Power measurement on FPGA-Xc3s4000-FG676

Status
Not open for further replies.
W

Wesley90

Junior Member level 1
Joined
Jan 29, 2013
Messages
15
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,281
Activity points
1,410
Hello,

I need to measure the power consumed by the FPGA for any design loaded onto it.

The PCB for this chip was custom designed by one of the students and i want to measure the power consumption of it.
It has separate supplies viz

1. Core voltage
2. Auxillary voltage
3. Internal logic voltage
which are plugged into the FPGA as small circuit boards (3cms*3cms).

Which is the best ( most accurate) current sensing IC that I can use for this purpose?
And also how do I connect this IC in serial with the CORE voltage supply.?

Thanks in advance
 

Hi,

there are current sensing amplifiers. Read about them

*
you need to decide
* if you want to measure only the (dissapated) power for the FPGA,
* or additionally all the power that leaves the FPGA on it`s IO lines

Klaus
 

Hello,

Thank you I want to measure the total power consumed by the FPGA for any design loaded onto it. So i do not know if I have to measure the power across the 3 supplies (Core, Auxillary voltage & Internal logic) separately or to measure the power directly across the main power supply (DC). I am new to the field of power measurement on FPGA. So kindly help me in deciding across which supply should I measure.

Thanks in advance
 

Hi,

Power_in = Voltage1 x Current1 + Voltage2 x Current2 + Voltage3 x Current3

Power_out = VoltagePin1 x CurrentPin1 + ... + VoltagePin_n x CurrentPin_n

*****
If you only want to know_Power_in then it is easy.
Please clarify this.


Klaus
 
I want to know if this POWER_IN varies from design to design (coded in Verilog).

If yes, then I would want to measure only the power_in.

I want to measure power_in which gives me the total power consumption of the FPGA for a design since I have no external peripherals connected to the board.

What is Power_out? (I am guessing that it is the power consumed by the peripheral devices connected to the FPGA)
 

Wesley90 said:
I want to know if this POWER_IN varies from design to design (coded in Verilog).
Click to expand...

I do not know the Xilinx software, but at least in Altera it is possible to evaluate this by using specific tool available on the manufacturer's IDE. It is an approximation, but if the objective is not to know the exact value but just to analyze differences in distinct code implementations, I think the simulation is the most recommended method since the effective measurement may be subjected not only to the error of measurement itself, but also to the variation of the ambient temperature, which may impacts on the result as well.
 

Yes
Xilinx has a power estimation tool. The task here is to find out the error between the ESTIMATION and the ACTUAL POWER CONSUMPTION (For 2 different designs).

Eg.

DESIGN 1: ESTIMATED POWER (W) versus MEASURED POWER(W)-(I get some %error)

DESIGN 2: ESTIMATED POWER (W) versus MEASURED POWER(W)-(I get some %error)

Hence, I need the circuit to measure the power at 25 degree celcius. It would be of great help if you can let me know across which pins should I connect the CURRENT SENSE AMPLIFIER to measure the overall power consumption.
 

Hi,

I assume the power tool calculates only internal power consumption (= dissipation).

Which is (referring to my post#4): Power_diss = Power_in - Power_out.

***
But maybe in the power estimation tool it´s possible to specify each I/O´s load: capacitive and resistive. I don´t know.


Klaus
 
Yes, This sounds perfect. I think the tool calculates only the internal power consumption. So, according to the post#4, I can connect the CURRENT SENSE AMPLIFIER across the CORE VOLTAGE SUPPLY so that I can get the total power consumption the same way the tool calculates.

Is it right?
 

Hi,

You can connect the current sense amplifier.
But all my posts say that this is NOT only the internal dissipated power. (Instead this measures the internal dissipated power PLUS the power at the I/Os)

Klaus
 
There is also the AUX supply that deals with a number of things like the PLL/MMCMs. On some parts they have a separate BRAM supply rail.

You need to connect to ALL supply voltages that the FPGA is using. The input voltage to the board won't be accurate as it will also contain any losses associated with the POL supplies and those can vary depending on the current draw on the supply (there are no 100% efficiency power supplies).

There is also the need for accurate switching information being added to the power analyzer spreadsheet. The best is accurate real world simulation vectors produced by a good testbench. A testbench that only toggle part of your design or doesn't make your design behave like it does when used will be worthless as a starting point for the PA spreadsheet. But for all I know maybe that is what you are trying to study?
 
Hello,

Yes. But, I have another problem. My FPGA operates at 100MHz. Should I choose a current sense amplifier which has a bandwidth equal to this or any current sense amplifier would do? Since I do not know if the FREQUENCY OF THE BOARD and the CURRENT SENSE AMPLIFIER are related.

- - - Updated - - -

Yes. I will be measuring all the voltages. But my research is limited to observe the error between the ESTIMATED and MEASURED values of power for which I might have to vary few parameters in the PA spreadsheet. It is best that I consider 100% switching activity?
 

Hi,

100MHz: For sure you don't need this bandwidth. And if you have, then it won't be used, because your capacitors next to the FPGA act like a low pass filter (with unknown fc).

What bandwidth to choose mainly depends on what time resolution you want when you display the resulting power.
Display a graph with a scope: Maybe some kHz are useful. I assume you want to see some details.
Display as numeric value: Maybe some Hz are useful. I assume you want to display a long time average.

I recommend to use a some kHz to some 10kHz filter in front of the amplifier(to ensure proper operation), and do the rest of the filtering after the amplifier (= averaging).

It is best that I consider 100% switching activity?
Click to expand...
Here I agree with ads-ee. Read post#11, especially last paragraph.
Your FPGA should perform real application operations.

Klaus
 

Hello,

I have found a way now. Thank you so much
 

Status
Not open for further replies.

Similar threads

Menu
Log in
Register
Cookies are required to use this site. You must accept them to continue using the site. Learn more…