Logo PTB

Measurement of the Impacts of Ionizing Radiation on 90 nm CMOS Circuits

General description of the project
The FATAL (A modelling framework for Fault-Tolerant Asynchronous Logic) project is aimed at developing a simulation model for asynchronous logic which takes the impacts of radiation effects on the circuit into account. On the basis of this model, it shall in future be possible to develop asynchronous logic which is resistant to radiation effects. By interaction with the crystal lattice of the silicon chip, high-energy particles generate free charge carriers which then lead to current and voltage pulses in the circuits. In the case of pulses, they are called Single Event Transients (SETs) and if a permanent state change is concerned, they are referred to as Single Event Upsets (SEUs). In a synchronous logic, the circuit is susceptible to disturbances by SETs only during the clock edge. For design and simulation software of synchronous logic it is, therefore, sufficient to know the occurring pulse width of the SETs. Therefore, extremely efficient design and simulation software is already available for the development of radiation-tolerant synchronous circuits. For this reason, in particular the pulse width of the occurring SETs has been determined in literature. In the case of asynchronous logic, there is no clock. These circuits are, therefore, susceptible to radiation-induced faults at any time. In addition, asynchronous logic uses voltage transitions for both handshaking and for the data. A SET is a voltage pulse which corresponds to two voltage transitions (0→1 and 1→0 or vice versa). These can easily be misinterpreted and, thus, lead to a fault. For the development of a model for asynchronous logic it is, therefore, necessary to know the exact aspect of the pulse shape and how these pulses propagate inside the circuits. The measurements carried out at PTB were aimed at measuring these pulse shapes directly on the chip and at investigating the propagation of the pulses inside the integrated circuit. In literature, no voltage measurements of SETs carried out directly on the chip can so far be found.

The simulation model
The final model must be as simple as possible to allow also complex circuits to be investigated in a reasonable calculating time. The verification and optimization of this simplified model requires, on the one hand, the carrying-out of measurements and, on the other hand, the realization of complex 3D semiconductor simulations. Especially the results of the long 3D semiconductor simulations of different basic circuits improve the understanding of the internal processes and allow the simple model to be verified. The results of the 3D semiconductor simulations depend, however, strongly on process parameters such as, for example, the doping concentration. For this reason, also this 3D model must be calibrated by means of measurements. This calibration requires an exact measurement of the occurring pulse shapes. We have already performed such measurements on the microbeam facility of the GSI with gold ions with an energy of 945 MeV. These energetic heavy ions generate large quantities of free charge carriers in the silicon chip and, therefore, lead to large voltage pulses. For a reasonable calibration it is necessary to carry-out additional experiments with particles which lead to clearly different ionisation densities and, thus, a clearly different quantity of generated charges. For the experiments at PTB, alpha particles with 8 MeV were used. The resulting Linear Energy Transfer (LET), which is a measure of the quantity of the generated charge, is smaller by a factor of approx. 170 than that of the gold ions (945 MeV).

The measuring system
The 3D semiconductor simulation has shown a strong dependence of the SETs on the position of the impact. For the measurements it is, therefore, necessary to exactly know the position of the impact. In addition, the voltage pulses caused by the ions shall not superimpose. It must, thus, be guaranteed that always only one single ion hits the chip at the same time. These preconditions are excellently met by the microbeam facility at PTB. Single ions can be positioned with an accuracy of approx. 2 µm to 3 µm.

The circuit nodes cannot, however, be simply connected to an oscilloscope, as the additional load of the oscilloscope would distort the pulse and - beyond that - impair the function of the circuit. For this reason, buffer amplifiers were integrated directly at the chip to load the internal circuit nodes only minimally. Their output can be connected directly to the oscilloscope. The chip itself is bonded directly to a high-frequency printed circuit board to distort the pulse shapes as little as possible. As to the connection to the oscilloscope, attention was also paid that efficient and short cables with an optimal impedance were used.

Figure 1: Measuring setup with measured SET on the oscilloscope

During the measurement, the surface of the chip was scanned with the microbeam. As soon as a pulse could be seen on the oscilloscope, it generates a trigger signal. This trigger signal stops the microbeam and the signals recorded with the oscilloscope are stored together with the current beam position . This allows the impact place of the ion to be assigned - after the measurement - to each measured pulse with an accuracy of approx. 3 µm. After an event has been recorded, the circuit is newly initialized, and the microbeam is started again.

Results
Figure 1 shows the beam line of the microbeam facility, the positioning device, the HF-PCB and the oscilloscope with a recorded SET. Figure 2 shows an enlarged section with the HF-PCB and the microbeam exit nozzle. In spite of their high technical complexity, the measurements were extremely successful and provided an important and firm basis for the model development and its verification.

Figure 2: HF-PCB and microbeam exit nozzle

This work is supported by the Austrian Science Foundation (FWF) under project number P21694.

Contact persons:

Opens window for sending emailU. Giesen(at)ptb.de, Department 6.5, Working group 6.54,
Opens window for sending emailM. Hofbauer, K. Schweiger, H. Dietrich, H. Zimmermann, Institute of Electrodynamics, Microwave and Circuit Engineering, TU Wien
Opens window for sending emailU. Schmid, Institute of Computer Engineering, TU Wien