Design of a 5M Resistor

[chikaofili]

Hello,
I am a newbie in Analog layout design. I have to make a resistor (using nwell or the other types) that will fit a space of 100u by 8u (thats the maximum space available. Is it possible? and any suggestions? It does have to be accurate. It has to be at least 5M for the resistor to work.


I am using TSMC 0.18um and working with cadence at the moment.



Any suggestion will be appreciated

Thank you!

 

[erikl]

... a space of 100u by 8u It has to be at least 5M for the resistor to work.
I am using TSMC 0.18um ...

If you don't have to match this resistor with another one, you can probably do it: With a min. resistor width of 0.18µm you can get 22 lanes within a space width of 8µm (lane width = lane distance; check your DRC rules!), and about 555 ⃤ (squares) per 100µm space length, which makes 22*555=12210 ⃤, say at least 12000⃤. 5000000Ω / 12000⃤ = 417Ω/⃤ , i.e. with a ≥ 500Ω/⃤ resistor value this is possible. A non-salicided nwell resistor will probably offer such a value -- may be a non-salicided poly resistor as well, check your PDK doc.

An nwell resistor has non-negligible temperature and voltage dependencies -- also check for those.

 

[chikaofili]

Thank you for your response.
After checking the DRC rules. I realized that only an nwell will work best for this case.

I have a stupid question: Does the nwell resistor have to obey the minimum dimensions of a typical nwell? Because I was looking at the CMCLayout reference and the widths of the resistors drawn were much smaller than that specified in the logic layout rules for a typical nwell.

If you don't have to match this resistor with another one, you can probably do it: With a min. resistor width of 0.18µm you can get 22 lanes within a space width of 8µm (lane width = lane distance; check your DRC rules!), and about 555 ⃤ (squares) per 100µm space length, which makes 22*555=12210 ⃤, say at least 12000⃤. 5000000Ω / 12000⃤ = 417Ω/⃤ , i.e. with a ≥ 500Ω/⃤ resistor value this is possible. A non-salicided nwell resistor will probably offer such a value -- may be a non-salicided poly resistor as well, check your PDK doc.

An nwell resistor has non-negligible temperature and voltage dependencies -- also check for those.

 

[srivatsan]

I would think that POLY based resistor would be apt too even though its silicided.

Irrespective of the material chosen i.e. WELL/POLY, it is recommended to do more than the minimum width owing to etching and granularity/deposition undersize./oversize. For example, 0.18um poly might be around 0.175 in which case the resistance increases than expected. If the width is around 0.5um, if its 0.495um, the value does not change that much (% difference).

 

[chikaofili]

Thanks for your response.

But the thing is, I am not really concerned about the accuracy of the resistor value. But I need the resistor to be REALLY BIG (at least 5M).

It is just used for the output stage of an amplifier.
But I am restricted with a 100u by 8u.

When I tried using the Pcells from CMC. I could only get about 1M using poly.

But If I have to design with an nwell. I am not sure how much the width has to be because the minimum width of an nwell is much bigger than 0.18um and I will not be able to fit more than 5 lanes in a width of 8um

 

[erikl]

Does the nwell resistor have to obey the minimum dimensions of a typical nwell?

You always have to respect the DRC rules.

I am not sure how much the width has to be because the minimum width of an nwell is much bigger than 0.18um and I will not be able to fit more than 5 lanes in a width of 8um

You don't need an individual nwell for each resistor lane: you can use a common nwell for all lanes (in series), so you can get 22 (or so) meandering lanes of 0.18µm width into an 8µm wide nwell. Connect this nwell to the highest positive potential (VDD).

 

[leo_o2]

Choose higher sheet resistance poly resistor. Some process provides such options like 5K/sq or 10K/sq.
Please discuss with the foundry.