5890 TCD Baseline noise

[aidnai]

Hello all,
I am doing initial setup of a GC-TCD. The instrument is a used 5890 with a split/splitless injection port as well as a gas injector valve.

The baseline seems noisy to me, so I wonder what is normal. As I investigated, I found that most of the noise can be eliminated by turning off the GC oven (I assume the oven fan is the problem). Does this indicate an electrical problem? can I do something to isolate the oven fan?


This is a screen shot of the noisy baseline with oven on, and the quiet baseline with the oven off. mV scale visible on the side.

Other parameters:
det a temp: 175C
column + makeup flow rate: 6.7mL/min
Ref flow rate: 27.5mL/min
Helium carrier

Many thanks,
AP

[Consumer Products Guy]

Adnoid -you need to run a real life sample. You may end up being thousands of times higher in signal than your worst noise shown. So you may just be worrying about nothing. These days one can magnify anything a zillion times to look like a mountain range.

I just went in and looked at one of our 5890 data files (which is FID). Our noise (top to bottom) is about 5 Area Counts on the scale, more than yours, and we get great chromatograms and performance.

[aidnai]

Thanks for your reply.
My SNR is not even 1000:1 for a full scale peak (noise >1mV, full scale is 1000mV) and I am trying to quantitate some peaks that are roughly 5% of full scale. That gives me a signal to noise of about 50:1 in those cases. I'd like to do better.

I notice your scale is counts, mine is mV. Trying to estimate from a printout, my noise is about 200 counts top to bottom.

AP

[AICMM]

aidnai,

In my opinion, you have too little flow for the 5890 TCD. Try pushing up the make-up flow (assuming the column flow is where it should be) and then increase the reference flow to something like 2-3 X of that. See what happens with that. I would shoot for 30 mL/min total flow and then I would set the ref for something like 60 mL/min. The price you pay is sensitivity. However, TCD is not all that sensitive to begin with.

Best regards,

AICMM

[Consumer Products Guy]

Thanks for your reply.
My SNR is not even 1000:1 for a full scale peak (noise >1mV, full scale is 1000mV) and I am trying to quantitate some peaks that are roughly 5% of full scale. That gives me a signal to noise of about 50:1 in those cases. I'd like to do better.

I notice your scale is counts, mine is mV. Trying to estimate from a printout, my noise is about 200 counts top to bottom.

AP

Aidnai - I'd try what AICMM suggested (we haven't run our 5890's TCD in a decade). Signal-to-noise of 50:1 may not be "ideal" but can work fine. In our HPLC work we can frequently be down that low.

[aldehyde]

I don't run 5890s much but I can usually achieve 0.15 DU* ASTM noise with a 6890 using TCD.

(*for agilent TCDs one display unit = 25 uV, look at the scale on the chromatogram screenshot)

FYI, you can collect noise for awhile and then in data analysis mode make sure to set your menus to full (check the bottom entry on the view menu and make sure it says short menu), then go to report > specify report, set report style to Performance+Noise. Then go to report > system suitability > edit noise range and give it a time in minutes (like if you collect 10 minutes of noise set it 0 to 10) and then report > print report.

On page 2 of the report you'll get noise information:



and for comparison here is my chromatogram:



HP-5 column, inlet 250 C and 25 psi, constant pressure. Oven temp isothermal 100 C. TCD 250 C and 20 mL reference/2 mL makeup. Most of the spikes are people walking by and stuff.

Are you sure your column wasn't touching the metal inside the oven? That can cause weird noise. You might want to cap off your detector so you don't have to worry about anything but the detector and the electronics. If you dont have a cap just cut the column an inch or so under the detector and cap with a septa.

I would also turn the filament off and watch the signal, which will be measured at 0 uV but will still jitter around a bit from dark current. If you have a problem with your detector electronics you may still observe noise.

[aidnai]

Hi all,
Thanks again for the replies.

AICMM -- your suggestion succeeded in lowering my signal, but not my noise >_<
I have a .53mm column x 30m, so I'm running it at 5mL/min with a head pressure of only about 3psig. I wanted to keep the make up low as possible but not get into situations where there is unswept dead space (my make up is about 2mL/min) ... I'm used to packed columns myself so the numbers feel weird to me but I was going off of agilent guides while choosing rates/pressures.

I really believe the problem is electronic in nature, as evidenced by the drastic reduction in noise when the oven/oven fan is turned off.

Aldehyde -- Thanks for looking into things for me. I did feel like the noise should be lower, you pretty much confirmed it. I did the noise determination you recomended and got the following results

My scale is still 'counts' whatever the hell that is, instead of 25uV or something I can understand. As close as I can tell from my printouts, 1000mV is equivalent to 200000 counts, giving a conversion of 25uV -> 5 counts. That would (correct me if I'm wrong) make your PtoP noise 1.1675 counts vs. my 270.9896 counts ...

I checked the column, it isn't touching anything except its cage (I kinda think you had a packed column in mind anyway). I also checked the baseline with the detector off -- dead flatline. While probing around, I found that some of the TCD filament/temp sensor leads are losing their insulation - i can actually see bare wire on a few. Should I consider re-wrapping them?

AP

[GasMan]

Can you check where the exit of your TCD is. You say that you have a used 5890. It is possible that the original GC had the TCD and FID in series, which meant that the exit of the TCD is inside the oven. This would have then been connected to the FID by a short restrictive jumper tube. If the exit of the TCD is still in the oven, and is not connected to a FID,then you will get a noisy baseline when the oven fan is rotating due to small pressure differences inside the oven.

Gasman

[aldehyde]

If you're able to re-insulate the wires carefully that would be a good idea, it may not be causing your problem but it could lead to other issues in the future. Be careful, 5890 wires are very brittle and if you do break them it can be near impossible to remove the heater cartridge from the detector weldment without destroying things.

[aidnai]

I reinsulated the wires carefully -- no effect on the noise. Then I checked back on this forum: OH MY GOD GASMAN YOU FOUND THE ANSWER!

So simple, but I doubt I would have ever thought to check it out. You are right, the TCD vent was piped into the oven, when I piped it out of the oven the noise was substantially reduced.

Thank you everybody for your time, thank you especially gasman for getting the right answer!