Date: Thu, 03 Aug 2006 16:49:24 -0400 From: Kevan Hashemi To: Christoph Amelung CC: Michael Bradshaw Subject: Driver Power Supplies Dear Christoph, Mike was able to re-produce your power cycle problem this morning. We now believe we understand what is going on. Here we describe the A2037 power supplies, and how they turn on: http://www.bndhep.net/Electronics/A2037/M2037.html#Power%20Supplies We produced the 15V switching on and off problem by trying to heat up a piece of wire with an A2053 RTD Head. As soon as we apply +15V to the wire with the A2053's heater circuit, the +-15V supply turns off. While it's off, the +5V supply is still on, and the +-15V lines drift to roughly 1V. The internal reset lines of all devices will drift to 1V. The 3.3V logic supply inside the A2053 and all other radiation-resistant devices comes from a drop-down regulator powered by 5V and biased by 15V. When the 15V drops down to 1V, the required biase disappears, and the 3.3V logic supply drops to around 1.5V. Despite dropping to 1.5V, the logic supply is still great enough to retain register settings, including the settings that caused the A2053 to heat a piece of wire. The power-up reset circuit, made of a resistor and capacitor, puts 1.5V on the reset line, so none of the registers get reset. Only if the logic supply drops to zero, and then comes up suddenly, will the reset line be able to do its work. Half a second later, the DC-DC converter on the driver tries again to turn on the +-15V supplies. It succeeds at first, but then the A2053's internal logic supply jumps up to 3.3V as its regulator bias returns, the heater turns on again, and the +15V supply gets shorted once again, and we begin the whole cycle anew. The cycle cannot be stopped with a sleepall, because none of the devices will respond to LWDAQ commands when their LVDS receiver chips, which run off the 3.3V supply alsoo, are getting only 1.5V power. The only way to stop the cycle is to turn off the head power, wait, and then turn it on again. Of course, in a heavily-loaded system, it's possible for a turn-on to slow, in which case the device reset lines don't work all the time, and we can get random registers being set on power-up. If there are enough of these set, switching on enough +15V power flow, the +15V supplies might be dragged down, and we enter the cycle. In the heat, the DC-DC converters will get hotter, and their switching may be slower, in which case the reset lines will be even less effective. If we had proper reset monitors on the devices, we would be okay. But I had to get rid of them for radiation resistance up to 30 krad. So, what to do about it? I think you are doing the best you can, but I remain surprised taht "off 5000 on" does not solve the problem on its own, at the end of each Acquisifier cycle. You say you need "off 2000 on 2000 off 2000 on". Can that be right? Yours, Kevan -- Kevan Hashemi, Electrical Engineer Physics Department, Brandeis University http://www.bndhep.net/