Browse Prior Art Database

Methods and Apparatus for Measuring Local Random Variability of A Process Using Latch-Based Sense Amplifier Test Structure

IP.com Disclosure Number: IPCOM000186160D
Original Publication Date: 2009-Aug-11
Included in the Prior Art Database: 2009-Aug-11
Document File: 7 page(s) / 119K

Publishing Venue

IBM

Abstract

Neighboring device random variation increases with technology scaling and limits circuit design. Measurement, characterization, and modeling of local random mismatch are crucial for robust circuit design and manufacturing yield. Conventional scheme for measuring local random variations requires lots of full device current-voltage characteristics and complicated data manipulation and analysis due to indirect nature of the technique. Other schemes including Ring-Oscillator (RO)-based test structures, depend on both random variation and correlated systematic (spatial) variation. RO-based test structures measure the “average” effect (or delay) of a large number of devices, and hence depend on the “correlated” spatial variation of the devices. In contrast, this invention can directly measure individual device random variation using latch-type sense amplifier-based test circuit structure. Summary of Invention Core Idea: Latch-type sense amplifier-based test structure for measuring and characterizing random variation with the following features: the measured circuit response depends only on random variation, not on correlated spatial variation (e. g. as in RO based test structures); the input offset voltage (min. for correct operation) of the sense amp based circuit provides a direct measure of random variation. Main advantages over previous know solutions: The proposed scheme has the following advantages: direct measurement of random variation; accurate, simple, fast, low-cost test structure and methodology; in-line measurement. The proposed scheme can be used in the present and future process lines and circuit designs where the device random variation is a limiting factor. Accurate characterization of random variation will facilitate technology development, and improve design robustness and manufacturing yield.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 35% of the total text.

Page 1 of 7

Methods and Apparatus for Measuring Local Random Variability of A Process Using Latch-Based Sense Amplifier Test Structure

    In the proposed scheme, a latch-type sense amplifier (used in high performance SRAM) based test circuit is used to measure the local random variability of a process (Fig. 1). The minimum offset voltage (min. for correct operation) required between IN1 and IN2

p

                                                  rovides a direct measure for the local random variability. Fig. 2 shows the basic operation of the proposed scheme.

SAE

P1

P2

N2

SAE

O2

N1

Devices under Test

O1

IN2

NI2 NI1

IN1

VDD-Δ

VDD

CLK

SAE

Fig. 1. Schematic of the proposed sense-amplifier based test circuit

1

Page 2 of 7

SAE

O2

Correct Operation

O1

O1

Incorrect Operation

O2

Fig. 2: Waveforms of operation of the sense-amplifier circuit.

    When SAE is low, the nodes O1 and O2 are pre-charged to VDD. When, SAE is raised high, since VIN1=VDD > VIN2 (=VDD-) O1 discharges at a rate faster than O2. If O1 reaches below the trip-

p

oint of the inverter P2-

N

                 2, the node O2 switches back to "1" and O1 goes to "0". However, if there is a mismatch in the threshold voltage of NI1 and NI2 (i.e., Vt(

N

I1) > Vt(

NI2)),

even if

VIN1 > VIN2 , node O1 can switches back to "1" and O2 can go to "0" resulting in an incorrect operation. A higher value of is required to avoid this incorrect operation. Hence, minimum required for the correct operation (defined as input offset voltage) is a direct measure of the local mismatch between the strength of the devices in the proposed circuit. Fig. 3 shows the distribution of the input offset voltage assuming random local process variations (simulation using predictive 70nm BPTM models). It can be observed that, the global variation in the devices (in other words, if the device variations are correlated) does not impact the distribution of the offset voltage. Hence, the proposed scheme can be used as direct measure of the local random mismatch among two devices in a very close proximity.

2

Page 3 of 7

Fig. 3: Distribution of offset voltage showing the impact of local variation.

    The proposed scheme measures the mismatch between the driver transistors NI1 and NI2. The key thing to note is the presence of several driver devices in the right branch. These devices are designed of same sizes and placed at different spatial distance from each other. This allows us to characterize the spatial dependence of local random variability. Moreover, different sizes of the driver transistor pair (starting from minimum size) can be used to characterize the area (or size) dependence of local variability.

    The incorrect operation in the circuit can be caused either by mismatch in NI1 and NI2 (driver offset) or by the mismatch in P1-P2 and/or N1-

N

                                 2 (latch offset). Driver offset results in difference in the initial discharging current of N1 and N2 (i.e. (I1 - I2)) whereas latch off...