#1517

Episode 1517
are Darlingtons dead? are FETs better?
Be a Patron: / imsaiguy

[@JonDeth]

I had to revisit this one now that I got some work done with single package Darlingtons. I started with an MPSA13, and a couple days ago swapped it from an audio circuit for an MPSA14 with no other changes, The gain difference was massive! For audio as I already mentioned, they're still highly applicable. For a simple circuit such as for headphones or a line driver, they're a convenience compared to an audio grade op-amp build. I paid 6 cents per Darlington lol. I'm running ridiculously high resistor values for my divider, so thermal stability isn't an issue; 1 mil. and 200K(what I had available). I have 150K, but even though it switched on, it's bandwidth was absolutely in the toilet with the input source. Unbeatable for the cost.

[@ke9tv]

One nice thing about the MOSFET is that the gate will tolerate a lot more reverse voltage than the base-emitter junction of a BJT will. One nice thing about the Darlingtons is that they have much better behaviour in the linear region, for when a saturated switch isn't what you want.

[@copernicofelinis]

When both BJTs have a gain of 1 the configuration is called Simpleton, not Darlington.

[@eeengineer8851]

The last time I used parts like this on a new design was around 2000. Used them to drive the coils on 25A power relays and they were driven from 5V logic in a 12v automotive type power controller. In addition to the high saturation voltage another commenter noted, these are relatively slow switching if that matters in the application. When driving from logic, one also needs to be careful to have a base pull down resistor when the product may need to operate over a wide (including hot) temperature as leakage current from the driver can turn these partly on when you want them off. If I was designing a product to do the exact same function now as that relay board, protected high side drivers like those from Infineon in SMT packages would be used instead of relays. All the digital logic on the control side would be 3.3v.

[@uni-byte]

The Darlington is the better choice for linear designs (audio and such).

[@tvelektron]

I like bipolar in general for output stages because they are more robust for esd and overvoltage. At least they do not need a lot of protrction circuit compared to mosfet...

[@ChrisSmith-tc4df]

The 2N7000 is possibly a 40+ year old part, and there are much better options now. It's gate threshold could be as high as 3.0V, so even 5V logic may struggle to drive it to saturation if the ID current demand is too high. Its RDS is also high by modern standards.

[@Broken_Yugo]

I think a fairer comparison, if we want to talk modern parts and low costs, almost has to look at the SMD only offerings at this point, I'd start with whatever small mosfet every other cheap gizmo defaults to using these days. I'm not sure when new parts stopped coming out in a TO-92 package, but I'm fairly certain it was a quite a while ago, especially in terms of semiconductor development.

[@benjaminpajk1299]

Resistors are put on mosfet gates to reduce the ringing when hard switching the FET on/off. A mosft gate has some capacitance and there is some inductance in the gate wire that makes a nice LC oscillator. A resistor in series dampens the oscillations.

[@RexxSchneider]

Also, the input capacitance of a 2N7000 at around 50pF max isn't such a problem as we find in power mosfets (often around 1nF), but it still presents a direct short initially to the output of a microcontroller, and it's really not sensible to risk damaging the MCU output for the sake of something like a 220R resistor.

[@miroslavstevic2036]

@@RexxSchneider Is there any practical evidence that the gate capacitance will actually damage MCU or MOSFET when driven directly? What about inside of a MCU, it's all directly connected gates, nobody use resistors there. The output driver of an MCU is slow and non-ideal voltage source: it has rise time, resistance, inductance. 50pF is minuscule. I've used small MOSFETs directly connected to various MCUs - many thousands of devices, 24/7, for years - and never saw any problems or damaged parts.

[@RexxSchneider]

@@miroslavstevic2036 And yet there are many mosfet driver ics being manufactured and sold. Why do you think that is if you can simply connect the output of an mcu into a capacitance without any problem? The internal mosfets in an ic are tiny devices with very low current handling and miniscule capacitances. They are very different from a discrete power mosfet. Take a common logic-level mosfet like the IRLZ44 and you'll see a typical input capacitance of 3.3nF. That's over two orders of magnitude bigger than the typical input capacitance of a small signal mosfet like the 2N7000 (20pF). You might be happy to drive your mcu into directly into several nanofarads without any current limiting, and you might get away with it, but anyone designing a commercial application will know that's a failure waiting to happen.

[@lumpyfishgravy]

The big problem with Darlingtons is their large saturation voltage of around 1V. At today's supply voltages of 3.3V or 5V, that's a big chunk. I've seen people use ULN2003 Darlington arrays in a 4 alkaline cell system for relay coil drive - and it conks out much earlier than a FET-based design would.

[@thomasives7560]

BJT darlingtons have quite a bit more linearity than MOSFETs, so they might be better in amplifier circuits - that said, a JFET is probably a better choice for a very high sensitivity amplifier, since they are lower noise. Another application well suited to a darlington BJT is a sensitive current-amplifier (nobody uses those any more) where you have some sort of current source that varies slightly, such as in a circuit where the active element has variable resistance/conductance, where you want to amplify slight variations in current (resistance) to a larger linear signal. Again, a JFET transconductance amplifier would probably be better, but might use more parts. Great video, thanks for all the information about the fiddly bits that make technology work. Cheers!

[@gkdresden]

Take care if you compare data sheet frequencies. The on / off switching time means that the device really carries a considerable part (90%) of its specified drain source current. A transit frequency means its gain bandwidth product. Especially for Darlingtons it means that at the given frequency the beta is only 1 and no more 10000 or more. Darlington transistors are relatively slowly. If you want use a beta of 1000 from the original 10000 the maximum frequency is only 125 kHz. If you want to use the full beta of 10000 you can only operate at 12.5 kHz. You can't even use the full beta within the audio frequency range.

[@JonDeth]

For musician's audio gear this actually makes it ideal, especially for musical instruments. I greatly favor traditional semiconductors over vacuum tubes, and one of the most endearing qualities of vacuum tubes is their cut-off in most popular preamp stages is around 10 Khz without intentionally adding staging to compensate. I'm a shred and instrumental guitarist, and the abundance of very high, even order harmonics in high gain guitar rigs creates what seems like white noise and strangles clarity when you plow through very complex and fast note patterns. I was aware of the limits in bandwidth as the tradeoff when you drive these suckers into the top of their operating specs, but your post told me I made the exact right choice in my recent change of directions. 12.5Khz as the organic limit is perfect for my current applications.

[@stamasd8500]

Darlingtons have drawbacks too which may or may not be important depending on application. Saturation voltage is increased compared to a single transistor, and they are also much slower than one. Because of that they are best suited for low-frequency switching applications. Forget anything above 100kHz or so. :)

[@danielmantione]

The price comparison is nonsense, a reel of 2N7002 can be very low cost when purchased in quantity, i.e. on LCSC they trade below $0.01. Few MOSFETs are sold in TO92 (Darlingtons score a point here), but even the 2N7000 can be bought for much less than the price discussed here. At realistic prices, MOSFETs are cheap enough to not worry about their cost. Also, 2N7000/2N7002 is kind of obsolete. Something like a BSS138 has significantly better specifications when working with logic level voltages.

[@SeanPorio]

I know in general for audio signal path purposes a bipolar/darlington is generally preferred to avoid the larger parasitic capacitance of a MOSFET. I think JFETs tend to have better capacitance as well but that’s generally a more specialized part as well and not perfect for every application.

[@Enigma758]

Assuming base resistors for both transistor types, you'll also need a pull down resistor for the FET to discharge the gate capacitance, so that's one extra part.

[@whatelseison8970]

I wish they made a TO-92 version of the AO3400/3401 SMD Mosfet's. The 00 is n-channel and the 01 is p and they both are rated for 30V, 5.8A 😲 and turn on fully from as little as 1V, so you can be sure they will be hard on even at 3.3V. They're also dirt cheap. You could probably fit 20 of em inside a TO-92 package but that's kinda the problem -- some assembly is gonna be required to get them on a breadboard.

[@deepblueskyshine]

Back in the days of empire of evil as Reagan put it almost all of eastern european countries have their semiconductor production. Some of them developed their own, some of them licensed western designs and soviets mostly stole from the west. As I remember czechs were doing quite a lot of their own, pols were licensing newest western designs, and in my native Bulgaria they were licensing older designs and "developed" from them. In the 80s transistors were not a tecnilogy that needed buying know how, but they reflected the development history in their part designation codes, i.e. poles used BC500 small signal transistor designation, czechs used their own, and bulgarians used a mix of soviet commonwealth designation standard mixed with western origins, i.e. I have some 40 years old 2T3109 which are further development from BC109 and 2T3109C have their beta ranging from 800 to 1200 not being darligton, but a regular design. Used microamp or less biased these trasitors have not only high gain, but were also low noise with higher bandwith product (than needed for sound frequency range they were mostly intended for and worked nice on AM and IF), so I gues here is a good answer to the question about transistor gain in itself. Darlingtons have their role back in the days, and can still do the same today, though having good alternatives.

[@itoffice]

Nice. Can you use that for wireless sensing the "hum" from 50/60Hz cables? - as a cable sensor you use on walls, to see where there are cables with current attached to the grid. - you know, not to put a nail in a hot cable when you want to hang up a painting on the wall.

[@JonDeth]

You could. There's many ways to do what you're asking with a variety of semiconductors without wandering into a complex circuit design. An inductor of very fine wire, an amplification circuit, a bandpass filter running roughly 30 to 90Hz for the Q, and you can affix an LED, a buzzer etc. for when your electromagnetic coil has a shift in induction from the cross talk between electrical signals. The most basic single transistor small signal amplifier with very little additions can be designed and built to do what you desire. In a single afternoon in fact.

[@rtos]

Low power darlingtons are pretty much obsolete; however power darlingtons are still very useful. In a pinch you can always use two low power transistors to function as a darlington. In product design you'd probably stay away from using special or hard to get parts, which the low power darlingtons are now likely to fall under.

[@byronwatkins2565]

A 10 MOhm base resistor would perform just as well. Also, turn-on time is a digital (switching) spec. Analog applications would always have the gate biased above the threshold voltage making the "turn-on time" much less.

[@stevepence9869]

Always enjoy your they're great

[@DiaconescuAlexandru2024]

If I gotta use BJT's, I prefer using the sizlak pair over the darlington pair, as it's saturation voltage is 0,7V-0,9V compared to darlingtons which are 1,4V. The sizlak pair still multiplies the hfe of the transistors with eachother, so it has the same gain as a darlington.

[@alexjenner1108]

I'm not sure if you can get an off the shelf *Sziklai* pair device, having both npn and pnp in one package might be more expensive, but good for some discrete component designs.

[@DiaconescuAlexandru2024]

@@alexjenner1108 true, I've only used discrete Sziklai pairs, haven't found one that's in the same package. Also ironic how I can never spell the name right even though I'm 1/4th hungarian and my entire dad's side of my family is hungarian. 😆

[@planker]

Excellent. I'm on Ebay to get me a few of these.

[@alklapaxida850]

gota chuck in some TIP122 & TIP127 in the mix.. i still got got em hanging around

[@zyeborm]

Through JLC they have a list of "basic parts" that are always loaded in their pick and place machines. No extra charge to use them. 30v 5A sot23 pack n and p channel fets were 1.3 and 1.5 cents each in qty 1 😂. They had transistors (not Darlingtons) for slightly less, but i kept to the fets to keep the bom count down lol. It was worth the extra literal half a cent just to simplify things a little for me lol

[@ricksampson6780]

I believe so.

[@RexxSchneider]

There's not really the difference in current handling capability between the MPSA14 and the 2N7000 that you think. You were quite right to observe that power dissipation is also a limiting factor, but you have to actually explore that to make a meaningful comparison. The MPSA14 has a maximum power dissipation without heatsink of 625mW at 25°C, and a maximum saturation voltage of 1.5V at 100mA, so on the generous assumption that Vce(sat) doesn't increase much at higher currents, that indicates a maximum continuous current of about 400mA. You'd have to operate it pulsed at no more than 80% duty cycle to achieve the 500mA it is specified for. The 2N7000 can deliver 200mA continuous, but 500mA pulsed, and with an Rds(on) of less than 10ohms, the power dissipated is at most 400mW continuous. Interestingly, the On-semi datasheet (dated Jan 2022) for the 2N7000 claims " They can be used in most applications requiring up to 400 mAdc and can deliver pulsed currents up to 2 A." Which is directly contradicted by the information it provides on the following pages :o

[@bigjd2k]

You can also get Supergain single transistors, which compete against small MOSFETS. NXP made loads of them!

[@aduedc]

Darlington have problem with high frequency in CE mode. However, you use a cascode transistor to resolve some of the problem However in CC or emitter follower mode is superior over MOSFET. Because the output impedance is 1/gm in this configuration; therefore, it is traditionally used in the push pull class AB output stage or output of LDO or other voltage follower applications. Note that gm of MOSFET is low. Therefore, in order to be used for these applications, they are used in "Super Voltage Follower" topology -- which has a limited voltage range -- or "Flipped Voltage Flower" topology -- which could have stability problem.

[@aduedc]

**** I recently discovered a forgotten Transistor UJT, which is very similar to JFET but needs positive Gate (Emitter) biasing and also could generate negative resistance ****

[@RexxSchneider]

@@aduedc When I first started dabbling with electronics (60-odd years ago), the 2N2646 unijunction transistor was a staple means of making a pulse generator with just a couple of resistors and a capacitor to set the timing. It was later superseded by the programmable unijunction transistor (PUT - which was actually a 4-layer device with three junctions), and then the 555 arrived and nobody used UJTs any more. I still have a few in my parts box somewhere.

[@mr1enrollment]

Devil in the details. Gain, Voltage, Speed, Power. ($$)

[@OneBiOzZ]

I still use darlingtons when triggering relay coils and simple brushed motor drivers, i find them to just be overall more robust, however that is about it. The BoM cost is typically offset by the fact that we buy many many many many more times the amount of general purpose mosfets than we do transistors.

[@joeteejoetee]

The 2N700X N-Channel Enhancement mode FETs have an internal diode intrinsic to the substrate that prohibits using them in some applications. Example would be a series pass battery charger (or switch) and when you turn it "off" the battery could back-flow thru and discharge into your charging circuit - 3-Terminal regulator for example, so don't forget to add a diode across the 3-Term-Reg!

[@absurdengineering]

Transistors have typical BE reverse breakdown of 5-7V so they have an intrinsic reverse diode too. It’s just a Zener :)

[@gmanshackshack6822]

I still have a BUNCH of ULN2003's in my parts bin, just waiting to find a use.

[@Daysofpavara]

How much voltage drop is Dralinton transistor ?

[@jeffschroeder4805]

Is the capacitance of the FET what slows down the response time?

[@rogergreenwood1536]

Exactly, 55pF is huge in RF land, but of course many other FETs exist.

[@PracticalCat]

I use darlingtons for water sensors from time to time.

[@davidahmad6090]

Surely mosfet is much more vulnerable than bjt technology and less robust, in terms of ESD

[@Peter_S_]

I layed out a board with a ULN2003 a couple days ago. When you have 5V around, it's been a workhorse for driving relays small steppers for almost 50 years. The built in reverse diodes save on BOM cost and I don't know of a decent MOSFET based replacement.

[@RexxSchneider]

The TPL7407LA is a pin-for-pin replacement with benefits in reduced power lost and in reduced leakage.

[@Peter_S_]

@@RexxSchneider 🏆 Thank you, kind sir. Too bad they don't seem to have a TPL7408 as well.

[@UpLateGeek]

I'm currently working on a design for a "vintage look" VFD clock, which requires an anode voltage of around 24V DC, so you need some kind of level shifting between the logic and the display. Using Darlington array ICs it's just 6 chips at a cost of around $0.43 each from the mouse-based electronics supplier, compared to a field of 42 discrete MOSFETs, which would not only cost more, but also take up a lot more board space. So there's definitely still places where a Darlington makes more sense. Anyway, I would still at least put a pulldown resistor on the gate of the MOSFET, just to make sure it turns off cleanly. And you can also see gate ringing at higher frequencies, so you might also need a gate drive resistor, or possibly even a dedicated gate driver IC. A Darlington will also dissipate more heat at the same load because of that voltage drop, because the MOSFET Rds(on) can be very low (in the mOhms for some parts), the package temperature isn't usually the limiting factor in switching applications.

[@RexxSchneider]

For $0.67 from the same supplier, you can get TPL7407LA, which is seven mosfets and associated cmos circuitry in a single package, designed as an improved drop-in replacement for the ULN2003A, so an extra $1.20 on your BOM. The Darlington advantage is rather less than you think nowadays.

[@UpLateGeek]

@@RexxSchneider That's very interesting, I didn't come across this in my searches, but I'll definitely check it out.

[@t1d100]

Interesting

[@jack002tuber]

I always wondered if you can't make a darlington from two NPNs

[@IMSAIGuy]

yes

[@gyrgrls]

Gain of 100 or 101? I always thought along the lines of beta plus one.

[@frankowalker4662]

Darlingtons are still very useful.

[@andymouse]

I think so to, especially in audio.

[@frankowalker4662]

@@andymouse It's a 2 stage pre-amp in one package.

[@bobkozlarekwa2sqq59]

Could you create a super darlington using three or more transistors, or cascade two darlingtons?

[@RexxSchneider]

You could, but the emitter leakage current of the first stage would be multiplied by thousands by the next two stages, making it really hard to turn off, especially under hot conditions. Also the base would need 2V or more to turn on properly, and the maximum saturation voltage would likely be as much as 2V as well, causing power dissipation problems.

[@absurdengineering]

It’d need to be a Sziklai pair or triplet to be of use in modern low voltage circuits.

[@clivebradley2633]

You may have made the case for switching circuits, but perhaps you should have considered their relative usages in analogue circuits as well ? And you haven't mentioned the disadvantages of gate and miller capacitance at all.

[@xenoxaos1]

Definitely a place for a uln200x.

[@RonDogInTheHouse]

Darlington can be used as an analog amplifier, MOSFET, not so much, it makes a better switch.

[@dixsusu]

Hell no !

[@markhall3323]

Coming next…are resistors obsolete…etc etc (no)

[@adrianramos2229]

Arduino-land xdddd

[@andymouse]

Nah I love 'em !....cheers.

[@simontay4851]

Are they obsolete? Absolutely not. Especially TO-220 power darlington's. They are used a lot in audio amplifiers.

[@1shARyn3]

Hm! So here, you go and create a new protoboard, yet continue using the old standbys. Hmmm

[@IMSAIGuy]

I don't shoot videos in order

[@pidaras_pidarasina]

Cutoff frequency is low which makes them very niche, like audio amplifiers applications. FET has more, but I couldn't see the cutoff frequency of this specific FET transistor.