 |
Reply
| | | From: Jaom (Original Message) | Sent: 3/29/2008 3:02 PM |
Why is it enough with one standardsolution (point) if the sample and the standard has approximetly the same concentration (the conc. of sample is known, like in control analysis in medicines)...but one need a calibration curve if the sample has a concentration that lies in a wide interval? |
|
 First
Previous
2-4 of 4
Next
Last
|
|
Reply
| | From:  ·Steve· | Sent: 3/29/2008 9:59 PM |
I think one standard is sufficient if, as you say, the samples being measured are close to the control value. It is important to know that the instrument response is consistent if it is an instrumental analysis such as colorimetry, and to know that the instrument response varies predictably with small variations from the control value. This would be determined using multiple standards over a wider range. But even for a one point calibration, several identical controls should be measured to check for precision (reproducibility) of the measurements or experimental method.
Often, sample values have a wider range, so, again if it is an instrumental analysis, it is important to know what the instrument response is over a range that covers all of the sample values. In part of the range, the instrument response may be linear (e.g., obey Beer's law for colorimetry measurements), but at other sample concentrations, the instrument response may become nonlinear. Using multiple samples over a wide range, the instrument response is measured and a best fit line is calculated in the calibration curve.
Makes sense. If you were using a bathroom scale to weigh objects, and the objects you were weighing were about 50 pounds give or take only a little, you could check the balance response with a real 50 lb weight and see if it is reading a little high or low and correct the readings accordingly. But the error would likely be considerably different than, say, plus or minus 1 lb if you were weighing a 300 lb object and a 5 lb object, unless the scale response is remarkably consistent over such a wide range. Most instruments are not so consistent. Instead of plus or minus one pound error, with large weights the error may be plus or minus 5 pounds. So what you would do is, weigh multiple objects of known weight over the range from about 1 lb to 350 lb and make a graph of the scale reading vs. the true weight. After you weigh an object, you would find the scale's value on the Y-axis of your graph and read the corresponding X-value which would be the true weight.
Steve
|
|
Reply
| | | From: Jaom | Sent: 3/30/2008 2:56 PM |
Im doin a lab calculating conc. acetylsalicylic acid with Uv/Vis Abs. spectrophotometer in headache pills. So using one standard solution will yield a "perfectly linear relationship", ...and give higher precision, as long as the concentrations are similar, obey Beers law? Conc. do not shift considerably. (Using a curve you need to find the best fit line, and this is hightens the precision for samples in a wider interval.) This thing has always been some kind of an achilles heel to me  |
|
Reply
| | | From: ·Steve· | Sent: 3/30/2008 8:38 PM |
>> So using one standard solution will yield a "perfectly linear relationship", ... << Well, no, with only one standard you would have a calibration graph with only one point! You would have no idea if the relationship is linear or not. So beforehand, in a separate calibration procedure, the instrument response should be checked over a range of concentrations. After that, if you know that your sample concentrations will fall within the linear portion of the graph, you could do a "quick" absorbance-concentration calibration with only one standard and use A = mC, C = A/m, where c is the concentration and m is a constant, to calculate the concentrations of the unknown samples from their absorbance readings. (In a Beer's law plot, m would be the slope of the line if linear.) So, for any careful work of this nature (and analytical work is inherently "careful"), you should always check the instrument response over a range of known concentrations. It would be risky to blindly use only one known concentration for the calibration if you had not already verified that this was valid (within the range of your sample concentrations) by earlier testing the instrument with a range of concentrations. Let's say you use only one standard for the calibration, after verifying that a one-point calibration will be sufficient for your particular samples, but you unknowingly make an error in measuring the amount of the standard. Because of that, all of the sample determinations would be in error. To avoid this, you should repeat the calibration to see if you are getting consistent results. If you have to do repeated calibrations with one standard, why not instead calibrate using a range of standards? You are doing the same amount of work but getting more information (a complete Beer's law graph) for your effort. If you are using only one of each standard, then some of the points may be a little off of the best fit line in the graph. But by using a best fit line, you are averaging in a sense too.  If you want to really do it right, take multiple measurements of standards at different concentrations, average each absorbance at a given concentration, and then make your A vs. C plot.  One example I have seen using a one-point reference is in chromatography. A single known amount of a reference compound is loaded with the sample and the amounts of one or more compounds in the sample are determined by comparing their peak areas in the chromatogram with the peak area of the known substance. Obviously, that reference compound must be measured very accurately! Any error in measuring the reference translates into an error in the calculated amounts of the other substances. And, in the manner discussed above, the validity of basing the concentration calculations on only one reference must be verified beforehand. So, in your experiment determining the amount of acetylsalicylic acid in different analgesics, you should make a normal, multi-concentration Beer's law plot! That is overall the most reliable method. You would be "taking a chance" if, knowing nothing about the instrument, you use only one standard for the calibration.  Steve |
|
|
First
Previous
2-4 of 4
Next
Last
|
|
Free Forum Hosting
Important Announcement
Return to General