3.1.1. Receiver with a high-pass discriminator

The structure of the receiver channel design described in paper I is shown together with a photograph of the circuit in Fig. 9. The receiver consists of two identical channels, one for the start pulse and the other for the stop pulse, integrated on the same chip so as to minimise errors caused by delay variations due to changes in temperature, supply voltage or process parameters. An external avalanche photodiode (APD) is used to convert the optical pulse to a current pulse. The noise level at the output to the preamplifier is 1 mV and the maximum linear amplitude is about 1.5 V. Thus the transimpedance preamplifier limits the dynamic range of its input signal to about 1:150 (SNR>10). A current mode gain control cell (CMGC) has been used successfully between the photodetector and the preamplifier to enlarge the dynamic range of the input signal up to 1:620. The cell acts as a current buffer with variable attenuation, and the change in attenuation causes only a little variation in the propagation delay and bandwidth (Gilbert 1968, Vanisri & Toumazou 1995, Ruotsalainen et al. 1999b). This property is very important in time-of-flight rangefinding, as any change will directly affect the distance result, with 6.7 ps corresponding to 1 mm. After the CMGC cell, the current pulse is converted to a voltage pulse in a low-noise transimpedance preamplifier and then amplified in two double-stage voltage amplifiers (Cherry & Hooper 1963) which have offset compensation circuitry with three selectable offset voltages. A further reduction in the dynamics of the signal is achieved with a voltage mode gain control cell (VMGC, R-2R ladder) between the preamplifier and the voltage amplifier. Finally, the high-pass timing discriminator generates a logic-level timing pulse for the TDC. The chip was implemented in an AMS 0.8 µm BiCMOS process.

The operation of the gain control is shown in Fig. 10, in which the x-axis shows the current pulse amplitude at the input to the receiver and the y-axis the voltage pulse amplitude at the input to the timing discriminator.

The signal is first attenuated with voltage mode gain control in order to keep the SNR as large as possible, and current mode gain control is used when the transimpedance preamplifier begins to saturate. The figure shows that the 1:600 dynamic range of the input signal has been reduced to 1:13, which is acceptable for the timing discriminator.

Measurements performed using the circuit as a part of a laser radar device are summarised in Table 1.

The total accuracy, taking into account the walk error (signal varies in a range 1:620) and jitter, is about ±4.7 mm (average of 10 000 measurements), and the temperature drift is less than ±3 ps (±0.5 mm) in the temperature range from –5°C to +40°C. The accuracy can be further improved by compensating for the dominant error sources, i.e. the walk error and voltage mode gain control error, by using a correction table together with information on the amplitude of the signal, measured with the peak detector included in the chip. The current consumption is ~ 50 mA from a 5 V supply and ~ 3 mA from a –10 V supply.

Electronic gain control structures can be used to replace optical ones in some applications, which along with the small size and low power consumption achieved by the use of integrated electronics, simplifies the construction of the resulting rangefinding devices.

3.1.2. Receiver with a leading edge discriminator

The structure of the single channel receiver described in paper II is shown in Fig. 11. The circuit is implemented in an AMS 0.8 µm BiCMOS process and is intended for an industrial application in which the dynamic range of the input signal is large (as high as 1:4000) and the measurement has to be made with a single optical pulse without any prior information about the signal amplitude. This means that gain control structures are useless. The circuit employs a leading edge discriminator with a constant threshold voltage. The amplitude of the received pulse is measured with a peak detector and the information is used to compensate for the walk error. Fully differential structures were used in the receiver channel and in the timing discriminator in order to reduce sensitivity to disturbances, improve common mode rejection, power supply rejection and linearity, reduce inductance of the input bond wires due to mutual inductance etc.

The measurements were performed using the circuit as a part of a laser radar device, which achieved centimetre-level single-shot accuracy over a wide dynamic range. The accuracy of a single measurement is better than ±45 mm, including walk error (amplitude of the input signal varies in the range 1:4000), drift with temperature (ambient temperature varies in the range 0°C to +50°C) and jitter (σ-value, SNR > 35). Due to the wider bandwidth, 250 MHz, the jitter is smaller than that of receiver channel using a high-pass timing discriminator described in section 3.1.1, but the walk error is larger because of the different timing discrimination method.

The walk error measured at three temperatures, after compensation, is shown in Fig. 12. The origin of the walk error was not completely understood at the time of writing paper II, but it can be easily explained in terms of the theory presented in section 2.1, as the uncompensatable walk error, ±30 mm, agrees well with the walk error simulated in that section. The parameters used in the simulation are equal to those of the receiver channel and to the properties of the optical pulse used in the measurements.

The circuit contains real-time calibration and testing structures to ensure the correctness of the distance result. The circuit is well suited for applications in which distance has to be measured with a single pulse, the amplitude of which cannot be predicted from previous measurements. The measurement results are summarised in Table 2.

A photodiode with a large area (diameter ~500 µm) is usually needed in industrial measurement applications, and this brings a capacitance of about 2 picofarads into the input to the preamplifier. In addition, the interconnection between the photodetector and the receiver includes structures such as ESD protection, an I/O pad and PCB wiring, which entail an additional capacitance of several picofarads. The large capacitance generates a voltage noise peaking in the preamplifier and calls for a careful design in order to ensure stability with a large bandwidth.

3.1.3. Combining a timing discriminator with TAC

As the comparator proved to be the source of the walk error, a new topology was developed in which timing discrimination and time interval measurement are combined and no comparator is used at all (Paper III). The timing point of the pulse is produced using a C-R high-pass filter and is free of walk error. A fast comparator is usually then used to detect the timing point and to produce a logic-level signal for the TDC, which can be a time-to-amplitude converter (TAC) followed by an A/D converter, for example, or a TDC based on digital delay cell topologies.

In the new method presented in paper III time interval measurement is based directly on linear signal processing of the timing pulses of the receiver channel. The idea is to steer the input transistors of the TAC directly with the variable slope analogue signal which is usually connected to the comparator. Thus the input transistors of the TAC are used as “analogue switches” instead of merely as current switches. It will be shown in the following paragraphs that even if the amplitude of the input signal varies, the time-to-amplitude conversion will give the same result, i.e. the method is free of walk error.

A coarse schematic diagram of a differential pair which serves as the core element of the TAC is shown in Fig. 13. Time to amplitude conversion is produced by discharging a capacitor C1 with a current I_b for as long as the input signal V_in is positive. Before each measurement the level of V_out is reset to a positive supply with a switch S1. The change in the amplitude ΔV_out, which corresponds to the time interval, is measured with an A/D converter.

The immunity to walk error is presented in Fig. 14 with the aid of a large signal transfer function for the differential pair. The input signals to the TAC V_in are shown as a function of time in the bottom left panel of Fig. 14. The start edge of the time interval to be measured is constant, whereas the slew rate of the stop edge varies as a function of the amplitude of the received signal. The input voltage signals are converted to the output current signals I_out shown in the top right panel of the figure using the large signal transfer function of the differential pair shown in the top left. Finally the output currents are integrated as a function of time, shown in the bottom right corner. The integral corresponds to the voltage drop at the output, ΔV_out. As can be seen in the figure, the result of the integration is independent of the slew rate at the stop edge, which means that the distance result does not depend on the amplitude of the received signal (which causes variation in the slew rate of the falling edge) and therefore has no walk error.

In order to be able to make use of the proposed method, the receiver channel should produce the pulse presented in the bottom left panel of Fig. 14 from the original unipolar timing pulse. This is achieved by the receiver channel construction shown in Fig. 15. For simplicity, the receiver and signals are drawn in a single-ended form, whereas the implementation of the test chip was differential.

A bipolar pulse is generated from the unipolar input pulse using a similar high-pass filter to that described earlier in section 2.2. In order to achieve a large gain-bandwidth product and to process the pulses and the ramp signal linearly, the signal swing in the blocks is limited to ±400 mV and the gain of the amplifying stages is set to about 2 with emitter degeneration resistors. To pick up the rampsignal around the zero-crossing point of the bipolar signal, structures have to be used which will force the signal to be positive before the pulse and negative after the pulse (the positive and negative pulse shaper blocks in Fig. 15). This can be done using the circuitry shown in Fig. 16. The upper part of the diagram shows a limiter block with a force function and the lower part the latch circuitry which detects the edge of the bipolar pulse and controls the force function. The level of input signal which changes the state of the latch can be set with the adjustable bias current source.

The receiver (and also the latch) must be reset before measurement, whereupon the latch forces the output of the circuitry to be positive (+400 mV). When the input signal exceeds the preset offset voltage of the latch, the state of the latch changes and the output of the circuitry begins to follow the bipolar input signal. In the same way, but in opposite order, the negative part of the bipolar pulse is formed with similar circuitry (negative pulse shaper). The output of the circuitry first follows the input signal, and when the negative part of the bipolar signal crosses the negative offset voltage of the latch, the output is forced to be negative (–400 mV). Finally the start signal is included and the signal shown in the bottom left panel in Fig. 14 is produced for the time to amplitude converter.

The voltage drop at the output to the TAC is measured with an A/D converter. If a resolution of one millimetre (~7 ps) is needed and a 12-bit A/D converter is being used, the maximum integration time will be about 27 ns. This will result in a measurement range of about 4 metres. The range can be increased without detracting from the resolution by using a coarse counter to count the periods of a reference clock and interpolating the timing point inside the clock period using the scheme presented. Two TACs are needed for this kind of structure, the rising edges of their input signals being produced by the start and stop pulses and the falling edges by the reference clock.

The operability of the idea was verified by implementing a test circuit in a 0.8 µm BiCMOS process (paper III). The measured walk error is shown in Fig. 17 and the time result varies ±3.5 ps (±0.5 mm in distance) with an input signal dynamic range of 1:21. The error is smaller than the ±12 ps walk error of the high-pass timing discriminator in a dynamic range of 1:10. The dynamic range of this new method is limited by the finite bandwidth and slew rate of the receiver channel. The duration of the “ramp signal” changes from about 2 ns to 100 ps.

The single-shot precisions of the conventional channel and the new channel are of the same magnitude, but the scheme involves no very fast, power consuming switching operations, which means that the circuitry produces less disturbances. The idea enables walk error to be reduced, especially at small signal amplitudes and allows the level of integration of the system to be increased still further.

The new scheme was not used in the laser rangefinder prototype devices whose construction is described in the following section, because a new test circuit should have been designed to enlarge the measurement range by means of a reference clock and interpolation inside the clock period. As the receiver channel with a high-pass discriminator and a separate TDC chip were already available, these were used in the prototypes.