Recently, owing to integrated circuit technology progress and innovations in circuit design techniques, more and more systems can be fabricated in a single chip (SOC) that make integrated biomedical device smaller and have possibility to be implanted to supplement, or even entirely replace biomedical operation. Several implantable microelectronic systems already exist [1] (Fig. 1.1): pacemakers are used to replace a defective natural cardiac pacemaker [2] (Fig. 1.2), cochlear prostheses are used to help provide a sense of sound to a person who is partially or profoundly deaf, visual prostheses are used to restore eyesight to the blind [3] (Fig. 1.3) and neural recording and stimulation systems are used to treat patients with Parkinson‟s disease and epilepsy [4]. This thesis focuses on neural recording and stimulation systems‟
power supply design. Fig. 1.4 is the overall system‟s architecture. The mainly function of this system divide into two part. The first part is amplifying the neural signal by amplifier, then converting the analog signal to digital signal and then transmits the information through the RF circuit. The second part is use DSP to analyze the digital signal and then use analyzed result to determine whether the patient needs stimulating.
Since biomedical device need to implant into the human body and implanted medical device are vital to the patients, the implantable medical device most have a long term reliable power supply for implanted medical device. However, currently battery technology cannot stand for an enough time even for small power
consumption implanted pacemaker [5]. Therefore, wireless power transmit to medical device must be considered. Several possible wireless power transmission method are summarize in table 1.1 [6] [7] [8] [9] [10]. Depend on different power consumption on implanted medical device, different wireless power transmission method would be adopted. Since neural recording and stimulation systems power consumption may up to ten mW, thermoelectric and far-field electromagnetic energy transfer methods which still cannot offer such power level on recent research would not be a proper method for wireless power transmission. Secondly, because neural recording and stimulation systems need to implant into human brain on a fix position to sense the neural signal and stimulate human brain, it may unsuitable use vibration and heel strike to generate power from human motion or walking. Third, transmission efficiency should be high to avoid excess human exposure to electromagnetic wave, therefore optical recharging which power transmission efficiency is poor would not be a good candidate. Summarizing above reason, near field inductive link power supply become most widely use method to power the implanted device from medium power level to high power level [1].
Near-field inductive link power supply majorly can divide three different parts:
near field coil, rectifier and regulator (Fig. 1.5). For near field coil design, coil‟s size which be confined by the implanted physical size in human body is an important design factor that mainly determined the power efficiency and the transmission distance. However, operating frequency, structure of coil and inductance value also have impact on transmission efficiency. Designer must carefully trade off the above factors to make a high efficiency transmission. For rectifier part, passive and off chip schlocky barrier diode used to be a suitable choice to implement [9]. However, the off chip schlocky barrier diode have large dropout voltage 0.3V, which will result large power consumption especially in low voltage high current system of the neuron
recording and stimulating and the off chip schlocky barrier diode cannot integrate with other circuit. The first on chip rectifier was proposed [11]. Then more and more on chip active rectifiers have been developed [12] [13]. The on chip active rectifiers using four switching power mos which can effective reduce the dropout voltage when current conduct. The experiments verify that the new on chip rectifier not only have better AC to DC conversion efficiency, but also have smaller area and suitability for system integration. After the rectifier converse the AC power into DC power, the regulator need to connect the rectifier output since the voltage output voltage of rectifier is really noisy. For regulator part, high efficiency low dropout regulator is the most choice for most biomedical system. However, since high resolution and multi channel simultaneous neuron signal sensing application need, the noise of power supply for analog circuit like pre-amp and analog to digital conversion must be less than before works.
1.2 Motivations and Main Result
In CMOS active rectifier, the proposed comparator circuits compensate both turn-on and turn-off delays and make active rectifier operated at a higher frequency, (FCC approved ISM band13.56 MHz) than that in the previous work (1.5MHz)[14].
Since the large power transistor delay is compensated by the proposed comparator circuits, the larger output current (20 mA) than that in the previous work (2mA)[13]
can be achieved for implantable epilepsy detection and stimulation applications. The deep n-well parasitic diodes of power NMOS and additional start control circuit make the active rectifier start up robustly. The fabricated CMOS active rectifier can be operated at high frequency (13.56MHz) and deliver 20mA output current while maintaining the power conversion efficiency of 84.8% and provide low output peak-to-peak ripple (14mV) than those in the previous works(≈100 mV [14], ≈ 200mV[13]). In CMOS LDO regulator, the fabricated CMOS LDO regulator has
-58.49dB PSR up to 10MHz that better than previous work (-40dB at 1MHz)[36] [18].
The fabricated CMOS LDO regulator deliver 20mA output current while maintaining the lower dropout voltage (200mV) than the previous work (600mV) [18]. The single dominant pole compensation [28] of LDO regulator consumes the higher quiescent current 98uA than previous work [36][19] when deliver the same amount of output current. The additional ESR zero can be avoided. This makes LDO regulator have the smaller transient load regulation of 5.75mV than that in the previous work (15mV)[19]
when the load current rises from 0mA to 20mA in 10ns. In the inductive link power supply, the fabricated inductive link power supply has a lower peak-to-peak output ripple of 11.5mV and higher power conversion efficiency of 76.3% than those in the previous work(80mV, 75.6%)[35].
1.3 Thesis Organization
In chapter 1, the background has been introduce in section 1.1, including the different purpose biomedical devices, the neural recording and stimulation system‟s overall architecture, the distinct wireless power transmission methods and the overall near-field inductive link system architecture. Then the motivation and main result of this work are proposed in section 1.2. The section 1.3 is thesis organization.
In chapter 2, the whole system and the individual blocks design would be discussed detailed. In section 2.1, the whole system design consideration would be discussed. In section 2.2, the efficiency optimum equation derivation, coil structure design and the IEEE safe limit to human exposure computing method would be described. In section 2.3, the rectifier transistor level operation, the start-up mechanism, the latch-up consideration and the new proposed delay compensated comparator operation would be described detail. In section 2.4, the stability, the compensation methods, transient load regulation and the high PSR technique of low dropout regulator would be discussed.
In chapter 3, the whole chip layout consideration would be described in section 3.1. In section 3.2, the chip performance measurement setup would be described. In section 3.3, the experiment results of each blocks and whole chip would be demonstrated. In section 3.4, the whole chip experiment results and each blocks experiment results will compare with other recent research results and discuss the distinct circuit design techniques.
In chapter 4, the results of the near-field inductive link power supply of this work would be summarized. Furthermore, the chapter ends also discuss about the possibility of improvement to the future near-field inductive link power supply design.
Energy source Power density efficiency distance implant
Thermoelectric [7] 20uW / cm2 N.A. N.A. YES
Optical recharging
[8] 4.4mW / cm2 20% 2mm YES
Vibration
(human motion) 4uW / cm3 N.A. N.A. NO
Heel strike (walking) 10-700mW N.A. N.A. NO
Near-field inductive
energy transfer [9] 140 mW / cm2 75% 7mm YES
Far-field
electromagnetic energy transfer [10]
3.5 mW / cm2 0.08% 1.75cm YES
Table 1.1 Power sources options
Figure 1.1 the implanted medical devices for various applications.
Figure 1.2 the visual prostheses are used to restore eyesight to the blind.
Figure 1.3 the overall system‟s architecture of neural recording and neuron stimulator.
The system is fully power by wireless inductive link.
This Work
Figure 1.4 the near-field inductive link overall architecture.
CHAPTER 2
SYSTEM DESIGN AND SIMULATION
In this chapter, the near-field inductive coils, the on chip active rectifiers and low-dropout regulator will be discussed in section 2.1, 2.2 and 2.3 respectively. In section 2.4, the whole chip architecture and design consideration will be mentioned.
In section 2.5, the post-simulation of the individual blocks and whole chip will be described.
2.1 Near-field coil
For near field coil design, coils‟ size which be confined by the implanted physical size in human body is an important design factor that mainly determined the power efficiency and the transmission distance. However, operating frequency, structure of coils and inductance values also have impact on transmission efficiency.
Designer must carefully consider the trade off on the above factors to make a high efficiency transmission. The following sub-section will derive the coils‟ efficiency equation; describe the coils‟ design flow and calculate the human exposure to EM field.
2.1.1 Transmission Efficiency Analysis
The fundamental inductive powering circuit is show in Fig. 2.1 [20]. The primary part use series resonance capacitor instead of parallel resonance capacitor for matching needy. The diode voltage conversion is define to be A = Vo / Vpk, where Vpk
is peak a.c. voltage across the tank circuit L2C2 and Vo = Vpk - Vdiode. When diode
dropout voltage is much smaller than Vpk and RoCo >> 1 / f , then A ≒ 1 and Vpk ≒ Vdiode. The equivalent ac load resistance Rac which will dissipate an amount of a.c.
power equivalent to the d.c. power in Ro is:
(2.1)
The equivalent ac parallel resistance can be transform into equivalent ac series resistance [21]:
(2.2)
The w is the radian frequency 2πf. The Fig. 2.2 is the equivalent circuit of secondary part. The total equivalent series resistance in secondary tank circuit is R2+RL, where R2 is the parasitic series resistance of the inductor.
The equivalent resistance Re, reflected back into the primary coil, is:
(2.3)
where is the mutual inductance of the coils and Q1 = wL1 / R1 and Q2
= wL2 / R2 are the unloaded Qs of the primary and secondary coils respectively.
From the primary equivalent circuit, show in Fig. 2.3, the circuit efficiency at resonance can be determined:
(2.4) and
(2.5) and
(2.6)
In inductive link wireless power, the maximum value of k is determined by coil size and spacing. For maximum efficiency d η / d R2 equals zero, thus:
(2.7)
Substituting equation 5 into equation 4, the optimum efficiency of the circuits is given by
(2.8)
Equation above indicates that optimum efficiency ηopt, increases as k2Q1Q2 increases, which is the same conclusion with recent research [22]. However, the coils‟ Q is limited, since coils tend to have great Q when coils‟ sizes which be limited by implant body size become large. Therefore, we design the coils must have the highest unloaded Q and k in a small space to achieve the efficiency wireless power transmission.
2.1.2 Near-Field Coils Design
Once the system application decide the implant physical size constrain, the coupling coefficient (k) can be maximize by proper choosing the outer diameter size of coils, and the Q factor can be maximize by proper choosing the coils‟ structure and coils‟ material. Since Q factor and coupling coefficient are increase as the outer diameter increase, the first step of efficient near-field coils design maximizing the coupling coefficient would not conflict the design parameter of maximizing the Q factor. Once we maximize the coupling coefficient by deciding the primary and secondary coils‟ outer diameter, that recent research have been widely studied [23]
(Fig. 2.4, Fig. 2.5), we can maximize the Q factor by using allowable wide copper metal and low loss coil structure. The coil structure mainly divides into three parts:
multi layer cylindrical coil (Fig. 2.6), single layer cylindrical coil (Fig. 2.7) and spiral coil (Fig. 2.8). The different structure have different Q factor, because the skin effect
and the proximity effect [24].
Skin effect is usually described as the tendency of current to flow primarily on the surface (skin) of a conductor as frequency increases. Because the inner regions of a conductor are thus less effective at carrying current than at low frequencies, the useful cross-sectional area of a conductor is reduced, thereby producing a corresponding increase in resistance (Fig. 2.9) [21].
Proximity effect is a phenomenon of current crowding at a surface of a conductor because of current flowing in a nearby conductor. The reason of current crowding at a surface of conductor is the nearby conductor current flow constructs magnetic field and makes a perpendicular force to the moving electron in the conductor. When electrons crow at a surface of a conductor, the effective resistance increases (Fig. 2.10) [21].
Summarizing above physics phenomenon, the single layer cylindrical coil have the greater Q factor than multi layer cylindrical coil and spiral coil. The reasons are the multi layer cylindrical coils have worse proximity effect than single layer since its second layer conductor overlap the first layer, and the spiral coil wind conductor circle smaller and smaller since a smaller circle have smaller Q factor than a bigger circle. However, the single layer cylindrical coils have the largest size for a particular inductance. Therefore, secondary part coil is not suitable using single layer cylindrical coils since it‟s too thick. The spiral coil is the better option for secondary coils, because spiral coil so thin that make coil implant into human body possible. For primary part, it have different consideration to secondary part. The coil on the primary part is outside the body and coils‟ thickness would not be a big problem, so single layer cylindrical coil would be a good option.
After decided the coil structure, the conductor material must be properly selected.
Since effective series resistance of coils needs to be decrease to achieve high quality
factor (Q), the conductor must have high electrical conductivity. If cost doesn‟t consider, silver would be the best option since silver have the highest conductivity.
Copper conductor coat with silver were be the second option which use a little silver to decrease the surface resistance of conductor. Since electrons tend to gather on the conductor surface, this approach may be a good option. The cheapest and widely adopting option is simply using solid copper conductor. There are also have many previous researches using litz-wire to make low resistance conductor [9] (Fig. 2.11).
The litz wire is a wire which composes by many thin wires. Since electrons move on the conductor surface, litz wire can increase the surface area. However, when operating frequency increase further (above 2MHz), irregularities in stranding and capacity between the individual strands result in failure to realize the ideal conditions in litz conductor [24][25]. Therefore, the biomedical inductive link which must operate in ISM band (6.78MHz or 13.56MHz) is not suitable using litz wire as the coils‟ conductor.
After maximize coupling coefficient and Q factor, input resistance matching and optimizing R2 value are the next step. The R2 value and input resistance can be adjusted by change winding turns on primary and secondary coils. First, the R2 value increases as winding turns increase since wire become longer. Once we decide the R2 value, secondary coil inductance value can be decided, because Q factor (wL2 / R2) have unremarkable dependence on coil winging turns. After R2 optimizing, input resistance can be tuned. The input resistance is Re+R1 which also increase as winding turns increase. The primary coils‟ Q factor (Q1) also has unremarkable dependence on coil winding turns. If power amplifier output resistance is 50 ohm, then we can use below equation to decide the R1 value:
(2.9)
(2.10)
Since ac equivalent loading (Rac), coupling coefficient (k), quality factor (Q1, Q2) and secondary coil effective resistance (R2) are not depend on primary coil winding turns, we can adjust the primary coil winding turns to increase or decrease R1 value that makes input resistance equal to 50 ohm without changing other parameters. Once R1 value is decided, the L1 value also is decided since Q1 factor unchanged remarkable. After the last step of input resistance matching, the near-field coils can efficient transmit wireless power from primary to secondary.
Summarizing all above design consideration, we can derive an efficient coil design flow for neural recording and stimulation systems.
1. Decide the size constrains and transmission distance for implant condition.
For neural recording and stimulation systems, implant size cannot excesses 2cm diameter, thick 0.5cm and transmission distance is about 0.6cm to 1cm.
2. Maximize the coupling coefficient (k) through Figure 2.5.
Select z = 1cm, doutR = 1.5cm and doutT = 4cm result coupling coefficient
=0.11. (Since dinT and dinR are not remarkable influence coupling coefficient, we reserve these variables for Q factor maximize and R2 optimum.)
3. Maximize the coil quality factor (Q).
Use single layer cylindrical coil for transmit coil instead of the spiral coil.
(AWG = 18, Q about 160)
Use conductor in receive coil as wide as possible under size constrain. (AWG
= 25, Q about 100)
4. Calculate ac parallel equivalent resistance (Rac), R2 OPT and R1 value.
Before calculating ac parallel equivalent resistance, the rectifier electrical performance and system power consumption (after rectifier) need to be ready.
Since the active rectifier (which is the most efficient rectifier in recent research) has AC to DC 85% efficiency and dropout voltage 0.3v, the system power consumption need 40mW (2V, 20mA), we can calculate the Rac by following step:
(2.11)
(2.12) (2.13)
(2.14) (2.15)
Once R1 value is decided, L1 value would be decided and can be measured by experiment.
5. Calculate and connect the needy resonance capacitor (Cres) to the primary and secondary coils.
Since coils have parasitic capacitance, resonance capacitors which need connecting to the coils must minus the parasitic capacitance from total capacitance.
(2.16)
(2.17)
The Cparasitic value can be obtained by measurement.
2.1.3 IEEE Safe Limit to Human Exposure to EM Field
When 13.56MHz sine waves transmit from primary coil to secondary coil, the magnetic field variation generator by primary coil would heat the human skin [26].
Since high temperature on human skin may damage the human cell, magnetic field should be low enough such that heat would not cause any damage on human body.
The roughly calculate to power dissipation on human skin can be showed below [27]:
(2.18) Then magnetic fields generator by primary coil is:
(2.19)
Then electric fields induce by magnetic field on skin is:
(2.20)
Then the power loss due to skin absorption can be calculated:
(2.21)
(2.20)
2.2 CMOS Active Rectifier 2.2.1 Active Rectifier Operation
The conceptual circuit of the proposed rectifier is show in Fig. 2.12 [14]. The input sine wave voltage is generated from the inductive link receiver coil. The pMOS power transistors Mp1, Mp2 form a cross-coupled pair such that both Mp1 and Mp2 function as switches with only a small |Vdrop, p| drop when either of them is turned on. In addition, a 2-terminal comparator CMP1 (CMP2) controls the switching of nmos power transistor Mn1 (Mn2) by comparing voltage Vin1 (Vin2) with ground potential. Transistor Mn1 (Mn2) will only be turned on when the voltage Vin1 (Vin2) < 0 V to maintain the unidirectional charging current flow. The Mn1
(Mn2) operates in the linear region as a switch with small Vdrop, n drop during conduction. As a result, the active rectifier voltage conversion ratio can be derived as:
(2.21) include: reverse leakage current which would happen when power mos turn off delay, comparator power consumption and buffer power consumption.
The active rectifier steady-state operation method is illustrated in Fig. 2.13. The operation steps can be divided into three steps during the negative half cycle Vdrop, n. Since both Mp2 and Mn1 operate in triode region, the large power mos size of Mp2 and Mn1 make switching resistance small, and the corresponding small |Vdrop, p|
and Vdrop, n. Therefore, voltage conversion ratio is maximized since |Vdrop, p| and Vdrop, n are minimized. When the voltage Vin reach the peak value, Vin voltage starts to
and Vdrop, n. Therefore, voltage conversion ratio is maximized since |Vdrop, p| and Vdrop, n are minimized. When the voltage Vin reach the peak value, Vin voltage starts to