czhe
Newbie level 5
- Joined
- Mar 19, 2007
- Messages
- 9
- Helped
- 0
- Reputation
- 0
- Reaction score
- 0
- Trophy points
- 1,281
- Activity points
- 1,360
Hi,
I have some data for the dynamic power of an ASIC for a given workload. These measurements go through a range of voltages (all other factors are constant, like frequency, etc). Is it possible to develop a model to fit this set of data with the traditional formulae used for switching and short circuit power, i.e:
P = alpha*Ceff*V^2*f + K*(V-Vt)^3*f
The first part is for switching power, which seems to be the most commonly adopted version. The second part is borrowed from Veendrick's model (http://www.acsel-lab.com/Projects/lowpower/references/Veendrick_84.pdf).
What I've been trying to do is fit a curve onto the data for all the voltage points and solving for (alpha*Ceff), K, and Vt. I can come up with an equation similar to above that has very little error but I'm wondering if this is the correct/best way?
If anyone has any suggestions or experience, please let me know your thoughts.
Thanks,
czhe
I have some data for the dynamic power of an ASIC for a given workload. These measurements go through a range of voltages (all other factors are constant, like frequency, etc). Is it possible to develop a model to fit this set of data with the traditional formulae used for switching and short circuit power, i.e:
P = alpha*Ceff*V^2*f + K*(V-Vt)^3*f
The first part is for switching power, which seems to be the most commonly adopted version. The second part is borrowed from Veendrick's model (http://www.acsel-lab.com/Projects/lowpower/references/Veendrick_84.pdf).
What I've been trying to do is fit a curve onto the data for all the voltage points and solving for (alpha*Ceff), K, and Vt. I can come up with an equation similar to above that has very little error but I'm wondering if this is the correct/best way?
If anyone has any suggestions or experience, please let me know your thoughts.
Thanks,
czhe