native devices as input diff pairs

[szekit]

anyone has experience with that? how was it?

my bandgap is higher than it should be on silicon (and poor PSRR) and suspect it is due to the usage of native N devices as the input pairs on the op amp. (conventional bandgap).

thanks!

 

[jgk2004]

I have just made a sigma delta converter with all my amplifiers having native Nmos for input pairs for CMR reasons. Everything works fine and as to what I see I am having no problems. I am wondering if this is common practice thou.. Has anyone else made amplifiers with Native inputs and had any problems?
Jgk

 

[tuvialiran]

So far, we heard only about the advantages of native devices - low Vt, better matching, lower noise. The only disadvantage is the lower BVDSS, which requires large gate length, as defined in the design rules.

If so, why analog designers are using standard NMOS in most cases, while the use of native devices is considered as non-standard?

Can any one explain this conflict? Are there any disadvantages to these devices?

 

[jgk2004]

Like you said, the one disadvantage is you can't use minimum gate length. The other disadvantage, is that sometimes they aren't always available. I think thats why people don't always use them... As for the statement, is it considered a non-standard... I guess you could say YES, because some designers, and I mean experienced ones, don't know what they are and then just don't use them.....

If you got them, use them when you can!

Jgk

 

[tuvialiran]

Thank you. Now, I would like to progress one step forward.

Does anyone have know-how and experience in using native NMOS devices in sub-threshold region?

 

[jgk2004]

Well yes of course! They work great! I use them all the time as input pairs in my amplifier designs.. in weak inversion to have higher GM and also use them in wide swing current mirrors.

like I said.. if you have them use them!

Jgk

 

[dick_freebird]

If you want to use them subthreshold, you would need a good idea of
how high the leakage floor can be, and respect that by at least a
decade. Otherwise you may operate in a place where you have zero gm
(gate voltage has zero effect on drain currrent).

These devices can be poorly modeled and controlled, and the light
(or un-) doped channel is prone to wander more than explicitly shot
implants as well as being sensitive to local charging effects. What your
foundry signs up for, in these respects, is on you to determine.

 

[tuvialiran]

Thank you Dick

Can you explain the relation between lightly doped channel and wander?

In addition, can you explain what is the charging effect, and why these devices are more sensitive to these effects?