BYU Home page BRIGHAM YOUNG UNIVERSITY
Department Title
Navigation Menu


Power Studies

Study Results

The results of the two different power measurement tools are presented here. Along with callibration results, we discover the power costs of TMR. It is also interesting to note how sensitive power measurments are to design placement. Different placements of each incrementer and counter design are used to emphasize this sensitivity.

In the results that follow, both the JPower and XPower tools were used to measure the power of a design at frequencies ranging from 10 MHz to 70 MHz. The results of these measurements are plotted in the graphs that follow. In each graph will usually be two pairs of results. The lower red and blue line are the results for the design without the use of TMR and the top red and blue lines are the results with TMR. All blue lines correspond to JPower results, while all red lines correspond to XPower results.

Also included in the following results are tables which summarize important information for each design. One of the pieces of information summarized is the slope of each line in the graphs. This slope reports the increase in mA per 1 MHz for a given design. Another important result is the TMR vs. non-TMR power ratio for a given design.




Incrementer Results
Effects of placement on power measurements on TMR designs was primarily studied with the first incrementer design (incrementer replicated 72 times with each incrementer output feeding an IOB). Xilinx's auto placement results were first examined. Then three different hand placements were examined. The first hand placement was puposefully a poor placement. The second placement improved upon the first placement and the third placement was an optimized hand placement. The three hand placements of the TMR designs are demonstrated as follows:




Power Measurment Graphs
Power measurments were taken for each of these placements as well as for Xilinx's auto-placement. Measurements were taken while the design ran at frequencies from 10 MHz to 70 MHz and are shown in the graphs which follow. Remember that blue lines correspond to JPower results while red lines correspond to XPower results. Also remember that in each graph, the bottom set of lines correspond to non-TMR results while the top pair of lines correspond to TMR results.


Table Summary Results
The following table summarizes the information from the above graphs. Remember that the slopes reported are in units of mA per 1 MHz.


   Auto-Place Place 1 Place 2 Place 3
No TMR TMR TMR TMR TMR
JPower Slopes 1.51 7.37 10.65 6.15 4.78
XPower Slopes 1.54 5.23 6.20 5.21 4.76
JPower TMR vs. Non-TMR Ratio 4.88 7.04 4.06 3.15
XPower TMR vs. Non-TMR Ratio 3.40 4.04 3.39 3.10
Size (LUTs) 576 1728 1728 1728 1728


The blue and red shaded rows of the above table identify how sensitive power measurment can be to design placement. It also shows how the accuracy of XPower measurements are dependent on placement. For design placement 1, JPower reports that the cost of TMR is about 7x the power of non-TMR, but XPower reports a cost of only about 4x. Note that with the optimal placement, both JPower and XPower report a cost of about 3.1x.

Related Links