SCAN more sensitive than the SIM

[Absolutdude]

Hello,
I work with an Agilent 5975c in EI-Mode and want to calibrate my standards.

I started with the Scan-Mode and then switched to Sim. (I also do Sim and Scan at the same run)
The Scan shows higher abundances as the SIM and I wonder why?

I choose the right fragments for the SIM. I checked it at the data-library and also in the actual Scan-run. I also tried to take just the highest fragment and e.g. the five highest fragments.

But the Scan-Mode always shows higher abundances.

Do you know how this can be?

[dblux_]

Hello,
I work with an Agilent 5975c in EI-Mode and want to calibrate my standards.

I started with the Scan-Mode and then switched to Sim. (I also do Sim and Scan at the same run)
The Scan shows higher abundances as the SIM and I wonder why?

I choose the right fragments for the SIM. I checked it at the data-library and also in the actual Scan-run. I also tried to take just the highest fragment and e.g. the five highest fragments.

But the Scan-Mode always shows higher abundances.

Do you know how this can be?

Have you tuned masspec ?

[R13]

First: do Scan and SIM at the sime time just in case of some problems - as exception. Doing both requires lots of time for the istrument to gather data - less sensitivity, etc.... When You know which ions to collect - swich to SIM and do not collect Scan data. I think this is the source of Your problem.

Do not collect too many ions at the same time in SIM - 2 or 3 is the best choice, make time segments if more ions (compounds) are needed.

In the SIM parameters -- selecting "Low resolution" might increase sensitivity.

If You have many ions in SIM - make Dwell time shorter (default value of 100 ms is for one ion). That might increase noise, but chromatographic peaks will be sharper.

[Don_Hilton]

In setting up a SIM mode acquisition, set the dwell time long enough that you are not oversampling the peak, but short enough that you are under sampling the peak.

In oversampling the peak, you are sampling many times at short duration. But, you increase the total dead time while the instrument is switching between ions. This results in loss in sensitivity because the instrument is spending too small a portion of the time actually acquiring signal.

In under sampling a peak, you have long dwell times and little dead time (good), but switch few enough times that only a few data points are gathered across a peak (bad). In this case, the acquisition fails to get a good representation of the peak shape - and such measurement is highly variable.

The rule of thumb is that it takes at least 8 points to properly sample a Gaussian peak. So, sample a bit more than that and you should be good. This if you have a peak 20 seconds wide, aim for the complete acquisition cycle to be 2 seconds. This would give 10 points across the peak. And set this up for a peak that is not overloaded on the column. Overloaded peaks tend to become broad. Slightly oversampling a large peak is less of a problem than undersampling a small peak for a less concentrated sample.

As pointed out before, dwell times do not have to be the same for all ions. You increase the length of the dwell time of a particular ion relative to the others to increase the sensitivity of the instrument to that ion at the expense of the others.

And - last edit - the common usage of SIM mode acquisition is one strong ion that can be used for quantification and two ions that can be used for confriming identity (by examination of the ratio of confirming ion intensith to the quant ion). The addition of addtional ions for a given compound does not add much and decreases sensitivity.

And, the quant and connfirming ions are selected, not just on size, but on the abiltiy to obtain an interference free trace in your samples. Example: with TMS derivatives m/z 73 is often the strongest ion in the spectrum - and in many cases it is by far the strongest ion. But becaue this ion is seen in "everything" in the chromatogram, it is by and large useless. As a geneal rule higher mass fragments are a better place to look for unique traces than low mass fragments.

[tlahren]

I have also noticed that it seems when running dual SIM/Scan acquisition the calculated time/cycle (s/scan) is incorrect. It seems to estimate the same as it would for a SIM only method i.e., you have to reduce your ion dwell times more in SIM/Scan than in SIM only to get the same SIM counts per peak. In SIM/Scan every other data point acquired is a SIM data point and the others are Scan data. Basically, I've found that when running dual SIM/Scan it is hard to get the same sensitivity that you would using SIM only.

Has anyone else seen similar with this or contrary?

[Yama001]

I recall seeing a note about the incorrect cycle time for the scan/sim mode in Agilents setup. It looks like they just never told the software how to handle it.

An old Finnegan instrument manual from the seventies claimed the relationship between dwell time and signal to noise is a square root type of thing; i.e. a 100x increase scanning a particular ion will yield a 10x increase in S/N. Easy to lose the benefits as you squeeze more ions in.

[lmh]

the square-root thing is nothing to do with chromatography, it's just statistics in action, and it's only half the story.

There are two sorts of error when sampling across a peak, errors in the y-direction and errors in the x-direction (area of a peak is, of course, affected by both). It's easiest to think about errors in the x-direction for a hypothetical square-wave shaped peak. If we managed to synchronise our detector such that it always started to measure just as the peak began, we'd have an uncertainty on the peak's width, because the peak may have stopped anywhere between our last non-baseline peak, and the first base-line peak. This error is uniformly distributed, not normally distributed. If you were to decrease the time between points linearly, you would see the error increase linearly to a point, instantly drop, and then increase linearly again, but each saw-tooth would be a little smaller than the one before.

This average size of this error is linearly related to the sampling time. Half the sampling time means half the error.

The Finnigan square-root on S:N ratio is, I suspect, another thing altogether: it's looking at the y-error. Each point is an estimate of the height of the peak (at that time). The more estimates you have, the better your knowledge of the peak height. If each measurement has a standard deviation (which is constant, and is "noise"), then our best estimate of the peak's height ("signal") is the standard error of lots of measurements, and the standard error decreases with the square root of the number of measurements we make (SE = sd/sqrt(n)), so 1/100th the scanning time means 100 times as many points, which means a 10-times better SE estimate of the y-measurement of the peak.

Another way to look at it, is that there isn't much difference between calculating a peak-area and calculating a mean of a load of measurements. A mean is the sum of lots of values divided by n. An area is a weighted sum of lots of values multiplied by their spacing. The error of a mean is therefore influenced by the same things as the error in an area, and we can apply everything we know about standard errors to the calculation of peak areas.

[tlahren]

lmh-

Very good explanation. Thank you.

Ty