RF Noise Characterization for SOI MOSFETs

Note that in the above derivation, we have neglected the contribution from S_ig and S_igd.

From Equs. (3-8) and (3-9), we can see that except S_id, the trans-conductance g_m would play an important role in determining both intrinsic NF_min and R_n. The temperature dependence of g_m for devices with different channel lengths is shown in Fig. 3-9. It suggests that g_m decreases with temperature at a rate larger than that for S_id . Therefore, according to Equs. (3-8) and (3-9), both NF_min and R_n would tend to degrade and become larger with increasing temperature as shown in Figs. 3-10 (a) and (b), respectively.

3.4 RF Noise Characterization for SOI MOSFETs

Figure 3-11 shows the noise factor  for both the bulk and SOI devices. It shows that, in the medium-long channel devices (L0.36μm),  seems to remain the same for both SOI and bulk devices. However, the SOI devices would have an increasing  as the channel length shrinks. Two mechanisms may contribute to this phenomenon: floating body effect (FBE) and self-heating effect (SHE) [20]. Due to the floating body structure of the SOI nMOSFET, there is a potential barrier between the source and the body region. Therefore, the holes generated by impact ionization [19] at high drain bias condition can be easily trapped in

42

[19] at lower V_GS (V_dd 2), FBE would have a larger impact on the excess noise at lower VGS.

On the other hand, as V_GS increases, the DC power and therefore the temperature of the SOI MOSFET increases due to the so-called self-heating effect [18][19]. This effect is caused by poor thermal conductivity of the buried oxide, which is about two orders of magnitude less than that of the silicon [18][19], and the lattice temperature would play an important role in determining the SOI MOSFET noise characteristics [8]. Besides, the noise arising from the neutral-body resistance should be enhanced by the elevated lattice temperature and its contribution to the channel noise S_id may have to be considered. However, since the effective mobility and hence channel conductance should be decreased accordingly, the excess noise caused by SHE would partly counterbalanced by the reduction of channel conductance.

This captures the slight increase of  at high V_GS (see Equ. (3-2)). It is worth noting that since the SHE may reduce the body potential by inducing more diode leakage [20], the excess noise caused by FBE at high V_GS could be further alleviated.

Figure 3-12 shows the temperature dependence of  for both SOI and bulk devices.

Since the FBE can be eliminated at high temperature [17], the channel suffering less FBE would have decreasing  with increasing temperature. This is especially obvious at low VGS, where FBE dominates the excess channel noise behavior. For bulk devices, since they suffer neither FBE nor SHE, they have the similar  over the whole temperature region.

Finally, we compare NF_min and R_n for the SOI and bulk devices for a given DC power consumption. Figure 3-13(a) and (b) respectively show the comparison of S_id and g_m versus current for a given drain voltage (V_DS 1.0V). Because SOI device has larger S_id and lower g_m than the bulk counterparts in our experiments, referring to Equs. (3-8) and (3-9), it is expected that it would has worse NF_min and R_n as shown in Fig. 3-14(a) and (b), respectively. It is worth noting that the extrinsic parameters, such as gate capacitance and terminal resistances would not significantly contribute to the deviations, since both devices

have been checked to have similar C_gg (see Fig. 3-13(c)) and terminal resistances (shown in Table 3-2) for each temperature.

It should be noted that we have neglected the neutral-body effect on the RF characterization in this section. This is because the previous chapter has demonstrated the insignificant neutral-body effect on the RF small-signal characteristics of SOI MOSFETs except the output admittance. Besides, the body trans-conductance and drain leakage current have been presented to have significant effect mostly on the low frequency noise behavior due to its low-pass nature [21]. Note that at the very high frequency, the neutral-body resistance Rb would be equivalently parallel to the channel resistance and can contribute to the output noise current associated with the drain terminal. However, its thermal noise contribution is at the level of about 4k_BT R_b 1.6610^22A² Hz (for R_b 100Ω), and can be neglected compared with the extracted S_id (see Fig. 3-13(a)).

3.5 Summary

We have investigated the temperature dependence of S_ig , S_id and S_igd* for the

medium-long RF MOSFET. S_ig and S_igd* are found to have positive correlation with ambient temperature, while S_id has negative one due to much lower channel conductance at higher temperature. For L0.12μmdevice, however, since g_d0 does not decrease with temperature as much as that for both L0.24μm and L0.36μm devices, S_id relatively

44

contribution to S_id may not be significant.

Finally, since the trans-conductance decreases with temperature in a rate higher than that for S_id, both NF_min and R_n would increase accordingly. Our experiment also shows that the SOI device has worse NF_min and R_n due to the larger S_id and lower g_m than the bulk counterpart.

References

[1] A. J. Scholten, L. F. Tiemeijer, R. van Langevelde, R. J. Havens, A. T. A. Zegers-van Duijnhoven, and V. C. Venezia, “Noise modeling for RF CMOS circuit simulation,”

IEEE Trans. Electron Devices, vol. 50, pp. 618–632, Mar. 2003.

[2] A. Pascht, M. Grozing, D. Wiegner, and M. Berroth, ‘Small-signal and temperature noise model for MOSFETs’, IEEE Trans. Microw. Theory Tech., vol. 50, no. 8, pp. 1927–1934, Aug. 2002.

[3] C.-H. Chen, M. J. Deen, Y. Cheng, and M. Matloubian, “Extraction of the induced gate noise, channel noise, and their correlation in submicron MOSFETs from RF noise measurements,” IEEE Trans. Electron Devices, vol. 48, no. 12, pp. 2884-2892, Dec.

2001.

[4] K. Han, H. Shin, and K. Lee, "Analytical drain thermal noise current model valid for deep submicron MOSFET's," IEEE Trans. Electron Devices, vol. 51, no. 2, pp. 261~269, Feb. 2004.

[5] K. Han, J. Gil, S.-S. Song, J. Han, H. Shin, C.-K Kim, and K. Lee, "Complete high frequency thermal noise modeling of short-channel MOSFETs and design of 5.2 GHz low noise amplifier," IEEE J. Solid-State Circuits, vol. 40, no. 3, pp.726~269, Mar. 2005.

[6] M. J. Deen, C.-H. Chen, S. Asgaran, G. A. Rezvani, J. Tao, and Y. Kiyota,

“High-frequency noise of modern MOSFETs: Compact modeling and measurement issues”, IEEE Trans. Electron Devices, vol. 53, no. 9, pp.2062-2081, Sep. 2006.

46

and bulk-silicon MOSFETs for RF applications, ” IEEE Trans. Electron Devices, vol. 55, no. 3, pp. 872-880, Sep. 2008.

[10] S. M. Nam, B. J. Lee, S. H. Hong, C. G. Yu, J. T. Park and H. K. Yu, "Experimental investigation of temperature dependent RF performances of RF-CMOS devices," in Proc.

6th Int. Conf. VLSI CAD (ICVC’99), 1999, pp. 174–177.

[11] Y. S. Lin, “Temperature dependence of the power gain and scattering parameters s11 and s22 of an RF nMOSFET with advanced RF-CMOS technology,” Microw. Opt. Technol.

Lett., vol. 44, no. 2, pp.180–185, Jan.20, 2005.

[12] M. Emam, D. Vanhoenacker-Janvier, K. Anil, J. Ida and J.-P. Raskin, “High temperature RF behavior of SOI MOSFETs for low-power low-voltage applications,” in Proc. IEEE Int. SOI Conf., 2008, pp. 139–140.

[13] S.-C. Wang, P. Su, K.-M. Chen, C.-T. Lin, V. Liang, and G.-W. Huang, ‘Temperature dependence of high frequency noise behaviors for RF MOSFETs,” IEEE Microw.

Wireless Comp. Lett., vol. 18, pp. 530-532, Aug. 2008.

[14] A. van der Ziel, Noise in Solid State Devices and Circuits, Now York: Wiley, 1986.

[15] S.-C. Wang, P. Su,K.-M. Chen, K.-H. Liao, B.-Y. Chen, S.-Y. Huang, C.-C. Hung, and G.-W. Huang, “Comprehensive noise characterization and modeling for 65-nm MOSFETs for millimeter-wave applications,” IEEE Trans. Microw. Theory Tech., vol. 58, no. 4, pp. 740–746, Apr. 2010.

[16] J. Jeon, I. Song, I. M. Kang, Y. Yun, B.-G. Paark, J. D. Lee, and H.Shin, “A new noise parameter model of short-channel MOSFETs,” in IEEE Radio Freq. Integr. Circuits Symp., Jun. 2007, pp. 639–642.

[17] S. C. Lin and J. B. Kuo, “ Temperature-dependent kink effect model for partially-depleted SOI NMOS devices,＂ IEEE Trans. Electron Devices, vol. 46, pp.

254–258, Jan. 1999.

[18] L. Su, J. Chung, D. Antoniadis, K. Goodson, and M. Flik, “Measurement and modeling of self-heating in SOI NMOSFETs,” IEEE Trans. Electron Devices, vol. 41, p. 69, Jan.

1994.

[19] P. Su, K.-I. Goto, T. Sugii, and C. Hu, “Enhanced substrate current in SOI MOSFETs,”

IEEE Electron Device Lett., vol. 23, no. 5, pp. 282–284, May 2002.

[20] P. Su, S. K. H. Fung, S. Tang, F. Assaderaghi, and C. Hu, “BSIMSPD: A partial-depletion SOI MOSFET model for deep-submicron CMOS designs,” in Proc. IEEE Custom Integr.

Circuits Conf., 2000, pp. 197–200.

[21] W. Jin, P. C. H. Chan, S. K. H. Fung, and P. K. Ko, “Shot-noise-induced excess low-frequency noise in floating-body partially depleted SOI MOSFETs,” IEEE Trans.

Electron Devices, vol. 46, pp. 1180–1185, June 1999.

48

Table 3-1 Extracted g_d0, C₀ and their normalizations with respect to cases at -40℃ for the bulk MOSFET. (L0.36 μm )

 

^K

T T

 

K g_d0(mS) g_d0 C₀(fF) C ₀

40

T  ℃ 233 ^1 112.4 1 520 1

0

T ℃ 273 1.17 96 0.85 517 0.99

100

T ℃ 373 1.60 66 0.59 508 0.98

200

T ℃ 473 2.03 51 0.45 506 0.97

Table 3-2 Extracted R_s, R_d, and R_g for both the SOI and bulk devices. (L0.12 μm )

SOI BULK

()

Rs R_d() R_g() R_s() R_d() R_g()

 23

T ℃ 0.1 1.7 1.9 0.1 1.5 2.0

100

T ℃ 0.1 1.8 2.2 0.1 1.8 2.2

200

T ℃ 0.1 2.2 2.3 0.1 2.0 2.5

50

0 1 2 3 4 5 6 7 8 9 10 11 12

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6

T = -40C T = 0C T = 100C T = 200C

S

ig

(10

-22

A

2

/Hz)

Frequency (GHz)

Figure 3-1 Induced gate noise (S_ig) versus frequency for the bulk MOSFET under different temperatures. (L0.36 μm , and V_GS V_DS 1.2V)

0 1 2 3 4 5 6 7 8 9 10 11 12 1.0

1.2 1.4 1.6 1.8 2.0

T = -40C T = 0C T = 100C T = 200C

S

id

(10

-21

A

2

/Hz )

Frequency (GHz)

Figure 3-2 Channel noise (S_id) versus frequency for the bulk MOSFET under different temperatures. (L0.36 μm , and V_GS V_DS 1.2V)

52

0 1 2 3 4 5 6 7 8 9 10 11 12

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6

T = -40C T = 0C T = 100C T = 200C

S

igd*

(1 0

-22

A

2

/Hz )

Frequency (GHz)

Figure 3-3 Correlation noise (Sigd*) versus frequency for bulk MOSFET under different temperatures. (L0.36 μm , and V_GS V_DS 1.2V)

-50 0 50 100 150 200 0.0

0.2 0.4 0.6 0.8 1.0



,



,



, and c

Temperature (



C)





,



,



, and c c/j





Temperature (



C)

Figure 3-4 Model parameters  ,  ,  and correlation coefficient c versus temperature for the bulk MOSFET. (L0.36μm , and V_GS V_DS 1.2V)

54 for different temperature and gate bias conditions for the bulk MOSFET. (L0.36μm )

(a) (b)

(c) (d)

-50 0 50 100150200

Noise Factor, 

VDS = 1.0V

-50 0 50 100150200 0.5

Temperature (C)

-50 0 50 100150200 0.5

Figure 3-6 Temperature dependence of  for bulk devices with different channel lengths.

56 -50 0 50 100150200 0

-50 0 50 100150200 0

Temperature (C)

-50 0 50 100150200 0

Figure 3-7 Temperature dependence of g_d0 for bulk devices with different channel lengths.

-50 0 50 100150200 -50 0 50 100150200

0.8

Temperature (C)

-50 0 50 100150200 0.5

Figure 3-8 Temperature dependence of S_id for bulk devices with different channel lengths.

58 -50 0 50 100150200

0

Trans-conductance, g m (mS)

-50 0 50 100150200 0

Temperature (C)

-50 0 50 100150200 0

Figure 3-9 Temperature dependence of g_m for bulk devices with different channel lengths.

-50 0 50 100 150 200

NF min@8GHz (dB)

-50 0 50 100 150 200

Temperature (C)

-50 0 50 100 150 200

-50 0 50 100150200 0

-50 0 50 100150200 0

Temperature (C)

-50 0 50 100150200 0

60

Noise Factor, 

Figure 3-11 Noise factor  for both SOI (symbols) and bulk (lines) devices with different channel lengths.

0 20 40 60 80 100 120 140 160 180 200 0

1 2 3 4 5

VDS = 1.0V

L=0.12m V_GS = 0.8V

V_GS = 1.0V V_GS = 1.2V

Noi s e Fac tor, 

Temperature (  C)

Figure 3-12 Temperature dependence of noise factor  for both SOI (symbols) and bulk (lines) devices.

62

S id (10-21 A2 /Hz)

Drain Current, I_D (mA)

Trans-conductance, g m (mS)

Drain Current, I_D (mA)

0 20 40 60

NF min@8GHz (dB)

Drain Current, I_D (mA)

Drain Current, I

D (mA)

64

Chapter 4

RF Noise Modeling and Characterization for SOI Dynamic Threshold Voltage MOSFETs

4.1 Introduction

Due to its larger current driving ability with low leakage current, the dynamic threshold voltage (DT) MOSFET is attractive for low power applications [1]. Hence, the DC characteristics and modeling of the DT MOSFET have been widely studied since its introduction [2]-[4]. Moreover, the temperature effect on its DC characteristic has also been well investigated [4].

Several optimized SOI- or bulk-based DT MOS fabrication processes with improved performance have been demonstrated [5][6], and its ability of radio-frequency (RF) applications with high cut-off frequency ( f_t) and maximum oscillation frequency ( f_max) has been reported as well [7]-[9]. However, the temperature effect on the RF characteristics of DT MOSFETs is rarely known.

In this chapter, we will first conduct RF small-signal modeling for the SOI DT MOSFET and demonstrate a practical extraction method to facilitate the extraction work with physical accuracy. Based on the small-signal model structure, the RF noise model for the DT MOSFET will be built, and this model is shown to well capture its RF noise behavior. Besides, the accuracy of some important model parameters will be examined by comparing them to those of the standard conventional devices with different channel lengths at various bias conditions.

Finally, we will give an experimental investigation on the RF small-signal and noise characteristics of SOI DT MOSFETs, including their temperature dependences [10]. To avoid a large leakage current flowing through the source-body junction, a DT MOSFET is usually biased in the low gate overdrive (V_GT) region. Therefore, we will be dedicated to examining the RF small-signal and noise characteristics under this regime.

4.2 Devices and Experiments

The RF SOI DT MOSFETs used in this work were fabricated using UMC 65nm SOI technology. These RF devices were laid out in the multi-finger and multi-group structure with the following denotations: L for channel length, W_F for finger length, N_F for the number of fingers, and N_G for the number of groups (total gate width W W_F N_F N_G).

On-wafer 2-port common-source high frequency S and noise parameters were measured using ATN NP5B noise measurement system with Cascade microwave probes.

Besides, to eliminate the inevitable parasitic accompanied with the probing pads, the S parameters of devices’ corresponding open dummy were measured and then used to perform the S and noise parameters de-embedding procedure.

Figure 4-1 shows the temperature dependences of threshold voltage (V_T) extracted by the constant current (I_th 50nAW L) method. Due to the negative temperature coefficient of the device’s Fermi potential [4], V_T has the negative temperature dependence for devices with different channel lengths.

4.3 RF Small-Signal and Noise Modeling

The RF small-signal and noise equivalent circuit suitable for the DT MOSFET modeling and characterization will be described in this section. Then, a set of simple and analytic expressions of Y parameters beneficial to the model parameter extraction will be presented accordingly [11].

66

analytic two-port admittance ( Y ) parameters can be derived when the effect of series resistances compared to access body resistance (R_b) can be neglected. Following especially shows the expressions benefiting the parameter extraction:

   

procedure shown in Fig. 4-4 is then proposed. Compared to the method proposed in reference [12], which needs some parameters determined from DC characteristics, our extraction method relies only on local optimizations using definite RF fitting targets to obtain all model parameters, so the excellent modeling results with less than 10% relative root-mean-square errors for each real and imaginary part of Y parameters, as shown in Fig. 4-5, can be expected. For the reader’s reference, the extracted model parameters are listed in Table 4-1.

Besides, as shown in Fig. 4-6, based on the RF small-signal equivalent circuit, the RF noise equivalent circuit can be built by adding the corresponding noise current sources. In this noise equivalent circuit, i_d stands for the intrinsic channel noise current, and the assumption

that the high-frequency prominent drain-induced gate noise can be neglected is adopted. This assumption had been shown to be reasonable especially for deep sub-micrometer devices [15], and its validity will also be examined in Chapter 6. Furthermore, the noise current sources related to series resistances and access body resistance are considered as thermal noise current sources (i 4kT R, R : resistance value). Finally, the inherent shot noise current caused by the source-side junction current is estimated using shot noise current formula (i_j,sb  2qI_b  2qI_g ).

The only one unknown model parameter i_d can be directly obtained by optimizing the four measured high-frequency noise parameters (minimum noise figure NF_min, equivalent noise resistance R , magnitude of the optimum reflection coefficient _n _opt , and phase of the optimum reflection coefficient _opt). The good noise modeling results are shown in Fig.

4-7.

4.3.2 Verification of the Extraction Results

To further examine the accuracy of the modeling results, some important model parameters versus VDD for different channel lengths are examined. Note that we let



V



V VDD

V_GS  _BS  _DS  to keep the device operating in the saturation region. Figure 4-8(a) shows that compared to the standard device, the DT device has larger trans-conductance (g ) _m

68

channel resistance (R ) for shorter channel devices in Fig. 4-8(b). _ds

Besides, lower threshold voltage also increases the channel charge, and hence increases the intrinsic capacitance [1]. Therefore, as shown in Fig. 4-8(c), the DT device would have larger gate-to-source capacitance (C ) than the standard one. Figure 4-9(a) shows that the _gs body trans-conductance (g ) tends to increase with _mb VDD. However, in the low-voltage regime where the DT device normally operates, compared to g , its value is small and hence _m its contribution to the total device performance could be negligible.

Finally, the source- and drain-side junction capacitances (C_j,sb and C_j,db) as well as

access body resistance (R ) versus _b VDD are examined. In Fig. 4-9(b), C_j,sb tends to exponentially increase as VDD increases due to the nature of its forward-biased diffusion capacitance, while C_j,db shows less bias dependence. Besides, decreasing channel length can

help decrease C_j,sb, but increase C_j,db. Figure 4-9(c) shows that R may decrease with _b increasing VDD, which results from the abundant positive charge supplied by the external DC source through the body contact. This figure also supports that because the shorter device has a smaller cross-section for current flowing into the body, it has larger R . Note that all _b the channel length dependences for C_j,sb, C_j,db, and R become weak for channel length _b below 0.12 μm .

4.4 RF Small-Signal Characterization

In this section, using the extraction methodology proposed in the previous section, we will study the temperature dependences of the extracted small-signal parameters for the RF SOI DT MOSFET.

4.4.1 Temperature Dependences of Small-Signal Parameters

Figure 4-10 shows the temperature dependences of gate-to-source capacitance C , _gs channel resistance R , and trans-conductance _ds g for the DT MOSFET. Lower threshold _m voltage at higher temperature can induce more charges in the channel and hence larger C _gs [1] and lower R as shown in Figs. 4-10(a) and (b), respectively. This also results in _ds positive temperature dependence for g in the low _m VDD regime as shown in Fig. 4-10(c) [4]. At high VDD, however, the lower mobility at higher temperature would degrade g , so _m g tends to decrease with increasing temperature in the high m VDD regime [4][17]. Also note that at all bias conditions where saturation holds, C shows much less temperature _gs dependence than g . _m

The temperature dependences of inherent body-related parasitics of the DT MOSFET are shown in Fig. 4-11. Due to the more leaky behavior encountered in source-to-body junction at higher temperature, the source-to-body junction capacitance C_j,sb would increase with temperature. On the other hand, compared to C_j,sb, the drain-to-body junction capacitance

db

Cj_, shows less temperature dependence (see Fig. 4-11(a)). Besides, at higher temperature and larger VDD (and hence, larger V ), more charge would be injected into the body _BS region through source-to-body junction, and this could contribute to the observation that the body resistance R tends to decrease with increasing temperature and _b VDD as shown in

70

figures of merit used to characterize the RF performance of a device. To derive out simple and analytical equations for analysis, the series resistances have been omitted at the moment, and also de-embeded from the maseured data for model comparison. Based on the equivalent circuit shown in Fig. 4-6 without considering series resistances R , _s R , and _d R , the _g f _t and f_max for the DT MOSFET biased in the low VDD regime can be approximately expressed as the following equations [18].

 

gs The approximation in Equs. (4-8) and (4-9) holds in the low VDD regime, where

1

mb

m g

g , R_b R_i1, g_mbR_b C_j,db C_j , and R_bC_j 1 around f_max, and the good modeling results for f and _t f_max in the low VDD regime are shown in Figs. 4-12(a) and (b), respectively.

Equation (4-8) implies that the inherent body-related parasitics of the DT MOSFET would have little influence on f . In the low _t VDD regime, since g tends to increase _m with temperature, f would have a positive temperature coefficient as shown in Fig. 4-13(a) _t for VDD below 0.4V.

On the other hand, Equation (4-9) implies that the body-related parasitics would degrade

fmax through the degradation factor _DT , which is about 3/4 and almost bias and temperature independent as shown in Fig. 4-14. In addition, due to the less temperature dependent behavior of R_ds R_i (also shown in Fig. 4-14), f_max tends to have the same temperature dependence as f (see Fig. 4-13(b)). That is, in the low _t VDD regime, both

fmax and f would tend to become larger at higher temperature. _t

4.4.3 Series Resistance Effect

In the previous sub-section, we have focused on the ‘inner’ device performance without considering the series resistance effect. To judge the series resistance effect on the overall performance and complete the characterization, this effect will be considered in this sub-section. Besides, to facilitate the examination of the temperature effect, we have normalized the related parameters with respect to their corresponding values at T  25C in the following discussions.

Figure 4-15 shows that the series resistance has much more significant effect on the unilateral power gain U (or f_max) than the short-circuit current gain H₂₁ (or f ) at _t

V 3 .

0

VDD . Compared to the series resistance, the much larger input and output impedance in the low VDD regime would dominate H₂₁ , and hence, f . The little series resistance _t effect on f can be also deduced in Fig. 4-16(a) and (b), where _t f has nearly the same _t temperature coefficient as g for each channel length device. This coincides with the _m

72

fmax.

4.5 RF Noise Characterization

4.5.1 Channel Noise and Equivalent Noise Resistance

The extracted power spectral density for the channel noise current i (denoted as _d S ) _id is shown in Fig. 4-18, and usually expressed as follows [19].

4 _{B d}0

id k T g

S   (4-12)

where J/Kk_B 1.3810^23 is Boltzmann constant, T is the ambient temperature in Kelvin,

0

g is the channel conductance at zero drain-source voltage, and d  is noise factor. Besides, reference [20] has shown that  has the weak temperature dependence, and the temperature

g is the channel conductance at zero drain-source voltage, and d  is noise factor. Besides, reference [20] has shown that  has the weak temperature dependence, and the temperature

在文檔中 先進金氧半場效電晶體考慮溫度相依之高頻小訊號及雜訊特性分析 (頁 58-0)