MOS process variation

"slow" or "fast" refers to the process, not individual devices.

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | |

| 1962 | I love to cook with wine. Sometimes I even put it in the food.

Yes, "slow" and "fast", particularly WRT the subject line, refer to the processing of the wafer. Wafers can be intentionally processed for "fast" transistors, possibly at the cost of power, yield, or both.

There are also "fast" and "slow" transistors within a wafer (low Vt vs. high Vt, Thin vs. thick gate ox., for instance).

--
  Keith

With shitty processing. Wafers tend to be relatively uniform across a wafer, and quite uniform across a single die. Otherwise analog stuff would never work ;-)

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

But could transistors with different characteristics be _intentionally_ put on a die? I think that's what Keith is getting at.

Thanks, Rich

Not at all. It is done quite intentionally.

It depends on the parameters of the transistor in question. If one needs a *FAST* transistor, one selects a thin-ox with a low-Vt. If it needs high voltage on the gate, thin-ox doesn't work so well. If it doesn't need speed, one chooses a transistor with thick-ox and high-vt (saves power). ...and all variations inbetween. That's one of the knobs (powering of gates is another biggie) the timing folks twist.

Not everyone wants every transistor to be equal. There are good reasons to tweak individual transistors.

--
  Keith

Sure. It's done _all_ the time in advanced processes. One selects the parameters needed for each transistor, if need be.

--
  Keith

I've just not seen "fast" and "slow" as descriptors for such processing. I deal with multiple device TYPES all the time.

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

half-fast ?:-)

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

We don't really call them "fast" or "slow" either. Rather "High Vt"/"Low Vt" or "Thin-Ox"/"Thick-Ox" (or something inbetween). Depending on the application, either are used for fast/slow choices. ...or maybe "fast/not-fast" or "fast/who cares" are better choices of descriptors. ;-)

--
  Keith

Aw, Jim. You peeked at our spec! ;-)

--
  Keith

Of course. Analogue design is a *trade off* of various parameters. There is no "one size fits all". For example, right now I am designing a HV cmos amplifier using extended drain transisters to obtain the required

50V voltage rating. Extended drain transisters require several times more area then a conventional device, so one only uses them when there is HV. In addition, a penalty of HV rating is lower speed, e.g you cant usually use the minium process gate length.

Other variations are "natural devices". For an nmnos, a natural can have a Vt of around OV. This is useful in source followers. It can allow the output to be pulled much closer to the supply. Some devices are isolated in a well. This avoids the backgate bias increasing the threshold voltage as you can connect the backgate to the source, e.g. in diff pairs. These devices are larger then non isolated ones, hence again a trade of size verses performance.

I could go on, but am sure one gets the picture...

Kevin Aylward snipped-for-privacy@anasoft.co.uk

SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design.