Search Results

Integrated circuits continue to advance, but many of the highest-performance analog front ends still begin with a discrete JFET. Why? Because when you're trying to capture extremely weak signals, the input stage often determines the performance of the entire system. Engineers continue to choose discrete JFETs for applications requiring: Ultra-low noise High input impedance Low input capacitance Excellent linearity Whether designing photodiode amplifiers, hydrophone preamplifiers, precision sensor interfaces, or instrumentation front ends, these characteristics can make the difference between detecting a signal—or missing it. That's why devices like the LSBF862, LSK170, LSK389, and LSK489 continue to be designed into new systems where signal integrity starts at the input stage. 👉 For free samples or to discuss your application with our engineering team, complete the form HERE!

There are a lot of circuits and applications of discrete and integrated component combinations solving a myriad of tasks defined as solutions to a problem. Dad used to tell me that, "The problem isn't the problem; the definition of the problem is the problem." Dr. Gene Slottow at U of I would say, "When you think you're done with the design, throw it out and start over as your unyielding self-critic until you can't anymore, then proceed." Few first circuit iterations are "the one." This was one of those iterations. Gain control circuits typically use a differential front end and use it as a pivot point as a low-noise front end for an Op-Amp and its inherent massive loop gain, feedback, and convolved transitions of nonlinear elements in between. Compensation is often dicey because of the amount of loop gain being thrown out, as well as the group delay to the feedback node. Audio doesn't need the PPM level of a typical op-amp having >120 dB of gain. Audio requires linearity and acute transient response without overshoot. With that in mind, illustrated here is an often-repeated textbook differential input stage with its transconductance gain controlled by emitter current modulation. Preceding an Op-Amp, the low-noise LS312 bipolar pair input voltage must be less than 14 mV for linear operation before the square-law difference becomes nonlinear and it starts to become a comparator. These collector currents source into a discrete differential transimpedance amplifier with the relatively low loop gain of <60 dB. This particular circuit was originally designed to replace the TA7136 in Boss DS-1 guitar pedals with 9 V battery operation and would work down to about 6.5 V—dead battery mode. This Op-Amp has a similar asymmetric output impedance character to the TA7136, where slew rate going negative is less than the slew rate going positive, producing slight even-order, octave-related harmonics. When the amplifier's midpoint voltage is stabilized, it operates linearly open loop with a gain of about 1000, 1 V/mV, and with very low distortion. The LSK389 JFET front end provides high input impedance and linear gain to the single PNP Gm output stage. This 5-component Op-Amp is ideal for battery audio work with only 1 mA operating current. The VCA operates on 9 V and about 2 mA. Higher input voltages simply require a divider to keep input V <20 mV without emitter degeneration. Differential gain can be modified by a resistor between Q1 and Q2 emitters. There are no sacred cows in Analog design. As usual... — Kirkwood Click HERE For free LSK389, LSK170, LSK189, LSK489, or LSBF862 samples, product information or design assistance.

"In 1979, we teamed up at DEST Corp. to design the first desktop OCR machine for scanning standard letter and legal pages with a linear Reticon stored-charge diode array. Internal to that sensor were two identical 2048-pixel diode strings. Light would deplete the diode pixel reset charge and be read as exposure versus time through a charge integrator. Because the CMOS diode-select shift registers created substantial charge injection noise, the dark string was differentiated with the active string for noise cancellation and then sampled. This sample-and-hold circuit had to be fast and free of artifacts. Unlike many feedback S&H circuits, this is an errorless open-loop stored-voltage sampler with no loop recovery or compensation time. A capacitor, shunt-switched by DMOS, stores the dark voltage value before a scan so differenced pixel voltages are DC referenced. Pixel voltages are then capacitively coupled into a high-speed, high-impedance JFET buffer to the sampling switch. DMOS parts are often substrate-biased to lower inter-terminal capacitance, resulting in reduced charge injection when used as switches. Buffered pixel voltages are switch-sampled through a DMOS part compensated for blow-by capacitance in the off-state, charge injection cancellation, and constant video tracking of source-to-substrate bias. The gate switching voltage of -0.7 V to +5 V also tracks the source voltage for constant charge management. The sampled pixel voltage is then buffered through a JFET-controlled variable gain amplifier for setting an automatic gain control to keep text video at a constant amplitude. Circuits like this use bipolar, JFET, and DMOS parts for their unique and diverse properties. The Linear Systems DMOS used here was originally made by Signetics at the time, the JFETs were from Siliconix, and the bipolar dual, an MP352, was from an earlier John Hall company, Micro Power Systems, before LIS. Overall, the signal-to-noise performance was over 70 dB and, moreover, was very repeatable with medium-tolerance components in manufactured quantities in the thousands. This circuit is just one example of the many uses of FETs as switches, amplifiers, and variable resistors. In other portions of this system, an SD5000 quad DMOS was used for isolation and integrator reset on 5-picocoulomb charge integrators following the Reticon RL2048 outputs. Matching of the quad DMOS on die made charge injection cancellation relatively simple and repeatable. I've now been using these parts for the last 45 years and will continue to do so." - Kirkwood Rough Why This Design Still Matters Today While OCR technology has evolved dramatically since 1979, the fundamental analog design challenges remain the same: maximizing signal integrity, minimizing noise, managing charge injection, and preserving accuracy when processing extremely small signals. Kirkwood's experience highlights an important principle that continues to drive modern circuit design: selecting the right device technology for each function. JFETs remain valuable for their high input impedance and low noise performance. DMOS devices continue to excel in high-speed switching applications where low capacitance and controlled charge injection are critical. Bipolar devices provide precision gain and matching characteristics that complement both technologies. Many of today's advanced systems—including image sensors, medical instrumentation, industrial sensing, scientific equipment, and precision measurement systems—still rely on these same analog fundamentals. At Linear Integrated Systems, we continue to develop and manufacture JFET, DMOS, and bipolar technologies that enable engineers to solve challenging signal chain problems across a wide range of applications. Interested in learning more about our analog portfolio or discussing your application? Contact our engineering team or request samples using the links below.

Field-effect transistors are foundational building blocks in modern electronics, but not all FETs are optimized for the same tasks. While MOSFETs dominate digital circuits, power management, and switching applications, JFETs continue to play a critical role in low-noise, high-impedance analog front ends. Understanding the strengths of each technology can help designers optimize system performance and avoid unnecessary tradeoffs. Understanding the Differences Both JFETs and MOSFETs are voltage-controlled devices, but their operating principles differ significantly. A JFET controls current flow using a reverse-biased PN junction. A MOSFET uses an insulated gate separated from the channel by a thin oxide layer. These structural differences directly influence noise, capacitance, linearity, leakage current, and application suitability. Where JFETs Excel JFETs are often the preferred choice when preserving weak signals is the primary design objective. Ultra-Low Noise Performance Thermal channel noise is the dominant noise source in JFETs and is closely related to transconductance. Higher transconductance results in lower input-referred noise. The monolithic dual LSK389 achieves: 1.3 nV/√Hz input noise at 1 kHz 1.5 nV/√Hz input noise at 10 Hz Very low 1/f noise Elimination of burst, or popcorn, noise through 100% noise testing High Input Impedance JFETs offer extremely low gate leakage current, making them ideal for high-impedance signal sources such as: Piezoelectric sensors Electrometers Condenser microphones Photodiodes Hydrophones In high-impedance applications, even small leakage currents can significantly degrade performance. JFETs help preserve signal integrity where every femtoamp matters. Excellent Linearity JFETs exhibit highly linear transconductance characteristics, enabling low-distortion amplifier designs for demanding analog applications. Matched Differential Front Ends Monolithic dual JFETs, such as the LSK389, provide closely matched electrical characteristics and thermal tracking, simplifying the design of low-noise differential amplifiers. By paralleling matched JFET pairs, designers can achieve input-referred noise levels as low as 0.7 nV/√Hz in differential amplifier architectures. Where MOSFETs Excel MOSFETs are the preferred technology when switching efficiency, scalability, or integration density are priorities. Common applications include: Power management Battery-powered devices High-speed switching Motor control Digital integrated circuits Power conversion Beyond Switching: Radiation Sensing MOSFETs also enable specialized sensing applications. P-channel MOSFETs such as the 3N163 can function as RADFETs, or radiation-sensitive field-effect transistors. Radiation exposure creates electron-hole pairs within the gate oxide, causing measurable threshold voltage shifts that correspond to accumulated radiation dose. Unlike many sensing technologies, RADFETs can operate without a continuous power source and retain exposure information as a non-volatile analog memory. Studies using the 3N163 demonstrated radiation sensitivities of approximately 33 mV/Gy in unbiased operation and 62 mV/Gy when biased. Selecting the Right Technology Design Priority Recommended Technology Lowest input voltage noise JFET Lowest 1/f noise JFET High-impedance sensor interfaces JFET Differential analog front ends JFET Power switching MOSFET Battery-powered systems MOSFET Radiation sensing MOSFET Digital integration MOSFET The Bottom Line The question is not whether JFETs are better than MOSFETs, or vice versa. The right choice depends on the application. When the design challenge involves extracting weak signals from high-impedance sources while minimizing noise and distortion, JFETs remain the preferred solution. When switching efficiency, power handling, or large-scale integration are the priorities, MOSFETs are often the better choice. Understanding these tradeoffs enables engineers to select the right device technology for optimal system performance. Related Resources What Is a JFET? Why JFETs Still Matter LSK389 Application Note: Dual Monolithic JFET for Ultra-Low Noise Applications 3N163 Application Note: Radiation Sensor Design and Applications

When engineers design front-end circuits for microphones, hydrophones, piezoelectric sensors, MEMS devices, and other high-impedance signal sources, noise is often the first specification they look at. And for good reason. Weak signals can easily be overwhelmed by amplifier noise, making low-noise devices essential for preserving signal integrity. But noise is only part of the equation. In many high-impedance applications, input capacitance can have an equally significant impact on overall performance. The Challenge with High-Impedance Sources High-impedance sensors behave differently than low-impedance signal sources. As frequency increases, input capacitance interacts with source impedance to influence bandwidth, phase response, and distortion performance. Junction capacitances within active devices are inherently nonlinear, which can introduce unwanted effects as signals become larger or frequencies increase. For applications such as: Condenser microphones Electret microphones Piezoelectric sensors Hydrophones MEMS sensors Electrometer inputs Precision instrumentation both low noise and low capacitance become critical design requirements. Looking Beyond Noise Many designers focus exclusively on achieving the lowest possible noise figure. However, reducing capacitance can often simplify circuit design while improving real-world performance. Lower capacitance helps maintain high input impedance across a wider frequency range and can reduce distortion mechanisms associated with nonlinear device capacitances. In some applications, it may also reduce the need for additional circuit techniques used to manage capacitance-related effects. The result can be cleaner signal acquisition, wider bandwidth, and more straightforward circuit implementation. The LSK489 Approach The Linear Systems LSK489 was developed specifically for applications where designers need an exceptional combination of low noise and ultra-low input capacitance. This monolithic dual N-channel JFET combines: 1.8 nV/√Hz typical noise voltage 4 pF typical input capacitance Excellent device matching and thermal tracking High common-mode rejection performance Extremely low leakage current Available in TO-71, SOIC-8, SOT-23, and DFN packages The combination makes it particularly well suited for demanding sensor and instrumentation front ends where both signal integrity and source loading are important considerations. Applications That Benefit Microphone and acoustic sensing systems are excellent examples. Whether working with condenser microphones, electret elements, underwater acoustic sensors, precision vibration transducers, or MEMS devices, designers often face the challenge of extracting extremely small signals without compromising bandwidth or introducing distortion. The same considerations apply to modern instrumentation and electrometer circuits, where preserving the original signal is often just as important as amplifying it. The Takeaway Low noise will always be important. But when working with high-impedance sources, capacitance deserves equal attention. The best front-end performance often comes from balancing both characteristics rather than optimizing only one. For designers building the next generation of microphone preamps, acoustic sensing systems, hydrophones, MEMS interfaces, and precision instrumentation, understanding the role of input capacitance can be the difference between a good design and a great one. Interested in learning more about the LSK489? Visit the LSK489 product page HERE!

“My daughter, during her graduate studies in renewable energy applications, once needed a tester for solar panels to find their maximum power point under real-world conditions. So, this is what dads do. The resulting tester demonstrates how different semiconductors can be selected for highly specialized tasks. For instance, the JFETs used here serve three uniquely different functions. J3 is used as a variable-resistor gain element to regulate the control oscillator output level. J2 is used for polarity-control switching of a phase demodulator. J1 functions as a floating integrator reset switch, where charge injection is referred to the op-amp integrator’s low-Z output. Further down the line, a pair of LS312 bipolar duals are arranged in a classic National Semiconductor multiplier circuit for power calculation. Matching here is essential for achieving reasonable multiplier accuracy. The design objective is to apply an increasing low-level, low-frequency sine current load to the solar panel up to the point where the panel voltage begins to dip under load. An 88 Hz frequency was chosen because it is non-synchronous to nearly everything, while also being slow enough to avoid panel charge trailing effects (slew-rate related), minimizing gating error during demodulation. As long as increasing current results in increasing power, the circuit behaves as a positive feedback loop. A phase demodulator feeds the signal through an integrator until an increase in current produces a decrease in power level. At that point, the integrated phase demodulator signal reverses and becomes negative feedback. The balance point between increasing and decreasing current load on the panel occurs at the maximum product of current and voltage — the Maximum Power Point. Power per unit area of the panel is normalized to the 1 kW/m² standard, where quantum efficiency can then be determined using a separate radiant energy measurement alongside the panel. That portion of the circuit is not shown, as it was added later as an upgrade. Of course, a few op-amps and a PIC processor could also perform these functions — but that’s not who I am. To this day, integrated and discrete components remain complementary elements in control systems and analog signal processing, and that will foreseeably continue well into the future." - Kirkwood Rough Designing precision analog control, low-noise sensing, or measurement systems? We’d love to talk about your application. Click HERE or call (510) 490-9160.

In some JFET op amps such as the AD743, the input capacitance is in the order of 18 to 20 pF. In comparison, with an LSK489 dual FET, the input capacitance is in the order of 3 pF, which will be suitable for low noise photodiode applications. In this section we will see why it is important to have low equivalent input noise and low input capacitance in a photodiode preamp. A simple photodiode is shown in Figure 1 below, which uses an op amp. In the photodiode amplifier below, when light is shined onto the photodiode, current is generated by the photodiode, PD1. As configured with the cathode of PD1 connected to the (-) input terminal of U1, Vout generates a positive voltage proportional to the amount of light into the photodiode. Also shown in Figure 1 are the equivalent capacitances from the photodiode, Cpd, and (-) input terminal, Cin(-), which are connected in parallel. To minimize Cpd, the photodiode capacitance, the anode of PD1 is connected to the minus 12 volt power supply for maximum reverse bias to lower its junction capacitance. For example, if a BPV10 photodiode is used, Cpd is about 2.7 pF at 12 volts reverse bias. At a lower reverse bias voltage such as 1 volt, Cpd is about 7 pf. For low noise considerations, these two capacitances, Cpd and Cin(-), should be low as possible. The reason is that the equivalent input noise density voltage, Vnoise_input of the op amp will be amplified in the following manner at Vout, neglecting any noise current from the photodiode: Vout_noise for a bandwidth of 1 Hz = (Vnoise_input) √(1+(ωRFCt)2 ) + √4kTRF (1) Where ω = 2πf, RF = feedback resistor, k = 1.38 x10-23 Joules per degrees Kelvin, T = 298 degrees Kelvin Ct = Cpd||Cin(-) = total capacitance at the (-) input terminal, and Ct = Cpd + Cin(-) √4kTRF = thermal noise voltage of the feedback resistor RF for a bandwidth of 1 Hz. As we can see from the equation above, the output noise, Vout_noise, goes higher if Ct is increased. In designing a low noise transresistance preamps the goals are to: Minimum equivalent input noise voltage. Equation (1) above shows that the output noise voltage is dependent on the equivalent input noise voltage, Vnoise_input. Minimize noise current from the (-) input because the noise current at the input will form a noise voltage across the feedback resistor. Generally, a JFET is desirable for the (-) input because of its low gate noise current. Minimize the capacitance from the (-) input to ground. The equation (1) shows that more noise is generated at the output when the capacitance, Ct = Cpd + Cin(-), at the (-) input terminal is increased. Use as large value RF as possible. At first glance, it would appear increasing the resistance in RF would increase the output noise because of the resistor’s thermal noise. This is true but the signal amplification from the photodiode is increased more so that results in a net increase in signal to noise ratio when RF is increased in value. For example, doubling the value in RF increases the resistor noise from RF by √2 = 1.41 while increasing the photodiode signal output voltage by 2. Thus, there is a net gain of √2 or + 3 dB, in terms of signal to noise ratio in this example. In Figure 1, the typical input capacitance, Cin(-) at the (-) input of an FET op amp is about 18 pf. To lower the capacitance of the op amp, a low capacitance and low noise JFET is used as a buffer or source follower to the (-) input. See Figure 2. A low noise JFET such as an LSK189 is configured as a source follower, with a source biasing resistor R3. In terms of input capacitance at the gate of J1 with an LSK189, it is about 3 pF from the gate to the drain, which is much less than the 18 pF of Cin(-). Capacitance between the gate and ground due to the gate to source capacitance approaches zero. This is because the source follower configuration provides substantially the same AC voltage at the gate and at the source, which substantially cancels out the capacitance between the gate and the source. The source follower also greatly reduces the capacitance seen at the gate to ground even when the source is driving signal into a capacitive load, Cin(-), the capacitance at the (-) input terminal of the op amp U1A. It should be noted that the source follower circuit may add some phase shift to the overall amplifier circuit. To ensure phase margin and no oscillations, a resistive divider R1 and R2 is used. With the values given at 3900Ω for R1 and R2, the equivalent feedback resistance is [1 + (R2/R1)] x RF = 2 x 1MΩ = 2MΩ, the same resistance value shown in Figure 1 for RF. If the (+) input of the op amp in Figure 2 is grounded, such that Voffset = 0 volts, Vout will most likely have a DC offset. To “zero” Vout when there is no signal from the photodiode, a clean DC voltage, Voffset may be applied to the (+) input of U1. Op amp U1A = AD797 has an equivalent input noise voltage of 0.9 nV/√Hz and J1 =LSK189 with an equivalent input noise voltage of 1.8 nV/√Hz, the total equivalent input noise voltage is about 2.0 nV/√Hz. This is lower than a very low noise JFET op amp such as an AD743 that has 3.2 nV/√Hz. Note that bipolar input stage op amps AD797 with 0.9 nV/√Hz has lower equivalent input noise voltage than 2.0 nV/√Hz, but the AD797’s input noise current is too high and are not suitable for amplifier circuits with large value feedback resistors (RF) in the MΩ such as the circuit shown in Figure 1. Note that J1 may be substituted with an LSK170 (0.9nV/√Hz ) if a slight increase in capacitance from the gate to ground is acceptable. This FET has about half the equivalent input noise of the LSK189.

Accurate simulation depends on accurate models. That’s why we’ve released new SPICE models for the LSK389, now covering all A, B, C, and D grades. These updated models are derived from real device characteristics, with key parameters like VTO and β refined based on IDSS and transconductance. This gives designers a much closer representation of how each grade actually behaves in circuit. What this means for you: More accurate circuit simulation across all LSK389 grades Better predictability in low-noise and precision designs Improved confidence before moving to hardware The LSK389 remains a go-to solution for: Ultra-low noise front ends Differential amplifier stages Precision instrumentation Audio and sensor applications If you're designing with low-level signals, model accuracy matters. These updates help bridge the gap between simulation and real-world performance. Interested in trying them out or need help selecting the right grade? We’re happy to help.

Detecting weak signals starts at the analog front end. Modern sensing systems depend on capturing extremely small signals amid noisy environments. Whether the input comes from acoustic sensors, radio frequency (RF) devices, or other sensor types, the design of the front end determines the overall system’s ability to perform accurately and reliably. Precision analog components like the LSK389 dual JFET transistor play a vital role in this process. Their ultra-low noise and tight matching make them ideal for applications requiring sensitive signal detection and conditioning. Why Analog Front End Design Matters The analog front end (AFE) is the first stage in a sensing system where the raw signal is captured and prepared for further processing. This stage often deals with signals that are extremely weak and buried in noise. If the front end cannot cleanly capture these signals, no amount of digital processing or advanced algorithms can recover the lost information. For example, in acoustic sensing systems such as microphone arrays, the front end must amplify faint sounds without adding noise. The quality of this initial amplification directly affects the accuracy of beamforming and sound source localization downstream. Similarly, in RF detection, the front end must handle very low-level signals from antennas and convert them into usable data. Any noise or distortion introduced here reduces the system’s sensitivity and reliability. The Role of the LSK389 Dual JFET The LSK389 is a precision dual JFET transistor designed specifically for low-noise, high-sensitivity applications. Its key features include: Ultra-low noise voltage and current Tight matching between the two JFETs in a single package High input impedance suitable for sensor preamplifiers Stable performance over temperature and time These characteristics make the LSK389 well suited for: Acoustic sensor preamps and microphone arrays Low-level signal detection and conditioning Differential front ends in high-sensitivity measurement systems Using the LSK389 helps engineers build front ends that preserve the integrity of weak signals, enabling more accurate sensing and analysis. Applications That Depend on Clean Front Ends Several fields rely heavily on precise analog front end design to capture subtle signals: Acoustic monitoring and beamforming systems: Detecting faint sounds in noisy environments for surveillance, wildlife monitoring, or industrial noise analysis. Industrial and environmental sensing: Measuring low-level chemical, pressure, or vibration signals that indicate system health or environmental changes. RF signal detection and analysis: Capturing weak radio signals for communication, radar, or spectrum monitoring. Advanced security and surveillance systems: Identifying subtle acoustic or RF signatures related to unauthorized activity. Anti-drone detection: Combining acoustic and RF sensing to detect and track drones based on their unique signal patterns. In all these cases, the front end’s ability to minimize noise and distortion directly impacts the system’s effectiveness. How Acoustic Detection Benefits from Low-Noise Front Ends Microphone arrays used for acoustic detection rely on very low-noise front ends to pick up subtle sound signatures from a distance. The cleaner the signal at the start, the more accurate the downstream processing such as beamforming, source localization, and classification. For instance, detecting a quiet drone or a distant vehicle requires capturing signals that are often below the ambient noise floor. The front end must amplify these signals without adding noise or distortion. The LSK389’s low noise and matched dual JFETs help achieve this by providing a clean, balanced input stage. The Front End Sets the Stage for AI and Algorithms While AI and advanced algorithms often receive the spotlight in modern sensing systems, their success depends on the quality of the input signal. If the analog front end fails to capture the signal cleanly, no algorithm can recover lost information or correct poor data quality. This means investing in precision analog components and thoughtful front end design is essential. It ensures that the data fed into AI models or digital processors is accurate and reliable, leading to better detection, classification, and decision-making. Final Thoughts on Analog Front End Design The analog front end is the foundation of any modern sensing system. Detecting weak signals in noisy environments requires ultra-low noise, tightly matched components like the LSK389 dual JFET. These components enable clean signal capture for acoustic, RF, and other sensor inputs.

“There is sound and then there is the sound of sound.”™ In designing discrete element electronic audio circuitry, there will inevitably be a need to split a signal to a 0° and 180° phase differential. Either to drive a balanced signal to a transformer, a line signal to an XLR connector, or a differential voltage to drive a push-pull stage within a power amplifier stage, or on and on and on. Though not always needed for the task, a phase splitter will often require a zero volt DC offset, particularly for transformers to eliminate core bias and XLR drives where a transformer is the terminating end. Many high end audio systems seem also to need a zero offset differential input. When designing a differential Bipolar or JFET stage with current sourced emitters or sources, the attendant circuitry to make it a zero bias differential output can become a concern, “I’m being nice”. The single transistor splitter having a source/emitter resistor and a drain/collector resistor works but has the additional circuitry to reference outputs to ground. In the effort to minimize active components in the audio path and retain as high a linearity/fidelity for my designs, the circuit here is was what became a solution for many applications. Though the signal output of each arm of this differential splitter is -4db down from the input signal it is flawlessly the same, though inverted, and referenced to ground by an offset potentiometer. The POT places the source voltage at exactly -5V and the collector at +5V where the resistor string for both terminations have 1mA as does the JFET current. R8 & R9 combine R string current with the FET current to set 5 V from each rail for the JFET S&D. Choice of JFET depends on the load impedance and Gm required for an application of the design. Those choices determine linearity/fidelity for a signal range needed for the use. VDD&VSS can be anything up to the S-D compliance limit but I’ve found the sweet spot for a LSK170A to be 10 V. the resistor string for both arms is a 3:2:1: Ratio making the choice of Id simple. For lower Noise, Parallel JFETS such as Dual LS parts work well and offer greater drive current, such as the LS844s. An earlier circuit in these posts uses this configuration for the zero DC pass through audio application. This circuit and a traditional splitter are shown. Frequency and voltage response of input/output and gate to source voltage are shown. For this configuration, Gm shift/input voltage should be noted for the even order harmonic content on the output signal giving it a Vacuum tube like influence.” - Kirkwood Rough Low distortion JFET phase splitter using LSK170A, generating clean ±phase outputs with zero DC offset for high fidelity audio designs Input vs. ±phase output showing precise 180° phase splitting with matched amplitude and low distortion Frequency response showing consistent gain and phase behavior across bandwidth with well matched ±phase outputs

When people talk about AI, the focus is almost always on GPUs, CPUs, and advanced nodes. But that’s only part of the story. Behind every AI system is a broader layer of semiconductors responsible for signal integrity, timing, and control — the functions that allow high-performance compute to actually work in real systems. AI Performance Depends on More Than Just Processing Power Modern AI hardware is pushing limits across the board: Faster data movement Higher board density Tighter signal margins Greater power and thermal constraints At these levels, small electrical effects matter. Switching speed, capacitance, and signal loss are no longer secondary concerns — they directly impact system performance and reliability. Where Discrete Devices Make the Difference This is where high-performance analog and discrete semiconductors come in. Devices like high-speed DMOS switches are designed to support: Nanosecond-level switching speeds Extremely low capacitance for high-frequency operation Minimal signal distortion Efficient, simple drive requirements For example, devices in the SD/SST211 series offer: Turn-on times around 1 ns Reverse capacitance as low as 0.2 pF Performance optimized for fast, low-glitch signal paths These types of components are commonly used in: High-speed analog switching Sample-and-hold circuits DAC deglitching Timing and signal control functions They may not define AI systems — but they are essential to making them perform. The Analog Layer Is Becoming More Critical As AI infrastructure scales, engineers are placing greater emphasis on: Signal integrity at higher speeds Reliable, repeatable analog performance Long-term supply and second sourcing Flexibility for system-level optimization This is where precision discrete semiconductors play a growing role. Where Linear Integrated Systems Fits In At Linear Integrated Systems, we focus on ultra-low-noise JFETs and high-speed DMOS devices designed for demanding analog and mixed-signal applications. Our role isn’t at the processor level — it’s in enabling the performance, stability, and longevity of the systems around it. Because in the end, AI performance isn’t just about processing power, it’s about everything that supports it. Let’s Connect If you’re working on high-speed, low-noise, or signal-sensitive designs, we’re happy to help. Whether it’s: Finding a second source Replacing an obsolete part Improving performance in an existing design Or reviewing a new application Feel free to reach out or start a conversation HERE. We’re always open to discussing your design challenges.

High-voltage startup problems can quietly kill performance and hardware. This week’s Design Spotlight dives into a ±1kV soft-start circuit that solved a real-world failure in electron microscope HV amplifiers. “A HIGH VOLTAGE SOFT START CIRCUIT FOR THE MASSES “There are times where a system design doesn’t work when the anticipated design conclusion is missing a critical element. This happened when an electron microscope’s High Voltage amplifiers were turned on and capacitive input surge currents caused their power supplies to fold back and quit till reset. The first solution attempt was a simple R-C delayed gate voltage applied to a series HV MOSFET, which self-destructed as they went out of the safe operating VxI product range. So this circuit was contrived to solve this issue by employing a discrete transistor VxI multiplier to control the MOSFET turn-on power limit curve. Initial voltage drop across the MOS pair is sensed and applied to a differential current sensing bipolar pair controlling FET turn-on with a second bipolar pair forming a multiplying transconductance amp. This amp drives the MOSFET gate where device dissipation is then limited to a value of 100 watts when current is multiplied against the FET voltage drop for control. D6 provides a latching current to force this amp to provide full gate voltage at the turn-on cycle end. Though conceptionally simple this type of circuit required tight matching of bipolar parts used to form a predictable multiplying function, and, of course the LS312 and LS352 worked perfectly for the job. Since the HV amplifier loads had bipolar high voltages, negative and positive soft starts were made. The circuit is the same but all semiconductor polarities are reversed. HV FETs were cascoded because 1KV FETs weren’t available then; later converted to a single HV FET. Included is a plot showing input turn-on voltage, load voltage, and the load charging current. Pictured is the final board having plus and minus 1 Kv soft start circuits. This is one more example of a circuit favoring the application of precision discrete bipolar parts over an integrated solution. Circa 1998. - Kirkwood Rough” This kind of controlled VxI limiting approach is still highly relevant today in high-voltage and sensitive analog systems. Can you see which side is negative and which side is positive? It's staring at you. Working on high-voltage, low-noise, or precision analog designs? We can help with device selection, matching, and long-term supply. Contact Us Figure 1: High Voltage Soft Start Circuit Figure 2: Response Plo Figure 3: Prototype High Voltage Soft Start Circuit Board