[SOLVED] Effect of current direction on mismatch within current mirror

[Pat_Mustard]

I'm looking for a bit of depth to my understanding of matching techniques.

As I understand it to achieve a well matched mirror it is important to match the direction of current flow in the reference device and the mirrored device, i.e. if you have two transistors sinking 1uA drain current option 1 below would be better as current would flow left to right in both the reference and mirror

Device 1 Device 2

1. Drain Gate Source Drain Gate Source

2. Drain Gate Source Source Gate Drain

Option 2 allows us to share the sources however current would flow from left to right in the reference, and right to left in the mirror. I understood that as the silicon is crystaline direction of current flow is important. I thought it was due to impedance in the active area but don't understand how the impedance can be different when the current is taking the essentially the same path, albeit in another direction.

Can anyone explain the mechanism that means current direction should be uniform in current mirrors? Many thanks

 

[erikl]

Because of photolithographic invariance (PLI), s. this excerpt from Allen-Holberg's book:

 

[Pat_Mustard]

So implant shadowing is definately one reason, thats good I have an explanation.

I have also read that the crystaline structure of silicon has an effect. I can understand why this would lead to mismatch between devices that are at right angles to each other, however I don't understand how the crystal nature of silicon would cause mismatch between two devices orientated identically? Does anyone have an explanation why it could, or is it simply not the case and only reason for matching current direction given in Erikl post.

Cheers

 

[erikl]

So implant shadowing is definately one reason

Essentially due to the usually adjusted implant angle of 7° (against the vertical, into <100> silicon).

I don't understand how the crystal nature of silicon would cause mismatch between two devices orientated identically?

Never heard of. Can't imagine any reason for mismatch due to symmetry in the same crystal orientation.

 

[Pat_Mustard]

Cheers for the input Erikl, I can only think the crystal structure idea is when a device is placed perpendicular to the reference and I have got mixed up.

Many thanks.

 

[erikl]

I can only think the crystal structure idea is when a device is placed perpendicular to the reference and I have got mixed up.

Yes, possibly. Orthogonal devices surely would show mismatch, not only from different crystal orientation, surely from missing PLI (different shadowing).

 

[dick_freebird]

If you want matching in the end, begin with absolute geometric
identicalness. Anything else and you'll just spend intellectual
effort on fooling yourself.