statistics because there's a lot of activity there. Then, that's fine. The bottom segment, you could count it for days and you are not going to get that good accounting statistics, but you are safe because you don't really need it because there is very little activity in that part and when you fold in the total activity, that error is not very important. problem is, we don't know what segment it comes from. The Only DRI knows this, so we have to send the results to DRI and they look at their master log book to see where it comes from, and they fold in all of the errors and say, "Ah ha. This is fine," or “Ah ha, we need more counting on, say, a 10 middle sample somewhere," so they will get back to us and say (that's DRI, 11 it looks like ORI), DRI tells the lab to recount sample such and such, and 12 we'J] take that and recount it for a thousand minutes, 13 back to computer analysis, and follows through the flow diagram in the same 14 way. 15 Is that ok?" and they will say “yes" or "no." 16 are stored. 1? Then we send another letter to DRI saying: The cesium analysis is nice in 18 always have that dirt, no matter what. 19 if you like. 20 that it's and then it goes "This is our new value. If it's okay, the samples nondestructive. You'l] You can count it hundreds of times The plutonium is different. A word about turnaround time. If we have 100 sites to be analyzed and 21 each site has four segments, that's four hundred samples to be analyzed for 22 cesium. 23 tor days for the cesium and if you add in an extra 50% for QA, duplicate 24 samples, split samples, whatever DRI wants to send in, plus some time for .25 computer down-time, detector down-time, that gives you 150 detector days. We can do essentially four a day per detector. That's 100 detec- A 1,000-minute count takes up an 26 Now you also have 1,000-minute counts. 2? entire detector day effectively, and it's hard to know how many of these we 28 will need. An initial estimate might be 20% of the samples may need to be 160