DSC (Digital signal controller, digital signal controller) is one kind face the high-end embedded system’s most advanced monolithic control processor. Has the quicker processing speed based on floating point construction’s DSC, needs the procedure storage capacity are less, supports the higher paraffin being helpful to save the power loss the computation algorithm, simultaneously further expanded system’s performance. The floating point programming is quicker than the fixed-point programming’s speed, SoC (system-on-a-chip, on piece system) the integrated way can the effective control panel level space, part quantity and the integrated system expenses.
The duty which undertakes along with the embedded system is getting more and more complex, no matter is reduces the power loss or realizes such as the automobile guidance and so on new function, they need to have the high performance control processor. Reduced the energy consumption the urgent demand to affect each kind of equipment which we used, including some user uncommon equipment. For example, according to the famous energy and the automated technology supplier ABB group’s analysis, the electrical energy which the industry electrical machinery consumed already achieved the entire profession to use the electric power 2/3. Although may use the speed change control technology to further raise electrical machinery’s operating efficiency, but present only then 1/20 electrical machinery used truly have supported this function the numerical control electronic technology. For all this, changes speed the electrical machinery to compare the similar constant speed electrical machinery the economical electrical energy to be equal every year to 10 power plant output, but these 10 power plants will discharge the approximately 68,000,000 ton carbon dioxide gas. Obviously, the high-quality motor control technology is reducing the energy consumption and the protection environment aspect has a bigger superiority.
Introduces the high-level control technology to reduce the energy consumption
Renewable energy source, for example solar cell and wind power turbine, is also needs to use the high-level control technology “the green” the energy. Is the same with the electrical machinery, these energy collecting device also needs to use the high-level numerical control technology to raise the efficiency. For produces the battery board either the turbine the primitive electrical energy real-time transforms uses in the appropriate electrical energy which the battery storage or the equipment movement need, we must use certain control technology. Because the sunlight gentle breeze intensity of pressure will change unceasingly, therefore needs to use the complex algorithm to guarantee that they can output the biggest energy throughout; Because the renewable energy source system can simultaneously turn on the electrical network, therefore uses the real-time control technology regarding the protection device is also essential.
Figure 1 F28335 floating point DSC contains one to use in processing the data transmission 6 channel DMA controller
For example, TI Corporation promotes F283x floating point DSC (as shown in Figure 1) has the performance to be high, characteristics and so on software development convenience, are suitable for the electrical machinery and the operating control, the trailer-mounted radar system and take the solar cell array and the wind turbine as the origin renewable energy source generating system. This section of 32 DSC series processor highest operating frequency may reach 150MHz, the peak value operating speed may reach each second 300,000,000 floating point operation (MFLOPS).
Compared fixed-point DSC TMS320F28x saying that after using the floating point construction, the DSC performance has promoted 50%. This a series of DSC contains a 6 channel DMA controller, can reduce the processor nucleus greatly the data transmission supervisory work, its software update function is helpful to the user develops the identical origin conveniently the floating point and the fixed-point code.
Floating point calculation
F283x the series DSC main performance superiority attributes to its numerical calculus form, this kind of form simplified the programming greatly, reduced the code execution time, reduced the word length. In other DSC and in the DSP chip the universal use fixed-point format can only express that the integer (does not have decimal point positive number and negative number), therefore involves time the decimal computation must use other expression method. On the contrary, the floating point form can express in the wide range real number (a mixed decimal number). The internal expression decimal as well as the wide numerical range means that the processor can carry out the calibration operation highly effective (e.g. in control algorithm common multiplication, division and trigonometric function operation).
Figure 2 F283x 32 word length’s first 8 use in expressing the index. F283x digital signal controller’s 32 float the braille
Regarding 32 fixed-point constructions, its expression integer’s scope for - 231~231, has a sign bit. Although this scope is very big, but works as when system execution massive calibration operation, this numerical range will use up very quickly, causes the value to surpass 32, will thus have the overflow of register. The procedure through carries on the half adjust or the interception to this kind of value, can process this kind to be called the saturated question, but has sacrificed the computation precision. Moreover, the procedure may divide the long number the number which Cheng Duoge short can calculate section by section, each time processes 32. Behind this kind realizes the method to be able to maintain the entire precision, but because the processor is busy with motion and the memory value section, the estimated performance reduced. At the same time, procedure scale also because will need the extra instruction increased.
In this case, the floating point construction demonstrated its performance value. As shown in Figure 2, in F283x, 32 word length’s first 8 uses in expressing the index, other 23 use in expressing the mantissa, 1 serves as the sign bit. Although the index does not have the sign bit, but carries on standardized displacement processing in the operation to the preserved index, enables the index to be able to cover the negative number and the positive number scope finally.
Thus, saves in 32 to float in the braille the number standardized scope for ±1.738~±3.438, surpasses 32 fixed-point numbers greatly the scope. Because the negative exponent expresses the score, therefore this scope has covered the minimum number to the limit big number. The such great scope is very difficult to have the saturated problem again, therefore, the procedure has avoided either through the half adjust or the interception sacrifice precision, either carries on the staging treating through the use attachment cycle and the instruction to the great value to sacrifice the performance and the memory is in a dilemma the question.
Processor performance enhancement
In table 1 lists the operation showed the floating point construction is the advantage which the DSC performance brings. The first row is the control system commonly used four kind of arithmetic operation, the trigonometric function operation and two algorithms (the fast Fournier transformation and infinite impact response).
Table 1 fixed-point and floating point construction performance benchmark test comparison
The second row is carries out the periodicity which on fixed-point DSC the corresponding operation needs, the third row is the periodicity which carries out on floating point DSC needs. Last row is the second row with third row result ratio, expresses the two’s relative performance. The floating point processor execution enumerates the speed which the student movement calculated compared to fixed-point processor quick 2~3 times, carries out the FFT algorithm pick-up also in this scope, the IIR pick-up is slightly low, but was still quicker than the fixed-point processor.
Generally speaking, the control algorithm can obtain the higher performance promotion compared to the signal processing algorithm, are many because of the control routine execution’s basic mathematics operation, needs to carry on the fixed-point conversion and saturated processing. Even if so, the floating point construction obtains the periodicity is also much lower, has obtained 50% performance promotion equally to all signal processing test order.
In each kind of application, these benchmark test procedure transforms the concrete application procedure. Fine, the multi-dimensional control technology may apply in the robot and CNC (computer numerical control) a kind of equipment. The servo electrical machinery driver’s efficiency will have further enhancement, can realize PLC (power line control) and other high-level algorithms. The solar energy and the wind energy invertor and the UPS can obtain the higher energy conversion efficiency, further reduces each kilowatt electrical energy the production cost, and can control the different disposition the solar energy board or the wind power turbine.
Software development superiority
The floating point form is also advantageous to the simplified code compilation and the debugging. The floating point number representation compares the fixed-point representation nature regarding mathematics operation, therefore is more direct to higher order language’s support. When code after compilation and confirmation debugging, may induct floating point DSC to carry on the further test and the final production directly.
Comparatively, in aims at the fixed-point controller development code time, must on PC the compilation and the debugging routine carries on the confirmation, then must act according to the stricter hardware fixed point to express that the method compiles the code. This increased the code development cycle greatly, moreover, once carries on the transformation after the code is unable to carry on the retroversion.
Generally, the development personnel can not but simultaneously compile fixed-point the code and the floating point code, this existence possibly mutually confuses danger. The F283x controller only needs to develop the floating point code, thus simplified the code performance history greatly, has saved the development time, enhanced the software reliability.
In the existence cost restraint’s situation, may as develop the platform take the floating point controller to carry on the prototype and the early edition design first, then changes to the fixed-point controller to carry on the quantity to produce the manufacture, uses this development strategy to have the obvious superiority. Uses the C compiler and IQ the Math tool can translate the floating point and under very conveniently the fixed-point two ways the same sound code, thus supports this strategy. F283x DSC is one kind of economical controller solution, it is the field first section uses the SoC integration the floating point controller. Rises along with the floating point construction and the fixed-point construction’s cost intersection to more and more high level, the high-quality system to save the cost not to need to carry on the modification. The performance was higher, develops the easier floating point construction cost already to be able to accept for more and more applications, urges the people to carry on the innovation research and development unceasingly in the embedded system control domain.
