1. Home
  2. Field-Programmable Gate Arrays
  3. Power Analisys with MicroBlaze

Power Analisys with MicroBlaze

I'm having some problems in simulating the power consumption of a design I worked on for university. I'm using Xilinx EDK 6.3, Xilinx ISE 6.3, Modeltech Modelsim 6.0 SE and a Spartan 3 Starter Kit.

The design uses just the MicroBlaze and the controller for the SRAM on the board, it synthetizes without problems and it does what it should do, but I'm having troubles in making the results of XPower 'accurate'.

Here is how I proceed:

1) creating a timing simulation from Platform Studio, 2) running ModelSim from Platform Studio 3) typing the following

do system.do vsim -sdftyp system.sdf system vcd file system.vcd vcd add system/* run 2usquit -f

4) importing the generated .vcd file together with the desing .ncd file and the physical constraints .pcf file in XPower.

After that XPower tells me that only 88% of the signals have been set and only 67% of the signals toggle and that the results are 'inaccurate'.

I would like to have at least 'reasonable' results, is anybody able to give me a hint on how to improve the accuracy? Am I omitting something?

Thanks in advance for any advice.

kl31n

Reply to
kl31n
Loading thread data ...

design I

on the

but

file and

and

to give

Reply to
Peter Alfke

First of all thanks for the answer.

The decision of using a software approach is due to the fact that I've to estimate the power consumption of the FPGA only(not of all of the board, that's to say I've to ignore the consumption of the SRAM, all of the LEDs, the RS-232 driver and any other device around which isn't the FPGA itself) and that the measuring equipment that I've at university isn't refined enough to allow the measurement of an SMD.

kl31n

Reply to
kl31n

The XPower results are dependant on the toggling information coverage from your VCD file, which is dependant on other factors like your software running on the Microblaze processor and/or external events (like interrupts, network packets etc.) . To give you a simple example you can have a counter/timer in your Microblaze system but your software doesn't enable the counter/timer (is not running) so it will not take much power (as if it was running) try to write a test software which enables all the peripherals in your system (dummy functions) to force a better coverage in your VCD file.

But the mother of "accuracy" still remain the lab measurement as Peter stated, you can "ignore" the peripherals power (sram, leds) by measuring the current for the Vccint only (assuming that 1.2V, Vccint goes only on your device under test) but you need to insert your amp-meter by cutting the Vccint track (some boards are provided with a jumper for this purpose, to insert an amp-meter)

Aurash

kl31n wrote:

--
 __
/ /\/\ Aurelian Lazarut
\ \  / System Verification Engineer
/ /  \ Xilinx Ireland
\_\/\/
 
phone:	353 01 4032639
fax:	353 01 4640324
Reply to
Aurelian Lazarut

Hi,

Try to simulate a little more than 2 us. In the first 2 us after reset not much is toggled.

Göran

kl31n wrote:

Reply to
Göran Bilski

I tried up to 100 us, but no improvement in accuracy. I forced the sys_clk_pin signal to be a 50MHz clock and the sys_clk_rst signal to provide for a warm reset, but still no improvement in accuracy.

Thanks,

kl31n

Reply to
kl31n

Il Mon, 24 Jan 2005 13:05:13 +0000, Aurelian Lazarut ha scritto:

The software I've written is already a simple version of the custom DSP I've designed(the software has been written just for the power analisys). It doesn't use interrupt, it doesn't even use functions. There's just the main with inside an infinite loop. The only peripheral used is the EMC. I forced the sys_clk_pin signal to be a 50MHz clock and the sys_rst_pin to provide for a warm reset. I cannot understand what keeps the accuracy low as the whole thing is very simple.

Thanks,

kl31n

Reply to
kl31n

Do you have any printout in the code? If the program starts with some printouts, then it will wait on the UART to transmit the characters before doing something useful.

Göran

kl31n wrote:

Reply to
Göran Bilski

No, it doesn't have any sort of I/O to the RS-232 port. The app used for the power analisys has been taylored just for that purpose and all of the data come and go on the SRAM(they aren't meaningful data, as I said the app is taylored only for power analisys). As I said in the reply to Aurelian the only peripheral which is used is the EMC.

Thanks,

kl31n

Reply to
kl31n

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.