC's in our lab,

an Agilent 1260 series with a variable wavelength detector
and a 1100 series with a deuterium diode

both have a ODS hypersil reverse phase ( 10 microns particle size and both same lengt and diameter ) column installed

we run test on XXX( patented )

= 3-[3-(3,5-ditert-butyl-4-hydroxyphenyl)propanoyloxy]-2,2-bis[3-(3,5-ditert-butyl-4-hydroxyphenyl)propanoyloxymethyl]propyl] 3-(3,5-ditert-butyl-4-hydroxyphenyl)propanoate

and other additives . . .

with methanol as mobile phase, both systems operate on the same parameters ( flow, pressure, temp, wavelength, internal standard etc )

now here is the problem;


we put the exact same sample in each LC system and they each produce very different results, even though the methods on the software integrations are the same ,

the average difference between each system's series of injection
is about 1.3 area %

however, each series of results produced by each LC has a very small standard deviation; wich would mean that each LC system is very accurate in itself and NOT compared to another system running the same sample


anyone who has any tip or clue or even the slightest idea please reply

[Archer]

Something is definitely wrong with one of your instruments. Have you tried using the same column on both instruments? Are your guard cartridges both new?  If your peaks are tailing more with one column then the integrator may be cutting the 1.3% off your analyte. Poor peak shape can be a result instrument specific factors too.

This really highlights why you need QC checks with certified (or at least standardised) materials on any instrument.

[pottesjaak on den hoak]

thanks for your reply ,

both systems run on brand new columns, packages of the columns were open and installed the very same day on each LC's

the retention times dont differ much at all , the columns are not deteriorated and the peaks that each instrument shows have a good symmetry ( .90 and .92 )


and yet for all 8 injections on instrument A shows more than +/- 1 area % difference
in comparison for the 8 injections on instrument B

 the lab manager wants to keep the difference down to .3 % max to be acceptable for the products we analyze

that is how i got to a dead end;

[Archer]

How can both instruments have very low standard deviation and yet one gives ± 1% compared to the other? That's statistically impossible. You are saying that you have good precision on both instruments but the precision is poor.

Does one instrument give consistently lower results each time i.e. one instrument is lacking accuracy?

Have you tried producing a calibration curve of the analytes to see whether this makes a difference i.e. is the linearity similar on both instruments?

Could you present your data? Peak areas for analyte and internal standard for both instruments.

[DrCMS]

So two different detectors and two different sets of results I wonder what might cause that?

[Archer]

In principal a diode array should give the same results as a variable wavelength instrument if they are both calibrated correctly. It's not like using an LC-UV and a LC-QTOF at parts per trillion concentrations and wondering why the latter give much better sensitivity.

Both detectors work on the same principal just one has a filter before the sample and the other diffracts the light after the sample.

If an internal standard is being used then, while the response from absorbance may change from one instrument to the next, this should not matter becuase so will the extinction coefficient for both analytes.

Provided the instruments are calibrated correctly they should give the same results for a standardised material.

That being said, if we consider this post on another thread:

http://www.chemicalforums.com/index.php?topic=68971.msg253296#msg253296

perhaps there is a way around it,

I don't know what software your system runs on,

But whenever I encounter callibration problems i tend to cheat and outsmart the system,
if there is a mode in your software, you could manually integrate your peaks let it calculate the area for you,
and by writing down that very area, you can establish a series of different areas of different concentrations.
which you can fill down in you callibration table

that is only if the software supports manual integration ( and other ways of data analysis )

I think the problem may lay somewhere between the data-system and the chair!

[MOTOBALL]

Although "everything is the same except the detector", there is another difference--the actual needle/injection system.

Are these filled-loop systems (which really should give identical injection volumes) or syringe systems that may be variable from instrument to instrument ??

Is there a pre-injection needle wash vial on each instrument ??

[Arkcon]

Is there a pre-injection needle wash vial on each instrument ??

Additionally, if the needle wash has been neglected on one instrument, the seal material may have dried out and become faintly leaky.  Not enough to make a puddle, but enough to cause loss that affects reproducibility.

For that matter, there isn't a big, visible air bubble in the syringe, right?  Because that will definitely affect injector reproducibility.  Hopefully, you're able to run an injector purge, to remove possible air bubbles from the loop.  That's an important part of daily maintenance, when reproducibility is very important.

[Archer]

Problems with the needles do cause poor reproducibility but, apparently this is not the problem:

however, each series of results produced by each LC has a very small standard deviation; wich would mean that each LC system is very accurate in itself and NOT compared to another system running the same sample

Presumably by "accurate" the OP means good precision.

If the SD of multiple samples is very small then it's not an instrumental thing, it's a data system / calibration issue.

[MOTOBALL]

You might try making say 8 x 100 uL injections of water from a pre-weighed, capped vial. Re-weigh vial to determine wt and vol removed (ca. 800 mg and uL)

Repeat on second HPLC; is the difference ca. 810.4 mg and
uL (1.3%) or larger ??  If yes, the area discrepancy is probably (at least partially or totally) injection-related; if no, then integration parameters are also contributing, as per Archer.

[Archer]

 both systems operate on the same parameters ( flow, pressure, temp, wavelength, internal standard etc )

Injection volume issues shouldn't arise if an internal standard is being used. 100μl or 98.7μl, the relative responses between the anslyte and the internal standard, in principal, should be identical.