It's been some four years since the first set of circuits, long time for another installment. This page contains my computerized electronic drawings from 2004 to 2007.
I would like to think these express my developing electronics skills. Many of them relate to my investigations at the bleeding edge of bandwidth, for instance the pulse generators, high bandwidth amplifiers, logic stages and such. Some of these schematics appear on other pages; I'm not sorting them out, but most are indicated. Most of them should be of better quality than last series, i.e. complete and probably functional. There are still mistakes, so again I offer these with no guarantee. I will offer some description on each this time. (Of course, I offer no guarantee that my *description* is at all correct either...) So without further adieu, the list.
Statistics: 104 images, 455KiB, average size 4.4KiB. Organized mostly alphabetically within categories. This file (index.html), 37KiB.
The classic joke.
A pretty typical differential amplifier and emitter follower (in this case, quasi-complementary) output stage. As I recall, this, and the three below, were part of a conversation I started concerning this circuit.
Using a complementary-differential input stage, this circuit ensures a 1.2V offset at the output. Oops. I may've drawn this after learning about the complementary differential stage, which has some esoteric uses, but isn't appropriate here.
I believe this design was suggested as a development of the preceeding one. Instead of a complementary differential stage, the emitters are broken and returned to the supply rails (with some bypass, assuring some AC gain). The collectors "tug" on the emitters of a complementary pair of current sources, biased with LEDs. The symmetrical drive and class A conduction should make this gain stage quite fast. (The complementary transistors should be hfe matched for best HF performance and low distortion.) Since the gain stage is not differential and inverts the signal, shunt feedback is used to set overall amplifier gain.
Also pretty typical, this design was drawn for single supply use. Beware the turn-on settling of single supply circuits...
This was drawn sarcastically in response to a conversation about high voltage gain stages.
Textbook 555 PWM circuit. Supply protection provided for automotive application. The MOSFET and snubber are intended for a beefy inductive load, such as a DC motor up to about 1/3 HP. The snubber (R || C) really should return to the +12V rail; then, a smaller resistor could be used to quickly dissipate more flyback energy.
The complete (discrete) schematic for an analog sampler device. The block diagram is fundamentally the same used in all sampling oscilloscopes (be they digital or analog). The overall effect is that the input is frequency-shifted down to a multiple of the sweep frequency. This circuit, as shown, will only trigger reliably up to a few hundred kilohertz, but a more sophisticated circuit (namely, a sharp comparator and S&H) is suitable up to 50GHz in the most advanced oscilloscopes today.
A method to switch AC using MOSFETs. An isolated gate driver could be used instead of the battery and switch.
This circuit consists of a small high voltage inverter, current-limited linear regulator and edge triggered avalanche breakdown pulse generator. In effect, a discrete realization of Linear Technology's pulse generator listed in AN-47. Functionally, the 2N3904 breaks down around 100V, exhibiting negative resistance. If a capacitance is attached (5pF plus an optional pulse line, an open 50 ohm cable), that negative resistance can discharge the capacitance very quickly, typically under 5ns. The 2N2369 is also suitable as an avalanche device. The sync isn't very stable, and would probably benefit from higher voltage or isolation (a pulse transformer could couple the BFQ241 directly to the 2N3904's base). With harmonics into the GHz band, pulse quality depends heavily on wiring.
These two related cicruits are the result of contemplating a current-mirror-loaded differential amplifier with symmetrical loading and outputs. Of course, the first is pointless, as both In1 and In2 have a diode drop above COM, but by adding additional outputs, the effect should be that turning one current mirror on turns the other off, producing fast, balanced operation. But for one side to turn on, the connection from the other side limiting its current must also turn off, so the improvement (if any) may not even apply.
Textbook example of bootstrapping a biased emitter follower.
A symmetrical differential amplifier, with common-mode servo controlling bias voltage. (-Out + +Out) / 2 should remain constant. Addition of a buffer (e.g., emitter follower) to each side should give a balanced (bridgable) power amplifier, taking care to allow feedback to adjust AC balance as well as quiescent operating point. Half the time, I can convince myself this circuit will be stable...
Bipolar equivalent of a CMOS inverter stage.
I used this circuit, built on a solderless breadboard, for one summer's duty amplifying my computer / radio audio. Even with op-amps, the headphones output was prone to power supply ripple, unfortunately. Supply voltage was a single 15V: notice the two capacitors effectively splitting the supply into +(+V/2) and -(+V/2). This approach effectively eliminates the turn-on POP associated with most capacitor-coupled single-supply amplifiers, but produces some "weird" saturation waveforms at low frequencies (because the capacitors charge to one side or the other).
A roughly constant voltage is developed across the transformer, so long as it's not loaded so heavily as to induce current limiting. The current-limiting PNP transistor must be rated to dissipate full current in case the load is shorted. Notice the unbalanced inverter which assures some amount of shoot-through on switching; since the supply is current-limited, this isn't an issue.
A simple current amplifier, based on resistive current sources mirrored. This circuit has excellent bandwidth, high voltage gain, good linearity and terrible thermal stability. Shoot-through is a major source of heat when driven fast. Unfortunately, the stability issues aren't easy to fix with this topology.
A representative diagram of two op-amp-stabilized MOSFET current sinks (if Vout is grounded and Vin is the current output terminal). On the left, R1 and R2 form a voltage divider, so overall gain is Iout = Vs / Rs = VIn * R2 / [Rs * (R1 + R2)], whereas on the right, R1 and R2 form a voltage divider within the feedback loop, so that overall gain is Iout = VIn * Rs * (R1 + R2) / R2, that is, the contribution R2 / (R1 + R2) is inverted by the negative feedback, so the difference in gain between these circuits is [R2 / (R1 + R2)]2.
This, and the two drawings below, belong to my scope mod project. This design boasts JFET buffered inputs, differential input (the 2N5179s and 2N3904 current sink form a long-tailed pair, so common mode signals are rejected), high voltage output and common mode servo biasing.
This is the above circuit, as built. I omitted the JFET buffers, opting to take the LTP bases out on shielded twisted pair. Some values are different, reflecting my parts supply and experimental goals. Unfortunately, the 2SC1569, despite a claimed fT ca. 100MHz, is only good for full output up to a few MHz. The 2N5179s were prone to UHF oscillation (notice the 10pF capacitors added in an attempt to stabilize the circuit), a somewhat slower type such as the MPS3563 shown below, or even regular old 2N3904s, is quite sufficient.
The revised model changed to one buffered input (the other is available for biasing, e.g. position control), a somewhat slower differential amplifier (MPS3563 fT ≥ 600MHz) and instead of emitter follower drivers (suffering from instability), less coupled common emitter stages and a cascode output stage. The design output rail is only 100V (50V shown), as I found the Heathkit's CRT deflects at about 10V/div, so only 60Vp-p is needed to fully cover the graticule. This reconsideration allows much higher output current and much faster transistors to be used (the 2SC3597 is a high-bandwidth video amplifier, fT = 500MHz). The actual measured bandwidth drops off around 20MHz, and as built, has about 10% of trash on the waveform: not quite oscilloscope grade material, but certainly a good start. Note: circles represent wires taken off the PCB; components past circles were breadboarded for testing.
Two edge-triggered one-shot 555 timers wired in series, so the LED lights for a certain duration only a certain time after the switch is closed.
In boredom one day, I breadboarded this circuit. The 2N3646 is similar to a 2N3904, but old -- the stock I had on hand at the time was probably 30 years old! This circuit had good bandwidth, but dreadful gain (less than the gain of 4 the shunt-feedback resistors suggest), I'm not sure why.
Diodes store charge when forward-biased. When reverse-biased, the charge takes some time to clear the junction. This reverse recovery has a current vs. time profile depending on the doping profile of the junction. In abrupt and hyperabrupt diodes (varactors, more commonly used for their steep junction capacitance vs. reverse voltage profile), the charge is cleared uniformly for some time, then -- in just a few picoseconds for the fastest diodes -- the junction suddenly stops conducting. The dI/dt causes a large induced voltage across any nearby inductance, making a visible blip on the oscilloscope. Your average diode doesn't work too well at this, although some high voltage diodes like the 1N4007 (which have a profile more like a PIN diode) have been known to work. I haven't seen any. The sharpest pulse I observed came from a schottky barrier diode, which was probably due to the steep capacitance vs. reverse voltage response rather than any junction recovery (after all, schottky diodes have no minority carrier charge as such).
In the above circuit, forward bias (about 1A) was provided by a 15 ohm, 20W power resistor. Supposedly, applying forward bias for a short period of time gives better results, so I created this double pulse generator to excite a somewhat different pulse head. I got similar results as before.
A simple RS flip-flop, in traditional RTL (Resistor-Transistor Logic). Resistor values depend on supply voltage; example values might be +V = 5V, R1 = 470Ω, R2 = 4.7k, R3 = 2.2k.
These circuits come from my flip-flop tutorial.
This circuit limits load current to a specific value (depending on the shunt resistor). Since it responds within microseconds, a bypass capacitor of perhaps several nF would be a good addition, after the 100 ohm resistor. This circuit can easily be adapted to other voltages and layouts; it was designed specifically for a 50V supply.
These drawings come from my sample-and-hold page.
Expanding the last drawing above, this sampling head uses current source biasing and open-collector switches to enable/disable the diode bridge, connecting Input to Output bidirectionally and symmetrically. SRD is a Snap Recovery Diode, which is intended to reverse-bias the schottky bridge in under a nanosecond, allowing this circuit bandwidth into perhaps the GHz range using otherwise regular parts.
A crude buffered, half-wave diode gated sampler. Note: complementary SAMPLE signals must be used.
A relatively conventional amplifier, this was found to give a slew rate around 20V/μs, which is comparable to such op-amps as the TL072, which is reasonably fast.
Two current mirrors with a JFET and resistor between them. This motif, with a bipolar transistor, appears in several circuits on this list, such as the Fast Function Generator and the H Bridge Connection.
Flyback supply concept.
This circuit forwards the rising edge and inverts the falling edge of the input square wave, thus doubling its apparent frequency. Values of R and C depend on frequency, supply voltage and bias current. Because it depends on frequency I added "not general".
A typical function generator circuit, this consists of a current mirror (the top mirror sends double current to the bottom mirror), JFET follower, differential hysteretic comparator and an output buffer. A triangle wave is seen at JFET source, which can be trimmed into a sine wave, hence this is a function generator.
A simple, versatile H bridge connection. For inductive loads, antiparallel diodes are recommended across the output transistors.
This circuit accepts input from an off-the-shelf ATX style computer power supply and regulates it down to a regulated voltage or current, suitable e.g. for electrolysis experiments.
An improved version, this circuit does not require an isolated 12V supply and is simpler in design. Warning: both this circuit and the above should have adequate compensation, inside each CCM and overall.
Adjusts a car's idle fuel rate depending on RPM and temperature.
Comes from my page on testing inductors.
Another joke (a gate used with annoying signals).
The challenge was, how to light two LEDs, only one or the other at any given time, using just resistors and a switch. This seems to work on paper.
Appropriate methods to power long strings of LEDs.
LED1 lights and LED2 turns off when Switch is closed. LED2 is operated by shorting it out, so current consumption will be quite high.
This was a response to an automotive LED flasher question. CD4022 is an octal counter (cf. the classic 4017 decade counter), so the left and right banks alternate at 1/4th clock frequency. The two 1N914 diodes act as a diode AND gate, so the outputs blink four times left then four times right. The 2SD1273 was chosen for its superhigh hFE and power capacity. Alternately, a MOSFET could be used, or the signals could be amplified with another transistor stage and the loads switched with a conventional bipolar transistor. A darlington isn't really suitable due to the forward voltage drop. Warning: supply filtering should be provided (diodes, TVS, large filter capacitor, etc.) to protect the digital components from automotive switching noise.
A MOSFET switch connects a charged (open) transmission line to the output, thus sending a square pulse of half the peak voltage and length corresponding to the electrical length of the delay line. Notice the MOSFET is "flying", so a diode is provided to clamp the flyback current which builds in the coax transformer.
Regulated boost converter runs motor. Warning: not current limited, so a heavy load or excessive voltage demand WILL cause the MOSFET to fail. A clamp could be applied to the error amplifier's output to prevent it from exceeding the triangle wave's peak voltage.
Very, very versatile circuit. Many an online question I have answered with this circuit, or an equivalent (e.g., using the 555). An RS flip-flop (which could be implemented with just two transistors) controls state, while discharge and "comparator" transistors control the voltage on the timing capacitor CT. Using the Trig input as shown, the flip-flop is edge-triggered, so the circuit can be used as a one-shot timer, sweep generator (particularly with a constant current charging CT), variable delay (RT (or a current source) variable, or changing the 3.3V zener), or with Trig wired straight to the transistor (omitting the 220pF and 1N914), a missing-pulse detector as well. Typical performance with jellybean transistors and resistors as shown has rise/fall time and propagation delays on the order of 100ns. Large capacitor values with small charging currents easily yield delays of several minutes.
Textbook op-amp circuit.
A simplified approximation of the reactive components that make up a transformer.
How to wire FWBs in parallel. Only two resistors are needed per bridge, because only two diodes are conducting at any given time.
This circuit probably won't actually work; the resistors are wrong. Note that, when a comparator is on, the 10k and 4.7k form a resistor divider that only allows the CMOS gate's input to reach 2/3 +V. The correct wiring, which is on my original induction heater control circuit, has the resistors going to +V. (Note: that circuit is itself faulty, as I soon realized I needed the greater phase margin of a type II over the type I shown.) That said, operation is as follows: the comparators, 100pF capacitors and CD4001B inverter-wired-NORs constitute rising-zero-crossing pulse generators, one for each input, A and B. These pulses, which are perhaps 400ns long (roughly 100pF * 4.7k), set or reset the RS flip-flop. As a result, the flip-flop stays in the high or low state an amount of time corresponding to the delay between edges, in other words, it is a PWM signal proportional to phase. The output filter removes [most of] the switching, leaving a continuous output.
An analog realization of Pong is conceptually quite easy: only switched current sources/sinks to make the "ball" (voltage on a capacitor) move, and to set the trajectory of the ball, sample-and-holds can be used to maintain a constant, arbitrary current. This circuit is the simpler axis, the Y axis, which bounces between walls and has variable rate from the sampler.
This circuit operates by using pulses to switch a flip-flop into one state or the other; as such, it can be used from DC to approximately 4MHz. However, if the state changes internally (due to desat, RFI, etc.), there is no guarantee that anything will continue to work right. In practice, I found what appeared to be random fluctuations in the inverter output waveform while using this on my induction heater, which went right away when I switched back to a full pulse coupled driver.
A standard differential amplifier, with quasi-complementary output stage. 4-0976 is a house-marked power transistor of roughly Vcbo = 100V, Ic > 2A, hFE ~ 50, fT ~ 3MHz, of which a lot were used to drive a dot-matrix printhead. The 33pF capacitor is necessary to compensate for L, an artifact of breadboarding; it wasn;t found necessary on PCB. As of writing, this is my current preferred medium power amplifier circuit, offering about 10W from a somewhat lower voltage. Linearity and stability are excellent, and bandwidth is sufficient (about 100kHz, one of the slower circuits I've made).
A tutorial of different rectifier circuits, illustrating their similarities.
When driving a MOSFET where high bandwidth and linearity are necessary, this class A, constant-current-biased, complementary emitter follower driver circuit could be used.
A tutorial on forward/flyback inverter methods.
This circuit comes from my induction heater project.
Amplitude detector switching a load.
A power emitter follower.
This power supply is featured in the avalanche pulse generator circuit above.
Common-emitter gain stage followed by emitter follower, this amplifier suffers from low PSRR. It was operated from the computer's 12V supply, which I found is amazingly noisy for a regulated supply.
A classic circuit, a self-excited class C oscillator/amplifier.
A typical (open-loop) realization of a TL494 based half bridge switching power supply. Gate drive is supplied by an open-collector H bridge, so a reactive circuit (LC filter and FWB) is necessary to suppress flyback current in the transformer. Output coupling capacitance as shown is excessive, but wasn't found to be harmful.
Thanks to the high bias current and overall simplicity (only five transistors and two diodes in the functional loop), this triangle wave generator is able to run very fast. With Ct = 0 and the circuit breadboarded, this circuit runs at over 30MHz. With Ct = 680pF and room temperature about 60°F, it was found adjustable from 1Hz to 4.5MHz in one range. Linearity is excellent with sufficient capacitance (Ct > 200pF or so). The squarewave at the 2N4403 collector has rise/fall times on the order of 20ns.
This logic output stage is simple and reasonably effective, having lower offset voltage than an unbiased complementary emitter follower. Such a circuit could be used to drive medium-sized MOSFET gates, for instance. This circuit is wired for 12V RTL operation, and shows rise/fall time around 50ns into 50 ohms.
These three circuits come from my SMPS tutorial using the UC3842.
A DC-coupled, differential, buffered preamp with active tone stack. DC bias is dubious and AC stability and tone effectiveness are suspect, but this is a good example of DC coupled tube circuits.
These three circuits are in a series concerning my development of a (possibly world's first?) class D tube amplifier. And just to be fun, it's using forgotten compound tubes, in this case, 6X8 (triode-pentode with common cathode) and 38HE7 (beam tetrode plus damper diode). The first diagram is representative of the topology, which shows a short-tailed-pair comparator forming an RC oscillator, another STP forming a PWM comparator (the RC oscillator produces a roughly triangular waveform of about 20Vp-p). The resulting PWM'd square wave is sent to a pentode, which has low saturation voltage (~50V). An LC filter and reaction diode (with Vf ~ 30V) strip the PWM from the audio, which is transformed to speaker impedance conventionally. PWM doesn't affect the characteristic impedance tubes offer, so an output transformer is still necessary, but the low saturation voltage and high peak current capacity of sweep output tubes allows revolutionary efficiency (up to 90%) and incredible power output (perhaps over 200W from a pair of 6LQ6). A side effect of the high current (over 1A peak per tube for beefy types e.g. EL519), very low primary impedances are necessary (~600 ohms), making these OPTs somewhat easier to wind.
Comparing the first two to the last, it was found that a gain stage and cathode follower were necessary to develop sufficient voltage and drive current to "switch" the 38HE7. The output network also changed, since the earlier filter was fallacious. Since output bias current is supplied by a resistor, this could be referred to as Class D-A, since the output operates constant current, but the active device is switched. Efficiency is similar to class A for obvious reasons, but the switching device doesn't dissipate much. The full circuit generated about 3.5W at nearly 25% efficiency, which is essentially theoretical for resisitve-loaded class A. Linearity was also quite excellent.
Although sweep tetrodes offer extremely good performance in saturation voltage and peak current, if plate voltage drops too low, screen current goes up astronomically (easily 200mA+!). A method to limit saturation that doesn't depend on load current or input waveform would be very useful. In solid state, a schottky diode (named the Baker clamp) is used to divert base current to the collector, reducing storage time (as used in, for instance, the 74S and 74LS series "schottky" logic chips). Screen voltage is often above plate voltage, and in sweep tubes, not by too much (Vg2 = 100~200V), so a simple diode could be used to divert screen current as shown, thus reducing peak screen dissipation. On triodes, the control grid is significantly below the saturated plate voltage (by perhaps 75V), so diverting grid voltage is more troublesome. As shown in the second diagram, a current sink could bias a constant voltage drop, and a diode would then be able to divert grid voltage, so long as the downward current is greater than the grid drive's current capacity.
Examples of biasing a tube for constant current draw using the traditional op-amp-and-follower current mirror circuit. In the top right, the op-amp responds to cathode current slowly (C1 should be placed between output and -in, as shown next), while R4, a grid-leak resistor, allows C2 to couple in the signal voltage. In the lower right, the signal travels through the op-amp, which has unity gain at signal frequencies and therefore is functionally equivalent to the earlier circuit, but some may have concerns over the signal passing through an op-amp anyway. In the lower left, a voltage feedback loop is added, demonstrating a servo-biased low impedance tube output stage.
I used this simple lash-up to plot tube curves. A high voltage transformer provides sweep.
I'm not sure what I drew these for. They appear to be regular sorts of stages.
One way to convert Frankenhouse to a SET amplifier. Notice the bottom 6V6 is not driven, it provides a constant current draw to ballast the triode 6V6 across the transformer. Power output will be perhaps 10dB down, as 6V6s make very bad triodes.
Examples in a series of possible developments incorporating a MOSFET follower into the μ-stage. For ultimate bandwidth (in excess of 200kHz), the Corrected Mosfet Driver is used here.
The classic RC oscillator.
So named because the pentode is on the bottom. Biasing would be better served from a voltage divider. High gain (potentially as high as the pentode's "mu", which is typically around 2000) is expected, as well as moderate driving capability (the triode acts as a cathode follower). Linearity is questionable, as pentodes are usually accompanied by negative feedback, reducing gain and distortion.
Revision 3's circuit with the tone control and preamp stage scratched out.
This all-tube servo biased μ-stage allows use of an ungapped OPT for power output. It is drawn by the triode, which provides gain for the tetrode or pentode, which acts as a cathode follower output with current Iq. Since Iq > It is probably true, a sink carrying Iq must be provided (preferrably something better than a resistor, which would waste half the output power).
A rearrangement of a typical SRPP circuit. I think. I can't quite convince myself this is right...
Perhaps the most beautiful circuit, this circuit comes from my flip-flop tutorial.
A tube PWM generator circuit, consisting first of a hysteretic comparator (CCS loaded, although the 6HS8 has poor plate curves for this purpose), dynamic CCS (the 6AB4, ala μ-stage) and discharge switch (12B4). This is followed by a PWM comparator with just enough positive feedback to increase its gain just to the edge of hysteresis. Incidentially, this was drawn years before my Compound Class D amp above.