Agilent 3396 Integrator: some peaks not being analyzed

[IdiotFool]

Hi, all. I was hoping someone could help me with this... Occasionally, my integrator fails to recognize something that is obviously a peak; despite a sharp, pronounced, defined shape, no retention time is recorded and no area % is displayed. After a series of analyses of the same sample, though, the peak may be recognized.

Is there some setting I don't know about that is making the integrator ignore these peaks? Any help would be appreciated.

Thanks!

[mbicking]

We are getting to a time when there are only a few people left who know what this system actually is; most have retired. But since you have a functioning unit, I can try to help you.

I have not used this actual unit, but the programming is similar to what later appeared in ChemStation. There are several integration parameters that can all cause the problem you describe. You will have to look at the settings and make adjustments.

Threshold/Slope Sensitivity - reduce this value to make it more sensitive for smaller peaks.

Peak width - this should be set to the peak width (usually at half height) for your most narrow peak.

Area/Height Reject - the system will not report values that fall below these limits.

Note: The names may have changed but look for these kinds of settings.

[IdiotFool]

Thanks, mbicking. The only setting that might be the culprit would be the threshold sensitivity. If I understand things correctly, then Peak Width and Area Reject settings only come into play after a run is complete and during the peak calculation phase. The problem I'm having is that the peak is not identified during the run, itself. Retention time is not recorded for what is obviously a peak.

Recently, I was comparing temperature methods in the GC to ensure that a higher temperature ramp after 5 minutes would not cause issues in the chromatograms. To be thorough, I am even comparing solvents that should be unaffected, such as Acetone.

I'm using the same integration method for both heating methods.

The left image is my standard heating method while the right is my "fast method:(please excuse the images; I don't have access to a scanner and had to use my cell phone to photograph them):


For whatever reason, the peak was not identified during the run of the "fast" method. I've had this occur on both methods and repeating typically has the integrator correctly identify the peak. Overlaying the chromatograms, I see no discernable differences between the runs. Peak height and width and retention time are identical.

Bizarre.

Since I can't get the screw-up to be reproducible, I won't really be able to tell if what I did worked... still, I'll mess with the threshold and see what happens. Thanks for the help; if you think of anything else, please, let me know.

[mbicking]

Are there any other settings on the integrator? You may have to print them out and list them here. I guessed at the most likely options.

[IdiotFool]

Are there any other settings on the integrator? You may have to print them out and list them here. I guessed at the most likely options.

When I list the run parameters, the following prints off:

PEAK CAPACITY: 1205

ZERO - 0, 1.5
ATT 2^ - 6
CHT SP - 0.5
AR REJ - 3500
THRSH - 0
PK WD - 0.04

ATT 2^ is the vertical scale, CHT SP is the speed the chart moves per minute and the last three are self explanatory.

Outside of these parameters are Integration functions. Set baseline now, at next valley, at all valleys... draw horizontal baseline, etc. Most I don't think are the culpret, but I do see automatic solvent detection... Might the problem be that the integrator is assigning various peaks as solvent peaks, which are then not recorded?


Thanks, again.

[mbicking]

Compare the peak width for your actual peaks with the setting of 0.04. If they are much smaller that may be one problem. Try reducing the value.

Also, I am assuming that the peaks that are not detected are not significantly smaller than the detected peaks, but you could always try reducing the reject value.

And although it may seem counter-intuitive, try increasing the threshold. Using 0 can cause problems sometimes.

Have you considered switching to a computer based system? It really is time. You are wasting considerable time and resources with this approach.

[IdiotFool]

Compare the peak width for your actual peaks with the setting of 0.04. If they are much smaller that may be one problem. Try reducing the value.

Peak width at half height for the acetone above is about 0.5

Also, I am assuming that the peaks that are not detected are not significantly smaller than the detected peaks, but you could always try reducing the reject value.

Same size. Area for the run that was detected is 102,209,520. This is not a small peak.

And although it may seem counter-intuitive, try increasing the threshold. Using 0 can cause problems sometimes.

I'll give it a shot. The next time a peak is skipped, I'll save the chromatogram and reintegrate using various settings to see what, if anything, get's affected.

Have you considered switching to a computer based system? It really is time. You are wasting considerable time and resources with this approach.

If I had the resources, I would switch to Chemstation in a heartbeat. Unfortunately, it's not in our budget. Because nobody has been here with enough knowledge to properly run a GC in at least five years, the equipment hasn't been used at all. It'd be a hard sell... especially given that I've only been here three months.

Thanks, again.

[Consumer Products Guy]

You should be able to program in stuff like threshold, Integrator off, Integrator on, peak width, etc. all at chosen times with that integrator.

Do you mean a #3396 integrator? I have pdf files for that, like operating manual.

[IdiotFool]

You should be able to program in stuff like threshold, Integrator off, Integrator on, peak width, etc. all at chosen times with that integrator.

Do you mean a #3396 integrator? I have pdf files for that, like operating manual.

Yes, I meant 3396. Yes, it is capable of more robust integration methods than what I'm currently using. I haven't really had the time to play with it; I've got it doing what I need it to do with occasional hiccups I've been unable to correct.

[AICMM]

Idiotfool,

Some suggestions. Peak width will certainly affect peak integration (not just calculation.) Keep in mind that peak width will grow as the chromatogram proceeds. Therefore, you should sit it at one value at the beginning of the run (barring the solvent) and then increase it (double typically) at some later point in the run (about 1/2 way or so.) From the size of the peak you are looking at, I would actually try increasing the initial peak width.

Also, List Time to see if you have any time table events.

Best regards.

[IdiotFool]

I'm trying to wrap my brain around how this integrator works. This is my take...

Every signal that the integrator recieves from the detector is compared to the previous signal value. As a peak rises, the slope is positive and the integrator thinks: "Here's a peak!" Once the slope becomes negative, the integrator knows that the max-height has been reached and it records the retention time above the peak and continues charting. Once the signal returns to the baseline OR there is a positive slope detected, the area under the peak get's assigned to that retention time.

Upon finishing a run, the integrator goes through and ensures that each peak it detected meets the requirements laid out by the operator. If peak widths are too small or areas are too low, the area is not counted. These are all post-run processes.

Where I'm having trouble is that the integrator occasionally fails to identify the peak during the run; no retention time is assigned, so no area is stored. These are highly visible spikes, so it's odd that they're missed. What's more, running the sample again, the integrator may identify the peaks.

I can't see how the value of the peak width could affect the integrators ability to identify a peak during a run... especially given that the width of the peak is MUCH broader than the minimum setting. Couple that with the integrator identifying tiny peaks that are far more narrow, and things become truly perplexing.

If my understanding of how this integrator works is incorrect, please, let me know.

Thanks!

[skunked_once]

Your understanding of how the integrator works is correct. It has been a few years since I operated a 3396 but I think you may be on to something with the automatic solvent detection.

but I do see automatic solvent detection... Might the problem be that the integrator is assigning various peaks as solvent peaks, which are then not recorded?

Be sure that this function is turned off. I don't remember if this is an [INT] function or an [OP] function. Consult the manual for details.

[skunked_once]

IdiotFool and others,

Here is a great article on how peaks are integrated.

http://chromatographyonline.findanalyti ... rticle.pdf

For more information on peak integration, go to the LC.GC web site (a great sponsor) and type "peak integration" in the search box.

[Russ]

Do you have any timed events? Try LIST TIME ENTER. You might try enabling the peak tick marks: TIME 0 INTG() 8 ENTER. This will show if it is detecting the start of the peak but fails in other peak detection criteria.

[IdiotFool]

Well, I had another screwy chromatogram. After saving, I played with several values and re-integrated; none of the settings made a difference save changing the threshold. A 0 threshold value did, indeed, seem to confuse the instrument, as setting the value to -1 or 1 both worked to identify the peak that 0 would miss.

Negative values confused it in a different way, so now it is set and left at 1. Fingers crossed, there will be no more issues.

Thank you, all, for your helpful discussions.