next up previous contents
Next: Simple Model of an Up: Power Dissipation in RSFQ Previous: Power Dissipation in RSFQ

Introduction

Power dissipation in a typical RSFQ circuit comes from two main sources. First, power is dissipated in the shunt resistor when the junction is in the resistive state during generation of an SFQ pulse. Total energy tex2html_wrap_inline1829 dissipated during this process of a tex2html_wrap_inline1573 phase jump is of the order of

  equation317

Here tex2html_wrap_inline1489 is the critical current of the junction and tex2html_wrap_inline1835 is the magnetic flux quantum. As discussed in the Introduction, tex2html_wrap_inline1489 cannot be made smaller than a certain limit ( tex2html_wrap_inline1839 ) set by thermal fluctuations [1] and, therefore, nothing can be done to further minimize tex2html_wrap_inline1829 (if we do not go after the reversible computation [41] which is slow and hardware consuming). When a part of an RSFQ design operating at clock frequency f, Josephson junction lets through of the order of f SFQs per second and dissipates power

  equation323

The main source of power dissipation, however, is the second contribution: DC power dissipated in the biasing resistor. Biasing of the junctions is essential for RSFQ devices. Consider for example a stage of JTL in Fig. gif.

  figure327
Figure:   JTL with a biasing line

Instead of the short-hand symbolic notation for the biasing current sources which was used throughout the Introduction and is convenient for circuit simulation, schematic in Fig. gif is closer to the physical layout with biasing resistors (see Fig. gif) connected to a common voltage source V. Bias current tex2html_wrap_inline1849 should be smaller than the critical current tex2html_wrap_inline1489 of the junction but close enough to it so that junction is capable of making a tex2html_wrap_inline1573 jump in phase under the influence of an incoming SFQ pulse. Correct operation of the circuit limits variation of the bias voltage V (and, hence, of the tex2html_wrap_inline1849 ) to a certain range

equation333

roughly limited from above by the tex2html_wrap_inline1859 and from below by the minimum bias when a junction still can make a tex2html_wrap_inline1573 phase jump under the influence of an SFQ pulse. This variation can be found in simulation and measured in experiment and usually is presented in the form of a relative bias margin tex2html_wrap_inline1863 :

  equation341

Empirical rule of thumb in RSFQ is that it is hard to make a complex circuit with more than a tex2html_wrap_inline1865 bias margin, so we can crudely accept that tex2html_wrap_inline1867 . Although other biasing schemes may be possible, the one depicted in Fig. gif seems to be the most practical, especially for large designs. Power dissipated in biasing resistor is simply

equation358

Traditionally, all RSFQ layouts were done for the value of bias voltage close to tex2html_wrap_inline1869 which is roughly the gap voltage of the Nb. The reasons for this choice were that, first, sufficiently large bias voltage ensures that all the effects of junction interaction with the bias line (which are the main topic of this chapter) are minimal and, second, this voltage could, in principle, be stabilized by the current step at tex2html_wrap_inline1871 of an unshunted Josephson junction. This choice of bias voltage, however, results in a very large dissipation of DC power tex2html_wrap_inline1873 as compared to the fundamental limit tex2html_wrap_inline1875 :

  equation360

The last numerical estimate was done for tex2html_wrap_inline1869 and tex2html_wrap_inline1879 . Eq. (gif) shows that, although it is impossible to push the bias voltage below the fundamental limit of tex2html_wrap_inline1881 , a considerable (of the order of 10) gain in power dissipation can be achieved simply by lowering the bias voltage V closer to that limit. The dangers of this straightforward approach include possible shrinking of the bias margins and interaction of the junctions through the bias line so that a detailed study of the processes in an RSFQ biasing line is necessary.


next up previous contents
Next: Simple Model of an Up: Power Dissipation in RSFQ Previous: Power Dissipation in RSFQ

Alexander Rylyakov
Fri May 23 18:57:25 EDT 1997