I’m “manufacturing” data again. The difference between manufacturing data and collecting data is, of course, that manufactured data has been “coaxed” in some way to make it look like you want it to. It looks “just like real data” and often we do this for the benefit of the customer, but it rubs me the wrong way on a fundamental level.

Basically, there’s a defect in these parts. I know it’s there. I can see indirect evidence of it through one of our other scanning techniques. For whatever reason, the part construction makes the echoes look really weird in this area, so it doesn’t “look” like a defect. To anyone with just a brief lesson on how this technology works, it looks like a defect-free area. The echoes I mention are waves on an oscilloscope. Generally, waves with positive amplitude are a good area, and areas with negative amplitude echoes are defects. The machine takes these waves and turns them into pixels to create an image one pixel at a time as the machine scans over the parts. However, the polarity detection in these things is crap. so if you have a waveform of an area that has a big negative lobe with a moderate positive trailing lobe, it detects it as positive.

So, what I’ve been asked to do is play with the gates (the limits set on the oscilloscope for where the machine reads data from) and “shave off” the trailing positive lobe. This will make the area look like a defect. I have no question that it is a defect, but I still think it’s wrong to do this. Presenting modified data to laypeople is wrong… even if it is to help them better understand it.

So here I go again violating my scientific integrity for this company. I’ll say it again… yay for leaving.

In other news, Bob’s boss (who is actually a really cool guy) is getting lunch for us. Free food is ALWAYS acceptable.

Categories: Uncategorized Tags:
  1. February 2nd, 2005 at 19:03 | #1

    You should refuse, let them fire you, then sue them and reveal their hoax to all the customers. Or you could pull a Fight Club and get an incredible severance package.

  2. February 2nd, 2005 at 21:21 | #2

    Is there a way to configure it to integrate across the period of the gate, and providing the polarity of the integral instead of doing instantaneous polarity detection? That way, you’re not throwing away data (which is, indeed, scientifically offensive), but gives the desired result (presenting the polarity of the largest component of the echo signal).

    /Andrew

    • February 2nd, 2005 at 21:33 | #3

      unfortunately, the machine already stresses the computer enough. the hardware is a resource hog. it runs an acquiris card to give a true o-scope readout on the screen. It controls 3 RF slice channels at once too. the best we can do is save each individual waveform instead of saving just a pixel with a color/brightness representative of the polarity/amplitude. that way we can go back and “rescan” the part virtually since we have all the waveform info. we can do FFTs on the data to play in the frequency domain and stuff. but that’s it, and that’s all after scanning is done. very little can happen in real-time.

      feh.

      moot point now anyways. w00t.

  3. February 3rd, 2005 at 18:27 | #4

    It maybe the lack of sleep or me getting sick…but i am so confused and lost in this entry..blah…

  1. No trackbacks yet.