I am using Magic to draw some transistors and create digital logic gates. While I was studying the theory about MOSFET I've always seen images like the one below.

enter image description here

In this picture, one can see that below the gate, in the channel, there is substrate. In other words, the n-diffusion is not continuous. But when we are using Magic, the diffusion we draw is continuous as shown in the following image.

enter image description here

In Magic, the intersection between polysilicon and a diffusion generates a layer named ptransitor/ntransistor, but it is not clear to me if this represents a channel or not. So I've tried to export to GDSII and check if Magic would generate a hole/space/channel in order to split the diffusion, but even in the generated GDSII the n-diffusion seems to be continuous like the image below. I expected to see three boxes: first a diffusion, then the polysilicon and then another box, but in the GDSII there's only a rectangle that crosses the polysilicon, just like in Magic. I view the generated GDSII in gds3d and KLayout.

enter image description here

What I want to know is the time that the channel is generated. I can understand that in Magic we draw the diffusion crossing the polysilicon as this helps in speeding the drawing and also may help the algorithm for circuit extraction. But I can't understand why there is no channel in the generated GDSII, which is supposed to be the file that we send to the foundry for fabrication. Does the foundry process the GDSII and generate the channel, or is the channel already in the GDSII and I am not seeing it?


1 Answer 1


If the process uses self-aligned gates, there's no need to specify the channel location. Everywhere the diffusion is covered by poly, there's a channel. The poly acts as a mask for the diffusion. Here's a Wikipedia page on the subject:


  • \$\begingroup\$ Thank you, @Adam. If I understood correctly, in the GDSII the diffusion is indeed continuous, but at fabrication time, the gate is formed first and only then the diffusions. Now the gate acts like a physical mask, thus eliminating the need for a logical mask generated by software. Is this correct? \$\endgroup\$
    – hdhzero
    Mar 22, 2016 at 6:06
  • \$\begingroup\$ Yes, that is how it works. \$\endgroup\$ Mar 22, 2016 at 6:55
  • \$\begingroup\$ @hdhzero Yes, that's correct. \$\endgroup\$
    – Adam Haun
    Mar 22, 2016 at 13:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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