Many embedded ARM processor systems are based on battery-powered power supplies. The power consumption of the processor is critical to the entire SoC chip. Therefore, the ARM processor's low-power advantages can fully save energy consumption. In conclusion, current current diagrams for typical power consumption do not depend on standard processes, standard sets, or workloads.
EnergyBench offers several tools that can be easily combined with economical hardware to measure typical power consumption using standard methods developed by E EM BC. However, in addition to processors, specific chip designs and peripheral modules integrated into the chip are also important factors that affect chip power consumption. Although many chip vendors will provide power consumption parameters in the product datasheet, these parameters are often not comparable. When designers try to compare different processors integrated into the SoC, it becomes very difficult to figure out what the true power consumption of the processor is. This is because suppliers often use typical power consumption parameters to describe their processors, but rarely indicate the processor's workload when these measurements are taken, which will be a key factor in determining energy and power parameters.
Many embedded ARM processor systems are battery powered. Therefore, ARM is recognized as a "low power leader" in the field of processors.
However, the power consumption of the system does not only depend on the processor. In addition, the specific chip design and the peripheral modules integrated into the chip will also affect the on-chip energy consumption.
Although many chip vendors will provide power consumption parameters in the product datasheet, these parameters are often not comparable. This is because vendors often use typical power consumption parameters to describe their processors, but rarely indicate the processor's workload when making these measurements.
In the past, the industry usually focused on the performance of the processor, but with organizations such as E EM BC developing various test benchmarks (such as benchmarks for automotive, consumer electronics, and networking applications), we can be more clear Understand the real situation inside the processor. As the power consumption problem is gradually becoming a focus in embedded applications, power consumption must be taken as an equally important indicator as a performance parameter when evaluating a processor. The ultimate goal is to help system designers get the best balance between performance and power in portable applications.
EEMBC's approach to this goal is to develop the benchmark software utility Energy B ench, which provides real data on energy consumption when the processor is actually operating. Designers can use the EnergyBench and EEMBC performance benchmarks simultaneously to compare the energy consumption efficiency of different processors when performing a series of standardized application tasks. When using EnergyBench to view the power consumption of a single device, it is clear that there is no so-called "typical power" because the average power varies greatly when running different E EM BC benchmarks. EnergyBench does not reflect the typical power of the processor, but it can get the typical power consumption value of some specific algorithm or application at a specific performance level.
Using the LabVIEW platform and data acquisition (DAQ) card from National Instruments (NaTIonal Instruments), EEMBC has successfully implemented EnergyBench. The DAQ card provides multiple differential measurement channels that allow simultaneous power measurements on multiple power inputs (capturing voltage and current for each measurement) and a single trigger channel. Any ARM processor or vendor that uses an evaluation board or its own hardware platform needs only to modify its board-level circuitry to achieve measurable power input lines and add shunt resistors.
EnergyBench can use the DAQ card to sample the voltage and trigger channels and write all sampled results to a file. A flexible trigger mechanism enables performance benchmarks and power measurements to be synchronized. This ensures that the measurements represent the power consumption of the benchmark code at runtime, not the power consumption in the reference initialization or retention record phase.
When running benchmarks and sampling power consumption, the reliability, repeatability, and consistency of results must be ensured, which is especially important for the popularity of the standard. EnergyBench uses a variety of methods to ensure the achievement of these goals:
1. Reliability:
In order to obtain accurate statistical results, it is usually necessary to sample at a frequency of 2 XN yquist or more, or at a random point.
The EnergyBench sampling module accepts the sampling frequency as an input, and then must call the module multiple times with different sampling frequencies. Sampling points generated by multiple sampling with unblurred frequencies during the reference run avoid any resonance with the reference execution. In other words, assuming that each benchmark's repetitive tests generally occur within periodic intervals, using frequencies that are not confused with cycles ensures that pseudo-random points are sampled at each repetition. This method is easy to implement and guarantees the accuracy of the statistical results. Using this flexible method, the frequency confused with the reference period can be easily detected because this will result in different results in one of the sampling frequencies. If this condition is detected, a new set of unconfused frequency is selected and tested until a valid result is obtained.
2. Consistency:
Since we can repeat the test as many times as we need and increase the sampling frequency arbitrarily, EnergyBench performs multiple samplings until the average power consumption can be obtained through accurate statistics. If the deviation of the value of each repetition is too large, the sampling frequency can be increased to increase the accuracy and reduce the deviation.
3. Repeatability:
The measurement process is repeated several times to ensure measurement accuracy and the standard deviation of the final result is calculated. It is also easy to get any detection deviation because each benchmark generates the average power consumption for each repetition of the benchmark. Of course, if it is assumed that the target device provided by the supplier is universal, then testing the device to get the result is considered credible. At the same time, EEMB has always had very strict certification procedures to ensure that the selected chip under test is representative.
For the same reason, all semiconductor manufacturers should have prior consideration for the potential for any minor deviations in the manufacturing process, and EnergyBench's testing for a wide range of potential applications can help manufacturers gain insight into how to select test chips and Test because these will eventually affect the power test results.
After multiple benchmarks have been run on a single benchmark and all measurement sample data is captured, the analysis module calculates the average power consumption for the benchmark based on these data. Based on the average power consumption value, the EEMBC power analysis module finds the minimum and maximum values ​​in the data samples by analyzing the samples captured in each test.
If the variation of a particular sampling frequency is too large, the user can increase the frequency and/or the number of reference repetitions until the above sampled data is stable enough so that the confidence interval of the average value falls within a specified tolerance (95%), thereby guaranteeing the data obtained. accuracy.
The final result of the EnergyBench test is the average power consumption value of the chip when running a benchmark test.
The EEMBC-approved EnergymarkTM test results are an optional parameter that the chip manufacturer can choose to provide to the customer along with the chip performance parameters as a reference to indicate processor power consumption.
The EnergyBench specification states that the device must be ready for at least 30 minutes with an ambient temperature of 70°F +/- 5°F.
These basic conditions are very important to ensure the consistency of the results. Because there is evidence that as the device temperature increases, its power consumption will increase.
The DAQ card allows and the Encoder B ench specification also requires measurement of all power rails on the processor. Obviously, for low-end ARM based devices with a limited number of pins, there are very few power rails to measure. High-end ARM based devices with multiple power inputs must run multiple benchmarks to capture complete processor power data. EnergyBench's Test Harness includes multiple executable files that can measure one, two, or three power rails simultaneously. For processors implemented using multiple power rails (ie, core power and I/O power), there are two ways to calculate the power consumption of the benchmark for each repeated test. In the first method, EnergyBench uses a DAQ card to measure up to three bowrails simultaneously. However, when using this method, since all channels must use the same sampling rate, it may be necessary to reduce the sampling rate of the DAQ card so that the host can follow the sampling process (since there is too much input data at the same time). In addition, the power rail can also be measured separately. The sum of the power values ​​of each power rail is the total cumulative power consumption.
Which method to use?First, some processors have more than three power rails. Even if three power rails are measured at the same time, it is still necessary to make a single measurement for a power rail or consider using a DAQ card on more input channels.
In addition, the sampling frequency should be proportional to the operating frequency of the processor to ensure that sufficient data is sampled for each reference during repeated tests. In order to accommodate GH z-class processors, a higher sampling rate may be needed so that the host only tests one power rail at a time.
To understand this method in depth, we consider many alternatives. Such as specifying the connection temperature when measuring power consumption and using high-frequency oscilloscopes and high controllable test environments. However, since we are not trying to describe the chip features themselves, but rather to determine the typical power consumption, we decided to use the available hardware and control the ambient temperature, not the temperature of the connection or a case. Another problem is that power consumption measurement targets target devices ranging from as low as 5MHz microcontrollers to the fastest processors on the market today. Therefore, it is also necessary to consider that the measurement can be repeated in multiple places and not in a specific environment, and the results are independently verified. With programmable DAQ, we can easily specify parameters (such as sampling frequency) and then permanently retain all captured data.
In conclusion, current current diagrams for typical power consumption do not depend on standard processes, standard sets, or workloads.
EnergyBench offers several tools that can be easily combined with economical hardware to measure typical power consumption using standard methods developed by E EM BC.
Safety Light Curtain,Safety Curtain,Laser Safety Light Curtain,Safety Optic Light Curtain,Security Light Curtain,Press Brake Safety Light Curtains
Jining KeLi Photoelectronic Industrial Co.,Ltd , https://www.sdkelien.com