From mcvittie at cis.Stanford.EDU Tue Jul 13 17:10:56 2010 From: mcvittie at cis.Stanford.EDU (Jim McVittie) Date: Tue, 13 Jul 2010 17:10:56 -0700 (PDT) Subject: STS2 -- Found Matching Network Cooling Problem In-Reply-To: Message-ID: Hi, 1. Coil Phase/Magnitude detector tuning Because of all the matching problems we have had with STS2, I took out the phase/magnitude detector and retuned it using a 50 ohm dummy load, the rf generator and an oscilloscope. The phase/mag detector is the sensor used by the matching network controller to auto match the plasma impedance to 50 ohms. I found it was quite a bit off. On putting the detector back into the system, I find that we can get significantly lower reflected power for the switched processes. However, the unswitched processes (O2 clean) now have a higher reflected power. I am also seeing heating effects in that the matching getting worst as I use the system. It seems that the phase/mag detector is warming with use. 2. Matching Network Box Heating Problem I think I have found the root problem for our matching problems. In checking why the detector is showing signs of heating, I found that the cooling fans on the box containing the coil matching network and the phase/mag detector were mount wrong. Both fans were blowing inward with no clear pathways for the air to get out. This explains why it has taken so long for the rf system to recover after heating from rf power drop outs from poor matching. I have reversed one of the fan. I am now wondering rf detector diodes in the detector may have been over heated and are part of our problem. Lets see how the improved cooling is working. If we are still matching seeing problems, I will replace the diodes. Jim -- -------------------------------------------------------------- James (Jim) P. McVittie, Ph.D. Sr. Research Scientist Paul G. Allen Building Electrical Engineering Stanford Nanofabrication Facility jmcvittie at stanford.edu Stanford University Office: (650) 725-3640 Rm. 336X, 330 Serra Mall Lab: (650) 721-6834 Stanford, CA 94305-4075 Fax: (650) 723-4659 From mcvittie at cis.Stanford.EDU Tue Jul 13 19:18:06 2010 From: mcvittie at cis.Stanford.EDU (Jim McVittie) Date: Tue, 13 Jul 2010 19:18:06 -0700 (PDT) Subject: STS2 -- Found Matching Network Cooling Problem In-Reply-To: Message-ID: Hi, I was wrong about the cooling. Changing the fan made no difference and there is an air pathway around the rf feed to the coil out of the matching box. Jim On Tue, 13 Jul 2010, Jim McVittie wrote: > Hi, > > 1. Coil Phase/Magnitude detector tuning > > Because of all the matching problems we have had with STS2, I took out the > phase/magnitude detector and retuned it using a 50 ohm dummy load, the rf > generator and an oscilloscope. The phase/mag detector is the sensor used > by the matching network controller to auto match the plasma impedance to > 50 ohms. I found it was quite a bit off. On putting the detector back > into the system, I find that we can get significantly lower reflected > power for the switched processes. However, the unswitched processes (O2 > clean) now have a higher reflected power. I am also seeing heating > effects in that the matching getting worst as I use the system. It seems > that the phase/mag detector is warming with use. > > 2. Matching Network Box Heating Problem > > I think I have found the root problem for our matching problems. In > checking why the detector is showing signs of heating, I found that the > cooling fans on the box containing the coil matching network and the > phase/mag detector were mount wrong. Both fans were blowing inward with no > clear pathways for the air to get out. This explains why it has taken so > long for the rf system to recover after heating from rf power drop outs > from poor matching. I have reversed one of the fan. I am now wondering rf > detector diodes in the detector may have been over heated and are part of > our problem. Lets see how the improved cooling is working. If we are still > matching seeing problems, I will replace the diodes. > > Jim > > -- -------------------------------------------------------------- James (Jim) P. McVittie, Ph.D. Sr. Research Scientist Paul G. Allen Building Electrical Engineering Stanford Nanofabrication Facility jmcvittie at stanford.edu Stanford University Office: (650) 725-3640 Rm. 336X, 330 Serra Mall Lab: (650) 721-6834 Stanford, CA 94305-4075 Fax: (650) 723-4659 From mcvittie at cis.Stanford.EDU Thu Jul 29 21:37:47 2010 From: mcvittie at cis.Stanford.EDU (Jim McVittie) Date: Thu, 29 Jul 2010 21:37:47 -0700 (PDT) Subject: STS2 Update -- Matching Network and Pulse Generator Message-ID: STS2 Users, I was hoping to be able to tell you that I had found a specific problem with the STS2 matching network, which would solve all the problems we have had with it. But I was not able to find such a cure. My conclusion is that the matching network is of an marginal design. Until we get a better long term solution, we have to be careful how we setup recipes and how we make changes to them. The tool is up and I am cuurently working with students to get their previous recipes back up and running. Here are the near term solutions for the matching problem. 1. If you want to use STS2, I need to modify your recipe so it work well with matching network. Please send me the name of the recipe you want me to look at. It can take me a hour to modify a recipe if it is new, so I need time to do these modifications. I usually need to run a recipe several times to check that it matching correctly. Please do not run unmodified recipes on the tool. You can tell if I have modified a recipe by looking at the version number on the 1st page of the recipe. The modified recipes have a "m1" version number. Note I have no problem taking users off qualified user list if they abuse the tool by not following instructions. 2. How are the reciped being modified? a. The new tolerance recipe (dated June 3) is added. These tighter tolerances cause the process to fault if such things as the reflected power goes too high. b. Addition of a zero power etch/dep step before the main etch/dep step. This step is needed to position the load and tune capacitors so they go in the correct direct during stablization step. The matching unit should not need this step but it has a bad hysteresis problem which makes this step necessary. c. Setting of specific pre-set load and tune valves on the matching recipe page. These values have to be tested under plasma conditions. d. Checking that the pulse mode box is checked on the low frequency generator page. Low frequency duty cycle setting --- if you are using one of the low frequency (LF) SOI recipes, you may or may not know that the LF (380 KHz) rf is pulse modulated at a frequency around 40 Hz and at a duty cycle between 33 and 20%. In particular, our system was suppose to have been setup some years ago by STS so the LF was on for 5 ms and off for 20ms ( a 20% duty cycle). In June STS told me that this pulse rate should be checked with a sensor on the rf line and not simply by looking at the pulse rate coming from the pulse generator, I found the pulse width coming from the pulse generator was indeed 5ms but the pulse width on the rf line was only 3 ms. The rf generator introduces about a 2 ms delay on turn-on but no delay on turn-off. This means that we have been operating at a duty cycle of only 12%. This difference (12% instead of 20%) appears to be important. I increased the duty cycle to 20% and it appears to be causing a significant decease in resist selectivity for one user who previously had a selectivity of 400 and now it is down to 300 and below, however, all the data is not in on this one. A year or so ago, a user developed a SOI recipe and found that he had to turn off the electromagnet because of a glass problems. Increasing the duly cycle to 20% would probably allow the electromagnet to be kept on for this recipe. There is a question about whether we should keep the duty cycle at 12% or increase it to the currently STS recommanded 20% to 25%. Alternatively, We could keep the pulse generator in the lab next to thed tool and allow users to change it for best process results. I also need to tell you about the failure of the pulse generator. Sometime between mid-June and mid-July the pulse generator failed. The failure was such that the duly cycle went to 100% (LF full on during the etch step). I have a temporary unit with an oscilloscope in place. It will be a few weeks we get replacement pulse generator. Thanks, Jim -- -------------------------------------------------------------- James (Jim) P. McVittie, Ph.D. Sr. Research Scientist Paul G. Allen Building Electrical Engineering Stanford Nanofabrication Facility jmcvittie at stanford.edu Stanford University Office: (650) 725-3640 Rm. 336X, 330 Serra Mall Lab: (650) 721-6834 Stanford, CA 94305-4075 Fax: (650) 723-4659