Calculate the LODs of a gas analysis

[thohry]

Hi all,
I am analyzing some gases and hydrocarbons in transformer oil by headspace and GC. Now I'd like to calculate the LOD of the method, let say for H2. According to some documents, the LOD of H2 is about 3 ppm and I have only standard mixes which contain about 100 ppm hydrogen. How can i make standards of 3 ppm or so to analyze for LOD calculation or if there are any other methods to calculate it ? Or I have to purchase the standards of 3 ppm ?
Could anyone give me some ideas ?

Thanks for any input

TH.

[rb6banjo]

To truly get the LOD, I would get a standard that is close to the detection limit and measure it. If it's not critical to know it exactly, you can measure the signal-to-noise ratio at 100 ppm, assume the response of the detector is linear, then use a theoretical S/N for legitimate detection (say 2 or 3) to estimate the detection limit. The math is a simple proportion. It would give you an estimate without having to purchase a standard. That can become a pricey venture. I've never had good luck making gas standards myself.

[AICMM]

TH,

What size loop are you using? Say you are using a 1 mL loop. Then try a 100 uL loop to get you to 10 ppm "on-column". That gives you a snapshot of S/N at much lower level.

Best regards,

AICMM

[Peter Apps]

Hi all,
I am analyzing some gases and hydrocarbons in transformer oil by headspace and GC. Now I'd like to calculate the LOD of the method, let say for H2. According to some documents, the LOD of H2 is about 3 ppm and I have only standard mixes which contain about 100 ppm hydrogen. How can i make standards of 3 ppm or so to analyze for LOD calculation or if there are any other methods to calculate it ? Or I have to purchase the standards of 3 ppm ?
Could anyone give me some ideas ?

Thanks for any input

TH.

How are you calibrating your measurement ? - presumably by running a series of dilutions from a known stock standard. Continue the dilutions until you cannot see a peak, or until the repetability in replicate analyses is worse than 33%.

Peter

[thohry]

Thanks all for giving me advice.
@ rb6banjo: I also think of your method, and it's quite simple. But I am not sure if the response is linear at concentrations close to LOD. May be I will try to get more diluted standard, say 10 ppm and try.
@ AICMM : I am using 1 mL loop and may have 100 uL loop. But I am not sure if we can get smaller size GC loops, say 10 or 20 uL.
@ Peter: I use only one point calibration. Dilute a gas standard is not as easy as liquid one.

Thanks again and I will try some how to dilute the standards I have.

TH.

[Peter Apps]

Thanks all for giving me advice.

@ Peter: I use only one point calibration. Dilute a gas standard is not as easy as liquid one.

Thanks again and I will try some how to dilute the standards I have.

TH.

How do you calibrate from a gas standard to content of gas in transformer oil when you are doing headspace of transformer oil ?

Peter

[rb6banjo]

Is your quantitative analysis based on that 100 ppm standard? If yes, then all of your results assume that your detector is linear at all points around your standard. If you're not sure it's linear at low concentrations, then basing your results on a 1 point calibration is pretty risky business.

[aidnai]

It is possible with mass flow controllers to dilute and mix gases fairly accurately. You would ideally buy a standard that is at the high end of your desired curve as well as a zero gas that is the same as the balance gas of the standard. Then you use two MFCs to control the flow of both gases. You can create a 10 point calibration curve easily this way, if you wish. The upfront cost is moderate, with the MFCs being probably $1k a pop, plus a system to communicate with and control them. But it can save you buying a lot of standards.

Accuracy is always percent of full scale flow, so choose your flow ranges carefully. Unless you pay extra you are probably looking at 2% full scale, which does add up with two MFCs particularly at the lower flow rates.

[thohry]

It is possible with mass flow controllers to dilute and mix gases fairly accurately. You would ideally buy a standard that is at the high end of your desired curve as well as a zero gas that is the same as the balance gas of the standard. Then you use two MFCs to control the flow of both gases. You can create a 10 point calibration curve easily this way, if you wish. The upfront cost is moderate, with the MFCs being probably $1k a pop, plus a system to communicate with and control them. But it can save you buying a lot of standards.

Accuracy is always percent of full scale flow, so choose your flow ranges carefully. Unless you pay extra you are probably looking at 2% full scale, which does add up with two MFCs particularly at the lower flow rates.

You mean a gas blender ?

Is your quantitative analysis based on that 100 ppm standard? If yes, then all of your results assume that your detector is linear at all points around your standard. If you're not sure it's linear at low concentrations, then basing your results on a 1 point calibration is pretty risky business.

The results normally is closer to the std than to the LOD and we assume it’s in the linear range.

How do you calibrate from a gas standard to content of gas in transformer oil when you are doing headspace of transformer oil ?

We just fill the emty vial with standard gas and analyze. I think the procedure have some error but acceptable.

[Peter Apps]

"We just fill the emty vial with standard gas and analyze. I think the procedure have some error but acceptable."

I see two potential problems; first the repeatability and accuracy of filling a vial with a standard gas mixture is critically dependent on the method you use and how good the operator's technique is, second your calibration does not take into account the partition of hydrogen between oil and headspace.

Peter