First, let me agree with the sentiment expressed that Howard has done everyone a real service in his clear and detailed analysis of digital vs. analog electronics. Thank you. Second, let me state my disclaimers. Cytomation has bought me lunch (more than once) and given me a travel mug. BD has bought me lunch (more than once) and given me a Swiss Army pocket knife. Unfortunately neither currently has made an offer I can't refuse, which makes life more difficult. It would be much easier to be owned by someone so you wouldn't have to think about things and come to your own independent decision. That being said, let me add a few comments to Howard's analysis. As data, Cytomation's mailing is not useful. There is no vertical or horizontal scaling, and the size of the left panel (DiVa) is smaller than that of the right (MoFlo), adding visual uncertainty to what is being demonstrated. Moreover, there is little said about the conditions under which the measurements are taken. This would include laser power, emission filters, laser focusing system, etc. Also, one could surmise that the DiVa data was taken at a meeting (FASEB), while the MoFlo data was taken either at the company or someone's lab, where there would have been much more time to align and optimize the instrument setup, etc. Lastly, if you look at the striping of the low peak on the DiVa, you see that they are close at the upper end, further apart in the middle, and close again at the lower end. From my experience with data modeling, striping increases in severity and doesn't usually have a local maximum as this seems to show. So I am unsure of what is being shown. Let's give everyone the benefit of the doubt and assume the conditions under which the data were taken were equivalent. Then what can be concluded from it? First notice that for the DiVa data all 8 peaks are on scale, which is not true of the MoFlo data. If we assume that there has been no manipulation of the data, then it says to me the scaling of the DiVa data is different than that of the MoFlo. In that case, then using the visual separation between peak 0 and peak 1 as a measure of the sensitivity, as Howard mentioned, doesn't work. In the 8 peak bead set, each peak has an equivalent number of beads. As one increases in intensity on the log scale, for the same number of beads, the peaks will be narrower and greater in height. That is, the peak of lower intensity will be broader and not as high. The height ratios and widths shown on the MoFlo data are typical of what one sees (although it appears peaks 7 or 8 my be clipped) for this beads set on FITC or PE collection channel in properly operating instruments. For the DiVa data peak 1 (the picket fence) is unusually high and broad, seeming to indicate it has many more events in it that the other peaks. Perhaps this could be an artifact of the striping, but I am hesitant to make any conclusions here. So if we ignore the lowest peak on both histograms, the two data sets visually appear similar to me, making it difficult to draw any conclusions at all about sensitivity of the optical system or processing in the electronics. Without more information on scaling, etc. it is difficult to say what the lower peak on the DiVa data set means. At the extremes, it could be either serious electronic design problems at BD, or creative artistry at Cytomation. Until more is known and demonstrated publicly about the DiVa, we will all be left speculating. -- Marty Gladstone Institutes Flow Core mbigos@gladstone.ucsf.edu 695-3832
This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:57:40 EST