OUTPUT CURRENTAND INPUTPOWERREGULATIONWITHAPOWER

CONVERTER

Dongsheng Zhou

Kent Kernahan

Milton D. Ribeiro

CROSS REFERENCE TO RELATED APPLICATIONS

[0001] The present application claims priority of U.S. provisional application no. 60/641,919, filed on January 5, 2005. The provisional application is incorporated herein by reference in its entirety, including its computer program listing appendix.

[0002] This application is related to U.S. Patent Application Serial No. 10/295,739 filed on November 14, 2002, by Kent Kernahan and John Carl Thomas, entitled "Switching Power Converter", now U.S. Patent No. 6,825,644 issued November 30, 2004 which is incorporated herein by reference.

BACKGROUND OF THE INVENTION

[0003] This invention relates to switching power supplies or converters. Some embodiments use a switching power converter in a constant current or constant power mode.

[0004] Efficient and effective backlighting of displays is important for personal digital assistant computers (PDAs) and other portable computers, cell phones, cordless phones, handheld game devices, and the like. White light emitting diodes (LEDs) are popular for backlighting, but must be white as perceived by the human eye to enable color graphics, such as pictures, to be the correct color. LEDs of other

colors may also require a certain power input for a specified color. In addition, many products require the ability to dim the display in response to ambient light, product mode (e.g. sleep, run, play a movie, etc.), user preference and the like. In some applications multiple LEDs are needed to uniformly illuminate a large display.

[0005] The color of some LEDs is affected by the current with which they are driven. For example, white LEDs actually emit a mix of blue and yellow wavelengths which human eyes perceive as white. The mix of the two wavelengths is affected by the current through the LED and the power delivered to the LED.

[0006] The current driver can be implemented as a switching power converter having an output terminal connected to the LEDs. The power converter keeps the voltage on the output terminal at a pre-calculated constant target value corresponding to the desired current. This voltage regulation to provide current control is not fully satisfactory however because the current through the LEDs can drift at a constant voltage due to heating of the LEDs, age, and possibly other conditions.

[0007] Fig. 1 shows a driver circuit that senses the LED current rather than output voltage. This is a buck converter circuit, described in "AN874 Buck Configuration High-Power LED Driver" (Microchip Technology Inc. 2003). Input voltage Vbatt is connected to one terminal of a PMOS switch 104 whose other terminal 110 is connected to induction coil Ll. The other terminal of coil Ll is the driver circuit's output terminal 120, connected to the anode of LED 130. The LED ' s cathode is connected to resistor R8 whose other terminal is grounded. The LED's cathode is also connected to one terminal of feedback current sensing resistor R6. The other terminal 140 of resistor R6 is connected to the non-inverting input of operational amplifier 150. The inverting input of amplifier 150 receives a voltage Vdrive. The amplifier's output is connected to the gate of transistor 104. A smoothing capacitor C3 and a diode D are also part of the circuit. Amplifier 150 turns the switch 104 on or off depending on the current through LED 130.

SUMMARY

[0008] This section summarizes some features of the invention. Other features are

described in the subsequent sections. The invention is defined by the appended claims which are incorporated into this section by reference.

[0009] Some embodiments of the present invention provide for one or more LEDs to be driven such that a steady light of the desired color and intensity is attained with good efficiency. Some embodiments regulate the current without sensing the LED current (i.e. using an open loop scheme). The driver circuit senses the output voltage and controls the converter's duty cycle to provide a steady current to the output terminal in each switching cycle.

[0010] While the driver senses the output voltage, the driver does not attempt to keep the output voltage constant. If the output voltage increases, the driver does not decrease the output current but keeps the output current constant. Similarly, if the output voltage decreases, the driver holds the output current constant. The output voltage is sampled simply to determine the converter's duty cycle needed to provide the constant output current.

[0011] In some embodiments, the driver circuit provides constant power to the driver circuit's output terminal in each switching cycle. The converter samples the output voltage but does not sample the power flowing through the LEDs.

[0012] In some embodiments, the driver circuit provides a controlled ramp-up mode when the circuit is first turned on, to enable a controlled increase in the LEDs' lighting intensity. The controlled ramp-up is desired to reduce the stress on the system and the user's eyes.

[0013] The invention is not limited to the features and advantages described above. The invention is not limited to white LEDs or to any LEDs or flat panel displays, and includes non-LED current or power regulators. Other features are described below. The invention is defined by the appended claims.

BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Fig. 1 is a circuit diagram of a prior art LED driver.

[0015] Fig. 2 is a block-circuit diagram of a constant current or constant power regulator according to some embodiments of the present invention.

[0016] Fig. 3 is a graph of a coil current versus time in some embodiments of the present invention.

[0017] Fig. 4 is a chart of the output voltage versus time in some embodiments of the present invention.

[0018] Figs. 5-7 are flow charts of duty cycle calculations in some embodiments of the present invention.

DESCRIPTION OF SOME EMBODIMENTS

[0019] The embodiments described in this section illustrate but do not limit the invention. The invention is not limited to particular circuitry, software, voltage and current values, or other parameters. The invention is defined by the appended claims.

[0020] Fig. 2 is a block-circuit diagram of a boost converter driver driving a LED array 130. An input voltage Vin is provided on a node 210 to one terminal of a coil Ll . The coil's other terminal 220 is connected to the drain of an NMOS switch 224 whose source is grounded (or connected to some other reference voltage). Terminal 220 is also connected to the anode of diode Dl. The diode's cathode node 230 provides the output voltage Vo to LED array 130. Node 230 is connected to one plate of an optional smoothing capacitor Cl whose other plate is grounded (or at some other reference voltage).

[0021] The voltage Vo on node 230 is sampled by a sample and hold (S/H) circuit 240 in controller 244 each switching period T (1.9 μs in some embodiments). The sampled voltage is digitized by analog to digital converter (ADC) 250. Input voltage Vin is also sampled each cycle T and digitized by the same or different sample and hold and ADC circuits in controller 244. The digitized voltages Vo, Vin are processed by microprocessor 260 (comprising an instruction execution unit 264 and a memory 270) to calculate the on and off times for switch 224 for each switching cycle T. (The on time is shown as tl in Fig. 3 described below; the off time is t2+t3.) The calculated

on and off times are provided to switch driver 274 to generate the signal SSET on the gate of transistor 224. Instruction execution unit 264 in processor 260 executes software instructions stored in memory 270. The software can be in the form of firmware or any other form. The digital values for Vo, Vin, and other parameters used by processor 260 are stored in memory 270. Memory 270 can be static random access memory (SRAM) or any other kind of memory known or to be invented. The invention is not limited to a particular architecture of processor 260.

[0022] Fig. 3 is a diagram of the coil current Icoil through coil Ll versus time in the normal operating mode (as opposed to a ramp-up starting mode described below). In each switching cycle T, transistor 224 is on for a time tl (also denoted LRST), allowing the current to ramp up to a peak value, as shown by line 310.1. Then the transistor turns off, and the coil current flows into the output terminal Vo (line 310.2). The current becomes zero some time before the end of the switching cycle (i.e. this is discontinuous current mode (DCM)). The time period during which the current decreases to zero is denoted t2. The time from the end of t2 to the end of the switching cycle is shown as t3. Icoil=0 during t3. The converter duty cycle dl (the switch 224 duty cycle) is tl/T by definition.

[0023] The time tl is computed by processor 260 to provide a constant average current Io to terminal Vo. Io is defined as the charge provided to the Vo terminal 230 in one switching cycle T divided by T. Thus, Io is kept at a target value "Itar" specified by the system designer and provided to controller 244.

[0024] The time tl is computed as follows. Since the coil current flows to terminal Vo only during the time t2, the charge provided to the Vo terminal in one switching cycle is the area Q2 under the line 310.2. This line has a slope -DV/L, where:

DV is the EMF (electromotive force) of the coil Ll during the time t2, i.e.

DV=Vo+Vd-Vin (1)

Vd is the voltage drop across the diode Dl ; L is the inductance of coil Ll . Of note, DV is positive since this is a boost converter. The area Q2 is thus:

Q2=l/2*DV/L*(t2) ² Therefore, ignoring converter efficiency for now:

I ₀ = ^ (t2) ² (2)

2TL ^{K J}

The time t2 can be expressed in terms of tl from the condition that the peak coil current is provided both at the end of tl and at the start of t2:

(Vin/L)*tl=(DV/L)*t2, and hence (3) t2=(Vin/DV)*tl (4)

Therefore,

Vin ²

Io=- (ti) ² (5) 271 * DV Hence,

_ ^2L * Io * (Vo + Vd- Vin) *T

Vin and the duty cycle

[0025] For Io=Itar, we obtain:

_ V2X * Mr * (Vo + Vd- Vin) * T tl (8)

Vin ^{y J} and the duty cycle

[0026] In each period T, controller 244 turns on the transistor 224 for the time tl computed according to (8) and turns off the transistor 224 for the rest of the T period.

[0027] To ensure the discontinuous current mode, the inductance L is chosen to so that the Itar value does not exceed the critical current Icrit defined as the output current Io in the critical current mode. Icrit can be found from the condition t3=0 (i.e.

tl+t2=T) and equations (1), (4) and (5). In the critical condition (see equation (4)),

T* DV (Vo + Vd)

(10) and therefore (equation (5))

(H) IfVd is negligible, we can write: _τ .. Vin ² * T _n Vin .

Icπt= (1 )

2I * Fo Vo

(12) From equation (10), and taking into account the converter efficiency (η), the following expression can be derived:

(13) where the converter efficiency η is defined as:

and Iin is the average input current in one T period. Generally in DCM (note equation (4)),

(15) Here the charge Ql is the area under the line 310.1 (Fig. 3). In the critical condition (see equation (10)),

_ T*Vin* (Vo + Vd -Vm) 2L* (Vo + Vd)

(16)

[0028] In some embodiments, the efficiency η is not measured during the

operation but is estimated in advance by the designer. The Itar value can be increased (for equation (9), for example, and other equations) to take into account the estimated efficiency value. The L selection to ensure DCM is described in more detail below.

[0029] In some embodiments, the constant current control as described above is enabled in response to the user turning on the display or in response to an enable signal (not shown) provided to controller 244. Controller 244 may use other control methods when the enable signal is deasserted. When the constant current control is first enabled, the output voltage Vo may have to be increased from 0 or some other value to an appropriate value InitVtar for the current control mode (see Fig. 4). For example, InitVtar should be above Vin for the normal boost converter operation. It is desirable to increase Vo at a controlled rate to reduce the stress on the system and the user's eyes. Because of parasitic effects, temperature and other factors, the exact output voltage Vo upon reaching the target current Itar is not known ahead of time. During ramp up (shown as the period from time 0 to a time Tramp in Fig. 4) controller 244 monitors Vo and compares it to the value InitVtar provided by the system designer. The time Tramp (Fig. 4) is defined as the time when Vo reaches InitVtar. The normal operation described above (determining tl from equation (6) with Io=Itar)) begins at time Tramp. Tramp can be many times larger than T. The controlled ramp up is achieved by keeping the output current Io at or below a predefined value Iramp specified by the designer. The value Iramp is provided to controller 244. (The value InitVtar provided by the designer can be stored in a memory location in memory 270, and can later be overwritten with a Vo value obtained after the time Tramp, so that if the converter is turned off and then back on again, the converter will use that Vo value instead of the InitVtar value.)

[0030] One embodiment is illustrated in Fig. 5. This is a flowchart of a program executed by processor 260. The voltages Vo, Vin are sampled every T period. As long as Vo is less than a value UNLV=Vm-2Vd (step 510), microprocessor 260 sets tl=0. Hence the switch 224 is kept off, allowing the voltage Vo to quickly increase towards the value Vin. The time when Vo reaches UNLV is shown as To in Fig. 4. Starting at To (step 520), processor 260 sets tl as follows (use equation (6) with Io=min(Iramp, Icrit)):

_ yJ2L * ItarRamp * (InitVtar + Vd- Vin) * T Vin

(17) where

ItarRamp=min(Iramp, Icrit). (18) Equation (17) can be derived from equation (6) if Io is replaced with ItarRamp and Vo is replaced with InitVtar. Icrit is determined from equation (11) or (13) for some hypothetical Vo and η values chosen by the designer to ensure that the calculated Icrit does not exceed the actual Icrit value for the actual Vo and η ranges obtained. In some embodiments, the Vo value is InitVtar and the η value is about 85%. In some embodiments, the value Iramp, and hence ItarRamp, is less than Itar.

[0031] When Vo reaches InitVtar (a value greater than Vin), then tl is calculated using equation (8), as shown at 530.

[0032] In some embodiments, Vin is 2.5-5.5 V. Vo is 5.0-30 V in normal operation (after Tramp). Vd=LO V; T=I.9μs; InitVtar=12 V; Itar=20 mA; Iramp=15 mA; L=2 μH; C=I μF. This values are exemplary and not limiting.

[0033] In some embodiments, the operation from To to Tramp takes place for at most a preset number RampCount of the, T periods. If the LED current is negligible at this time, the output current Io simply charges the capacitor Cl. Therefore, RampCount is calculated as: _D ^ . C{InitVtar -Vo) RampCount=— - (\9)

ItarRamp* T ^κ where C is the capacitance of CL

[0034] Fig. 6 shows a flow chart of a program executed by microprocessor 260 in each switching cycle to determine tl for the next switching cycle in one such embodiment. At step 610, Vo is compared to a preset maximum value Vom provided to controller 244. Vom=l 6 V in one embodiment. IfVo is not greater than Vom, then we go to step 620. IfVo is greater than Vom then we set an over- voltage fault flag (step 624) before going to step 620. The over-voltage fault flag can be used to provide an alarm signal to a designer or the system user. At step 620 we test for two

conditions. If the constant current algorithm has not been enabled or Vo is less than UNLV, then a state variable BOOST_STATE keeping track of the state of the circuit is set to 0 (step 630), and tl is set to 0 (step 634). Hence the signal SSET stays low (step 640) for the entire switching cycle, and switch 224 remains off. This operation occurs up to time To (Fig. 4). Of note, step 640 is performed by driver 274 which may or may not be part of processor 260.

[0035] If the constant current algorithm has been enabled and Vo is greater than or equal to UNLV, then we compare BOOST_STATE to 0 (step 650). BOOST_STATE=0 means that the calculation of Io, Vo, RampCount has not been performed for the ramp up. In this case, BOOST_STATE is set to 1 (step 654). At step 658, Icrit is calculated from the equation (13) by setting Vo=InitVtar. Io is thus set to ItarRamp (equation (18)). At step 662, RampCount is calculated from equation (19). Step 670 performs the tl calculation using equation (6), with the Io value provided at step 658 and with the sampled Vo value. At step 640, the signal SSET is generated based on this tl value.

[0036] If at step 650 BOOST_STATE is not zero (indicating that Io and RampCount have already been calculated for the ramp up), then two comparisons are made at step 680. IfVo is not greater than InitVtar and RampCount is not zero, then RampCount is decremented (step 684) and control passes to step 670 to calculate tl from equation (6). Here Io=ItarRamp (calculated at step 658). If at step 680, Vo is greater than InitVtar or RampCount is zero, then BOOST_STATE is set to 2 at step 688 (meaning that the ramp up mode should be terminated), Io is set to Itar (step 692), and control passes to step 670.

[0037] To ensure the DCM operation, the coil Ll is sized so that the target current Itar would be less than Icrit (equation (13)). This means:

XJ^ ηT * Vin ² * (Vo + Vd- — Vin)

(20)

2Itar*(Vo + Vd) 2

[0038] In DCM, the coil current Icoil may ring when it returns to zero. In some embodiments, to make it less likely that a T period may start during the ringing, the

coil is chosen so that Itar would be at most 25% of Icrit, and hence:

[0039] A suitable L value can be chosen by setting Vo and η to appropriate values to provide suitable operation for an expected range of the Vo and η parameters. In some embodiments, η=85% and Vo= 12 V.

[0040] The capacitance C of Cl is determined as follows. We assume that t2 is small compared to (tl+t3). Therefore, we can model the effect of the equivalent series resistance (ESR) of capacitor Cl is a voltage jump δV=(Itar*ESR) in normal operation. Since the converter delivers power to the output Vo only during the time t2, the voltage ripple δVo due to the capacitance is:

δVo=Itar*(tl+t3)/C=Itar*(dl+d3)*T/C (22) where d3=t3/T. The total voltage effect of the capacitor is δV+δVo. Therefore, we get:

δVlimit≥ Itar*(tl+t3)*/C+Itar*ESR (23) where δVlimit is the upper limit allowed by the designer for the output ripple voltage caused by capacitor Cl (δVlimit=20 mV in some embodiments). Noting that tl+t3=T-t2 and using the equations (4) and (8), we obtain:

(24)

[0041] For the C computation, Vo is set to some value to provide a suitable operation for an expected range of the actual Vo voltages. In some embodiments, Vo= 12 V.

[0042] In one embodiment the intensity of the LED emissions is varied to produce a dimming effect. Doing so by simply varying the current per duty cycle (e.g. reducing tl in each duty cycle) might cause a color shift in white LEDs. In some embodiments of the present invention dimming is accomplished by the technique of cycle skipping (setting tl=0 for some cycles). The technique consists of driving less

frequent pulses (tl>0), but any individual pulse event (i.e. any T cycle with tl>0) is at a current level Itar which will preserve the LED color. This has an additional advantage in that the FET 224 may be selected for best performance at Itar with a more uniform pulse width tl, thus preserving efficiency across a wide dynamic range of light intensity.

[0043] In some embodiments, controller 224 receives a signal identifying the cycle skipping number as the number of cycles to be skipped after each cycle with tl>0. In response, controller 274 enables and disables the driver 274 at a frequency determined by the cycle skipping number to provide the cycle skipping. Note also "TPS61060 TPS61061 TPS61062 CONSTANT CURRENT LED DRIVER WITH DIGITAL AND PWM BRIGHTNESS CONTROL" (Texas Instruments Incorporated, November 2004), incorporated herein by reference.

[0044] In some embodiments, the converter is controlled to provide a constant input power Pin rather than the constant output current Itar. The input power delivered in any cycle T is given by:

Pin = Vin*Iin (25)

The LEDs drop more voltage when cold, less when hot, but the emission of light is a factor of the power applied. Also, the source (not shown) of voltage Vin may be able to provide more power with the rising temperature, so the input power regulation reduces the waste of the input energy. Moreover, some devices (e.g. cellular telephone radio transmitters) may have to operate in a power range restricted by law. Constant power control simplifies meeting the legal requirements.

[0045] In some embodiments, power is only delivered to the load during the time period t2. The power developed during t2 is equal to the power delivered during tl. Iin is given by equation (15). Neglecting Vd, we obtain:

Here, Pin is the average power, i.e. Hn is the average current, equal to (Ql+Q2)/T. From equation (26), we obtain:

and

1 2L* _P in * (Vo-Vin) *T Vin \ Vo

Controller 244 obtains the target power value Ptar (which is provided by the designer) and uses the equation (27B) with Phτ=Ptar to calculate the time tl for each cycle T from the sampled voltages Vin, Vo.

[0046] Alternatively, for each T cycle n, dl=dl _(n ) can be calculated from the ratio dl( _n )/dl(n-i) where dl( _n-1 ) is the dl value for the previous cycle. From equation (27A) ₅

where Vin ₍ j), Vo® are the Vin and Vo values for the T cycle i for i=n-l, n. The same relationship holds for the tl values tlφ for the cycles i=n, n-1, i.e.:

Controller 244 uses this equation to calculate tl _(n) from tl _(n-1) for each cycle n (except the first one or more cycles of the constant power control operation, after the ramp up).

[0047] Fig. 7 shows a flowchart for one embodiment executed by controller 244. The flowchart is identical to that of Fig. 6, except that:

1. Step 692 is replaced with step 760, at which Pin is set to Ptar.

2. Step 670 (tl calculation) is replaced with step 770, showing that tl is calculated from equation (27B) or equation (29). Equation (29) can be used for any calculation after one or more T periods in which tl was calculated with equation (27B).

[0048] The constant power control embodiments may use the cycle skipping techniques described above for the constant current control embodiments.

[0049] Some embodiments of Figs. 6 and 7 omit the smoothing capacitor. This is done to provide a fast ramp up. In this case, C=O, so RampCount=0 (equation (19)),

so the ramp up mode is used only for one T period (the T period for which BOOST_STATE is set to 1 at step 654 of Fig. 6 or 7). Other embodiments omit the ramp up altogether.

[0050] The invention is not limited to the embodiments described above. Processor 260 does not have to be a microprocessor (e.g. within a single integrated circuit). Also, processor 260 can be any computing circuit, e.g. an analog computer or some other computer. Processor 260 can include multiple processors. The processor instructions can be stored on a data carrier such as a writable computer memory or a read-only memory. The data carrier can also be a physical electromagnetic wave transmitted through space or over a cable. Other embodiments and variations are within the scope of the invention, as defined by the appended claims.