JP2009223476A - Design method and device for multi-output power source circuit - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a design method and a device of a multi-output power source circuit by which the combination of voltage converters with high efficiency can be searched with small man-hours. <P>SOLUTION: In the design method of a multi-output power source circuit for supplying an output power source to a plurality of power source domains, a tree structure having the most satisfactory power efficiency is searched by repeating a process for generating a tree structure having a voltage converter at random between an input power source (root) and a plurality of output power sources (leaves), and a process for calculating the power efficiency of the multi-output power source circuit of the tree structure only by the number of times smaller than the number of whole tree structures. The search process is applied again to the input side tree structure having a plurality of input side voltage converters prior to the voltage converter which outputs an arbitrary output voltage in the tree structure found in the retrieval process, thereby achieving improvement to replace the input side tree structure with the input side tree structure having a more satisfactory cost function value detected by the retrieval. The improvement process is performed by successively moving an arbitrary output voltage to the input side. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a design method for a multi-output power supply circuit and a design apparatus for a multi-output power supply circuit, and in particular, generates a plurality of output power supply voltages from a single input power supply voltage by forming a plurality of voltage converters in a tree structure (tree structure). The present invention relates to a design method and a design apparatus for a multi-output power supply circuit.

  Large-scale semiconductor integrated circuit devices (LSIs) tend to be multi-powered due to the combination of multiple macros generated with different architectures and the high demand for power saving. . In other words, an LSI has a plurality of power supply regions (power supply domains) inside, and by controlling on / off of these power supplies, wasteful power consumption is eliminated and power saving is achieved. Therefore, a multiple output power supply circuit that supplies multiple power supplies to such an LSI is required.

  The multi-output power supply circuit generates, for example, a plurality of output power supplies corresponding to the power domain in the LSI, using a DC power supply in the system as an input power supply. One method for configuring a multi-output power supply circuit that generates a plurality of output power supplies from a single input power supply is to combine a 1-input 1-output voltage converter, such as a DC-DC converter. The 1-input 1-output DC-DC converter is a power conversion circuit that generates an output DC power supply having a second voltage lower or higher than the first voltage from an input DC power supply having a first voltage.

  Along with power saving of LSI, power saving of a multi-output power supply circuit is also required. Therefore, the multi-output power supply circuit is required to have low power, that is, a high ratio of output power to input power (high efficiency). On the other hand, the demand for space saving is also required that the number of DC-DC converters constituting the multi-output power supply circuit is as small as possible.

  However, there are many combinations of voltage converters that generate a plurality of output power supplies, and no effective method for searching for the optimum combination has been proposed yet. As a general theory, a method is conceivable in which power consumption is calculated for all possible combinations of voltage converters, and a combination with the smallest power consumption is searched from among them.

Regarding the layout method of a semiconductor device having multiple power sources, there are the following patent documents. This patent document discloses a method for facilitating the layout of a semiconductor device having multiple power supplies. A design method for a multiple output power supply circuit that generates multiple power supplies is not described.
Japanese Unexamined Patent Publication No. 2000-15018

  The method of searching for the optimal combination by calculating the power consumption for all possible combinations of the voltage converters described above is not preferable because it requires an enormous amount of man-hours. In particular, as the number of power supply domains of LSI increases, the number of required output power supplies also increases, and the number of possible combinations tends to increase astronomically.

  Accordingly, an object of the present invention is to provide a design method and a design apparatus for a multi-output power supply circuit that can search for a combination of high-efficiency voltage converters with less man-hours.

The design method of a multi-output power supply circuit that supplies output power to multiple power domains is as follows:
A plurality of voltage converters, each of which defines an input power source as a root and a plurality of output power sources as leaves, and uses any one of voltages from the input power source voltage to the output power source voltage as an input voltage and an output voltage; A tree generation step for randomly generating a tree structure between a root and a leaf, and a power function of the multi-output power supply circuit having the randomly generated tree structure or a cost function value including the power, the plurality of power supply domains The cost function value generation process that is calculated based on the required voltage and required current of each and the efficiency of each voltage converter is repeated a smaller number of times than the number of all possible tree structures, and the tree structure having the best cost function value is obtained. A search process to search;
For the input side tree structure having a plurality of voltage converters on the input side before the voltage converter that outputs an arbitrary output voltage among the tree structures found in the search step, the search step is performed again, and is better than that detected in the search An improvement process for replacing the input tree structure with a cost function value,
The improvement process is repeated by sequentially moving the arbitrary output voltage to the input side.

  According to the above multi-output power supply circuit design method, the search process performs the tree generation process and the cost function value generation process a smaller number of times than the number of all possible tree structures, and further the improvement process has a smaller number of voltage converters. Since a search process for finding a tree structure having a better cost function value is performed for the input side tree structure consisting of the above, a combination of voltage converters having an improved cost function value can be searched with a small number of man-hours.

  According to a more preferred aspect, in the cost function value generation step, the output node voltage converter is connected to the output node voltage converter based on the required voltage and required current of the plurality of power supply domains and the efficiency of the output node voltage converter that generates the output power supply. Calculate the output current of the input-side voltage converter that supplies the input voltage, and repeat the calculation to obtain the input-side output current based on the calculated output current and the output voltage and efficiency of the voltage converter up to the input power supply. The power output of the multi-output power supply circuit is obtained from the power efficiency of the multi-output power supply circuit based on the ratio of the sum of the power consumption of the output power supply and the power consumption of the input power supply.

  Further, the cost function value is preferably a value obtained by adding the reduction efficiency based on the number of the voltage converters to the above power efficiency. By using such a cost function value as an index, it is possible to search for fewer combinations of voltage converters in addition to higher power efficiency.

  Even if it is replaced with an input-side tree structure with a better cost function value in the improvement process, the output current of each voltage converter obtained by calculation from the output power source toward the input side in the cost function value generation process is wasted Without being used for the calculation of the power efficiency or power of the multi-output power supply circuit.

  According to a more preferred aspect, in the tree generation step, a plurality of output node voltage converters that respectively output the plurality of output power supplies, and the plurality of output node voltage converters and the input power supply are randomly inserted. A tree structure having a plurality of via node voltage converters is generated. That is, a relay node voltage converter between a plurality of output node voltage converters that respectively output a plurality of output power supplies and a single input power supply is randomly generated based on a random number or the like.

  Search for high-efficiency voltage converter combinations with less man-hours.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the technical scope of the present invention is not limited to these embodiments, but extends to the matters described in the claims and equivalents thereof.

  FIG. 1 is a diagram illustrating an example of a multi-output power supply circuit. FIG. 1 shows three types of multi-output power supply circuits PSC1, 2, and 3. The multi-output power supply circuit PSC1 includes voltage converters C1 to C4 that convert the voltage from the input power supply voltage 5V to the output power supply voltages 3.3V, 3.3V, 2.5V, and 1.3V, respectively. The multi-output power supply circuit PSC2 includes a voltage converter C1 that converts voltage from an input power supply voltage 5V to an output power supply voltage 3.3V, a voltage converter C5 that converts voltage from 3.3V to 2.5V, and 2.5V to 1.3V. And a voltage converter C6 that converts the voltage into a voltage. The multi-power supply voltage PSC3 includes a voltage converter C1 that converts voltage from an input power supply voltage 5V to an output power supply voltage 3.3V, a voltage converter C5 that converts voltage from 3.3V to 2.5V, and 3.3V to 1.V. And a voltage converter C7 that converts the voltage to 3V. As described above, a multi-output power supply circuit that generates four output power supply voltages from one input power supply voltage 5V can be configured with a tree structure (tree structure) of voltage converters, and there are a plurality of possible tree structures. .

  FIG. 2 is a diagram showing the definition of the multi-output power supply circuit. As shown in FIG. 1, the multi-output power supply circuit PSC designed in the present embodiment has a plurality of power supply regions (power supply domains) in an LSI to which power is supplied from one input power supply voltage P0 that is a power supply source. ) Generate an output power supply voltage to be supplied to D1 to D4. Therefore, in the multi-output power supply circuit PSC, the input power supply voltage P0 is defined as a root, the output power supply voltage to each of the power supply domains D1 to D4 is defined as a leaf, and the tree structure (tree) of the voltage converter C between the root and the leaf. Structure) is formed. Since it has a tree structure, one branch is not formed from a plurality of branches even if it is divided from one branch into a plurality of branches from the root to the leaves. Also, one branch is generated from one branch according to the voltage that passes through, and a voltage converter is arranged at the joint between them.

  Each voltage converter is a voltage converter that generates one output voltage from one input voltage. The multi-output power supply circuit has a tree structure of the voltage converter. The power consumption of each voltage converter is obtained by the product of output voltage and output current. The efficiency of each voltage converter is the ratio of the output power consumption to the input power obtained from the product of the input voltage and the input current.

  FIG. 3 is a circuit diagram illustrating an example of a voltage converter. Both voltage converters are DC-DC converters that convert the DC voltage Vin to the DC voltage Vout. The step-down DC-DC converter Csdwn includes a switched S1, an inductor L, a diode D1, a capacitor C1, and a load resistor RL, and an output voltage Vout is generated at the load resistor RL. The output voltage Vout is lower than the input voltage Vin. When the switch S1 is on, energy is stored in L by a solid line current. When the switch S1 is off, the diode D1 commutates the energy stored in the inductor L to the load RL like a broken line current. Then, the potential of the output voltage Vout is controlled by the ON / OFF ratio of the switch S1. Charges are accumulated in the capacitor C1, and fluctuations in the output voltage Vout are suppressed.

  On the other hand, the step-up DC-DC converter Csup also includes a switched S1, an inductor L, a diode D1, a capacitor C1, and a load resistor RL, and an output voltage Vout is generated at the load resistor RL. The output voltage Vout is higher than the input voltage Vin. When the switch S1 is on, energy is stored in the inductor L by the solid line current. When the switch S1 is off, the diode D1 commutates the energy stored in the inductor L to the load RL like a broken line current. Then, the potential of the output voltage Vout is controlled by the ON / OFF ratio of the switch S1.

  The multi-output power supply circuit of the present embodiment is a tree structure of a DC-DC converter as shown in FIG. However, the multi-output power supply circuit of the present embodiment can be configured with voltage converter circuits other than those shown in FIG.

  FIG. 4 is a flowchart showing an example of a design method for a multi-output power supply circuit. This design method is a comparative example to the design method of the present embodiment described later. In the design method of the multi-output power supply circuit of FIG. 4, the input / output data 10 required for the multi-output power supply circuit PSC and the efficiency characteristic data of the DC-DC converter constituting the tree structure (tree) are transmitted to the computer as the design apparatus. 12 is input. The input / output data 10 includes an input voltage 5V of the power supply source, an output voltage and an output current required for each of the power domains A to E. Furthermore, the efficiency characteristic data 12 of the DC-DC converter includes the power corresponding to the plurality of output currents Iout of the DC-DC converters a), b), c), d) and (d) which are generated in advance. Have efficiency.

  The computer generates all possible tree structures (trees) 14 from the given input / output data 10 (Sa). The tree structure (tree) is as shown in FIG. 1, and the tree structure (tree) 20 shown in FIG. 4 is based on the input power supply voltage P0, the power supply domains A to E as leaves, and the input voltage. The tree structure (tree) of DC-DC converters a), b), c), and d) for converting from the above to the output voltage has between the root and the leaf. In the tree generation step Sa, all tree structure examples (trees) 14 are generated. Next, the computer calculates cost function values with reference to the efficiency characteristic data 12 for all the tree structures (trees) 14 (Sb). The cost function value is the efficiency of the multi-output power supply circuit formed by the ratio of the output power to the input power of the tree structure (tree) 14. In this cost function value calculation step Sb, a power efficiency list 16 is generated for all tree structures (trees) 14. Finally, in the optimum tree selection step Sc, the computer selects the tree structure (tree) 18 having the highest efficiency from the power efficiency list 16 of all tree structures (trees) 14.

  The above multi-output power circuit design method has the following problems. In the case of the one-level tree structure (tree) shown in FIG. 2, the number of combinations of the voltage converters C that generate output power in each of the power domains D1 to D4 is the number of n divided or grouped methods. It is the number of bells (Bn) indicating the number corresponding to the total number. The number of bells is Bn = 1, 2, 5, 15, 52, 203, 877, 4140, 21147, 115975 where n = 1, 2, 3-10. In the tree structure (tree) for generating output voltages in a plurality of power domains, as shown by 20 in FIG. Therefore, in the case of a multi-level tree structure (tree), it is obvious that the number of bells when n = 10 is a combination of Bn = 115,975 or more.

  For reference, when the tree structure (tree) by the voltage converter is 2 hierarchies, n = 10 is 4,073,412 ways, and when it is 3 hires, n = 10 is 2,108,950,871 ways. This is easily verified mathematically.

  Therefore, as shown in FIG. 4, when all possible tree structures (trees) 14 are generated and the cost function values (for example, power efficiency or power) are calculated, a huge amount of man-hours are required. In future LSIs, the number of power supply domains tends to increase. For example, it is necessary to generate output power supplies for 100 power supply domains. As the number n of power domains increases in this way, the number of tree structure (tree) combinations increases dramatically, and the design method of FIG. 4 is not realistic.

  FIG. 5 is a diagram showing an experimental result by the inventor. The graph of FIG. 5 has an input power supply voltage of 5 V and an output voltage of 1.2 V, and for a multi-output power supply circuit having 10 leaves through 3.3 V, the process Sa of the design method of FIG. The power efficiencies of all tree structures (trees) considered by Sb are calculated and output as histograms. The horizontal axis is power efficiency, and the vertical axis is the number of tree structures (trees). The tree structure (tree) distributed on the rightmost side of FIG. 5 has the highest power efficiency.

  As shown in FIG. 5, it was found that the power efficiency of the possible tree structure (tree) is distributed in a relatively high region (64% to 83%). From this, it is not necessary to calculate the power efficiency (cost function value) of all tree structures (trees) that can be considered, but the tree structure (tree) is randomly generated and the power efficiency (cost function value) is calculated. However, even if a tree structure (tree) with higher efficiency is selected, it is understood that a tree structure (tree) having the highest power efficiency and a tree structure (tree) having power efficiency comparable to that of the highest power efficiency can be searched. it can.

  According to the distribution in FIG. 5, the tree structure solution with an efficiency of 80% or more and 83% or less occupies 9% of the whole. This means that the probability that the efficiency of one randomly generated tree structure has an efficiency (80% or more) within 3% of the maximum efficiency (83%) is 9%. Therefore, if a tree structure solution with higher efficiency is selected from the tree structure obtained by performing random generation twice, the tree structure has an efficiency within 3% (80% or more) from the maximum efficiency (83%). The probability of having 100% −91% × 91% = 17%. Similarly, it is 36% at 5 times, 61% at 10 times, 99% at 50 times, 99.99% at 100 times, and 99.99999999% at 200 times. In other words, the process of generating a tree structure at random and calculating the power efficiency does not necessarily have to be performed for all combinations, and if it is repeated a predetermined number of times, a tree structure (tree) having an efficiency within a certain range from the maximum efficiency can be obtained. Can be explored.

  FIG. 6 is a flowchart showing a design method of the multi-output power supply circuit in the present embodiment. The flowchart shown in FIG. 6 is an outline, and each process will be described in detail later. The tree structure is collectively referred to as “tree”.

  According to the design method of the present embodiment, input / output data 10 required for the multi-output power supply circuit PSC and efficiency characteristic data 12 of the DC-DC converter constituting the tree are input to the computer as the design apparatus. The The input / output data 10 includes an input voltage P = 5 V of the power supply source, and an output voltage and an output current required for each of the power domains A to E. Furthermore, the efficiency characteristic data 12 of the DC-DC converter includes the power corresponding to the plurality of output currents Iout of the DC-DC converters a), b), c), d) and (d) which are generated in advance. Have efficiency. The efficiency characteristic data 12 can be obtained by measuring or calculating the power efficiency of each voltage converter in advance by experiment or simulation. A DC-DC converter, which is a voltage converter, is a conversion circuit that converts an input voltage into an output voltage. The input voltage and the output voltage are equal to or lower than the input power supply voltage of the multi-output power supply circuit. That's it. The efficiency characteristic data 12 in FIG. 6 includes data on the input voltage and output voltage of each voltage converter.

  In the solution narrowing-down process S1, a tree is randomly generated from the input / output data 10 between the input power supply voltage P0 and the output power supply to a plurality of power supply domains, and a cost function such as power efficiency is calculated and randomly generated. Among the predetermined number of trees, the tree having the best cost function is narrowed down as a solution candidate tree 22. The number of randomly generated trees is sufficiently smaller than the number of all possible trees, based on the statistical properties obtained from the experimental results in Fig. 5, while the cost function is close to the best value. The probability is high enough. In an embodiment described later, this number is about 50 to 100 times even when the number of power supply domains is ten.

  The candidate tree is, for example, as shown at 20 in the figure, and has a tree structure in which the root consisting of the input power supply voltage P0 and the output voltage to a plurality of power supply domains are leaves. The tree structure has a plurality of voltage converters that generate one or more branches (single output voltage) from one branch (single input voltage). Each voltage converter is a voltage converter that converts either the input power supply voltage or the output voltage into one of the output voltages. Therefore, the voltage converter is arranged at the position of the via voltage composed of the output voltage.

  In the solution improvement step S2, the solution narrowing step S1 is performed on the input tree structure before the voltage converter that outputs an arbitrary voltage for the candidate tree narrowed down in the solution narrowing step S1, and the input side having a better cost function Explore the tree structure. Further, the above-mentioned arbitrary voltage is sequentially shifted to the input side until the input power supply voltage is reached, and the improvement of the input side tree structure based on that is repeated. As a result, the tree 24 having the best cost function is detected.

  In the design method of FIG. 6, the number of candidate trees 22 does not necessarily have to be one. A plurality may be used as long as the number is relatively small. For a plurality of candidate trees 22, it is desirable to improve the tree structure on the input side in the solution improvement step S2 and search for a tree with the best cost function.

[Solution narrowing step S1]
FIG. 7 is a flowchart of the solution narrowing-down step S1 of the design method according to the present embodiment. In the via voltage table creation step S1-1, a series of via voltages generated by voltage conversion by the DC-DC converter from the input voltage to the output voltage of the multi-output power supply circuit is generated for each power supply domain using random numbers. A specific method will be described later in detail. Then, in the tree generation step S1-2, a multi-output power supply circuit that satisfies the via voltage column of the via voltage table created in the via voltage table creation step S1-1 is generated using random numbers. Again, the specific method will be described in detail later. Generating using random numbers in both steps S1-1 and S1-2 means generating a tree at random.

  Next, in the cost function value calculation step S1-3, the cost function value of the multi-output power supply circuit using the tree generated in the tree generation step S1-2 is calculated. The cost function value will be described later in detail. For example, the cost function value is power efficiency, or a value obtained by adding the factor of the number of voltage converters to the power efficiency. In the best tree update step S1-4, the tree for the multi-output power supply circuit with the better cost function value is replaced with the best tree. Then, the above steps S1-1 to S1-4 are repeated a certain number of times, for example, 50 to 100 times when the number of power supply domains is 10 (S1-5).

[Via-voltage table creation step S1-1]
FIG. 8 is a diagram illustrating an example of the transit voltage table in the transit voltage table creation step S1-1. FIG. 9 is a flowchart of the via voltage table creation process. With reference to the via voltage table tbl [v, d] in FIG. 8, the via voltage table creation step S1-1 will be described along the flowchart in FIG.

  In the via voltage generation step S1-1, as shown in FIG. 8, from the input / output data 10 of the multi-output power supply circuit and the efficiency characteristic data 12 of the voltage converter (see FIG. 6), the via voltage table tbl [v, d ] Is created. Specifically, in the table creation step S1-1-1 of FIG. 9, each power domain name (A to E) of the multi-output power supply circuit and its voltage (3.3V) are arranged vertically in the table tbl [v, d]. , 2.5V, 2.5V, 2.5V, 1.2V), and the input voltage and output voltage of each converter are extracted from the efficiency characteristic table 12 of the DC-DC converter and written side by side. The lateral voltages (5V, 3.3V, 2.5V, 1.2V) in the table are arranged in descending order from left to right.

  Next, in the definite cell filling step S1-1-2, all the circles are written in the input voltage (5.5V) column of the multi-output power supply circuit (see 30 in FIG. 8). This is because the branch of the input voltage always exists. Further, for each row, a circle is written in the same via voltage column as the output voltage to each power supply domain A to E of the power supply destination of the multi-output power supply circuit (see 32 in FIG. 8). This is because there is always this via voltage. Then, “x” is written in the column of the via voltage less than the output voltage to each of the power domains A to E (see 34 in FIG. 8). This is because there is no need for this via voltage.

  Finally, in the random selection step S1-1-3, ◯ or X is written in the remaining cells using random numbers (see 36 in FIG. 8). In other words, ○ or × was written in the cell of the via voltage that was determined in the above process. Then, since the via voltage corresponding to the remaining cells may or may not exist, ○ or X is randomly written. As a result, the via voltage table tbl [v, d] shown in FIG. 8 is generated.

[Tree generation step S1-2]
FIG. 10 is a diagram for explaining the tree generation process. FIG. 10 shows the same via voltage table tbl [v, d] as in FIG. In this via voltage table tbl [v, d], the locations where the input voltage and the output voltage of the voltage converter constituting the multi-output power supply circuit exist are indicated by ◯. According to this table, 3.3 V of the power domain A is generated by a power converter that outputs 3.3 V from the input voltage 5 V, and 2.5 V of the power domain B is obtained from the input voltage 5 V to the transit voltage 3.3 V. It is generated by a power converter that outputs and a power converter that outputs 2.5V. The same applies to the other power domains C, D, and E.

  In this way, the via voltage where the circle exists except for the input voltage 5V of the multi-output power supply circuit needs to be generated by some voltage converter. When a plurality of circles exist in the same via voltage sequence, there are a plurality of options for the voltage converter that generates the same via voltage. For example, in the column of the via voltage 3.3V, there are three circles as indicated by 40 in the figure. The three groups 40 of the same via voltage can be generated by a single voltage converter that converts from 5 V to 3.3 V, or can be generated by two voltage converters or three voltage converters. Therefore, in order to generate a voltage converter that is necessary at random, it is only necessary to group the three groups of circles 40 at random.

  However, there are four circles in the column of the via voltage 2.5V. However, the group 44 is generated by a voltage converter having an input voltage of 5V and an output voltage of 2.5V, and the group 42 is generated by a voltage converter having an input voltage of 3.3V and an output voltage of 2.5V. Therefore, the groups 42 and 44 must be divided into different groups and each grouped randomly. Further, since the voltage converter converts a single input voltage into a single output voltage, when the three circles of the group 40 generating the input voltage are divided into different groups in the group 42, the output side Group 42 must also be divided separately.

  An example of generating a tree from the via voltage table tbl [v, d] in FIG. 10 will be described as follows. First, it is assumed that the group 40 is grouped into a circle belonging to the power domains A and B and a circle belonging to the power domain C by random grouping. As a result, as shown in the voltage converter group 40C in FIG. 10, the voltage converter C1 for the power domains A and B and the voltage converter C2 for the power domain C are generated. Next, considering the random generation of the group 42, the group 40 is grouped into two groups. Therefore, the voltage converter C3 for the power domain B is provided on the output side of the voltage converter C1, and the output side of the voltage converter C2 is provided on the output side. A voltage converter C4 for the power domain C must be generated respectively. In this case, the voltage converter group 42C has no room for randomly generating voltage converters. Further, considering random occurrence of the group 44, it is assumed that the circles belonging to the power domain D and the circles belonging to the power domain E are grouped into one group. As a result, as shown in the voltage converter group 44C, the voltage converter C5 for the power supply domains D and E is generated. Finally, the voltage converter C6 that generates the via voltage 1.2V has no room for random generation, and the voltage converter C6 is generated at the output of the voltage converter C5.

  As described above, among the plurality of relay voltages, the group of circles in the column of the same relay voltage is randomly grouped from the input voltage 5V side toward the output voltage 3.3V to 1.2V, and the grouped position is obtained. If a voltage converter is generated, a random tree can be generated. However, it is necessary to determine whether or not random generation is possible based on the regulation that the voltage converter generates a single output voltage from a single input voltage.

  In the multi-output power supply circuit shown in FIG. 10, if the voltage converter group 40C is grouped into a single voltage converter, the output-side voltage converter group 42C is allowed to be randomly grouped. In this way, it is affected by the grouping of voltage converters on the input side.

  11 and 12 are flowcharts of the tree generation step S1-2. FIGS. 11 and 12 are the same flowcharts, and descriptions of each step of the flowcharts are shown on the right side of each. FIG. 11 shows the first half, and FIG. 12 shows the second half.

  FIG. 13 is a diagram showing the progress of a tree generated by the tree generation step S1-2. 11 and 12 will be briefly described, and then a specific example will be described with reference to FIG.

  The flowcharts of FIGS. 11 and 12 show a processing flow for randomly generating the tree 46 from the via voltage table tbl [v, d] of FIG. An outline will be described along the steps in the figure.

  In step S1-2-1, work queues Qv, Qd, Qd_tmp and an initial tree are created.

  (1) The voltage work queue Qv is obtained by arranging the voltages from which the input voltage is removed among the voltages written in the horizontal direction of the via voltage table created in step S1-1, in descending order.

  (2) The power domain work queue Qd is a queue that stores only one set in which power domain names are written and arranged.

  (3) The power domain temporary queue Qd_tmp is a queue in which nothing is stored.

  (4) The initial tree is a tree in which leaf nodes having the power domain name of each power domain as a label have a common parent node.

  In step S1-2-2, one element is extracted from Qv and set to V. That is, the following processing is performed by sequentially extracting the via voltage from the input voltage side.

  In step S1-2-3, one element is extracted from Qd and set as D. Qd has the property that when the tree branches from the input voltage side, the branched branches do not merge again and branch further, so the set that can be further branched is an element.

  In step S1-2-4, the set D is divided into two. That is, the set D is divided according to XX in the column of the transit voltage V of the transit voltage table created in step S1-1. At this time, a set that collects the people who are ○ is Do, and a set that is × is Dx. Therefore, Do corresponds to the groups 40, 42, and 44 in FIG.

  In step S1-2-5, grouping is generated. That is, one set (grouping) obtained by dividing the set Do into subsets without leakage overlap is generated using random numbers. Let G (Do) be the set of subsets thus divided. This process corresponds to the random grouping process described above. This set G (Do) becomes an element of the power domain work queue Qd in the next cycle.

  In step S1-2-6, tree change (node insertion) is performed. That is, a parent node common to each element of the set G (Do) is inserted immediately above the leaf node. A voltage converter is arranged at the inserted parent node.

  In step S1-2-7, elements of set G (Do) and Dx are added to Qd_tmp.

  In step S1-2-8, if the work queue Qd is not empty, the process proceeds to step S1-2-3.

  In step S1-2-9, Qd: = Qd_tmp and Qd_tmp: = [] are updated.

  In step S1-2-10, if the work queue Qv is not empty, the process proceeds to step S1-2-2.

  By updating step S1-2-9, Qd: = Qd_tmp is set, and grouping is performed for each element of Qd in the cycle of the next flowchart. That is, the parent node is generated based on the grouping, and the branch of the tree is branched. Therefore, whether or not to branch further is selected by performing random grouping in the branched power supply node group.

[Specific example of tree generation]
Next, a specific example of FIG. 10 will be described along the above flowchart with reference to FIG.

  In step S1-2-1, work queues Qv, Qd, Qd_tmp and an initial tree are created. According to the specific example of FIG. 10, the work queue Qd: = [{A, B, C, D, E}], Qv: = [3.3V, 2.5V, 1.2V], Qd_tmp: = [], initial The tree becomes (a) in FIG.

[V = 3.3V cycle]
In step S1-2-2, one element is extracted from the voltage queue Qv and is set to V: = 3.3V. As a result, Qv: = [2.5V, 1.2V].

  In step S1-2-3, one element is extracted from the power domain Qd and is set to D: = {A, B, C, D, E}. As a result, Qd: = [].

  In step S1-2-4, among D = {A, B, C, D, E}, the columns in the via voltage V = 3.3V column of the via voltage table in FIG. , C, and x are power domains D and E, and are divided into a set of Do = {A, B, C} and Dx = {D, E}.

  In step S1-2-5, if the set Do = {A, B, C} is randomly divided into subsets without omission or duplication, the set of divided subsets is, for example, G (Do) = {{A , B}, {C}}.

  In step S1-2-6, as shown in FIG. 13B, a common parent node C1 is inserted into leaf nodes A and B, and a parent node C2 is inserted into leaf node C. That is, this corresponds to the fact that the group 40 in FIG. 10 is divided and the voltage converters C1 and C2 are generated.

  In step S1-2-7, each element of G (Do) and Dx are added to Qd_tmp, and Qd_tmp: = [{A, B}, {C}, {D, E}].

  In step S1-2-8, since Qd is empty, the process proceeds to step S1-2-9. In step S1-2-9, Qd: = Qd_tmp = [{A, B}, {C}, {D, E}] and Qd_tmp: = [] are updated. Furthermore, since the work queue Qv is not empty in step S1-2-10, the process proceeds to step S1-2-2. Qd = [{A, B}, {C}, {D, E}] is divided into parent nodes C1 and C2 in FIG. 13B, and the power domains are {A, B}, {C}, { D, E}. Hence, subsequent groupings must be made within these elements.

[V = 2.5V cycle]
In step S1-2-2, one element is taken out from the voltage queue Qv, V: = 2.5V, and Qv: = [1.2V].

  In step S1-2-3, one element is extracted from the power domain Qd, and D: = {A, B}, and Qd: = [{C}, {D, E}].

  In step S1-2-4, among D = {A, B}, the V = 2.5V column in the via voltage table is ○ for power domain B, and × is for power domain A. Do: = {B}, Dx: = {A}.

  In step S1-2-5, the grouping of Do = {B} is only G (Do) = {{B}}. In other words, there is no room for random occurrence.

  In step S1-2-6, the parent node C3 is inserted immediately above the leaf node B as shown in FIG.

  In step S1-2-7, each element of G (Do) and Dx are added to Qd_tmp, and Qd_tmp: = [{B}, {A}]. In step S1-2-8, since Qd is not empty, the process proceeds to step S1-2-3.

  Next, in step S1-2-3), one element is extracted from Qd and D: = {C}, and Qd: = [{D, E}].

  In step S1-2-4, among D = {C}, there is no C in the column of V = 2.5V in the via voltage table, and there is no x, so Do: = {C}, Dx: = {}

  In step S1-2-5, the grouping of Do = {C} is only G (Do) = {{C}}. In other words, random grouping is not possible.

  In step S1-2-6, the parent node C4 is inserted immediately above the leaf node C as shown in FIG.

  In step S1-2-7, each element of G (Do) and Dx are added to Qd_tmp, and Qd_tmp: = [{B}, {A}, {C}, {}]. In step S1-2-8, since Qd is not empty, the process proceeds to step S1-2-3.

  Next, in step S1-2-3, one element is extracted from Qd, and D: = {D, E}, and Qd: = [].

  In step S1-2-4, among D = {D, E}, there is no D, E, x in the column of V = 2.5V in the via voltage table, so that Do = { D, E}, Dx = {}.

  Assume that G (Do) = {{D, E}} is selected by random grouping of Do = {D, E} in step S1-2-5. That is, the two circles of the group 44 are grouped into one set.

  In step S1-2-6, a common parent node C5 is inserted immediately above the leaf nodes D and E as shown in FIG.

  In step S1-2-7, each element of G (Do) and Dx are added to Qd_tmp, and Qd_tmp: = [{B}, {A}, {C}, {}, {D, E}, { }].

  In step S1-2-8, since Qd is empty, the process proceeds to step S1-2-9, and Qd: = Qd_tmp = [{B}, {A}, {C}, {}, {D, E}, {} ], Qd_tmp: = [] is updated. In step S1-2-10, since the work queue Qv is not empty, the process proceeds to step S1-2-2. Qd = [{B}, {A}, {C}, {}, {D, E}, {}] is divided into {A}, { B}, {C}, {D, E}. Hence, subsequent groupings must be made within these elements.

[V = 1.2V cycle]
In the following, for V = 1.2V, the explanation will be simplified by paying attention to whether a tree update occurs.

  In step S1-2-2, one element is extracted from Qv, and V: = 1.2V and Qv: = [].

  In step S1-2-3, one element is extracted from Qd and becomes D: = {B}, Qd: = [{A}, {C}, {}, {D, E}, {}] .

  In process S1-2-4 to S1-2-8, B in row of V = 1.2V in the via voltage table is x. So Do: = {} so there is no tree update. At this time, G (Do) = {} and Dx = {B}. Since Qd is not empty, go to S1-2-3.

  Next, in step S1-2-3, one element is extracted from Qd and becomes D: = {A}, Qd: = [{C}, {}, {D, E}, {}].

  In steps S1-2-4 to S1-2-8, A in the row of V = 1.2V in the via voltage table is x. So Do: = {} so there is no tree update. At this time, G (Do) = {} and Dx = {A}. Since Qd is not empty, go to S1-2-3.

  Next, in step S1-2-3, one element is extracted from Qd and becomes D: = {C}, Qd: = [{}, {D, E}, {}].

  In steps S1-2-4 to S1-2-8, C in the column of V = 1.2V in the via voltage table is x. So Do: = {} so there is no tree update. At this time, G (Do) = {} and Dx = {C}. Since Qd is not empty, go to S1-2-3.

  Next, in step S1-2-3, one element is extracted from Qd, and D: = {}, Qd: = [{D, E}, {}].

  In steps S1-2-4 to S1-2-8, Do: = {} from D: = {}, so the tree is not updated. At this time, G (Do) = {} and Dx = {}. Since Qd is not empty, go to S1-2-3.

  Next, in step S1-2-3, one element is extracted from Qd, and D: = {D, E}, Qd: = [{}].

  In step S1-2-4, out of D: = {D, E}, the V = 1.2V column in the via voltage table is E, and X is D, so Do: = {E }, Dx: = {D}.

  In step S1-2-5, since the grouping is only {E}, G (Do) = {{E}}.

  In step S1-2-6, the parent node C6 is inserted immediately above the leaf node E as shown in FIG.

  In step S1-2-7, each element of G (Do) and Dx are added to Qd_tmp, and Qd_tmp: = [{}, {B}, {}, {A}, {}, {C}, { }, {}, {E}, {D}]. In step S1-2-8, since Qd is not empty, the process proceeds to S1-2-3.

  Finally, in step S1-2-3, one element is extracted from Qd, and D: = {} and Qd: = [] are obtained.

  In steps S1-2-4 to S1-2-7, Do: = {} from D: = {}, so that no tree update occurs.

  In step S1-2-8, since Qd is empty, the process proceeds to S1-2-9. In step S1-2-9, Qd: = Qd_tmp = [{}, {B}, {}, {A}, {}, {C}, {}, {}, {E}, {D} , {}, {}] And Qd_tmp: = [].

  In step S1-2-10, because the work queue Qv is empty, the tree generation step S1-2 ends. As a result, the tree in FIG. 13F is completed. This tree is the same as the tree 46 in FIG.

  As described above, the specific processing flow based on the algorithm for generating the parent node randomly described with reference to FIG. Then, random grouping is performed in FIGS. 13B and 13E.

[Cost function value calculation step S1-3]
Returning to FIG. 7, the cost function value of the multi-output power supply circuit corresponding to the tree randomly generated in the tree generation step S1-2 is calculated (S1-3). The first example of the cost function value is the power efficiency or power consumption of the multi-output power supply circuit. Higher power efficiency is better, and lower power consumption is better. Further, the second example of the cost function is a value obtained by subtracting a predetermined penalty corresponding to the number of voltage converters from the power efficiency of the multi-output power supply circuit. This is a value obtained by adding a predetermined penalty corresponding to the number of voltage converters to the power consumption of the multi-output power supply circuit. The second example is a cost function value taking into account the degree of integration of the multi-output power supply circuit.

  A method for obtaining the cost function value (power efficiency or power consumption of the multi-output power supply circuit) of the first example will be described. First, for the voltage converter that outputs the output voltage of the multi-output power supply circuit, the input power and the input current are obtained from the output voltage, the output current, and the efficiency of the efficiency characteristic data 12, and the input power and input of the voltage converter on the input side are obtained. Repeat the calculation to find the current until it reaches the input of the multi-output power circuit. Finally, the power efficiency of the multi-output power supply circuit is required from the ratio of the input power due to the input voltage and input current of the multi-output power supply circuit and the sum of the output power of the multi-output power supply circuit.

  FIG. 14 is a diagram illustrating a specific example of the cost function value calculation step. With reference to FIG. 14, how to obtain the power efficiency of the multi-output power supply circuit will be described. First, the voltage and current required by the power domain correspond to the output voltage and output current of the voltage converter on the output terminal side that generates the output voltage, and their product corresponds to the power of the voltage converter.

The multi-output power supply circuit PSC1 on the left side of FIG. 14 includes voltage converters C7 and C8. The power of each voltage converter C7, C8 is the product of the respective output voltage and output current, and is 3.3W and 2W. Therefore, the input power and input current of the voltage converters C7 and C8 can be obtained as follows.
(Input power) = (Output voltage) x {Σ (Output current)} / (Efficiency) (1)
(Input current) = (Input power) / (Input voltage) (2)
Here, Σ (output current) is the sum of the output current of the voltage converter. The efficiency is obtained from the efficiency characteristic data 12. The voltage converters C7 and C8 have only one output current.

As a result, both voltage converters C7 and C8 are required to have an input power of 3.5 W and 2.6 W and an input current of 710 mA and 520 mA, respectively. The sum of these input currents becomes the input current of the multi-output power supply circuit. Finally, the power efficiency of a multi-output power supply circuit is calculated as follows.
(Efficiency of multi-output power supply circuit) = Σ {(output voltage) × (output current)} / (input voltage) × (input current) (3)
Here, Σ {(output voltage) × (output current)} is the sum of {(output voltage) × (output current)} for each power domain.

  That is, according to the example of the multi-output power supply circuit PSC1 on the left side of FIG. 14, the efficiency is 86% as illustrated. The power consumption of the multi-output power supply circuit is found by subtracting the sum of the output power from the input power, and is 0.8 W.

  Similarly, the example of the multi-output power supply circuit PSC2 on the right side of FIG. 14 includes voltage converters C7 and C9. The input voltage, output voltage, and output current of the multi-output power supply circuit are the same as in the left example. Since the power of the voltage converter C9 is 2 W and the efficiency is 85%, the input power is 2.3 W and the input current is 710 mA. Therefore, the output current of the voltage converter C7 is 1710 mA which is the sum of 1 A and 710 mA, and the output voltage is 3.3V. Therefore, since the efficiency of the voltage converter C7 is 94%, its input power becomes 6.0W. Finally, the efficiency of the multi-output power supply circuit is 88% as shown. The power consumption of the multi-output power supply circuit is 0.7W. That is, the multi-output power supply circuit on the right side of FIG. 14 consumes less power and has better power efficiency.

Next, a method for obtaining the cost function value of the second example will be described. The cost function value is obtained as follows.
(Cost function value) = (Efficiency of multi-output power supply circuit)-(Constant) x (Number of voltage converters in multi-output power supply circuit)
Here, the constant is a parameter calculated depending on whether it is allowed to increase one voltage converter with an efficiency improvement of (predetermined%) or more. For example, (constant) = (predetermined%). Therefore, depending on this parameter value, if the number of voltage converters is large even if the efficiency of the multi-output power supply circuit is good, the cost function value becomes worse. That is, in this cost function value, the reduction efficiency corresponding to the number of voltage converters is added to the actual power efficiency. If the best tree is selected by this cost function value, it is possible to perform selection considering the circuit scale of the multi-output power supply circuit.

  Returning to FIG. 7, in the best tree update step S1-4, if the cost function value of the generated tree exceeds the best solution so far, the tree is replaced with the best solution.

[Repetition Count Determination Step S1-5]
FIG. 15 is a diagram for explaining the iteration number determination step S1-5 in FIG. The table shown in FIG. 15 shows the probability of finding the most efficient tree and the solution with the efficiency within 3% corresponding to the number of iterations of steps S1-1 to S1-4 in FIG. This is the same as described in FIG. According to this, the probability can be increased by increasing the number of iterations. The probability is 99% at 50 iterations, and the probability is 99.99% at 100 iterations. Therefore, in the present embodiment, in the iteration number determination step, the solution narrowing-down step S1 ends when the iteration number reaches 50 times or 100 times.

[Solution improvement process S2]
Thus, the solution narrowing step S1 of FIG. 6 is completed. Next, the improvement process S2 of the tree multi-output power supply circuit, which is a narrowed solution, will be described.

  FIG. 16 is a flowchart of the solution improvement step S2. In the solution improvement step S2, the solution narrowing step S1 is performed on the input side tree before a certain arbitrary via voltage among the multi-output power supply circuits corresponding to the tree narrowed down in the solution narrowing step S1, and the input side tree To improve. For example, it is desirable to change the arbitrary transit voltage from the transit voltage closest to the output voltage to the transit voltage on the input side sequentially, and it is desirable to execute the solution narrowing-down process S1 on each input side tree. Since the tree on the input side before any via voltage has a smaller number of squares in the via voltage table than the whole tree, the solution narrowing-down process S1 itself has less man-hours, and the number of trees that are randomly generated is also small. can do. Therefore, in the improvement step S2, it is possible to search for a multi-output power supply circuit with an improved cost function value with less man-hours.

  According to the flowchart of FIG. 16, in the solution improvement initial setting step S2-0, the routed voltage on the input side is extracted from the routed voltages in the routed voltage table to form a list Q set. In the example of FIG. 16, via voltages 3.3V and 2.5V are extracted and are elements of a list Q list. Next, in the tree cutting step S2-1, the voltage closest to the output side is extracted from the list Q and is set as the cutting voltage Vcut (S2-1-1), and the input side before the cutting voltage Vcut is cut to the output side from Vcut. (S2-1-2). Then, the via voltage table tbl2 is generated again for the input side (S1-1). In other words, a circle in the via voltage table is randomly generated in the square excluding the input voltage and output voltage portions.

  FIG. 17 is a diagram illustrating a first example of the tree cutting process S2-1-2 in the solution improving process. In this example, the cutting voltage Vcut = 2.5V. In the tree narrowed down in the solution narrowing step S1, the parent nodes C1, C2 output 3.3V, the parent nodes C3, C4, C5 output 2.5V, and the parent node C6 outputs 1.2V. Therefore, when the cutting voltage Vcut = 2.5V, the input side tree before the parent nodes C3, C4, and C5 becomes the re-search part 50, and the output side tree becomes the fixed part 52. Then, a corrected via voltage table tbl2 is generated by generating XX randomly for the tree of the re-search part 50. However, in the example of FIG. 17, XX is not changed. The disconnected portion of the output voltage 2.5V is named power domain B ′, C ′, D ′, E ′, and each output voltage is reflected in the via voltage table tbl2 as 2.5V.

  Then, based on the corrected via voltage table tbl2, the above-described solution narrowing step S1 (S1-1 to S1-5) is executed again. That is, the three circles in the column of the next via voltage 3.3V with the input voltage 5V are randomly grouped, and the circle of the column of the next via voltage 2.5V is randomly grouped for each grouping. Then, a tree with a further improved cost function value is searched.

  In the end determination step S2-6, it is determined whether or not the element of the list Q is empty. If it is not empty, the next element Vcut = 3.3V is extracted in the cutting voltage determination step S2-1-1, and the tree cutting process step S2- In 1-2, a new search tree is cut.

  FIG. 18 is a diagram illustrating a second example of the tree cutting processing step S2-1-2 in the solution improving step. In this example, the cutting voltage Vcut = 3.3V. The cutting method is as described in FIG. In this second example, the tree on the input side before the cutting voltage Vcut = 3.3V is the re-search part 54. Based on the corrected via voltage table tbl3 for the re-search portion 54, the solution narrowing-down process S1 (S1-1 to S1-5) is executed again. In this process, the three circles in the column of the next via voltage 3.3V after the input voltage 5V are randomly grouped. Then, a tree with a further improved cost function value is searched.

  Finally, in the end determination step S2-6, it is confirmed that the elements of the list Q are empty, and the solution improvement step S2 ends.

  FIG. 19 is a diagram illustrating an example of improving the power efficiency by the solution improving process. This improved example is an example of the cutting voltage Vcut = 2.5V in FIG. The input side tree before the cutting voltage Vcut = 2.5V becomes the re-search part 50, and the output side tree becomes the fixed part 52. In the figure, (A) shows each power efficiency before improvement, and (B) shows the power efficiency after improvement. Before the improvement of (A), the efficiency of the re-search portion 50 is 80%, the efficiency of the fixed portion 52 is 90%, and the overall efficiency is 72%. On the other hand, after the improvement of (B), the efficiency of the re-search portion 50 is improved to 85%, and the overall efficiency is improved to 76.5% even though the efficiency of the fixed portion 52 remains 90%. .

  In the cost function value calculation step S1-3 of the solution narrowing-down step S1, the output voltage and output current are set to voltages and currents required in each power domain, and the output of each voltage converter from the output side voltage converter toward the input side. Calculations were made to determine the input current from the voltage, output current and efficiency. In the solution improving step S2, the tree on the output side from an arbitrary cutting voltage is used as a fixed part. Therefore, the cost function value calculation result performed first can be reused in the fixed part, and there is no need to calculate again. . Therefore, in the improvement step S2, since the input side before the cutting voltage is used as a re-search portion, the improvement is performed, and therefore the man-hour can be reduced even in the cost function value calculation step.

FIG. 20 is a diagram illustrating the effect of the present embodiment. A comparison between the comparative example of FIG. 4 and the present embodiment of FIG. 6 is shown. In the case of the comparative example shown in FIG. 4, since all possible trees are generated and the best cost function value tree is selected, an optimal solution can be detected, but the amount of calculation is enormous. On the other hand, in the present embodiment shown in FIG. 6, a predetermined number of trees are generated, the best cost function value tree is selected, and a part of the input side is searched again. The cost function value of the circuit is only within a certain range as the optimal solution, but the calculation amount can be greatly reduced. For example, if there are 10 power domains, and the multi-output power circuit has a three-stage configuration, the number of trees considered in the comparative example of FIG. 4 is 2.11 × 10 ⁹ , whereas the embodiment of FIG. Then it can be reduced to several hundred.

  The multi-output power supply circuit design method described above can be realized by a multi-output power supply circuit design apparatus in which a multi-output power supply circuit design program is installed in a general-purpose computer. Such a multi-output power supply circuit designing apparatus is shown in the flowcharts shown in FIGS. 6, 7, 9, 11, and 16 together with input data, data Qd, Qv, Qdtmp being processed, tree data, input / output voltages, and current values. It comprises memory means for storing a design program based on it and a CPU for performing calculations.

  According to the above embodiment, it is possible to design a multi-output power supply circuit having power efficiency or cost function value that can be practically satisfied with less man-hours.

  The above embodiment is summarized as follows.

(Appendix 1)
In designing a multi-output power supply circuit that supplies output power to multiple power domains,
A plurality of voltage converters, each of which defines an input power source as a root and a plurality of output power sources as leaves, and uses any one of voltages from the input power source voltage to the output power source voltage as an input voltage and an output voltage; A tree generation step for randomly generating a tree structure between a root and a leaf, and a power function of the multi-output power supply circuit having the randomly generated tree structure or a cost function value including the power, the plurality of power supply domains The cost function value generation process that is calculated based on the required voltage and required current of each and the efficiency of each voltage converter is repeated a smaller number of times than the number of all possible tree structures, and the tree structure having the best cost function value is obtained. A search process to search;
For the input side tree structure having a plurality of voltage converters on the input side before the voltage converter that outputs an arbitrary output voltage among the tree structures found in the search step, the search step is performed again, and is better than that detected in the search A design method of a multi-output power supply circuit comprising an improvement step of replacing the input side tree structure having a cost function value.

(Appendix 2)
In Appendix 1,
A design method for a multi-output power supply circuit, characterized in that the improvement step is repeated by sequentially moving the arbitrary output voltage to the input side.

(Appendix 3)
In Appendix 1,
In the cost function value generation step, an input side that supplies an input voltage to the output node voltage converter based on the required voltage and current of the plurality of power supply domains and the efficiency of the output node voltage converter that generates the output power supply The output current of the voltage converter is calculated, and the calculation for further obtaining the output current on the input side based on the calculated output current and the output voltage and efficiency of the voltage converter is repeated until the input power supply, and the power consumption of the output power supply And calculating the power efficiency of the multi-output power supply circuit based on the ratio of the sum of the power and the power consumption of the input power supply, or the power of the multi-output power supply circuit based on the power consumption of the input power supply. Power circuit design method.

(Appendix 4)
In Appendix 3,
The cost function value is a value obtained by adding a reduction efficiency based on the number of the voltage converters to the power efficiency, and a design method for a multi-output power supply circuit.

(Appendix 5)
In Appendix 1,
In the tree generation step, a plurality of output node voltage converters that respectively output the plurality of output power supplies, a plurality of relay node voltage converters that are randomly inserted between the plurality of output node voltage converters and the input power supply, A method for designing a multi-output power supply circuit, comprising: generating a tree structure having

(Appendix 6)
In Appendix 5,
In the tree generation step, a relay node voltage converter between a plurality of output node voltage converters that respectively output a plurality of output power supplies and a single input power supply is randomly generated based on random numbers. Design method of output power circuit.

(Appendix 7)
In a multi-output power circuit design device that supplies output power to multiple power domains,
A plurality of voltage converters, each of which defines an input power source as a root and a plurality of output power sources as leaves, and uses any one of voltages from the input power source voltage to the output power source voltage as an input voltage and an output voltage; A tree generation for randomly generating a tree structure between a root and a leaf, and a cost function value including the power efficiency or power of the multi-output power supply circuit having the randomly generated tree structure are calculated for the plurality of power supply domains. Search for a tree structure having the best cost function value by repeating the cost function value generation calculated based on the required voltage and current, and the efficiency of each voltage converter, less than the number of all possible tree structures. Search means;
For the input side tree structure having a plurality of voltage converters on the input side before the voltage converter that outputs an arbitrary output voltage among the tree structures found in the search step, the search step is performed again, and is better than that detected in the search An apparatus for designing a multi-output power supply circuit, comprising: improvement means for replacing an input-side tree structure having a cost function value.

(Appendix 8)
In Appendix 7,
The design device for a multi-output power supply circuit, wherein the improvement means sequentially moves the arbitrary output voltage to the input side and repeats the improvement.

It is a figure which shows an example of a multi-output power supply circuit. It is a figure shown about the definition of a multi-output power supply circuit. It is a circuit diagram which shows an example of a voltage converter. It is a flowchart figure which shows an example of the design method of a multiple output power supply circuit. It is a figure which shows the experimental result by this inventor. It is a flowchart figure which shows the design method of the multiple output power supply circuit in this Embodiment. It is a flowchart figure of solution narrowing-down process S1 of the design method in this Embodiment. It is a figure which shows an example of the via voltage table by via voltage table preparation process S1-1. It is a flowchart figure of a via voltage table preparation process. It is a figure explaining a tree production | generation process. It is a flowchart figure of tree production | generation process S1-2. It is a flowchart figure of tree production | generation process S1-2. It is a figure which shows the middle progress of the tree produced | generated by tree production | generation process S1-2. It is a figure which shows the specific example of a cost function value calculation process. FIG. 8 is a diagram for explaining an iteration number determination step S1-5 in FIG. It is a flowchart figure of solution improvement process S2. It is a figure which shows the 1st example of the tree cutting process S2-1-2 in a solution improvement process. It is a figure which shows the 2nd example of the tree cutting process S2-1-2 in a solution improvement process. It is a figure which shows the example of an improvement of the power efficiency by a solution improvement process. It is a figure which shows the effect of this Embodiment.

Explanation of symbols

S1: Search process (solution refinement process)
S2: Solution improvement step 20: Multi-output power supply circuit with generated tree structure

Claims (5)

In designing a multi-output power supply circuit that supplies output power to multiple power domains,
A plurality of voltage converters, each of which defines an input power source as a root and a plurality of output power sources as leaves, and uses any one of voltages from the input power source voltage to the output power source voltage as an input voltage and an output voltage; A tree generation step for randomly generating a tree structure between a root and a leaf, and a power function of the multi-output power supply circuit having the randomly generated tree structure or a cost function value including the power, the plurality of power supply domains The cost function value generation process that is calculated based on the required voltage and required current of each and the efficiency of each voltage converter is repeated a smaller number of times than the number of all possible tree structures, and the tree structure having the best cost function value is obtained. A search process to search;
For the input side tree structure having a plurality of voltage converters on the input side before the voltage converter that outputs an arbitrary output voltage among the tree structures found in the search step, the search step is performed again, and is better than that detected in the search A design method of a multi-output power supply circuit comprising an improvement step of replacing the input side tree structure having a cost function value. In claim 1,
A design method for a multi-output power supply circuit, characterized in that the improvement step is repeated by sequentially moving the arbitrary output voltage to the input side. In claim 1,
In the cost function value generation step, an input side that supplies an input voltage to the output node voltage converter based on the required voltage and current of the plurality of power supply domains and the efficiency of the output node voltage converter that generates the output power supply The output current of the voltage converter is calculated, and the calculation for further obtaining the output current on the input side based on the calculated output current and the output voltage and efficiency of the voltage converter is repeated until the input power supply, and the power consumption of the output power supply A multi-output power supply circuit, wherein the power efficiency of the multi-output power supply circuit is obtained based on the ratio of the sum of the power and the power consumption of the input power supply or the power of the multi-output power supply circuit is obtained from the power consumption of the input power Design method. In claim 1,
In the tree generation step, a plurality of output node voltage converters that respectively output the plurality of output power supplies, a plurality of relay node voltage converters that are randomly inserted between the plurality of output node voltage converters and the input power supply, A method for designing a multi-output power supply circuit, comprising: generating a tree structure having In a multi-output power circuit design device that supplies output power to multiple power domains,
A plurality of voltage converters, each of which defines an input power source as a root and a plurality of output power sources as leaves, and uses any one of voltages from the input power source voltage to the output power source voltage as an input voltage and an output voltage; A tree generation for randomly generating a tree structure between a root and a leaf, and a cost function value including the power efficiency or power of the multi-output power supply circuit having the randomly generated tree structure are calculated for the plurality of power supply domains. Search for a tree structure having the best cost function value by repeating the cost function value generation calculated based on the required voltage and current, and the efficiency of each voltage converter, less than the number of all possible tree structures. Search means;
For the input side tree structure having a plurality of voltage converters on the input side before the voltage converter that outputs an arbitrary output voltage among the tree structures found in the search step, the search step is performed again, and is better than that detected in the search An apparatus for designing a multi-output power supply circuit, comprising: improvement means for replacing an input-side tree structure having a cost function value.