agilent detectors

[scio2]

Hi, can anyone tell me if Agilent VWD is a more sensitive detector than Agilent MWD.

I am having problems with transferring the LOQ from one instrument to the other.(better response on VWD)

Is there a factor of difference between the two detector types.

thanks

Scio

[unmgvar]

yes there is a difference.

the MWD is a PDA optics.
the VWD is a monochromator.

the sensitivity of the MWD is way smaller then that of a VWD because of that, the PDA optic type creates a lot more noise hence your problem of method transfer.

try reducing the "speed" at which you collect data, it might help.
increase the bandwidth used for collecting the data.
in last resort try using the ref. wavelenght.

unfortunately if your LOQ was very close to the lower limit in the VWD, you will probably not succeed

[HPLCCONSULT]

Yes, the two types of optical benches and detectors are different. You are comparing two different systems.

Some things to keep in mind:
(1) Which flow cells are you using in both detectors ? Compare volume and path length as these will make a big difference in signal. *Path length is key in this case.
(2) What is the bandwidth at the specific wavelength you are gathering ? The ideal bandwidth is determined by the actual optical block calibration, but it can be changed in the software to increase specificity or increase signal quantity. Set the bandwidths to be the same (just like the flow cell dimensions and path length) if running the same methods. This is not always possible, but get as close as you can. Signal wavelength, bandwidth, flow cell and pathlength should be the same if the method is supposed to "run" the same from the detector end. Ideally, the same detector is used. In your case, you will just have to get as close as you can.
(3) Do Not change (increase) the bandwidth unless you know (through the use of a scanning DAD) that nothing co-elutes near your peak(s) of interest and the regions "incorporated" will not add extra noise. If the area is clear around your peak (e.g. ~200 to 400nm) and you want to increase the "apparent" signal of the MWD you can increase the bandwidth by a larger margin. If your current bandwidth is 10nm (which means it is 5nm to the right and 5nm to the left of signal), you could increase it to 20, 30 or even 50nm. Be careful though, as you may also be adding in some noise, depending on where your actual signal apex is !
(4) "Speed" has little to do with sensitivity. An ideal collection rate can (and should) always be determined mathematically for the method. Quick Method: Use the narrowest "typical" real peak in the run as an example. Calculate the peak width in time units. Divide by 30 (20 is the standard, but we use 30 to play it safe) and find a collection speed that approximates that rate. In most case for runs at 1.00 ml/min you will find that a rate of 0.2 Hz (5/sec) is fine. Use a data collection rate that is appropriate for your sample. It could be much slower or faster than this example. Integration requires at least twenty good points to draw a line through and still have a useful statistical result. Note: Too few samples over time can result in inaccurate integration while extreme sampling (very high rates) will add nothing useful to the quality of your data and in the worst case may induce some noise and waste computer memory. Everyone's samples and methods are different so calculate it based on your actual run.
(5) Never use the reference wavelength (Always turn it OFF on both detectors initially!) without some serious thought as to how you are manipulating the signal. Few people have any understanding of this feature and by default you should seriously consider turning it OFF on all detectors that offer it (at least initially, until you fully understand how and when to use it). The 'Reference' feature operates by SUBTRACTING RAW DATA from your original signal and outputting only the result of the subtraction. Your original data is lost forever. If you want to subtract drift from your system simply set-up the MWD/DAD to gather and store a second signal during the same run. After the run is done, use the built in software spectral tools to subtract it from the original and see the result. This preserves the original data. *Reference can be useful if (a) you have used a scanning DAD FIRST to insure that the area where you might set the reference wavelength (and bandwidth) is clear of noise and peaks. (b) your baseline predictably drifts during the run (which makes integration less accurate). A well chosen reference, for certain methods, can reduce the drift over time resulting in a flatter baseline and better integration.

[danko]

Hi guys,

Scio2 is not using a diode array detector (DAD). MWD is a multiple wavelength detector, which is very, very similar to the single channel detector (Agilent calls it VWD or variable wavelength detector) but just offering 2 or 4 channels.
Otherwise I agree that there could be flow-cell dimensions, or bandwidth settings differences.
It should be noted too that enabling more than one channel on the MWD, results in sensitivity decrease.

Best Regards

[unmgvar]

this is from the agilent web site:

http://www.chem.agilent.com/en-us/produ ... fault.aspx

Diode array design enables simultaneous acquisition of up to 8 compound-specific wavelengths for increased sensitivity and selectivity.

the increased sensitivity part has we can see is not very precise, maybe if we read between the lines somewhere then it will look as if they are talking compare to their PDA detector, but defenetly not compare to any VWD.

however from looking at the site i saw that they have another feature that might help reduce noise depending on the application: slit control. it of course will have it's price as well

[Rob]

Based on a previous life as an application chemist for Agilent and now working as an independent consultant and provider of training in LC, LC/MS and CE, can I just correct/expand some of the comments I see here?

The Agilent MWD is based on the DAD, i.e. it is the same optical bench, same flowcells (ensure you have 10mm path for comparison with VWD), same spectrograph, same sensitivity - the difference is that the MWD does not have the spectral processing component. The VWD is a monochromator based instrument. Having more than one signal active in the MWD will NOT reduce the sensitivity since the detector element is actually a diode-array. Some other manufacturers use monochromator in their MWD and in that case the signal would be reduced because it is effectively time-sharing the detector element.

Years ago DAD (I've been using DAD since 1982) in general were less sensitive than VWD (by up to a factor of 4-5) but they got better and better and over the last few years the recent models have had similar sensitivity. So one parameter to be considered is how old are the instruments that you are comparing (and on the subject of age, are the lamps giving good output).

Although you don't usually think of bandwidth with monochromator instruments, they do have an intrinsic bandwidth and that is typically in the region 6-12nm depending on the manufacturer and model so setting a DAD or MWD bandwidth in that range is a good first step in the method transfer.

Assuming that the wavelength is set on the max absorbance you would find that increasing the bandwidth will result in a reduced signal (you are averaging in lower absorbance values). At the same time you are reducing noise to some extent as you are taking more values (error in measurement reduced by square root of number of determinations). There will be an optimum bw for best sensitivity i.e. Signal/Noise and this can be predicted for a given spectrum or found experimentally. However, unless you are optimising for trace level detection I would just stick with the narrow bw as described above.

The default slit width of 4nm can be opened up to 8 or 16 to let more light to the detector and hence reduce noise but you will need to determine if this gives the improvement you are looking for. Depends on your other parameters such as wv, bw, mobile phase, analyte concentration.

To set the correct data collection rate set the Peakwidth setting to be the same as your actual peak width and the system will take data at the appropriate rate. Too fast would increase noise, too slow would 'smooth' your peaks. Typically for LC on 4.6mm id columns you will have 0.1, 0.05 or 0.03min set - in any case make it the same as the VWD for direct transfer.

In addition to checking lamp energy, you might also want to check that your flowcell is clean.

Finally, having said that MWD/DAD and VWD were now equivalent in sensitivity it is worth noting that Agilent launched a new version (the "E" version) of the 1200 VWD in 2008 which moved the sensitivity goalposts again and leapfrogged ahead of the DAD/MWD by a factor of 3-5. It's a very sensitive detector and if that is the VWD you are using then you will have to wait for the DAD/MWD to catch up again!.

[HW Mueller]

Rob, some things you cleared up, but I am having trouble seeing a difference between bandwidth and slit width. In my understanding of spectral photometers you control bandwidth with the slit.

[Rob]

Hans, you are right but it is not the full story on a DAD (Agilent 1100/1200). There are two components to the bandwidth on a DAD signal: 1) the optical bandwidth due to the optical bench and in particular the size of the entrance slit allowing light into the spectrograph; and 2) the bandwidth due to averaging the signal over a number of adjacent diodes.
Imagine if the light entering the spectrograph was a single wavelength and the diodes had a resolution of 1 nm. The light would end up shining on a few adjacent diodes. If the entrance slit was narrow enough the "image of the slit" would shine on just one diode and you would have a spectral resolution of 1nm because you have an optical bandwidth of 1 nm. Now put all the wavelengths in and each diode nominally sees a single wavelength incrementing from one side of the array to the other. For instance you could choose to monitor the 250nm diode and you would have a signal bandwidth (resolution) of 1nm.
Now open the entrance slit wider so the single wavelength light shines on 4 diodes and you have an optical bandwidth (or resolution) of 4nm. If you monitored the same diode as before it would see a bandwidth of light of 4nm centred on 250nm and of course every diode sees 4nm bw. This is the default setting on the Agilent DAD - the slitwidth is 4nm.
Now let's turn to the diode-array itself and set the signal bandwidth. We can take the signal from any number of adjacent diodes (the "diodes" bandwidth) and sum them to get our output signal. With an optical bw of 4nm it makes sense to take at least 4 adjacent diodes, so we might set 250nm with bw 4nm and we would be detecting from 248-252nm. this would be narrow band detection. There could be advantages to making it wider so we might take 10 diodes which would be 250nm bw 10nm and we would detect from 245 to 255nm. We might want to detect in one signal all wavelengths in the range 220-280 nm, that would be signal 250nm bw 60nm.
To summarise: the "slitwidth" gives you the resolution of the spectra that you will obtain and the minimum possible bandwidth of the signals; the bandwidth defines the range of wavelengths encompassed by the output signal.
Opening the slitwidth wider will allow more light in and thus reduce noise but you will lose some spectral resolution. If you need high spectral resolution (rarely needed for UV spectra in the liquid phase) then close the slit down to 1nm.
Hope that helps.

[lmh]

Ah, right, you mean the slit-width is sort-of how well focussed it is? So, for example, the 450nm diode is just a diode that detects any light, and with a slit-width of 4nm, 450nm light will actually fall on the 450nm diode and on a couple of diodes either side. If I also set the spectral bandwidth in diode terms to 4nm, the instrument will average the data from the 450nm diode and its friends either side.
If both operate with sharp cut-offs, then 4nm slit-width + 4nm diode bandwidth = 8nm actual bandwidth compared to an optical slit/PM tube instrument??

[Rob]

Yes, that's about the size of it. Sure, an optical engineer might pick me up on a few approximations but this explanation works for the analyst. (Note, by the way, the 1nm per diode is approximate. In reality it's a little less (i.e. range 190-950nm over ca. 1000 diodes is ca. 0.8nm per diode) but 1nm is a useful working approximation.

[unmgvar]

Rob;
Still, scio2 problem is that he cannot achieve LOQ requirements.
so it means that is not getting enough peak signal compare to the amount of noise, which is not a surprise since he moved from a VWD to a MWD.

here are some agilent specs:
MWD: noise-± 0.8 x 10 -5 AU at 254 nm and at 750 nm; drift- 0.9 x 10 -3 AU/h at 254 nm
VWD: noise-± 0.15 x 10 -5 AU, at 230 nm drift-1 x 10-4AU/hr, at 230 nm

3 things important:
1. more noise and drift for optics of PDA
2. the difference is almost 2 times more noise without looking too much at the specs
3. If we do look at the specs we see that they are done at different wavelenght therefore different procedure, 254 nm is a less absorbing WL so the results for the MWD done on compared parameters to that of a VWD should be (and they are) far worst then what we see here.

Rob, based on your knowledge of Agilents optics what would you recommend him to do in order to try and reduce the amount of noise he is getting in order to transfer the method- or in other words get closely the same sensitivity?

[HW Mueller]

Rob, are you saying that if one sets the slit so that 4 diodes get light (4 nm), but sets the "bandwidth" on one nm then the light falling on three diodes is "not seen" (measured)? One would have the intensity (sensitivity) as if the slit width was set on one nm (low sensitivity) and resolution of 4 nm?

[Rob]

Unmgvar, as I said above it will depend on the relative ages and model numbers of the two detectors. The sensitivity specs you quote are for the latest model (G1314D standard version and the G1314E 160Hz fast version) which increased its sensitivity 3-5 times (depends on application) compared to the previous model which the MWD/DAD had caught up in sensitivity. If I remember correctly these new VWD G1314D&E versions hit the streets about October 2008. I don't have an old brochure but I think the VWD G1314C version noise spec was 0.75x10^-5AU.

Scio2, this is what I would suggest to get best sensitivity but this can only be a general guideline as I would need to see the spectrum of your analyte for finer tuning (I am assuming your wvl is between 230 and 300nm and you are using normal flow rates with 4 or 4.6mm id columns): Check the specs in your detectors' manuals to see what the relative sensitivities are likely to be. No point trying for the impossible. Make sure that you have UV lamp with good energy and acceptable noise. Make sure that the flowcell is clean and 10mm pathlength. Make sure that the visible lamp of the MWD is turned off and UV on. Set the analytical wvl same as VWD, set signal bandwidth to 16nm, set ref wvl to 360 with bw 60nm, set slitwidth to 8nm. Set peakwidth parameter to 0.1min or 0.05min depending on the width of peaks in your chromatogram. Further steps if that does not get you the desired result: open the slitwidth to 16nm, set signal bw to 32nm.

[Rob]

Rob, are you saying that if one sets the slit so that 4 diodes get light (4 nm), but sets the "bandwidth" on one nm then the light falling on three diodes is "not seen" (measured)? One would have the intensity (sensitivity) as if the slit width was set on one nm (low sensitivity) and resolution of 4 nm?

HW, hmmm, not quite. In any case, the detector will warn you if you try to set the bandwidth less than the slitwidth. Remember that all the diodes receive light because all wavelengths are present. I used a thought experiment in the previous post considering only single wavelength light when I talked about 4 diodes receiving light. Sorry for any confusion. Regardless of how many diodes you use for the signal bandwidth the effect of decreasing the slit from 4nm to 1nm would be to reduce the amount of light and so increase the noise - as sensitivity is S/N the sensitivity would reduce. The signal intensity measured is absorbance and this would not be affected by the intensity of light at a particular wvl(Beer-Lambert Law) as the same proportion would be absorbed. Independently of the slitwidth, increasing the diodes bandwidth does two things a) reduces noise because you have more readings to average; b) as you widen it, it will take into the average lower absorbance readings and start to reduce the signal.
This is quite difficult to explain without diagrams! In my training courses I cover how you optimise detection but using diagrams to explain makes it easier.

[HW Mueller]

Yes, all the diodes have light shining on them, but only the ones covered by the bandwidth setting are included in the intensity measurement? Your b) is a bit surprising; They average the signal of the diodes which are used, not add the signals? I think I see it, the I/Io term is doing the averaging automatically (if one uses 4 diodes then the intensity is I1+I2+I3+I4, but without sample the intensities are Io1+Io2+ . . . ).