Victory or Death!
Lately, things at work have not been going particularly well. My main project (since my arrival) has been to build a set of gradient coils which are the basis of a magnetic resonance microscope. My problem, of late, has been that for reasons unknown to me, my gradient coils have developed a nasty habit of killing themselves in the middle of a measurement. What does a dead gradient coil look like? Here's an early kill of mine:
Fig. 1: A destroyed coil. The red lines are wires which carry very large currents (up to 100 A) but for only very short times (no longer than one millisecond). The Aluminum frame transports the heat away from the wires and is connected to a water cooling system.
Notice the dark charring, the melted plastic, and the overall ugliness of the situation. When something like this happens, I have to scrape out all the bad stuff, replace everything and start over. It is a time-consuming and boring process and one at which I have become far too skilled. With each rebuild, I looked at the remains and try to diagnose the problem. For the longest time, I have assumed that the problem was of a physical nature -- the wires are under extreme forces inside the magnet so maybe they are being pulled apart. If you've ever had an MRI yourself, you'll recognize the loud clicking and popping noise that the machine makes -- this is the sound of the gradient coils warping, expanding and contracting over and over. Thus, I have spent a lot of time attempting to increase the physical robustness, adding epoxy, etc. and with each rebuild I think that maybe this time I have made it so physically secure that it will finally survive. Each and every time, I have been proven wrong.
Finally, I reached a point where there was no more improving to be done. I made a complete redesign, which will take the machine shop a few weeks to complete, and in the mean time started looking for more exotic causes of the problem. Based on a suggestion from Giovanni, I decided to look to see if the DAQ (the thing that controls the current pulses that pass through the gradient coils) was behaving exactly as it should. Given that our DAQ is very expensive and produced by the leading DAQ manufacturer in the world (National Instruments), there really isn't any reason to suspect it. Of all the parts of the system, it really should be the most accurate and the least troublesome. Thus, I only started testing it when I realized that I could test a couple other aspects of the system at the same time (things I thought were much more worrisome).
Testing involved having the DAQ send thousands and thousands and thousands of pulses, then monitoring the output and looking for bad pulses. I had the tests run overnight so that they wouldn't get in the way of the other people in the lab, and honestly, I figured that after sending the first 100,000 pulses and not seeing anything bad, what was the point? Really, I was just killing time until the new system was ready.
Anyway, this afternoon at 2:45 PM, on about the millionth pulse, I found my culprit. Without getting into the details, basically the DAQ decides to go crazy on random occasions and the result of this craziness is that it tries to send really large currents for way too long times. This is what causes my coils to explode and die. I honestly would never have believed this if I didn't have the full and complete proof right before my eyes.
So, what does this mean? Well, first, it means that I finally know what is killing all of my coils! Second, I know exactly what I have to do to solve the problem (I think ...), and it is almost ridiculously easy. Third, this means that I get to call up National Instruments and tell them what horrible, awful people they are what heaping piles of shit their hardware is. For those of you who know what jerks the National Instrument people are, you'll recognize that this is by far the best reward. And, finally, it means that I can stop wasting my time rewiring gradient coils and actually get back to doing real science.
It's time to get ill.
Fig. 1: A destroyed coil. The red lines are wires which carry very large currents (up to 100 A) but for only very short times (no longer than one millisecond). The Aluminum frame transports the heat away from the wires and is connected to a water cooling system.Notice the dark charring, the melted plastic, and the overall ugliness of the situation. When something like this happens, I have to scrape out all the bad stuff, replace everything and start over. It is a time-consuming and boring process and one at which I have become far too skilled. With each rebuild, I looked at the remains and try to diagnose the problem. For the longest time, I have assumed that the problem was of a physical nature -- the wires are under extreme forces inside the magnet so maybe they are being pulled apart. If you've ever had an MRI yourself, you'll recognize the loud clicking and popping noise that the machine makes -- this is the sound of the gradient coils warping, expanding and contracting over and over. Thus, I have spent a lot of time attempting to increase the physical robustness, adding epoxy, etc. and with each rebuild I think that maybe this time I have made it so physically secure that it will finally survive. Each and every time, I have been proven wrong.
Finally, I reached a point where there was no more improving to be done. I made a complete redesign, which will take the machine shop a few weeks to complete, and in the mean time started looking for more exotic causes of the problem. Based on a suggestion from Giovanni, I decided to look to see if the DAQ (the thing that controls the current pulses that pass through the gradient coils) was behaving exactly as it should. Given that our DAQ is very expensive and produced by the leading DAQ manufacturer in the world (National Instruments), there really isn't any reason to suspect it. Of all the parts of the system, it really should be the most accurate and the least troublesome. Thus, I only started testing it when I realized that I could test a couple other aspects of the system at the same time (things I thought were much more worrisome).
Testing involved having the DAQ send thousands and thousands and thousands of pulses, then monitoring the output and looking for bad pulses. I had the tests run overnight so that they wouldn't get in the way of the other people in the lab, and honestly, I figured that after sending the first 100,000 pulses and not seeing anything bad, what was the point? Really, I was just killing time until the new system was ready.
Anyway, this afternoon at 2:45 PM, on about the millionth pulse, I found my culprit. Without getting into the details, basically the DAQ decides to go crazy on random occasions and the result of this craziness is that it tries to send really large currents for way too long times. This is what causes my coils to explode and die. I honestly would never have believed this if I didn't have the full and complete proof right before my eyes.
So, what does this mean? Well, first, it means that I finally know what is killing all of my coils! Second, I know exactly what I have to do to solve the problem (I think ...), and it is almost ridiculously easy. Third, this means that I get to call up National Instruments and tell them what horrible, awful people they are what heaping piles of shit their hardware is. For those of you who know what jerks the National Instrument people are, you'll recognize that this is by far the best reward. And, finally, it means that I can stop wasting my time rewiring gradient coils and actually get back to doing real science.
It's time to get ill.
Comments
Post a Comment