Design Feature: May 9, 1996

Drivers for cold-cathode fluorescent lamps (CCFLs) are intriguing devices (Reference 1). They must produce fairly high-voltage (hundreds of volts rms), quasi-sinusoidal, adjustable, regulated ac voltages, usually at tens or hundreds of kilohertz. The drivers must operate from low-voltage, often unregulated dc sources. Most CCFL drivers deliver modest amounts of power—usually <1w, but must do so with high efficiency (;90%). Moreover, the operating conditions that produce the maximum electrical efficiency (P_OUT/P_IN) don't always produce the best optoelectrical efficiency (light out/P_IN). As a result, when developing CCFL drivers, and even when designing IC drivers into assembled products, achieving optimal performance requires making many challenging measurements. Understanding how to properly make these measurements is useful, even if you aren't designing or using CCFL drivers. If your job involves applying measurement technology, or if you are merely interested in electrical measurements, there's a good chance you can find other applications for CCFL-circuit techniques.

Establishing and maintaining accurate wideband ac measurements is a textbook example of a situation that requires careful attention to measurement. The combination of high-frequency, harmonic-laden waveforms and high voltage makes obtaining meaningful results difficult. Careful selection, understanding, and use of test instrumentation is crucial. You must think clearly to avoid unpleasant surprises.

The lamp's current and voltage waveforms contain energy over a wide frequency range. Most of this energy is concentrated at the inverter's fundamental frequency and immediate harmonics. However, for 1% measurement uncertainty, you must accurately capture energy content out to 10 MHz. Figure 1, a spectrum analysis of lamp current, shows significant energy at 500 kHz. Diminished, but significant, content is still apparent, however. This data suggests that the monitoring instrumentation must maintain high accuracy over wide bandwidth.

Accurate determination of rms operating current is important for electrical- and emissivity-efficiency computations and for ensuring long lamp life. Additionally, you may want to perform current measurements in the presence of high common-mode voltage (>1000V rms). The ability to make such measurements allows investigation and quantification of display- and wiring-induced losses, regardless of their origins in the lamp-drive circuitry.

Current-probe circuitry

Figure 2's circuit meets these measurement requirements. The figure shows a precision amplifier conditioning the output of a commercially available clip-on current probe. This configuration provides 1% measurement accuracy to 10 MHz. The clip-on probe is convenient even in the presence of high common voltages. The current probe biases IC₁, which operates at a gain of about 3.75. The probe's low-output-impedance termination eliminates the need for impedance matching. Additional amplifiers provide distributed gain, maintaining wide bandwidth with an overall gain of about 200. The individual amplifiers avoid any possible crosstalk-based error that a monolithic quad amplifier might introduce. Selecting D₁ and R_X and choosing D₁'s polarity trim the overall amplifier offset. The 100 Ohm trimmer sets the gain, fixing the scale factor.

The output drives a thermally based, wideband, rms voltmeter. In practice, the circuit is built into a 2.253131-in. enclosure, which BNC hardware connects directly to the voltmeter. There is no cable. Figure 3 details RF layout techniques used in the amplifier's construction. The result is a clip-on current probe with 1% accuracy over a 20-kHz to 10-MHz bandwidth. This tool is indispensable in any rigorously conducted backlight work.

Figure 4's circuit, a current calibrator, permits calibration of the probe and amplifier. IC₁ and IC₂ form a Wein-bridge oscillator. IC₄ and IC₅ rectify the oscillator output; IC₃ compares the rectified output to a dc reference. IC₃'s output controls Q₁, closing an amplitude-stabilization loop. The amplitude-stabilized voltage terminates in a 100 Ohm, 0.1% resistor to provide a precise 10.00-mA, 60-kHz current through the series-current loop. Adjusting the nominally 15-k Ohm resistor for exactly 1.000V rms across the 100 Ohm resistor trims the output.

Over one year, this current probe has shown 0.2% baseline instability with 1% absolute error. The sole maintenance requirements for preserving accuracy are to keep the current-probe jaws clean and to avoid rough or abrupt handling of the probe.

Voltage probes for grounded-lamp circuits

The high-voltage measurement across the lamp is demanding on the probe. The simplest case is measuring grounded-lamp circuits, in which the waveform fundamental is at 20 to 100 kHz, with harmonics into the megahertz region. This activity occurs at peak voltages in the kilovolt range. The probe must have a high-fidelity response under these conditions. Additionally, the probe should have low input capacitance to avoid loading effects that would corrupt the measurement. The design and construction of such a probe requires significant attention. Table 1 lists some recommended probes and characteristics. Almost all standard oscilloscope probes fail if you try to use them for this measurement.

Attempting to circumvent the probe requirement by resistively dividing the lamp voltage also creates problems. Large-value resistors often have significant voltage coefficients, and the resistors' shunt capacitance is high and uncertain. Therefore, avoid simple voltage dividers. Similarly, most high-voltage probes intended for dc measurement introduce large errors because of ac effects. The Tektronix P6013A and P6015 work well; the devices' 100-M Ohm resistance and small capacitance introduce low loading errors. The penalty for their 10003 attenuation is reduced output, but the recommended voltmeters accommodate the small signals.

All of the recommended probes are designed to work into an oscilloscope input. Such inputs are almost always 1 M Ohm paralleled by (typically) 10 to 22 pF. The recommended voltmeters have significantly different input characteristics. Table 2 shows higher input resistances and a range of capacitances. Because of this difference, the probe compensation must accommodate the voltmeter's input characteristics. Normally, you can easily determine and adjust the optimum compensation point by observing the probe output on an oscilloscope. Apply a known-amplitude square wave, such as that from the oscilloscope calibrator, and adjust the probe for correct response.

Using the probe with the voltmeter presents an unknown impedance mismatch and raises the problem of determining when compensation is correct. The impedance mismatch occurs at low and high frequencies. To correct the low-frequency effect, place an appropriate-value resistor in shunt with the probe's output. For a 10-M Ohm voltmeter input, a 1.1-M Ohm resistor is suitable. Maintaining a coaxial environment, requires you to build this resistor into the smallest possible BNC-equipped enclosure. Do not use cable connections; place the enclosure directly between the probe output and the voltmeter input to minimize stray capacitance. This arrangement compensates the low-frequency impedance mismatch.

Correcting the high-frequency mismatch is more involved. Problems result from the wide range of voltmeter input capacitances combined with effects of the added shunt resistor. Knowing where to set the high-frequency probe-compensation adjustment is confusing. One solution is to feed a known-value rms signal to the probe-voltmeter combination and adjust the compensation for a proper reading. Figure 5 shows a way to generate a known rms voltage. This scheme uses a standard backlight circuit reconfigured to produce a constant-voltage output. The op amp permits low R-C loading of the 5.6-k Ohm feedback termination without introducing bias-current error. You can series- or parallel-trim the 5.6-k Ohm value to obtain a 300V output. Stray parasitic capacitance in the feedback network affects the output voltage. Because of this effect, you should rigidly fix all feedback-associated nodes and components and build the entire circuit into a small metal box. This construction prevents any significant change in the parasitic-element values. The result is a known 300V-rms output.

Now, adjust the probe's compensation for a 300V voltmeter indication using the shortest possible connection to the calibrator box (for example, a BNC-to-probe adapter). This procedure, combined with the added resistor, completes the probe-to-voltmeter impedance match. If you alter the probe compensation (for example, for proper response on an oscilloscope) the voltmeter's reading becomes erroneous. Therefore, you may want to hide the probe when you aren't using it. It is good practice to verify the calibrator-box output before and after every set of efficiency measurements. Do this by using BNC adapters to directly connect the calibrator box to the rms voltmeter. Set the voltmeter to the 1000V range.

Voltage probes for floating-lamp circuits

Measuring voltages in floating-lamp circuits requires a nearly heroic effort. Floating-lamp measurement involves all the difficulties of the grounded case but also needs a fully differential input, because the lamp floats freely from ground. You must not only properly compensate the two probes, but also match and calibrate them within 1%. Additionally, to check calibration, you need a fully floating source instead of Figure 5's simple single-ended setup.

Figure 6's differential amplifier converts the differential output of the high-voltage probes to a single-ended signal for driving an rms voltmeter. If the probe compensation and calibration are correct, the amplifier introduces <1% error in 10-MHz bandwidth. Both probe inputs feed source followers (Q₁ through Q₄) via R-C networks that provide proper probe termination. Q₂ and Q₄ bias differential-amplifier IC₂, which runs at a gain of ;2. IC₁controls the FET dc and low-frequency differential drift. IC₁ measures a band-limited version of IC₂'s inputs and biases Q₄'s gate-termination resistor. This arrangement forces Q₄ and Q₂ to equal source voltages. This control loop eliminates dc and low-frequency error resulting from FET mismatches. Although you might want to use a monolithic dual FET, such devices cause excessive high-frequency errors.

Q₁ and Q₃ also follow the probe output and feed a small, frequency-dependent, summed signal to IC₂'s auxiliary input. This signal corrects for the high-frequency common-mode-rejection limitations of IC₂'s main inputs. IC₂'s output drives the rms voltmeter via a 20:1 divider. The divider combines with IC₂'s gain-bandwidth characteristics to give <1% error to 10 MHz at the voltmeter input.

To calibrate the amplifier, tie both inputs together and select RX (shown at Q₄), so that IC₁'s output is near zero. You may have to place R_X at Q₂ to make this trim. Next, drive the shorted inputs with a 1V, 10-MHz sine wave. Adjust the 10-MHz CMRR trim for a minimum rms voltmeter reading, which should be below 1 mV. Finally, lift the positive input from ground, apply 1V rms at 60 kHz, and set IC₂'s gain trim for a 100-mV voltmeter reading. As a check, grounding the positive input and driving the negative input with the 60-kHz signal should produce an identical meter reading. Further, known differential inputs at any frequency from 10 kHz to 10 MHz should produce corresponding calibrated and stable rms-voltmeter readings within 1%. You can correct readings outside this figure at the highest frequencies by adjusting the 10-MHz antipeaking trim. This adjustment completes the amplifier calibration.

The high-voltage probes must be properly frequency-compensated to give calibrated results with the amplifier. The R-C values at the amplifier inputs approximate the termination impedance for which the probe is designed. You must, however, precisely frequency-compensate individual probes to achieve the required accuracy. The probe characteristics make this exercise quite demanding.

Probing for answers

Figure 7 is an approximate schematic of the Tektronix P6015 high-voltage probe. A physically large, 100-M Ohm resistor occupies the probe head. Although the resistor has repeatable wideband characteristics, it suffers from distributed parasitic capacitances. These distributed capacitances combine with similar cable losses, presenting a distorted version of the probed waveform to the terminator box. When properly adjusted, the terminator box impedance-vs-frequency characteristic corrects the distorted information, presenting the proper waveform at the output. The probe's 10003 attenuation factor and high impedance provide a safe, minimally invasive measure of the input waveform.

The large number of parasitic elements associated with the probe head and cable result in a complex, multiple-time-constant response characteristic. Faithful wideband response requires the terminator box components to separately compensate for each of these time constants. Therefore, you must make no fewer than seven adjustments to compensate the probe to any particular instrument input. These trims are interactive, requiring a repetitive sequence before the probe is fully compensated. The probe manual describes the trimming sequence, using the intended oscilloscope display as the output. In this application, the ultimate output is from an rms voltmeter connected via the just-described differential amplifier. The meter display complicates determining the probe's proper compensation point but does not make the adjustment impossible.

To compensate the probes, connect them directly to the calibrated differential amplifier (see Figure 8) and ground the probe associated with the negative input. Drive the positive-input probe with a 100V, 100-kHz square wave that has a clean 10-nsec edge with minimal aberrations following the transition. (Suitable pulse generators include the HP 214A and Tektronix 106.) The absolute amplitude of the waveform is unimportant. Use an oscilloscope and a properly compensated probe to monitor two points: the input square wave and IC₂'s output in the differential amplifier. Perform the compensation procedure described in the Tektronix P6015 manual until both waveforms have identical shapes. When you reach this state, repeat this procedure with the negative-input probe driven and the positive-input probe grounded. This sequence brings the probe's interactive adjustments reasonably close to the optimum points.

To complete the calibration, connect the 50 Ohm precision termination (see Figure 6) and the rms voltmeter to the differential amplifier's output. Ground the negative-input probe and drive the positive probe with a known-amplitude high-voltage waveform of about 60 kHz, such as that from Figure 7's calibrator.

Perform very slight readjustments of this probe's compensation trims to get the voltmeter's reading to agree with the calibrated input. (Account for scaling differences; ignore the voltmeter's range and decimal-point indications.) To make this adjustment, use the trim or trims that have the greatest influence; you should have to make only slight adjustments.

Upon completing this step, repeat the procedure using the 100V, 100-kHz square wave, verifying input/output-waveform edge fidelity. If waveform fidelity has degraded, retrim and try again. You may have to repeat the procedure several times to achieve the required accuracy with both waveforms. Repeat this procedure for the negative-probe adjustment with the positive probe grounded.

Next, short both probes together and drive them with the 100V, 100-kHz square wave. Ideally, the rms voltmeter should read zero. In practice, it should indicate well below 1% of input. You can adjust the differential amplifier's 10-MHz CMRR trim (Figure 6) to minimize the voltmeter reading. Then, with the probes still shorted, apply a swept 20-kHz-to-10-MHz sine wave with the highest amplitude available. Monitor IC₂'s output with an rms voltmeter, ensuring that the output never rises above 1% of the input amplitude. (You have now made 14 interactive adjustments.)

Finally, apply the highest available-known amplitude, swept 20-kHz-to-10-MHz signal to each probe with the other probe grounded. Verify that the rms voltmeter indicates correct and flat gain over the entire swept-frequency range for each case.

If your measurements do not meet any condition described in the preceding list, you must repeat the entire calibration sequence. At this point, this calibration is, at long last, complete.

Differential-probe calibrator

A calibrator with a fully floating, differential output allows periodic operational checking of the differential probe's accuracy. You build this calibrator into the same enclosure as the differential probe.

Figure 9 is a schematic of the calibrator. The circuit is a highly modified form of the basic backlight power supply. In this circuit, T₁'s output drives two precision resistors which are well-specified for high-frequency, high-voltage operation. L₂, a wideband current transformer, monitors the resistor's current. L₂'s placement between the resistors combines with T₁'s floating drive to minimize the effects of L₂'s parasitic capacitance. Although L₂ has parasitic capacitance, the capacitance is bootstrapped to essentially 0V, negating its effect. IC₁ and IC₂ amplify L₂'s secondary output. IC₃ and IC₄ act as a precision rectifier. The 10-k Ohm/0.1-µF filter smooths IC₄'s output and closes a loop at the LT1172's feedback pin. As do other CCFL circuits, the LT1172 sets T₁'s output by controlling the drive to the magnetic multivibrator (usually called a Royer circuit, after its inventor, GH Royer).

To calibrate this circuit, ground the LT1172's V_C pin, open one of the connections to T₁'s secondary winding, and select the LT1004's polarity and associated resistor value for 0V at IC₄'s output. Next, put a 5.00-mA, 60-kHz current through L₂. (You can use the output of the circuit in Figure 8, rescaled for 5.00 mA.) Measure IC₄'s smoothed output (the LT1172's feedback pin), and adjust the output trim for 1.23V. Next, reconnect T₂'s secondary, remove the current-calibrator connection and unground the LT1172 V_C pin. The result is 500V rms at the calibrator's differential output. You can check this with the differential probe. Reversing the probe connections should have no effect; the readings should agree within 1%.

The differential probe and floating-output calibrator require almost fanatical attention to layout to achieve the performance levels noted. The wideband amplifier sections utilize RF layout techniques that Reference 2 documents reasonably well. Figures 8 and 9detail practical construction considerations related to parasitic capacitance.

Author's biography

Jim Williams, staff scientist at Linear Technology Corp, Milpitas, CA, specializes in analog-circuit and instrumentation design. He has served in similar capacities at National Semiconductor, Arthur D Little, and the Instrumentation Laboratory at the Massachusetts Institute of Technology, Cambridge, MA. A former student at Wayne State University, Detroit, Williams enjoys art, collecting antique scientific instruments, and restoring old Tektronix oscilloscopes.

References

1. Williams, Jim, A Fourth Generation of LCD-Backlight Technology, Linear Technology Corp, Application Note 65, October 1995.
2. Williams, Jim, High-speed amplifier techniques, Linear Technology Corporation, Application Note 47, August 1991.