I have designed a USB device based on the TMS320VC5506. I am using the TPS70245PWP to provide the 1.2 V core supply and 3.3 V I/O supply. The enables are set up to sequence the supplies so that 1.2 V is delivered immediately (EN2 tied to ground) and 3.3 V is delivered after 1.2 V has become stable (EN1 pulled low by a transistor controlled by VOUT2). In order to keep the device below the USB limit of 100 mA during startup, a third 5 V boost regulator is controlled via the DSP GPIO7 pin, with a 1 MΩ pull-down on /SHDN to ensure that the analog supply is off initially. The DSP firmware does not set GPIO7 high until the USB host has enumerated and configured the device. This all works fine.
Here's where the discussion starts. The analog supply is a MAX1595 step-up charge pump that can deliver 5 V at 150 mA from the raw USB bus power (which can drop as low as 4.01 V when a device is fed from an unpowered hub). Thus, the MAX1595 effectively has a 5 V supply. It's /SHDN input pin considers anything above 1.6 V to be logic high, so it's perfectly happy with the 3.3 V GPIO7 output. Leakage current when in shutdown mode is 0.1 µA, so that 1 MΩ pull-down only rises to 0.1 V, comfortably below the 0.6 V threshold for logic low on the MAX1595. At this point I have one concern and an actual problem.
My general concern is whether there is anything wrong with directly attaching GPIO7 to a device running on 5 V power. Maxim documents a leakage current when the input is low, as initially guaranteed by the pull-down, but says nothing about current flow when the input is a logic high. The data sheet only mentions that 1.6 V is the minimum threshold for logic high. My assumption is that no current will flow out of the MAX1595 /SHDN input, but I do not have a way to measure this at the moment. Any comments?
My problem is one that I should be able to fix, but I suppose it may be related to other problems I am experiencing. The MAX1595 runs hot, which I assume is because my analog circuits draw more than its rated 150 mA. It's hard to measure individual currents within my circuit, but estimations based on USB bus current delta when the analog supply is turned on and off indicate that I probably need to double that current rating to 300 mA. In addition to the heat, I also have problems whenever the DSP turns off the analog supply, with symptoms including error messages from the JTAG interface in CCSv4 complaining that EMU0/1 are not as expected. I assume here that there is probably a backwash of current when the MAX1595 is shut down, and that could be raising the GND voltage on my board. But there also seems the possibility that GPIO7 is having some trouble at this transition, although I don't know how to test that.
I will probably contact Maxim about this as well, to see what they have to say. However, if anyone here has comments on the above then I would appreciate them.
P.S. I marked this as a discussion rather than a question because I suspect there may not be a definitive answer forthcoming without help from Maxim.