I've got a custom board derived from the Spectrum Digital HD1080P EVM (aka DM6467T EVM). I got 5 PCBs made.
Due to lack of parts, I have the first three stuffed with DM6467's (without the "T"; specifically TMS320DM6467CZUT7). Everything worked great. While connected to CCSV4 with nand_flash_writer.out loaded and about to program NAND, my *system* power supply was providing 240mA at 12V. It was 160mA before "connecting" CCSV4. Note this configuration included an off-board power supply providing 1.2V to the core, and on-board crystals at 27MHz and 27MHz for DEV and AUX clocks. (In detail, one board worked great. The other two had fab problems.) Naked chip surface temp takes a while to get to 110 degF, staying under 120 degF during my test just now, when I got bored and went ahead and sent this post.
Now, I have the last two boards stuffed with DM6467T's (with the "T", specifically TMS320DM6467TZUTD1). Thing stink. While connected to CCSV4, when I load nand_flash_writer.out, it starts up in the "Running" state rather than "Suspended", and my system current is 480mA and rising. Naked chip surface temp is 177 degF and rising. Even before "connecting" CCSV4, my current began high, at 330mA. Remember, this is supply current for all four voltage levels and also converter inefficiency. Note this configuration includes an off-board power supply providing 1.3V to the core, the same onboard 27MHz crystal for AUX, but a different source for DEV. In the recent past I removed the 33 MHz crystal from the DEV clock input, routing into it instead the CDCE949 clock generated 27Mhz signal per the EVM reference design jumpering capability. I did this just in case the XDS510USB can't handle 33 MHz.
So, is this increased DM6467T current and chip surface temperature a given for this chip, versus the much lower values for the DM6467? Or is there something wrong with my board?
See more details at http://e2e.ti.com/support/development_tools/code_composer_studio/f/81/p/113356/401915.aspx#401915
Here are my individual voltage supply currents, for the TZUTD1 vs CZUT7 boards:
TZUTD1 CZUT7
1.2V 412mA @ 1.3V 330mA @ 1.2V
3.3V 532mA 132mA
1.8V 136mA 76mA
5V 570mA 300mA (note the 5V supply drives some 5V logic and also feeds the other voltage supplies)
12V 280mA 170mA (MAIN. this was measured much earlier and cooler. The 280mA corresponds to 350mA mentioned above. The 170mA ~= 160mA)
It seems that ALL of my currents are higher on the TZUTD1 board. Percentages are 1.2/1.3V: 125%, 3.3V: 403%, 1.8V: 179%. Obviously, the 3.3V supply has the biggest variation. This supplies both other chips and the processors "CPU_3.3V" signal from the reference design. Looking at the load monitoring resistor for CPU_3.3V, it appears that ALL of the 3.3V current is going to the core, not elsewhere. This is not true for the CZUT7 board. For it, the core is taking only 76mA, while the total is 120mA, suggesting that 44mA is going to other chips. Back on the TZUTD1 board (set), with two 0.025 ohm resistors, I actually measure 13.47mV downstream on the core, with 12.84mV upstream on the power supply. This says the core is using more current than the supply is providing. The difference must be an error in the 0.025 ohm resistors of at least 5%, allowing no power to the other chips. If the other chips are taking the same 44mA, then the error must actually be closer to 15%. Confusing. (FYI, the 13.47mV reading corresponds to 538.8mA, comparable to the 532mA in my table, the readings taken at different times.)
-Helmut
