The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Mixed-Signal Design >> Designing a VDD/2 reference (rail splitter)
https://designers-guide.org/forum/YaBB.pl?num=1261464901

Message started by jiesteve on Dec 21^st, 2009, 10:55pm

Title: Designing a VDD/2 reference (rail splitter)
Post by jiesteve on Dec 21^st, 2009, 10:55pm

Hi,

I'm designing a VDD/2 reference that splits a 1.5V supply. Error needs to be ~1mV or less, and the generator needs to be low impendance.

One way to do it:
I could use a simple resistor divider, but I need switches in series with the divider to turn it off when not in use... These switches will contribute a Resistance->Voltage error... Does anyone know of any schemes to try and cancel the error contribution from the switches?

If anyone can think of any other good schemes, lemme know.

Thanks!

Title: Re: Designing a VDD/2 reference (rail splitter)
Post by wave on Dec 22^nd, 2009, 9:44am

So does this mean your supply is as clean as 2mV, or are you drinking too much egg nog?

You probably need a trimmed BG Reference and a fraction of that.

:D
Wave

Title: Re: Designing a VDD/2 reference (rail splitter)
Post by jiesteve on Dec 22^nd, 2009, 10:49am

wave,

Actually the error should be specified from the ideal VDD/2 output -- this reference voltage needs to track the supply. So it can't be BGV-based. The error should be small.

Title: Re: Designing a VDD/2 reference (rail splitter)
Post by jiesteve on Jan 7^th, 2010, 5:41pm

I've convinced myself that the switch device doesn't make a big deal -- since its resistance is small compared to the ladder it has a negligible effect on the output voltage (vs process and vs temp)

I need very fine steps -- ~2mV range. With a 1.5V input voltage, that's 1.5V/2mV = 750 segments. I need the 2mV step size to only correct for variation from the ideal midpoint value, so quantization error is ~1mV max.

In order to cut down on the large number of (small L) segments, is it possible to group into fewer segments near the rails (with larger L) and use the smaller segments around the midpoint to correct for process variation? I am using poly resistors.

Anything I should watch out for here?

Thanks!

Title: Re: Designing a VDD/2 reference (rail splitter)
Post by loose-electron on Jan 8^th, 2010, 1:17pm

that is an awful lot of fanciness to get a mid scale voltage.

A thousand odd resistors will eat up a lot of space.

suggest you look at your system and ask why does a mid scale reference need to be so bloody accurate?

Title: Re: Designing a VDD/2 reference (rail splitter)
Post by eecs4ever on Feb 24^th, 2010, 11:59am

hi,

see attached picture

Title: Re: Designing a VDD/2 reference (rail splitter)
Post by eecs4ever on Feb 24^th, 2010, 12:07pm

to be able to turn off the R ladder, we need some series switches.

If this VDD/2 node doesnt draw DC power, then we can use a structure shown below.

As long as we make the nmos / pmos resistance small enough compared to the resistor R value. Only 10% matching between the N and P is required.

Title: Re: Designing a VDD/2 reference (rail splitter)
Post by jiesteve on Feb 24^th, 2010, 5:54pm

Thanks eecs4ever.

Right now what I have is just a very large NMOS at the bottom of the ladder (without the PMOS in your scheme). I like your scheme because if the N and P switches are sized to have roughly equal resistance, then the tap point is right at the center, whereas in my scheme I have to compensate for the NMOS IR rise. Are there any other advantages?

I'm wondering if your scheme suffers from more variation at cross corners (slow P, fast N and vice versa). I guess this is mitigated by having them quite large.

Title: Re: Designing a VDD/2 reference (rail splitter)
Post by eecs4ever on Feb 25^th, 2010, 9:17pm

Hi Jiesteve,

I think its fine either way, as long as your resistance is dominated by the resistor R. Keep the mosfets large so that its variation doesn't hurt your accuracy.

One thing to keep in mind is that mosfet do have gate leakage, especially for the thin-oxide devices in the lower technology nodes. When you are trying to lower the ON resistance, its probably you will need a very large transistor. simulate the gate leakage and make sure it doesn't hurt you. Gate leakage tend to have large variations from run to run and varies exponentially with temperature and voltage.

Title: Re: Designing a VDD/2 reference (rail splitter)
Post by loose-electron on Feb 26^th, 2010, 1:53pm

NMOS and PMOS do not scale together and do not track each other

AnalogDE wrote on Feb 24^th, 2010, 5:54pm: