Infotech Oulu Annual Report 2014 - Circuits and Systems (CAS-Oulu)

Academy Professor Juha Kostamovaara and Professor Timo Rahkonen,
Electronics Laboratory, Department of Electrical Engineering, University of Oulu
juha.kostamovaara(at), timo.rahkonen(at)


Background and Mission

The Circuits and Systems group consists of 26 researchers working at the Electronics Laboratory of the Department of Electrical Engineering at the University of Oulu. Its main activity is in the field of electronic and optoelectronic circuit and system design. The main interest of the group is devoted to certain novel devices, circuit topologies and functional units, although the group is also interested in applications, especially in the field of electronic/optoelectronic measurements and telecommunications.

The main research fields are:

  • time-to-digital converters and timing circuits
  • generation and detection of powerful and high-speed electrical and optical pulses/transients, and breakdown phenomena in semiconductors in general
  • development of pulsed time-of-flight laser range finding and time-gated Raman spectrometer technologies, especially for industrial applications
  • radio telecommunications, including linearization of power amplifiers, AD/DA conversion and baseband blocks, frequency synthesis.

Part of the group activities belong to the Center of Excellence in Laser Scanning Research (funded by the Academy of Finland, 2014-2019, During 2014 a FiDiPro Research Fellow Dr. Vassil Palankovski joined the group. This 3-year position is funded by TEKES.


Scientific Progress

In the following, some details and results of the work of the group are given in selected important research fields.

Time-to-Digital Converters

Algorithmic Time-to-Digital Converter based on Frequency Switching Interpolation

A two-stage TDC has been developed in 0.35 µm CMOS process. The TDC achieves a single-shot precision of 4.2 ps (rms) with a measurement range of 327 µs.

A coarse time quantization is done by a reference clock counter. The quantization error of the reference clock counter is then measured by two interpolators. These interpolators are based on a cyclic/algorithmic approach, thus they can achieve very high resolution. In each cycle, the time residue is quantized by a ring oscillator counter. The quantization error of the ring oscillator is then amplified by switching the frequency of the ring oscillator. The following amplified time residue is then measured again by the ring oscillator counter. This procedure continues as long as necessary to achieve the desired resolution.

A digital calibration scheme ensures that high accuracy and high precision is retained in various operating conditions. The interpolators provide a resolution of 0.61 ps, although the effective resolution is slightly worsened due to nonlinearities.

A die photo of the TDC is shown in Figure 1. A histogram depicting the measured precision of the TDC is shown in Figure 2.

Figure 1. Die photo of the TDC.


Figure 2. Single-shot precision of the TDC with and without nonlinearity correction.


Time-to-Digital Converter based on Time Domain Successive Approximation Interpolation

The proposed time-to-digital converter (TDC) aims at adjustable sub-ps-level resolution with high linearity in the ms-level dynamic range. To achieve sub-ps-level resolution with cyclic time domain successive approximation (CTDSA) within a clock cycle, the propagation delay difference is implemented by digitally controlling both the unit load capacitors and the discharge current of the load capacitance. The TDC uses only a CTDSA as an interpolator with a 5 ns dynamic range within a clock cycle for sub-ps-level resolution and a counter for ms-level dynamic range without a DLL. The layout design of the proposed TDC is in finalizing state and the design will be sent to fabrication in February 2015.

Time-to-Digital Converter (TDC) Based on Startable Ring Oscillators (SRO) and Successive Approximation

The operating principle in this TDC revolves around startable ring oscillators (SRO). SROs are used as coarse interpolators within the cycle of the reference clock, and the phase differences between the input hit signal and the reference clock are kept in the SROs for further interpolation. Thus, the SROs operate as time domain sample and hold circuits. The SROs are active only when the time interval digitization takes place to save power.

At first the hit signal, start or stop, is synchronized to the rising edge of the clock CLK to generate the signal SAMPLE, as shown in Fig. 3, to which to compare the state of the SRO started with the signal HIT. This state is the coarse interpolation result, i.e. how much earlier compared to the next rising edge of the clock the hit signal arrived. Furthermore, the first rising edge of the SRO preceding the rising edge of the clock is selected by a phase multiplexer with the rising edge of the clock as the residue for the fine interpolation. This residue, i.e. the delay between signals SRO_SEL and CLK_SEL in Figure 3, is then resolved with a cyclic time domain successive approximation method that performs a clock de-skew operation to this signal pair.

Figure 3. Operating principle as timing diagram.


The key building blocks of this TDC architecture are the hit signal synchronizer SYNC, SRO, register for storing the state of the SRO, thermometer to one-hot decoding logic that generates the select signals for the phase multiplexer, and the successive approximation block, shown in Figure 4. The dynamic range is extended by counters that are controlled by synchronized HIT signals.

Figure 4. Block diagram of the TDC architecture.


The building blocks of the TDC are currently being optimized and simulated, and the layout design of the chip begins in spring 2015.

Optical Receiver Circuits

An Integrated CMOS Receiver Channel with TDC for Pulsed Time-of-Flight

An integrated receiver channel has been designed for a pulsed time-of-flight (TOF) laser radar, based on a 0.35 µm CMOS process. The circuit is currently under manufacturing.

The main function of the receiver channel of the pulsed TOF rangefinder is to generate an accurate logic-level timing signal to a TDC. As the accuracy of better than 1 cm is aimed at, a specific timing point must be discriminated from the detected laser pulse (FWHM ~ 3 ns corresponds to 50 cm in distance). Simple leading edge detection would suffer from low measurement accuracy in the applications where the dynamics of the received echo varies a lot (> 1: 10 000). The new receiver channel architecture enables to detect two timing parameters, the pulse width and the slew rate of the detected pulse echo. These are measured with a multi-channel TDC and used for the walk error compensation. The receiver channel consists of a transimpedance pre-amplifier, voltage type post-amplifier, a threshold generator for the timing detection and two parallel comparators generating logic-level timing signals for the TDC.

The transimpedance and 3 dB frequency are 70 kΩ and of 230 MHz, respectively. The input referred current noise is around 100 nArms. An SNR of ~5…10 is needed for reliable distance measurement and thus the minimum signal current is about 0.5…1 µA. The uncompensated walk error is 2.5 ns in the range of 1: 100 000. The compensation curves are shown in Figures 5 and 6.

Figure 5. Compensation curve (walk vs. timing pulse width).


Figure 6. Compensation curve (walk vs. slew rate).


The layout of the receiver channel is shown in Figure 7.

Figure 7. Layout of the receiver channel.


The functionality of the receiver channel will be verified during 2015 by the measurements. A final goal is to integrate a high-performance receiver channel and a multi-channel TDC (time to digital converter) on the same IC chip. A miniature receiver channel/TDC chip will be realized to serve numerous applications in the field of pulsed of time-of-flight laser radars.

A Single Chip Laser Radar Receiver with a 9x9 SPAD Detector Array and a 10-Channel TDC

A time-gated SPAD receiver was designed using 0.35 µm High Voltage CMOS technology. The receiver consists of 81 SPADs in a 9x9 square array, but during each measurement only one 3x3 array is enabled to receive photons. To be able to achieve the highest possible fill factor, all 81 pixels share one common deep n-well, hence the same cathode bias, and the digital electronics needed for loading, quenching and buffering are placed in an array next to the array of SPAD detectors. Each SPAD has an active area of 24 µm x 24 µm. With this configuration, the fill factor for the 9x9 array and the 3x3 array are 43% and 51%, respectively.

The object will be illuminated by means of a unique pulsed laser emitting short 100ps (FWHM) and spectrally narrow pulses at an average energy of more than 1 nJ with a frequency of 100 kHz (up to 1 MHz). The pulsed laser also gives a triggering signal to the on-chip TDC, which serves as a start for measuring the laser pulse transit time. Block diagram the designed receiver is shown in Figure 8.

Figure 8. Block diagram of receiver.


At the beginning of each measurement, one of the 49 possible arrays of 3x3 gets selected. All 81 SPADs are kept quenched by a TDC generated pulse Quench. After arrival of Start signal, the 9 chosen SPADs are loaded by a short 1.5 ns Load pulse to get ready for detecting incident photons during the period of laser excitation. Upon detection of a photon, each SPAD sends a stop signal to the TDC. The on-chip TDC has 10 identical interpolators (1 start and 9 stop channels), and has a single shot precision of about 10 ps (sigma). It is able to measure the transit time of detected photons from all 9 SPADs simultaneously with a resolution of 10 ps. Then, after an adjustable time, the SPADs are quenched and the measured data can be transferred out of the chip in 8 parallel lines, after the rising edge of TDC Ready signal.

The circuit can be configured so that any of the 3x3 sub-arrays can be connected to 9 TDC channels. The developed circuit will be used in 1D radar measurements (single optical axis) and the motivation for the use of the above mentioned array structure in the detector is to relieve the requirements with regard to the precision of the optics and with regard to mechanical tolerances. In many-photon detection, the parallelism can be utilized also to separate the triggerings induced by photons hitting the target from random background induced triggerings. This circuit may pave way for the development of new generation of miniaturized pulsed time-of-flight laser radars.

This configuration of selectable sub-arrays and adjustable time windows gives rise to interesting applications. For example, the gating window can be shifted one clock cycle away from the start signal per measurement, until it covers the whole time range of 640ns. With this time gating approach, we are able to avoid background light and dark counts blocking the SPADs in measuring long distances. The width of the gate window can then be adjusted according to background light, to be able to increase speed by using wider timing windows (hence fewer measurement) in dark environments. Also, it is possible to dynamically change the 3x3 selected array during measurements to follow possible movement of the target image at the detector surface along the distance. The chip is currently under testing and is showing the functionality desired.

Generation and Detection of Electrical/Optical Transients

High-Speed/Energy QW Laser Diodes

A customized asymmetric waveguide laser structure for gain switching has been fabricated using a GaAs/AlGaAs Multiple Quantum Well active layer and its performance has been characterized. Figure 9 shows schematically the refractive index profile in the structure and the calculated transverse mode intensity profile (the structure supports a single transverse mode only). The structure belongs to the Broad Asymmetric Waveguide category in the sense that the mode overlap with the Optical Confinement Layer is smaller than that with the n‑cladding. The active layer consisted of five thin (4 nm thick) GaAs/AlxGa1-xAs (≈ 0.3) Quantum Wells, providing the operating wavelength of 808 nm. To help achieve a large da/Ga ≈ 3 μm, the position of the active layer was shifted away from the mode peak towards the p-cladding. The injection efficiency of the laser was measured as ≈ 0.75, and the internal loss, as ≈ 1.5 cm-1.

Figure 9. Schematic of the refractive index profile and the corresponding transverse mode intensity distribution in the structure analyzed. The active region is represented as a single layer with an averaged refractive index.


Single optical pulses about 100 ps long (FWHM), with an energy of ~ 1 nJ at l ~ 0.808 μm, have been achieved from lasers with a stripe width of 30 μm, using pumping pulses with an amplitude of a few Amps and a duration of ~ 1 ns from a compact MOS driver. Temperature performance of the laser has been investigated experimentally and theoretically. It has been shown that the inevitable performance degradation at high temperatures (either due to ambient temperature or Joule heating in the CMOS source) can be to a significant extent overcome by a relatively modest current increase.

The stripe width was chosen as 30 μm, as a compromise between the need for high power pulse generation and the need to keep the source dimension small. The ratio of the stripe width and the focal length of the transmitter optics defines the field-of-view of the transmitter and this should be kept small (e.g. ~1 mrad) to improve the spatial accuracy and reduce the level of background radiation seen by the receiver. The lasers were mounted n-side down on a ceramic sub-mount, which then was fastened onto the driver board at a right angle with no heat sink.

The laser diode driver used to drive the GaAs/AlGaAs Multiple Quantum Well semiconductor laser is based on an LCR transient pulse shape control. The principle of the driving scheme is shown in Figure 10. In this driver, a capacitor C is first charged to the Vbias voltage (20…150 V) and then rapidly discharged with a switch S1 realized with a MOS transistor. In this configuration the current pulse width is determined by the capacitance and the stray inductance of the current loop (µ(LC)1/2). Pulses with a width of ~1 ns and peak current of ~10 A are available with a MOS transistor switch. As the current pulses driven through the laser diode are short, the average current at a pulsing rate of, say, 100 kHz, is only 0.5 mA. Thus a pulsing rate of 100 kHz to about 1 MHz can be achieved. The achieved optical pulses of the QW laser diode and its injection current pulses with LCR driver using MOS switch are shown in Figure 11 a) and b).

Figure 10. LCR transient-based laser diode driver scheme.



Figure 11. a) Output pulse (red curve) versus injection current pulse (blue curve) for a QW laser diode with 30 mm stripe width and 3 mm cavity length; a) pulsing rate is 10kHz, b) pulsing rate is 1 MHz.


As an example of a practical laser diode transmitter realization, Figure 12 shows a module using a 30 mm / 1.5 mm (note cavity length difference compared to above) QW laser diode and a full-custom CMOS driver, and its output at the pulsing rates of 10 kHz and 1 MHz, respectively, driven with a ~2.5 A / 1 ns current pulse. The CMOS driver IC includes both a HV-MOS switch and a pre-driver so that as few as possible off-chip components can be used to fabricate a compact transmitter.

Figure 12. Construction and optical output of the 30 mm / 1.5 mm QW laser diode at 10 kHz and 1 MHz pulsing rates, respectively.


High-Power Pulsed Emitters for Sub-THz Imaging and 3-D Peculiarities in High-Speed Avalanching BJTs as High-Current Drivers for LDs and UV LEDs.

Si avalanche BJT’s have been most frequently used for nanosecond pumping of pulsed laser diodes, but operation principle of Si avalanche transistors at extreme current densities and with a switching time around 2-3 ns was absent until the last decade. First reliable 1-D and 2-D description of the process we made several years ago, while within last year we have experimentally proved that the parameters of short-pulsing avalanche switching cannot be explained (or predicted) without consideration of fairly complicated 3-D transient phenomena. Very recently, we have mainly finalized the problem in general by description of 3-D transient peculiarities during both delay and fast switching phases.

Much more impressive than avalanche switching in a Si BJT is that in specially designed and manufactured GaAs bipolar structures. Together with high-current/short pulse generation, a very promising (and apparently most important) application for the avalanche switching in GaAs BJT is the generation of pulsed broad-band terahertz emission.

Periodical nucleation and annihilation of ultra-narrow, powerfully ionizing “collapsing” domains is shown to cause the THz emission observed in our experiments. The task of design, development and investigation of high-power pulsed (ns/sub-ns) emitters for a new generation of active sub-THz imagers should be divided into several directions and stages, and this is the major part of our strategic TEKES project (MIWIM) started in 2014. The main direction is the design and development of BJT GaAs-based structures combined with properly designed sub-THz antennas, and using them together with miniature, fast, room-temperature quasi-optical detector based on a Schottky diode. The solution of a large number of related tasks is underway, and the first laboratory examples of transmission sub-THz imaging utilizing not only transmission intensity, but also propagation delay of the pulses across the object with temporal resolution of ~10-30 ps have been presented in several plenary and invited talks. The results from our first prototypes of millimeter-wave radars are at a very early stage however, before their public release as it concerns various technological and experimental details.

First examples of transmission images obtained in both attenuation and propagation delay modes, obtained using a prototype of our completely original sub-THz emitter in combination with commercial Schottky detectors are shown in Figures 13 and 14.

Figure 13. Two photos, X-ray transmission image, and two sub-THz transmission images of a tap. The upper sub-THz image, together with attenuation of the radiation in thicker parts of the tap, displays attenuation (red ring) near the edge of the object. This is caused by the diffraction and creates an erroneous idea on the tap shape if it is hidden in a box, or is a hidden part of a more complicated object. The lower image utilizing the propagation delay is free of this defect of the attenuation imaging. The total delay range between blue and red levels is about 25 ps.


Figure 14. Plastic explosives have chemical composition and THz absorption spectra somewhat similar to sugar and soap. In the photo (up/left) there is a cell from the organic glass filled with granulated sugar in which several pieces of lump sugar and one bar of soap are hidden. X-ray image (up/right) hardly recognizes the lump sugar, and does not show the soap at all. In the attenuation sub-THz imaging the boundaries are well resolved thanks to the diffraction, but the most correct impression on the object is provided by the propagation delay image (the lowest). Moreover, in very homogeneously looking soap bar the propagation delay image shows certain structure with the delay difference (yellow/dark red) of about 10 ps.

Pulsed Time-of-Flight Applications

A Laser Radar Utilizing Single Photon Detection

A recently developed compact laser pulser (with >1 nJ, ~100 ps pulse emission at >100 kHz) was tested in its target application of pulsed time-of-flight laser rangefinding using a single photon avalanche diode (SPAD) detector, see Figure 15 for the blocks of a single photon laser radar.

Test measurements were done outside in full sunshine, i.e., high background noise conditions (~100 klux). A black biplanar target at 50 m distance was used, with the rangefinder focused halfway between the measurement planes. The measurement result in Figure 15 shows the distribution of photon counts with a SPAD gate applied 10 ns before the signal. In this case 13% of reflected photons were detected from the target. The result also shows that if the measurement beam encompasses several targets, different planes separated by ~5 cm can be distinguished. A specific gating scheme was proposed to avoid blocking of the receiver in the case of high background illumination.

The especially high pulse energy of the developed laser pulser was shown to be beneficial and compares well to other compact laser pulsers. Also the pulsing frequency has been improved up to 1 MHz. The effective distance measurement rate in the above conditions can be even 10 kHz or more with cm-level precision to non-cooperative targets.

The measurement speed or maximum measurement distance can be traded off for 2D or 3D measurement capability using a detector matrix. The developed high power/high-frequency pulser is advantageous in this case because the reflected pulse power is spread over the detectors, and therefore less power is available for each individual SPAD.


Figure 15. Up: blocks of a SPAD based radar; below: distribution of measurement results from a target at 50 m in ~100 klux background illumination. Pulsing frequency 100 kHz, pulse energy 1 nJ.


Multiphase Time-Gated Single Photon Avalanche Diode (SPAD) Arrays for Raman Spectroscopy

Raman spectroscopy is based on inelastic scattering, or Raman scattering, of monochromatic light, usually from a CW (continuous wave) laser in the visible, near infrared or near ultraviolet range. Typically, the Raman spectrum is masked by a strong fluorescence background. As a result, in spite of the obvious advantages of Raman spectroscopy, this strong fluorescence background has so far restricted its use in potential applications in the fields of the agricultural, food and oil industries, security control and crime investigations, for example.

It is possible to suppress the fluorescence background to a great extent, if intensive short laser pulses are used to illuminate the sample instead of CW radiation, and by recording the sample response only during these short pulses. The suppression is due to the fact that Raman scattering is introduced immediately after the collision between the photons and the sample material, unlike fluorescence, which is emitted after a delay characteristic to the sample. Thus, by “time-gating” the measurement for only the period of the laser pulse, most of the fluorescence is blocked out from the recorded spectrum.

The block diagram of the pulsed Raman spectrometer and the principle of fluorescence suppression are shown in Figure 16. The material to be analyzed is illuminated with a laser emitting short 70-150 ps (FWHM) and spectrally narrow (~0.2 nm) high-intensity laser pulses. The SPADs of the detector array (SPAD-IC) are enabled by the trigger pulse from the pulsed laser just before each laser pulse, and the photons are counted during short time periods (Dt1, Dt2, Dt3, Dt4), in order to suppress the fluorescence and dark counts (DCR).

Figure 16. Block diagram and time gating principle of the proposed time-gated Raman spectrometer.


A multiphase time-gated 2*4*128 single photon avalanche diode (SPAD) array was designed for the above measurement environment. It was fabricated in a standard high-voltage 0.35 µm CMOS technology. As was mentioned above, the fluorescence photons can be excluded from the Raman result, however not completely. Some portion of fluorescence is occurring at the same time with the Raman scattering, and this residual fluorescence level can be estimated by determining the time constant of fluorescence by means of collecting fluorescence photons with post-Raman time gates. Figures 17 a), b) and c show the differences between the Raman spectra of olive oil sample measured using a) the multiphase time-gated 2*4*128 SPAD array with two time gates (Dt2 to collect Raman photons and Dt4 to collect fluorescence photons) and b) the 2x(4)x128 SPAD array with exactly the same time gate to measure Raman and fluorescence photons (two measurements with two time gate positions), c) a single SPAD element which was moved along the spectrum using a step motor (every spectral point was measured with the same SPAD and with the same time gating).

a) Raman spectrum of an olive oil sample measured with the 2x(4)x128 SPAD detector: time window2 (Dt2) for Raman, time window4 (Dt4) for fluorescence.

b) Raman spectrum of an olive oil sample measured with the 2x(4)x128 SPAD detector, using the same time window4 for Raman and fluorescence.

c) Raman spectrum of an olive oil sample measured with a single SPAD scanned over the spectral range.

Figure 17. Time-gated Raman spectra with different windowing arrangements.


As can be seen, the result shown in Figure 17 b, has a clearly improved signal-to-noise ratio compared with that in Figure 17 a. The improvement can be explained by the fact that the successive time windows, even at the same spectral point (e.g. time window2 and time window4), inevitably have some random variation, which modifies the local number of measured Raman and the fluorescence correction factor. The use of the same time window (and its related on-chip electronics) for collecting both the Raman and fluorescence photons improves the accuracy of the correction for the residual fluorescence, but unfortunately only at the cost of a longer measurement time and increased complexity in the system. Also, by comparing the results of Figures 17 b) and c) it is seen that the SPAD line array non-idealities do affect the result because the result shown in figure 17 c) shows basically the “ideal” result.

A time-gated 4*128 SPAD array with a 3-bit 512 channel flash 80 ps-TDC has been also tested. The flash type 512 channel TDC measures the arrival time of photons of every pixel (4*128) so that Raman photons and fluorescence photons can be distinguished at each of the spectral points, and thus the fluorescence background be excluded from the Raman result. As was mentioned above, some portion of fluorescence is occurring at the same time with the Raman scattering but this residual fluorescence level can be estimated and then subtracted from the result. The principle of this type of time of arrival measurement is shown in Figure 18. As can be seen the last four bins have larger width compared to the first four bins. In this way the measurement range could be expanded, but yet high resolution can achieved in the beginning of the range. Thus, Raman photons can be time-gated with high time resolution and longer fluorescence time constants can also be measured.

Figure 18. Principle of time of arrival measurement of Raman and fluorescence photons.


Figure 19 shows the Raman spectrum of cyclohexane measured with this SPAD array demonstrating the functionality of this time gating technique. Timing skews of the bins of the TDC were also measured in different spectral points. The result is shown in Figure 20 for the three first bins. As can be seen, bin width variation is quite small (s = 10 ps), and the timing skew as a function of a pixel i.e. spectral point is ± 70 ps.

Figure 19. Time-gated Raman spectrum of cyclohexane.


Figure 20. Timing skews of bins 2, 3 and 4.

A third version of the time-gated SPAD array has been also designed with the same technology as the earlier chips, however with a larger array of 16*256 pixels and a 3-bit 256 channel TDC. A test setup has been developed so that the single pixel can stimulated with a short laser pulse in the single photon mode and the time responses of the SPADs can be measured. Preliminary measurements have demonstrated better timing performance than the previous detector chips. The test setup is shown in Figure 21.

Figure 21. Photograph of the test setup.


Circuit Analysis and Linearization Techniques

General Distortion Contribution Analysis

The group has been developing a general distortion contribution analysis technique that operates on top of a standard harmonic balance simulation, and breaks any nonlinear distortion tone into contributions coming from different devices, current and charge sources (VCCS and VCQS), and band-to-band mixing mechanisms. The idea of the analysis is that when the designer knows where the distortion is coming from, he can affect it e.g. by proper harmonic terminations in the important nodes.

The analysis employs the fact that in AWR-Aplac circuit simulator we have access to each individual VCCS and VCQS, and to see the mixing effects, the non-linear source functions are replaced by multi-input polynomial functions, as this allows very easy spectral convolution to be calculated.

This year two technical improvements have been presented, in co-operation with Prof. Jose Pedro from Aveiro University, Portugal.

First, a new concept of port reduced distortion currents was proposed. The reason for this is that the internal structure of modern semiconductor device models varies a lot, and the model-intrinsic distortion contributions do not tell much to the user unless he knows the actual structure of the model. Instead, it is now proposed that for the distortion contribution analysis the entire device model is replaced by a non-linear y-parameter model that has nonlinear two-input polynomial VCCS and VCQS sources in the input (gate-source) port, and other two in the output (drain-source) port. This model is general enough to apply for any transistor model, and distortion currents generated by the input or output conductive or capacitive current sources can be easily interpreted by the users. Building a quasi-static model of such form is straight-forward, but also non-quasi-static (transit delay) effects were handled using an iterative fitting procedure.

The other big improvement is related into the fitting of the polynomial models. In a fairly linear amplifier the input and output signals are very similar, and their effects are difficult to separate, unless they have different spectral contents. When fitting the lumped nonlinear model we very often need a separate training signal that needs to cover the Vgs-Vds space thoroughly, but still resembling the real signal trajectory. Below in Figure 22 the top row shows a situation, where there is a 1-tone signal at gate and another 1-tone at another frequency in the drain. This makes very good coverage, but tends to over-emphasis the high-voltage, high-current region that is normally not visited at all. A better result is achieved by adding some amount of the input tone to the Vds spectrum, as it causes the dynamic load line to stretch in the direction of the load line (second row). Additionally, a small amount of 2nd harmonic at the drain makes the trajectory to cover the actual load line region even better (bottom row).

Figure 22. Actual and fitted MOS I-V curves for different Vgs-Vds excitation trajectories. Pale: the real Ids-Vds trajectory in a transistor; dark: the test signal trajectory used for fitting the polynomial models; thin line: real I-V curves; thick lines: fitted I-V curves.


Error Correction in AD Converters

The group has studied the periodicity and spur location in multi-bit sigma-delta converters employing data weighted averaging (DWA) for averaging mismatch errors. This study was now extended to quadrature sigma delta converters that employ complex-valued signal processing for picking IF signals and cancelling their image frequencies. The proposed generalized DWA-algorithm allows the placement of mismatch-cancelling zeroes to various frequencies, and this was seen to both reduce the spurious tones in the output, and reduce the sensitivity to IQ mismatch.

Biomedical Applications

In co-operation with Biophysics department, electronics for studying nerve signals of small insects has been developed for some years. In 2012 a 3D motion compensation setup was published (preventing a cockroach from piercing itself with the probes when breathing), and now a 16-channel nerve signal measurement IC was designed in a 0.35 um CMOS process, and tested.

This IC consists of 16 multiplexed charge amplifiers, two post amplifiers with programmable gain, a 10 bit successive approximation A/D converter, possibility to excite any probe with an injection signal (generating a nerve stimulus to the subject), and a 2-directional SPI interface to transmit the data. The circuit consumes 1.3 – 1.6 mW of power (depending on the bias and gain settings), delivers a buffered 3.5 Mbps stream of digitized input signals, and has ca. 8 uVrms input reduced noise. A photograph of the chip is shown in Figure 23. The test setup includes also a CPU card that receives the SPI bus signal, and transmits it to a PC via a USB bus.

Figure 23. Photograph of the 16-channel nerve signal measuring circuit. Preamplifiers are on the left, post-amplifiers in the center, and logic and AD converter on the right.





senior research fellows


postdoctoral researchers


doctoral students


other research staff




person years for research



External Funding



Academy of Finland

750 000


230 000

other domestic public

230 000

domestic private

15 000


15 000


1 240 000


Selected Publications

G. Duan, S.N. Vainshtein, J. T. Kostanmovara, V. E. Zemlyakov, V. I. Egorkin "Three-dimensional properties of the switching transient in a high-speed avalanche transistor require optimal chip design", IEEE Electr. Dev, Vol.61, No.3, pp. 716-721, 2014.

L. W. Hallman and J. Kostamovaara, ”Detection jitter of pulsed time-of-flight lidar with dual pulse triggering”, Rev. Sci.Instrum. 85, 036105 (2014);

S. Vainshtein, J. Kostamovaara, “Transmission subterahertz imaging utilizing milliwatt-range nanosecond pulses from miniature, collapsing-domain-based avalanche source”, chap-ter 23 (pp. 175-187) in “Terahertz and Mid Infrared Radiation: Detection of Explosives and CBRN (Using Terahertz)“, Springer, series: NATO Science for Peace and Security Series B: Physics and Biophysics, editor: Mauro Pereira (ISBN 978-94-017-8571-6, 194 p. 103 illus., 70 illus. in color), 2014.

I. Nissinen, J. Nissinen, P. Keränen, A-K. Länsman, J. Holma and J. Kostamovaara,” A 2x(4)x128 Multi-time-gated SPAD Line Detector for Pulsed Raman Spectroscopy", IEEE Journal of Sensors, vol. 15, no. 3, pp. 1358-1365, 2015.

J. Kostamovaara, J. Huikari, L. Hallman, I. Nissinen, J. Nissinen, H. Rapakko, E. Avrutin, B. Ryvkin, ”On laser ranging based on high speed/energy laser diode pulses and single photon detection techniques”, IEEE Photonics Journal, Vol. 7, Issue 2, April 2015, pp. 1-15, Digital Object Identifier: 10.1109/JPHOT.2015.2402129

J. Aikio, T. Rahkonen, J. Pedro : “Improved polynomial fitting technique for distortion contribution analysis”. IEEE Trans. on Microwave Theory and Techniques. DOI: 10.1109/TMTT.2014.2367530

M. Neitola, T. Rahkonen: “Generalized quadrature data weighted averaging”. IEEE Trans. on Circ. Syst.-II, vol 99 (Nov 2014). DOI: 10.1109/TCSII.2014.2368620

J. Aikio, T. Korkala, T. Rahkonen: “Analysis of band-to-band mixing distortion contributions in some usual circuit topologies”. Springer j. Analog integrated circuits and signal processing, May 2014. DOI: 10.1007/s10470-014-0320-2

J. Aikio, T. Rahkonen, V. Karanko: “Polynomial Fitting of Nonlinear Sources with Correlating Inputs”. Compel International Journal for Computation and Mathematics in Electrical and Electronic Engineering, Vol 33 no 4, pp. 1097 - 1106. DOI

T. Rahkonen, J. Aikio: “Analyzing Distortion Contributions in a Complex Device Model”. Compel International Journal for Computation and Mathematics in Electrical and Electronic Engineering, Vol 33 no 4, pp. 1264 - 1271. DOI

M. Tanveer, I. Nissinen, J. Nissinen, J. Kostamovaara, J. Borg and J. Johanson, ”Time-to-Digital Converter based on Analog Time Expansion for 3D Time of Flight Cameras”, Proceedings of the IS&T/SPIE Electronic Imaging Conference, February 2014, San Francisco , USA, 6p.

B. Lanz, J. Kostamovaara, “Current pulse length investigation toward optimal pumping of an asymmetric wave-guide laser diode structure aiming at optimized advanced gain-switched high peak-power picosecond pulses”, SPIE Proceedings of the Laser Technology for Defence and Security X (Conf. 9081), Baltimore, Maryland, USA, 5 - 9 May 2014

B.S. Ryvkin, E.A.Avrutin, J.T. Kostamovaara, D.V. Kuksenkov, ”High brightness picosecond pulse source by repetitively gain-switching a strongly asymmetric waveguide laser diode”, Proceedings of the SIOE’2014, Cardiff, 29 April-2 May 2014

B.S. Ryvkin, E.A.Avrutin, B. Lanz, J.T. Kostamovaara, ”Strongly asymmetric waveguide semiconductor lasers for picosecond pulse generation by gain- and Q-switching”, INVITED PAPER in the Proceedings of the 16th International Conference on Transparent Optical Networks, Graz, 6-10th July 2014

E.A.Avrutin, B.S. Ryvkin, J.T. Kostamovaara, ”Repetitively actively gain-switched strongly asymmetric waveguide laser diode for high brightness picosecond pulse generation, Proceedings of the 16th International Conference on Transparent Optical Networks, Graz, 6-10th July 2014

M. Kögler, L. Kurki, M. Tenhunen, R. Aikio, A. Härkönen, J. Kostamovaara, J.R. Tenhunen, “Fluorescence rejection in Raman detection using novel time-gated approach”, International Conference On Raman Spectroscopy (ICORS2014), 10 - 15 August 2014, Jena, Germany

L. W. Hallman, J. Huikari, J. Kostamovaara,”A high-speed/power laser transmitter for single photon imaging applications”, Proceedings of the IEEE SENSORS conference in 2014, 4p., 2014, 10.1109/ICSENS.2014.6985213

S. Kurtti, J. Kostamovaara, ”CMOS receiver for a pulsed TOF laser rangefinder utilizing the time domain walk compensation scheme”, Proceedings of the 20th IMEKO TC4 International Symposium and 18th International Workshop on ADC Modelling and Testing Research on Electric and Electronic Measurement for the Economic Upturn, Benevento, Italy, Sept. 15-17, 4p., 2014

I. Nissinen, J. Nissinen, J. Holma and J. Kostamovaara, ”A TDC-based 4x128 CMOS SPAD Array for Time-Gated Raman Spectroscopy”, Proceedings of the ESSCIRC 2014 40th European Solid-State Circuits Conference September 22-26, 2014 - Venice, Italy, 4p., 2014

I. Nissinen, J. Nissinen, J. Holma and J. Kostamovaara, ”Cross talk measurements of a time-gated 4x128 SPAD array for pulsed Raman spectroscopy”, In proc. Norchip 2014 conference Tampere, Finland, Oct 27-28, 2014, 4p. ISBN: 978-1-4799-5441-4

A. Mäntyniemi, J. Kostamovaara, “A time-to-digital converter (TDC) architecture based on a ring oscillator and successive approximation”, In proc. Norchip 2014 conference Tampere, Finland, Oct 27-28, 2014, 4p. ISBN: 978-1-4799-5441-4

O. Kursu, T. Rahkonen: “Integrated circuit for neural recording and stimulation”, In proc. Norchip 2014 conference Tampere, Finland, Oct 27-28, 2014, 4p.ISBN: 978-1-4799-5441-4

Last updated: 16.3.2015