What parts are “jellybeans”, and what are they like?
Normal FETs are N-channel enhancement-mode MOSFETs; JFETs, depletion-mode MOSFETs, and to some extent P-channel FETs are weirder.
G.M. Electronica’s power MOSFET list
Here’s a table of 16 popular FETs I found, with prices in quantity 1 from Digi-Key.
| PN | Vds | A | ohms | Qg (nC) | ¢ | W | type |
|--------------+-----+------+-------+---------+-----+-----+------------|
| 2N7000 | 60 | .2 | 1.9 | 2 | 36 | .4 | |
| 2N7002 | 60 | .115 | 7 | 2 | 38 | | |
| IRF630 | 200 | 9 | .4 | 45 | 86 | 75 | |
| IRF9630 | 200 | 6.5 | .7 | 29 | 151 | 74 | P-chan |
| IRLI630G | 200 | 6.2 | .400 | 40 | 229 | 35 | |
| IRLML6344 | 30 | 5 | .029 | 6.8 | 36 | 1.3 | |
| IRLML6402 | 20 | 3.7 | .065 | 12 | 40 | 1.3 | P-chan |
| EPC2036 | 100 | 1 | .065 | .910 | 97 | | GaN |
| SI3483CDV | 30 | 8 | .034 | 11.5 | 89 | 4.2 | P-chan |
| FQP27P06 | 60 | 27 | .070 | 43 | 134 | 120 | P-chan |
| NTD4906N | 30 | 54 | .0055 | 24 | | 2.6 | obsolete |
| IRF7307 | 20 | 4.3 | .140 | | 83 | | dual (P&N) |
| BSS138 | 50 | .200 | 3.5 | | 24 | | |
| CPC3703CTR | | | | | 70 | | depletion |
| 2N5457 | 25 | .01 | | | 230 | | JFET |
| 2N5458 | 25 | .01 | | | 230 | | JFET |
| SiS410DN | 20 | 35 | .0048 | 41 | 94 | 52 | |
| PSMN4R0-40YS | 40 | 100 | .0056 | 38 | 88 | 106 | holy shit |
| IRF540N | 100 | 33 | .044 | 71 | 145 | 130 | fuck |
| IRF9540N | 100 | 23 | .117 | 110 | 189 | 110 | P-chan |
| IRF9530 | 100 | 12 | .300 | 38 | 138 | 88 | P-chan SyC |
The IRF630 N-channel FET https://www.digikey.com/product-detail/en/stmicroelectronics/IRF630/497-2757-5-ND/603782 is a monster. It’s Digi-Key’s most popular IRF* part (twenty thousand in stock at 86¢ quantity 1, 39¢ in quantity 1000) and can switch 9 amps (with pulses up to 36 amps) at up to 200 volts in 20 nanoseconds. Even G.M. Electronica, which generally only carries obsolete parts, lists an IRF630 in their catalog. And the TO-220 package can dissipate 75 watts. It includes an antiparallel zener capable of about ten amps for overvoltage and flyback protection. Its main drawbacks for Arduinoish use seem to be that it doesn’t turn fully on until about 6 to 10 volts on the gate, its on-resistance is a high-for-a-power-FET 0.4Ω, and its max gate-source voltage is only 20 volts; but even at 4 volts, it’s ohmic with a reasonably low resistance up past two amps, maxing out at about three. Five volts gets you up to 11 amps. SyC Electrónica has it for US$0.75. It has a P-channel counterpart, the IRF9630, of which Digi-Key only has 2000 in stock; it’s slightly worse at 6.5A and 0.8Ω.
The logic-level counterparts to International Rectifier’s IRF parts are the IRL parts. One popular one is the IRLML6344 http://www.digikey.com/product-detail/en/infineon-technologies/IRLML6344TRPBF/IRLML6344TRPBFDKR-ND/2538162 (36¢ for one, down to 11.3¢ each) which switches 5A, 30V, at only 29mΩ, and turns on at 2.5V on the gate. Digi-Key has 650,000 of them in stock. A similar, slightly crappier, P-channel device is the IRLML6402 http://www.digikey.com/product-detail/en/infineon-technologies/IRLML6402TRPBF/IRLML6402PBFCT-ND/812500. The IRF630 itself has an IRL version, the IRLI630G, with the same 200V but a slightly lower current of 6.2 amps and like 70 ns of delay, but this is apparently a rare bird — Digi-Key charges US$2.29 for one.
The 2N7002 is Digi-Key’s absolute most popular FET, with 1.6 million in stock; it’s another N-channel FET that even G.M. Electronica carries; Digi-Key has 184 models, the most popular being https://www.digikey.com/product-detail/en/diodes-incorporated/2N7002T-7-F/2N7002T-FDIDKR-ND/1837287, which is 38¢ for one and 8¢ each for 1000. This is a smaller-signal part — although it can handle 60V, it can only do up to 115mA and 300mW, and its on-resistance is multiple ohms (like 7Ω). But it switches at 2.5V, also in the same 20ns as the IRF630. At 5V gate voltage it reaches 500 mA, which (because of the high resistance) will burn it up if you let it continue. The ON Semiconductor version http://www.digikey.com/product-detail/en/on-semiconductor/2N7002LT1G/2N7002LT1GOSCT-ND/917791 is cheaper, as low as 2.958¢ in quantity 1000.
SyC Electrónica doesn’t have the 2N7002 (though G. M. Electrónica does), but they do have the similar 2N7000 for US$0.12 http://www.sycelectronica.com.ar/articulo.php?codigo=2N7000. 60V, 200mA, 1.9Ω, switches on at 3V Vgs. eLemon carries it too at almost the same price http://www.elemon.net/BuscarSubRubros.aspx?Action=2&GrupoId=TR&RubroId=1710&SubRubroId=4. Even Tom Jennings recommends it!
The 2N7000/2N7002 datasheet from Fairchild doesn’t give its Qg directly, but it seems to be designed for a 10V Vgs and has typically 20pF and up to 50pF of input capacitance, which I guess would mean 500pC? But ST’s datasheet for their obsolete 2N7000 gives that for Qgd, plus another 800pC of Qgs, for a total of 1.4 to 2 nC.
The most popular non-2N7002 FET at Digi-Key is the EPC2036, which is a bumped gallium-nitride die (900μm square) with no packaging (to avoid bond-wire inductance!) https://www.digikey.com/product-detail/en/epc/EPC2036/917-1100-6-ND/5224979 for 97¢, down to 39¢. Needless to say, G.M. Electronica has never heard of it. It switches 1.7A at up to 100V with a resistance of only 65mΩ, which is supposedly about a 30 times higher breakdown voltage than can be achieved at that resistance with silicon, with a gate charge of only 910 pC, about an order of magnitude lower than MOSFETs, on a gate capacitance of about 80 pF. As a result, they can switch at over 10MHz. Apparently GaN power transistors like this debuted in 2010, which is why I hadn’t heard of them until now.
Digi-Key’s most popular P-channel MOSFET is Vishay’s SI3483CDV, with six hundred thousand in stock at 89¢, down to 41¢. It switches up to 8 amps at up to 30 volts, and can handle gate-source voltages of up to 20 volts. Being a P-channel device, the drain and source are backward from the more common N-channel type: the source is positive, drain negative, and the gate voltage is measured below the source.
Other candidate P-channel MOSFETs people on #electronics recommend for default use include the IRF4905 (94¢, -55V Vds, 0.02Ω, 74A, 200W, 2-4 V threshold voltage, fully on at 10 volts, 180 nC gate charge, 18 ns turn-on delay time, 99 ns rise time, 61 ns turn-off delay time, 96 ns fall time).
In P-channel MOSFETs, SparkFun recommends the FQP27P06 https://www.sparkfun.com/products/10349 for 95¢; Digi-Key has them https://www.digikey.com/products/en?keywords=fqp27p06 for $1.34 down to 61¢. They switch up to 27 amps at up to 60 volts.
Tom Jennings recommends the NTD4906N. G. M. Electronica doesn’t carry it, and Digi-Key marks it as obsolete. I don’t understand its datasheet, which has wildly different current and power numbers. I suspect I didn’t understand any of them.
For building an H-bridge, it’s often useful to use N-channel and a P-channel MOSFETs together, and sometimes they come in a package for this purpose; the IRF7307 is a popular such device at Digi-Key https://www.digikey.com/product-detail/en/infineon-technologies/IRF7307TRPBF/IRF7307PBFDKR-ND/1648232 for 83¢ down to 37½¢, with twenty-two thousand in stock. G. M. Electronica also carries it. It can switch 4.3 amps at 20 volts in 100ns; its resistance is 50 mΩ on the N-channel side and 90 mΩ on the P-channel side. This is convenient, but in the standard design, you kind of have to use the same supply for the gate as for the load.
For a MOSFET with lower switching voltage, people seem to suggest the BSS138. http://www.digikey.com/product-detail/en/on-semiconductor/BSS138LT1G/BSS138LT1GOSDKR-ND/1831753 lists it at 24¢ down to 4.6¢, switching 50V 200mA, with a high on-resistance of 3.5Ω. Sure enough, though, it passes almost 400mA at a 2.5-volt gate voltage.
Depletion-mode MOSFETs are apparently very exotic; Digi-Key’s most popular one is the high-voltage 70¢ CPC3703CTR, of which it has almost forty-eight thousand in stock.
Of all of these, the only ones I’ve seen mentioned in Horowitz & Hill are the 2N7000, the 2N7002, and the BSS138. Neither GM nor SyC has the BSS138. Horowitz & Hill also tout LND150, DN3435, BS170, MMBF170, 2N5457, 2N5458, 2N5459, and BS250 parts as popular.
The 2N5457 and 2N5458 are available at SyC, though not GM. Digi-Key has them for US$2.30 (!!) down to US$1.01 but they sound inferior to others mentioned above (25V, 10mA). They’re JFETs, though, so not directly comparable to the MOSFETs above.
All of the above are good only up to less than ten amps. The SiS410DN, by contrast, handles up to 35A; it’s a 20V N-channel MOSFET in a somewhat unusual package with almost 215,000 available at Digi-Key for 94¢ down to 43¢. Even more impressive, Philips/NXP/Nexperia’s PSMN4R0-40YS, which came out in 2010, does 100A at 40V, up to 106W, costs 88¢ down to 51¢, and has 77,000 available at Digi-Key. Here in Argentina, apparently the only power MOSFETs people sell are higher-voltage International Rectifier (Infineon) HexFET parts like the IRF540N — at Digi-Key, 10,000 in stock, 145¢ down to 66¢, 33A at 100V, nice low 4V threshold voltage, and available from both G.M. and SyC. SyC sells it for 60¢. The P-channel counterpart is the IRF9540N. All SyC has in stock is the IRF9530, which is rated for only 12A.
Amusingly, the TIP120 bipolar NPN Darlington Jennings recommends replacing with an unavailable MOSFET is still “Active” https://www.digikey.com/product-detail/en/fairchild-on-semiconductor/TIP120/TIP120-ND/1052441. It switches 60V at 5A and costs 61¢ down to 26¢, but as Jennings points out, its 1000:1 or 2500:1 current gain is vastly inferior to power MOSFETs’; its voltage drop is many times higher, especially at low current; and its switching time is vastly longer at around 1000 ns.
The ULN2003 costs US$0.40 at SyC: http://www.sycelectronica.com.ar/articulo.php?codigo=ULN2003 It’s an array of seven 500mA 50V 250ns current-sink Darlingtons with built-in flyback diodes. It’s like a seven-pack of somewhat lower-powered TIP120s.
There’s an 8-transistor version, the ULN2803, and (as I found in Microlens vibrating lightfield) there used to be a high-side switching version UDN2891, but it’s obsolete now. One alternative is the TI TLC59123 (US$1.81), which is an 8-output latching high-side driver, which has the advantage that it ignores its inputs except at clock edges. It’s only 13.2V, unlike the 60V ULN2003, but it’s also 500mA per channel, and it can supposedly be clocked at “up to 1 MHz” with clock pulses of 100ns and 100–200 ns propagation delays.
Aside from the TIP120 and ULN2003 mentioned above, we have transistors. For example, https://www.digikey.com/product-detail/en/diodes-incorporated/MMBT3904-7-F/MMBT3904-FDICT-ND/815727, the MMBT3904, for 12¢ down to 2½¢, a SOT-23 NPN small-signal transistor. That's the SOT-23 version of the 2N3904, thus the weird prefix.
SCRs, triacs, and occasionally diacs are the workhorses of powerline switching. They typically turn on in 100 ns. Unlike MOSFETs, they fail open rather than closed (I think), but they have the disadvantage that you can’t turn them off until the next zero-crossing of the waveform. Also, unlike MOSFETs and unlike relays (the other alternative for powerline switching), they have substantial voltage drops and so dissipate substantial power from the load current.
You might think that this would entirely prevent you from using them for switching DC, but the standard trick to solve this problem is to put an LC resonator either in series or in parallel with the thing so that the current drops to zero at some point, turning the thyristor off. This is called a “self-commutation circuit”.
Triacs are the usual thing to use for powerline switching, since they have a gate electrode like a regular transistor, but can handle power in either direction like a diac. A typical triac might be the 95¢ STMicroelectronics T405Q-600B-TR (“Applications: Mahjong machines, lighting dimmers”), which can block up to 600 volts; once you trigger it with 1.3V and 5 mA on the gate electrode, it can carry up to 4 amps (dropping 1.5 V or less), and will continue to do so until the load current drops below 10 mA.
Diacs have no trigger electrode; they’re triggered just by reaching the breakover voltage, like the no-longer-produced Shockley diode. A diac is like two Shockley diodes in antiparallel. A typical diac is the 49¢ Micro Commercial LLDB3-TP “silicon bidirectional diac”, which breaks over at 28–36 volts in 2 μs and can handle two amps. While it’s turned on, it drops 5 volts. There are higher-voltage diacs, as you would expect, but they’re all fairly low power.
SCRs, the classic thyristors, are unidirectional, like Shockley diodes, but have a gate electrode, like a regular transistor or a triac. A typical SCR is the Littelfuse SK055NRP, which costs US$4.81 and can handle 55 amps RMS when you turn it on (or surges of 650 amps), dropping 1.8 volts, and can block up to a kilovolt when off, leaking 30 μA. You turn it on by feeding 40 mA into its gate at 1.5 volts, and it turns off when the current drops to 60 mA. I’m not sure how much reverse bias it can handle.
The LM324, from 1975. Available (LM324D, LM324N) at G.M. and (LM324, LM324-SMD) at SyC, mentioned in Horowitz & Hill, etc., 35000 available at Digi-Key for 48¢ down to 12¢. A quad bipolar op-amp that runs on 3V to 32V good to gain-bandwidth of 1.2MHz, open-loop amplification of 100k×, not quite rail-to-rail, 10 mA output.
Horowitz & Hill recommends the TLC272 as a popular MOSFET-input alternative to the LM324 and the LF411 and LF412 as JFET-input alternatives. The LF411 (available from GM) touts being “pin-compatible with the standard LM741”, has 3 MHz of gain-bandwidth, sucks only 50pA from its input, costs US$1.19 down to 54¢ (but only has one op-amp on the chip), runs on up to 18 volts, 200k× open-loop amplification, ±12V output, like 20 mA output current. Digi-Key has 3000 available. The TLC272 (available from GM) spews out 30mA at up to 16 V and 2.2MHz of gain-bandwidth, sips an incredible 0.7pA of input current, and has two op-amps on a US$1.34 (down to 60.9¢) chip.
The LM741 itself (from 1968!) has stock of 68000, costs 66¢ down to 28¢, does 1.5MHz GBP at 80nA of input, 25mA of output, and up to 36V. I wasn’t quite sure what its major advantages over the LM324 are, but it seems to cost about five times as much. Fortunately, both chips are popular enough that people have discussed this very question previously; they claim the LM324 is noisy and has crosstalk, but runs off a single supply instead of needing positive and negative supply voltages.
Digi-Key’s most popular item in the op-amp category is the tiny 70¢–53¢ MCP6031T-E/OT, which is rail-to-rail and can spew out 23mA while taking in 1pA, but only up to 10kHz and 5.5V. It’s touted as having super low consumption. They suggest using it for battery current monitoring and “sensor conditioning”, whatever that is. They have almost 165000 in stock.
In non-super-low-frequency opamps, Digi-Key’s most popular item is the TSV321IDT, with 142000 in stock, but only available in quantity. The similar TSV324IPT, by contrast, is available in quantity 1 for 58¢, down to 24.9¢, with 162000 in stock. It’s a quad rail-to-rail 1.4MHz GBP op-amp chip with a 70nA bipolar-class input current, 80mA output current, and up to 6 V.
If we demand another order of magnitude in gain-bandwidth product, we exclude 85% of the op-amps on Digi-Key and the price goes up a little; the LM7321MF/NOPB from 2008 is US$1.83 down to 54¢, has 45000 in stock, and is a rail-to-rail 20MHz op-amp with 1100 nA input bias and 100mA output at up to 32V.
An additional order of magnitude excludes half of the remainder, and we end up with things like the OPA356AIDBVR, for US$1.91 down to 94¢, a 200MHz GBW (450MHz unity-gain!) voltage feedback amplifier (?) with rail-to-rail output, 3pA input bias current, 100mA output at 5V, and 80 dB (!) open-loop gain, 11500 in stock. This seems to be where modern op-amps are normally.
None of the popular Digi-Key opamps seem to be available locally.
The LM317 is a 100mA 71¢ linear regulator adjustable over 1.2 to 32 volts output and accepting 3.7 to 38 volts input. It’s 30¢ at SyC.
The 7805 is a 1A 62¢ linear regulator fixed at 5 volts output. There are other versions with other voltages, like the 7812, and the 79xx series is the negative-voltage counterpart. SyC has the 7805 for 35¢, and hasthe whole family, both negative and positive.
Viper-7 recommends instead the HT7333 and similar for low-current applications: "massively lower dropout voltage (around 1/10th) than 78xx regs, very low quiescent current (4uA), decent tempco (about 1mV per 2°C) and 0.2% line regulation."
Digi-Key’s most popular zener is the obsolete 50¢ MAZ30430ML, with 429000 in stock: 4.3V±7%, 200mW. Its cheapest available non-obsolete zener is the 5.1V 500mW 1N5231B-ND, with 16000 in stock, 11¢ down to 2.3¢; this is the 5.1V member of the 1N5221 series Horowitz & Hill use as their canonical low-voltage zeners.
Horowitz & Hill also suggest using the LM385 (1.23V, 63¢, 57000 in stock at Digi-Key, 10μA–20mA) instead of a low-voltage zener. Shirriff says the TL431 from 1978, an adjustable substitute for the LM385, is more common (42¢, 29000 in stock at Digi-Key, 2.5–20V, 600μA–100mA).
Viper-7 from Freenode ##electronics wrote a page about which 3500 electronic components he thinks a beginner should buy for US$110, with links to AliExpress vendors. It doesn’t go into a lot of detail in some cases, and I don’t want to just copy the list here, but I’ll point out a few notable things.
1500 of the 3500 parts are 1% resistors; historically 1% resistors were expensive, but they no longer are, so you might as well use them. However, he only recommends getting 30 different values to cover six orders of magnitude, which is five values per order of magnitude — presumably the standard 1.0, 2.2, 3.3, 4.7, and 6.8. This means you can hit any value to within ±10% with a single resistor, but if you need a more precise value you need to build a network.
In terms of capacitors (420 of them) he only recommends electrolytic and ceramic capacitors, and no large-value MLCCs — the ceramics only go up to 100 nF.
(He also recommends some inductors, of course, and 250 LEDs.)
He recommends relatively few discrete semiconductors, and unfortunately in some cases doesn’t say which, but does specifically mention the IRLZ44N and IRF4905 (10 each).
In power supplies, he recommends getting 200 (unspecified) fixed linear regulators, plus ten LM317s; the only voltage reference he recommends is a bag of 140 zeners.
The only connectors he recommends are breadboards, jumper wires, breadboardable through-hole screw terminals, and a MicroSD adapter.
There’s a section of recommended boards, including level shifters, Arduinos, serial adaptors, a NodeMCU, and an assortment of Arduino-targeted sensor boards.
The most interesting part is the “other ICs” section: ten 555s, five MCP23017 “16-bit I/O expanders”, two TLC5940 16-bit 120mA PWM drivers, five LM348A clones of the 741 quad op-amp, ten LM324 quad op-amps, ten TL084 JFET-input op-amps, and ten LM339 comparators. (This is not counting the microcontroller boards he also recommends.)
Viper-7’s a smart person with a fair bit of experience, so it’s really interesting to see what they chose to include — lots of op-amps and comparators, a few GPIO kinds of chips, and some 555s — and what he didn’t — switching regulators, voltage references, ADCs and DACs, class-D amplifiers, supervisory power chips, transistor arrays like the ULN2003, MOSFET gate driver chips, H-bridges, microcontrollers that aren’t already on a board, adjustable LDO regulators, memory, CMOS 555s, SSI or programmable logic.
More generally, he also doesn’t recommend a wide variety of discrete parts: no power resistors, resistors below 10 Ω, crystals, LED drivers (even WS2812s), film capacitors, tantalum capacitors, supercaps, relays, batteries, transformers (!), motors, or JFETs; no IGBTs, solid-state relays, or thyristors; no lasers, switches, speakers, microphones, or displays; no protoboards, solder, or solder braid; and not even any 2N7000s, though he says they're often included in BJT assortments, even though they aren't BJTs. (I want to be clear that I’m not disagreeing with their choices — on the contrary, I’m saying that if you disagree with them you should have a reason.)
All in all, a really interesting set of recommendations, one which I think will probably help a lot to make electronics as a hobby more accessible.