I'm studying the topic of cloning electronic circuits for my research. And I have a scenario as the following:

Chip A has two ring oscillators (RO1a and RO2a). And RO1a is faster than RO2a. Chip B has two ring oscillators (RO1b and RO2b). And RO1b is slower than RO2b. Now, if I want to clone chip A using chip B, I must somehow make RO1b becoming faster than RO2b.

One way I can think of is to shoot laser or UV light at the inverters of RO2b for a long time (of course, not too long because it will burn the inverters so bad that they will become nonfunctional). My reason for this approach is that laser can induce extreme heat. And if we subject the inverters to high temperature for a long period of time, it will age the transistors much quicker (I believe this is called aging effect in electronic circuits). As the inverters age, they will become slower. However, I am not an expert in semiconductor physics to really say with certainty that this approach will actually work. Can someone (who has studied a great deal of semiconductor physics) confirm to me that this approach will work or won't work? Can you also give me some references to justify your answer?

  • 2
    \$\begingroup\$ Why not use the laser to modify the trace impedance instead? \$\endgroup\$ Nov 6 '13 at 0:52
  • \$\begingroup\$ Can you elaborate more on how exactly the trace impedance can be modified by the laser? \$\endgroup\$
    – abc
    Nov 6 '13 at 2:50
  • \$\begingroup\$ I'm not sure I understand your "cloning"; normally the process is to effectively "photocopy" a chip layer-by-layer, which produces identical behaviour as if it came off the same factory production line. \$\endgroup\$
    – pjc50
    Nov 6 '13 at 13:57
  • \$\begingroup\$ (identical up to a small tolerance, admittedly) \$\endgroup\$
    – pjc50
    Nov 6 '13 at 13:58
  • \$\begingroup\$ I don't think "photocopy" a chip layer-by-layer will produce an identical copy of that chip. And the reason for it is process variation (it's a technical term and you can look it up). So even with the same design, same manufacture, same silicon wafer, the two chips you manufacture will always be slightly different. \$\endgroup\$
    – abc
    Nov 6 '13 at 17:00

It sounds like you want to use something similar, yet inferior, to laser trimming. Aging is a non-linear process and depends on your substrate and doping. Not only that, but typical accelerated aging processes (by heating) can take days. For instance, the graph below from this paper

enter image description here

That's BaTiO3 with different dopants.

It's hard to find much information about this, likely because it's not a very great idea.

  • \$\begingroup\$ Just for clarification: Which is not a very great idea? and why do you think it is not a very great idea? \$\endgroup\$
    – abc
    Nov 6 '13 at 3:15
  • \$\begingroup\$ It's not very selective, it reduces the chip life (that's the whole idea behind accelerated ageing), it's probably not going to be particularly precise or repeatable. A better solution is well-defined laser trimming where you will selectively cut elements with a laser to get the desired performance characteristics. Now, this is not done chip-wide, just on the relevant bits as it is very time consuming and hence expensive. \$\endgroup\$ Nov 6 '13 at 6:25
  • \$\begingroup\$ I looked up laser trimming and it seems like it's only used for increasing the resistance of resistor. But nobody ever used it for CMOS circuit. Does anyone know if laser trimming has been successfully used to alter the delay of CMOS circuit? \$\endgroup\$
    – abc
    Nov 6 '13 at 17:05
  • 2
    \$\begingroup\$ @abc It's becoming more clear that you don't understand enough about electronics for us to convince you that this is not a great idea. If you want to slow an RC circuit, what would you do to the R? \$\endgroup\$
    – Samuel
    Nov 6 '13 at 17:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.