You Can Have My LM386s When You Pry Them From My Cold Dead Hands

Everyone has a chip-of-shame: it’s the part that you know is suboptimal but you keep using it anyway because it just works well enough. Maybe it’s not what you would put into a design that you’re building more than a couple of, but for a quick and dirty lashup, it’s just the ticket. For Hackaday’s [Adam Fabio], that chip is the TIP120 transistor. Truth be told, we have more than one chip of shame, but for audio amplification purposes, it’s the LM386.

The LM386 is an old design, and requires a few supporting passive components to get its best performance, but it’s fundamentally solid. It’s not noise-free and doesn’t run on 3.3 V, but if you can fit a 9 V battery into your project and you need to push a moderate amount of sound out of a speaker, we’ll show you how to get the job done with an LM386.

Stuck in the Past

There are a lot of better audio amplifier chips these days if you’re looking for lower voltages. Cellphones and lithium-ion batteries, along with the overall trend toward lower voltages in gizmos across the board, have pushed chip manufacturers to do more and more with less and less. There are some great amplifier chips out there running on 3.3 V and 5 V instead of 9 V or 12 V.

These Guys Used To Walk the Earth Too via Nat'l Geo
These Guys Used To Walk the Earth Too via Nat’l Geo

In particular, there are a number of chips that run in “bridge-tied-load” mode, which means that it drives both sides of the speaker, which makes it louder for a given voltage and removes the necessity for a big output capacitor in the design. This is a win on all fronts.

Because these amplifiers are marketed toward use in tiny devices, the vast majority are in surface-mount technology (SMT) packages. With the exception of making heat dissipation a bit difficult, we’re big fans of smaller parts and not having to drill holes in home-made PCBs. If you’re not down with SMT, you’re going to have to catch up soon. For instance, our other favorite DIP chip amp, the TDA7052, has been end-of-lifed.

So for an SMT PCB design, the LM386 is dead. There are hundreds of few-hundred-milliwatt amplifiers out there that can outperform it. We’ve designed with the TPA321D, for instance, and it runs circles around the LM386, but it’s SMT. Maybe you’d like to point out your favorite grain-of-rice, few-hundred-milliwatt, 8 ohm speaker amplifier in the comments? Anyone want to buy a stick of LM386s off of our hands?

Good, Basic Design

Just kidding! The LM386 has its place — on the breadboard, in the one-off perfboarded circuit, or even free-formed with parts hanging off of it in mid-air. And as the granddaddy of DIP-format amplifiers, it’s not going anywhere. In contrast to other, supposedly superior, amplifier chips, the LM386 is still manufactured after (who knows?!) how many years. And the reason is not just the form-factor. It’s also a very solid design.

By Rohitbd CC BY-SA 3.0

In fact, it’s a classic push-pull amplifier. The basic design uses two output transistors, one for the positive half of the voltage waveform and one for the negative half. The problem with the basic design is crossover distortion, which can be reduced by biasing the transistors just into their operating region, or by using an op-amp to provide feedback and push them through the dead zone. The LM386 does both.

image006If there were no such thing as an LM386, you could take a very nice op-amp for the voltage gain stage and wrap up the output transistors in the op-amp’s feedback loop to handle the current demand. The op-amp will swing the output transistors around like wild to make sure that the output voltage is a scaled-up version of the input voltage, whatever the load on the outside. The better op-amp you use, the better the overall circuit will sound.

lm386-annotatedThat’s exactly what’s going on inside an LM386. The schematic, copied from the datasheet, is a simple differential amplifier (the mess of symmetric transistors on the left-hand side) that takes feedback from the output voltage on the right-hand side between the pull-up and pull-down power transistors. The diodes are there to bias the transistors just into conduction to help minimize crossover distortion. This is called class A/B operation, and depending on the audiophile in question, it’s second only to pure class A for sound quality.

In short, aside from the simplistic differential amplifier, the internals of the LM386 are essentially what you’d build anyway. No wonder it has stood the test of time: it’s a solid, basic design. Unfortunately, that’s not the same as saying that it’s easy to use.

The LM386 in Application

This Will Sound Horrible

Have a look at the “typical applications” section of the datasheet. What’s missing? The worst omission is decoupling on the power rails, but you were going to include that anyway, right? If you’re running on batteries with low internal resistance and short wires, 0.05 microfarads is fine. If not, decouple with at least 100 microfarads plus a 0.05 – 0.1 microfarad capacitor for noise immunity.

What else is missing? In a few of the examples, they’ve included a “bypass” capacitor on pin 7, but only in a few. Even when they do add it, it’s drawn as if it were optional. It is optional if you don’t mind the amplifier hissing like a mad cat. Otherwise, this is a good place for some capacitance: anywhere from 0.1 to 10 microfarads seems plausible. Another secret trick: grounding pin 7 can be used to mute the amplifier circuit when not in use.

We’ve also noticed, and we’re not alone, that the inverting input seems to be less noisy than the non-inverting. See how the datasheet applications ground the inverting input (pin 2) and put the signal into pin 3? Do exactly the opposite and you’ll reduce your noise floor even further.

Getting Closer…

An additional circuit is listed as being “with Bass Boost”. This circuit adds highpass-filtered (negative) feedback between the output and pin 1 which damps down some of the high-frequency hiss and adds a lot more to the bass and midrange. Since a common complaint about the LM386 is that it is prone to high-frequency hiss when it’s idling, adding about 5 dB more mid-range signal to that noise is a clear win. It’s especially welcome on the small toy speakers that are usually paired with LM386 circuits.

Finally, there’s the question of the snubber capacitor and resistor on the output (pin 5). In practice, we’ve included this some times, and not other times. We built up a test board with a jumper that puts the snubber in and out of the circuit for this article. We can’t tell the difference. Supposedly, if the amplifier is prone to wild self-oscillations, this should damp it. The datasheet authors wouldn’t add it if it didn’t help with performance or reliability, we just can’t verify which of these two it is.

Not missing in any of the examples is the absolutely massive 250 microfarad output capacitor. You need it, and it needs to be big if you want to pass any bass through it. With an 8 ohm speaker and a 250 microfarad capacitor, you’re still attenuating some of the bass: 1/(2pi250 uF * 8 ohms) = 80 Hz is already reduced by 3 dB and the low E on a bass guitar is another octave down from there. That tiny little speaker is probably not helping either. Use the bass boost circuit for any low end at all.

To recap:

  • More decoupling of the power supply: this chip can push peaky power, you need to feed it.
  • Bypass pin 7 for noise immunity. Ground it to mute the amp.
  • Use the inverting input.
  • Use the Bass Boost. Think of it as hiss-reduction.
  • The snubber. Do what you think is best. Retrofit if you need it?
  • Don’t forget the output capacitor. Bigger is bassier.
That’s the Right Way

The best design we’ve seen on the web? The 9 V battery Ruby guitar amp gets everything right. Because guitar pickups have a very low high output impedance, they also add a JFET preamp. We’d also use the bass boost option, but guitarists like their high-and-janglies and don’t seem to mind hiss.

Our Cold Dead Hands

The LM386 is a well-designed, basic workhorse that does a decent job when its hooves are kept clean and it’s well-fed. Aside from having a slow op-amp stage by today’s standards, it has decent performance. It can also sound horrible if you neglect it.

dscf9057Because it’s one of the classics, it’ll always be available in through-hole DIP format, so it’s easy to wedge into a breadboard or one-off designs. You’ll never have to worry about it going out of production or costing much more than a quarter. And it runs decently loud off of a 9 V battery, which is pretty convenient to just toss into your project alongside the 3 V that’s powering the logic. Keeping the power and logic supplies separate is always a win.

It’s not a modern chip, though. The modern chips have X times more stuff going on inside. Some of this is to increase audio fidelity by speeding up the op-amp. Some of this extra circuitry helps the chips remain stable even with fewer supporting parts. The killer innovation, and the one that leads us to use a modern chip in anything that’s actually designed instead of just lashed together, is driving the speaker in bridge-tied-load mode. BTL means no output capacitors and is louder to boot — loud enough that a higher voltage for the power amplifier may not be necessary after all, though you’ll still want to decouple the supplies well.

We’re not saying that the LM386 is the best amp of all time: it can be a bit noisy and it’s demanding. But with a little care, it can work out fine. It’s absolutely not our favorite amplifier chip, but we’d miss it if it were gone, and it would make our desert-island IC list unlike other parts of its generation such as the LM741 op-amp or the TIP120 transistors — they are old, but the LM386 is a classic.

137 thoughts on “You Can Have My LM386s When You Pry Them From My Cold Dead Hands

      1. Also no diodes across the input terminals to limit the maximum differential input voltage, so you can abuse it as a comparator without having to mind input current. Don’t forget the low in&output dropout at the bottom voltage rail -> easy single single supply use and can directly drive an external low side switch, since it can go below Vbe or Vth for pretty much every device. Sure a modern RRIO opamp can do it, it also costs 10x as much (still peanuts) and won’t be a ‘Well I had 1/4 left over and I needed a slow comparator, so, meh’ type of deal.

        For 70 euros you can buy 1000 of them from TI and turn them into a sculpture.

        The LM324 has quite a few useful features that you wouldn’t initially give it credit for.

        Now why people still use the uA741 is a mystery to me.

        1. gotta keep common voltage a few volt below Vcc or the output will reverse and be sure to have a load resistor on the out to one of the supplies or it will have a ton of cross over distortion. unless you are building a trillion of some gadget to be sold in poundland I don’t see why you wouldn’t find something else

    1. And LM318 for bandwidth, though there are much better faster amps at low costs today. And a lot of ICL7660 voltage inverters, to get a negative rail from an 8 pin DIP and 2 small caps. The 7660 saved a ton of time and money, not to mention generating the negative rail locally to an analog section.

    2. And that it will run down to 3Volts! but the output capacity is not symmetrical. The LM328 is probably better as a low power amp. The LM324 is very good for low voltage signal conditioning or even a differential input for a sensor. But if you need high accuracy the LM324 doesn’t cut because of it high offset voltage and drift.

    1. Well you could set one up with a large time constant C/R to turn on a resistive heating element on a few minutes in the quarter hour, so your hands would NEVER get cold. Then they’ll never get them, Muha, muhuhahaaa, MuhuahahahahahahahahaaaaaaAAAAAA!!!!

      1. If by 7400 you mean the old school TTL then no. But I am always using HC, HCT, AC, AVC, etc. in dip and smd. Now they have families that are single gates in an SOT23-5 package that you can put anywhere. I try to make it a practice to include some generic footprints in a design for future changes. It constantly surprises me that the younger engineers write code to execute a simple logic function.

      2. I read that and looked down, at the base of my monitor a 74HCT14E was lurking!

        Now, this particular component was manufactured long ago in an ancient time. But mouser has it in stock. Great chip for people who know their switches aren’t bouncing, they’re just analog.

    1. When I was reading this post, and the snubber on the output was mentioned, it reminded me of a time, maybe 30 years ago, that I made an amp using an LM380. I left out the snubber, and when I turned it on, it oscillated badly and drew so much current that it blew the top off the DIP. I added the snubber and it worked well for ages. I did use a current limited bench supply the second time I powered it up though!

  1. I almost never bother with milliwatt level amp chips, just usually do a common collector transistor amp. Thought there was a cheap TDA something or other in a SIP or ZIP with a heatsink mounting tab that is stereo, good for a few watts. Used those a couple of times, mainly because it’s what RadioShack had.

  2. I used an LM386 in 1983 for an 8th-grade science fair project – an amplified crystal radio. It only got one station – the transmitting antenna was about a half-mile from the school :-)

    While we’re at it, let’s add the 9V battery itself to the components-of-shame list. They’re terrible for energy density. If they weren’t less noisy than a pair of AAs and a DC-DC converter, they’d have disappeared long ago.

      1. I know, right? But pushing current through a resistance needs a voltage. And cheap step-ups are noisy.

        I’d absolutely pick a SMT bridged amp at 5 V over a big DIP running single-sided off of a 9 V battery. And certainly for a “design”. But for a quick and dirty hack, the LM386 and a 9 V is pretty simple.

        That said, I just built my son an MP3 player with toddler-friendly (read: arcade) controls, and did I use an LM386? No. The hypocrite! I used a $1 unit off of eBay (PAM8403) that makes toddler-friendly (read: not loud) levels off of a Lithium-Ion because I have to be able to recharge anything that he can leave on overnight.

        But there’s no good way to add a filter into the feedback path on the PAM part, for instance. And it breaks at 5.5 V, so battery choice is a bit limiting. (Lith-ion at 1W output is right in the sweet spot, non-coincidentally.) It needs a heck of a lot of decoupling to run, even if no gigantic output caps like the LM386. It’s better, but not without tradeoffs.

        If the LM386 were gone, I’d miss it.

    1. Yep 9v battery is the awkward member of the family. Manufacturers only seem to use it when they want us to endure hardship.
      Great story about the science fair-takes me back. I had a wind tunnel lol. Good times.

      1. That’s funny. Here in Germany, the 4.5v lantern battery was the classic tinkerer’s battery for decades. It wasn’t until the late 90s or 2000s or so, when it became obsolete and the 9v battery took over this role.

  3. Been thinking of doing a guitar amp with one of these chips. From the basic Ruby design, does anything need to be changed to allow for 12v operation? A standard 9v battery drains quickly and I plan on using a 3S1P 2200mah LiPo to drive the amp.

    Also, instead of outputting to a speaker, any mods needed to do headphone out?

    1. I made a guitar practice amp using the Ruby circuit and it worked perfectly at 12 Volts without issue. I have also used it as a headphone amplifier without issue but I suspect the purists would change the circuit to account for a pair of headphones – make the circuit stereo and change the output stage slightly to cover headphones.

      Good luck!

    2. Headphones are just speakers with higher ohm ratings, that only means they’re quieter at the same amp setting. A “headphone amp” is just an amp that has a suitable power range and a convenient volume setting. Or at the component level, a headphone amp is going to have a pass-through with an optional cutoff when the headphone is plugged in; purely convenience extras. Any amp that drives speakers will drive headphones fine.

      So for example, I’m using 32 Ohm headphones. If I use a regular 2W 8 Ohm amp, I’m only going to get 1/2 W or so out of it. Which is fine, 1/2 W right against my ear is louder than a 2W ambient speaker. So it will still sound louder on headphones than the speaker, even though you don’t use the full power rating.

    3. The LM386N-1, -3, LM386M-1, LM386MM-1 can run up to 12V so you would need a reasonably good regulator like a 7809 or LDO with enough current capacity. If you supply is guaranteed to always be under 12 Volts then your good to go.

      The LM386N-4 will run up to 18 Volts.

  4. Been thinking of doing a guitar amp with one of these chips. From the basic Ruby design, does anything need to be changed to allow for 12v operation? A standard 9v battery drains quickly and I plan on using a 3S1P 2200mah LiPo to drive the amp.

    Also, instead of outputting to a speaker, any mods needed to do headphone out?

    1. Use an LM386N-4. It has a max supply voltage of 18V, instead of 12V for the other model.. This has the side effect of higher power capability but also higher quiescent current. If your battery (nominal 11.1V) is guaranteed to stay below 12V you should be alright with a regular chip, as the devices are able to handle 15V (but may cease to operate or operate erratically until the voltage is lowered.) Alternatively you could use a regulator like a 7808 or 7809 (less common than 7805 or 7812 but still easy to find) with a loss of efficiency.

      As for the headphones, you will definitely need a series resistor. Either figure out the power rating of your headphones and do the math accordingly, or just start with a resistor (~10k) and go down until you are happy with the volume range.

    2. More voltage gets you more headroom for range of tone, not more volume and 12V is wonderful for guitar. The LM386 1-3 can handle up to 15V just fine and the -4 takes you up to 22V max or comfortably a pair of 9Vs for 18V.

  5. Smart highschool pupils in Romania used to build magnetically-induced earphones using the LM386, a big coil around the neck as inducer, and a tiny magnet inside the ear as the “vibration part”. The LM386 amp hooked to a smart phone was able to induce the sound into the ear using that tiny magnet. Result? Cheat the exams through a phone conversation with somebody else who transmit the correct answers/results for tests. There was no visible wire going into the ear and the sound was pretty damn good. LM386 is cheap and the resulting system is sometimes sold for around 50….100 bucks. And there are a lot of persons who wish for such a device to pass exams without even coming to school. When exams session come, such persons make a lot of money.

      1. Do they need to? If they’ve somehow got the exam paper in advance, it’s just a matter of “Question one…. A …. Question 2 …. 500mA ….” slow enough to copy down.

      2. It does work. It’s more widespread in faculties that are more theory-oriented. My last roommate was senior-year mechanics and he just quickly whispered what chapters i should read back to him. I should say though that now everyone uses wireless micro-earphones like in spy movies. You can buy them from the usual chinese suspects.

      3. It is working. Mike stays on a modified hands-free, and instead of speaker you have the coil.
        Now – for wireless (bluetooth) earphones the supervisors are trained to see them, but nobody will search for a tiny magnet inside the ear sitting on the ear drum. There are phone jamming devices for faculties and highschools. However, a lot of medical accidents happened because the magnet sit too long inside the ear, it got covered by wax (you know, the candle in the Shrek movie) and only doctors could remove it.

    1. How would you know if it is fake? What would it even mean when it is a part number that has lots of suppliers?

      I’d bet if you bought one of them and hooked it up, it would really be that part number. The LM-series chips originated with National Semiconductor, but most of the LM386s I see are made by Texas Instruments. I guess those are “fake?” Not genuine? Or do you mean unauthorized? Are those contracts even public?

      Remember, false accusations are… not truthful.

      1. There’s lots of fake everything on ebay. A part number is actually a specification number and not simply a part number. If it can’t live up the parameters of the part numbers’ specification then it’s definitely fake – *read* fake by definition.

        So how do you know it’s fake? …

        1) The color of the text printed on the top is different to the colors used by the manufacture. You may need 5700 degrees Kelvin LED lighting to detect this if you have ‘normal’ vision.
        2) You can see that the original top surface has been ground off by there being a difference between the texture of the top surface and the texture if the cavities or beveled sides of the chip that existed from the time of manufacture.
        3) The spec says it’s good for 10 Amps but it explodes at 2 Amps.
        4) It feels greasy and smells like grinding paste.
        5) The pins are spread apart from being re-worked.
        6) The original was $2.50 in 1000 unit quantities and the ebay product is $1.00 with free delivery.
        7) The ATmega328 you bought only has the FLASH storage of an ATmega168.
        8) The SID works fine but there are no filters.
        9) When you opened the box that should have had 6 precision ball screws in it, you only found chopped up Chinese phone books.
        10) Your credit card has been debited 15 times so far when you only ordered one product.

    1. Yeah. They were commonly specified in the schematics … in the 1970s … I think I have one or two saved up somewhere. On the other hand, they can be replaced with vacuum tubes. -runs away before SHTF-

    2. But “MPF102 is code for “generic n-channel FET”. Everyone specified it because it was available, and it was available because it was used in projects. You knew it wasn’t fussy, so just about any FET would work. If something better was needed, it would have been used.


      1. In this particular design, the MPF102 sounds nice, and a handful of n-channel FETs from my parts bin sounded yucky…

        Also remember that the best guitar gear is built using vintage parts that contain “mojo” and unicorn tears, and any part that is commonly available just can’t create “the tone”….

  6. Am I the only one slightly bothered by the second chip from the bottom left being upside down? Either way, I’ve been using the 386 since I was a kid and it’s a great, easy to use little chip.

          1. Hahaha, I seem to notice the weird stuff while more urgent things go over my head. Blessing and a curse I suppose; was only making a joke. ;) Great article.

      1. That’s a totally different beast, at least from looking at the datasheet. It’s not specced to run loads < 2k ohms. This is more of an op-amp than a speaker driver, IMO. (Never used one, just reading a PDF.)

        They say it can drive headphones, which is code for high resistance at low voltage / power levels.

        It's a fine-looking audio op-amp.

  7. But look on the bright side.

    Early audio output ICs were fairly complicated, a lot more external parts required. Almost not worth going to an IC.

    I remember building with one of those early ICs, mind of expensive, and I somehow destroyed it. The work done, I wasn’t going to spend the money again.

    Then the 380 came along, much simpler and it saw a lot of use. Then came the 386 in retrospect soon after the 380 but I recall there were a few years, and it took up less space and maybe was even simpler. So there may be better IC output amps, but they’d come a long way by the time the 386 came out.

    And they are easy to find, if I was stuck I must have scrap boards in the basement that would offer some ip.


    1. I while ago I got one of those synthesizers in a chip for Christmas. I hooked everything up and blew it out immediately. Turns out the adjustable power supply my dad had made with me many years ago that I dug out to get the 9V had no capacitor on the rectifier. Meter said 9V, scope said 14V half sine.I looked into buying a new one but it was pretty pricey for my age.

  8. For me, it’s the LM3909. That chip could flash an LED for a year on a single D-cell battery, and only required a single external capacitor to set the time constant. Pretty much worthless for any other application.

  9. I’m surprised that no one has mentioned the LM317. That’s one of my go-to parts as it is a quick and easy way to make a power supply for nearly any voltage. It’s not great, but it’s good enough.
    The LM78xx and LM79xx series chips are also nice to have on hand for the same reasons.

    1. The LM317 is better for lower voltages as it’s minimum offset is 1.2 Volt.

      However you can do exactly the same with a fixed voltage regulator. For example you can put an 8.8 Volt Zener in the ground path of a 7805 to make a fixed 13.8 Volt supply just as you would with a LM317 but with different value components as the offset if 5 Volt insead of 1.2 Volt.

        1. It’s definitely the most valuable mid range current source because of the abuse factor – and it’s internal thermal and current limiting.

          You can do much the same with discrete transistors but they go ‘pop’ when things go wrong instead of progressively shutting down.

  10. The TPA3122D2 is a 15W+15W class D amplifier that comes in a DIP Package. It runs circles around the LM386 in terms of output audio quality, is efficient, has two outputs & comes in a DIP package. It also costs around 4 times as much as an LM386 but is totally worth it in my opinion. It is my new go to chip for audio amplifiers.

    1. Thanks! I’ll put some in my next part order.

      There’s something psychological about the $1 part barrier for me. I know it’s irrational, but I’m totally happier making crappy projects with cheaper parts, even though I totally know how to extract the expensive ones from a circuit when I need it elsewhere…

  11. I’ve used the L272 with 5V (or 9V) more than the LM386, 4 resistors and a decoupling cap gives better audio performance than the LM386 with bridge-tied outputs (more power, lower distortion, etc.) also it’s useful as a general opamp.

  12. You can bridge the LM386, you’ll need two of them though. In fact anything with a single ended output (the other speaker lead grounded) can be bridged with another of the same type.
    Just make sure one of the amplifiers are inverting whilst the other is normal.

    Also found these chips on computer motherboards where a 741 or LM358 could of been used (the boards were Core2 era). Probably because that company had an entire cupboard brimming with these?

    I have a liking to the CS8571E, I don’t know what they are a clone of though. However I can’t find a datasheet in English

    1. Forgot to mention:
      I’m using two LM358 chips for the sub-woofer preamp stage in my multimedia laptop.
      One chip in unity gain to keep stereo separation and mixed into the other chip to boost and back into the other half of it self to filter the signal.

      A proper bodge job that relies on the amps #enable pin tied to the headphone #Internal-amp-enable pin.

    2. Yes/No …

      Yes you can bridge two but you have to compensate for the power change to keep it in spec.

      By bridging two chips (one inverted) you double the voltage which means you have 4 times the power. Also the impedance presented to the output pin from the speaker is half of what it was before.

    1. The thing is, at the same price a 32bit ARM uC is not better than an 8bit AVR. If you benefit from larger data width then ARM is better, if your data is actually 8 bit then the AVR is indeed better.

      I use both and often I’m paying a little extra for the AVR at the same flash size. The core is an old design, but the fab is modern and so are the accessories.

  13. How about the LM3909 LED flasher? That was a part that got used for all sorts of strange and unusual applications. Now, sadly, EOL.

    As for the TDA7052A, it’s a truly amazing chip given it’s one of the very few that is rated for operation down to -40C. Yup. If you need audio line drivers in mountaintop/arctic conditions, that is/was the go-to chip of choice. Puts a whole new twist on “prying chips from cold dead hands”, I think.

    1. haha yeh, normally the dead body falls and launches/drops item way way way before the body has a chance to freeze, but in minus 60 degrees your wet body will freeze solid before the wind has a chance to pick up (and knock you over) so yeh, 386 out of dead-people’s hands is very likely. PS class-D amps wont work in the arctic unless thawed-out before use, this is because class-D speaker amps rely on large, low-ESR caps, which wont work in that cold. if it pops, metal cap shrapnel could pose a short-circuit andor fire risk, as ironic as that is, in the arctic.

    1. I started with the PIC16C57 and PIC16C84 which were One Time Programmable (OTP) and a pain in the ass to test – debug – throw in bin – program new chip.

      The introduction of FLASH was a godsend.

  14. 74HC595, of course!

    But the LM386 is a ‘the best’ OP amp. It’s not the cleanest, or the most efficient, or the loudest… but it’s the most musical; which is what interests a large part of it’s modern user base. The Ruby, the Smokey… 10 components (that includes sockets, clips and a box), of which only one is active… and you can drive a pair of 4×12″ speaker cabs and have it sound loud and good.

    Beat THAT, LM4801.

  15. Man, random idea – why doesn’t someone repackage one of the modern class D amps in DIP? I’ve used them a few times, and a lot of them use remarkably few components to put out a full 3W (in 4 ohm) at 5V (seriously, fewer than the LM358). I’ve cringed a little inside every time I end up using a LM358 / random class AB amp for a one-off because it’s too annoying to deal with the class D packages. Those class Ds are really that wonderful. Literally the only painful thing about them was soldering the BGA-9 package – that and they were very sensitive to sudden transients from disconnecting the speaker (blew a few doing that). I think buying a more expensive model with that type of protection (like, $1 instead of $0.50) would have solved that, or adding some transient protection offboard.

    Hmm, maybe I should just make a DIP breakout board for those guys…

    1. TPA3125D2N from Texas Instruments comes in DIP-20 :) Single ended stereo. Almost $3 in single quantities, but that isn’t really very much considering that you don’t use more than a couple in a project. IMO it isn’t worth building your own audio junk if you’re trying to skimp, because cheap finished junk is so cheap these days.

  16. 386? maybe, how about these?
    tea2025b higher power and lower distortion, thermal gain fallback (not bias cut)
    358/324 opamp
    393/339 comparator
    34063 SMPS
    4066 analog mux/mix/volume/select/attenuate/other
    4040 12bit count
    4518 4bit dual count BCD
    40106 6bit logic invert with schmitt-trigger
    40194 4bit universal shifter, does: SIPO,SISO,PIPO, and PISO
    TIPx (both the darlington and regular)
    any NON-darlington transistor with darlington-LIKE gain (2sc3623 hFe=1k-3k)
    ULN200x seven of darlington common emitter but lower power-dissapation (PDIP)
    4148 diode, dumbest of the actives but the most usefull
    super-low-efficiency LEDs for use as zeners/voltage references without “spilling” too much light (watch the temp-co.)

  17. I’m slightly amused every time I see a project where some hobbyist is struggling to get his/her project going with an 8 pin uC in dip (Usually by adding shift registers or other extarnal components) while he/she could just as easily have used an LQFP or similar.
    (Is there a chapter about SolderAngst in the DSM-5?)

    And the TDA1543 is of course the best audio dac ever invented.
    It’s so cheap you can afford to wire up a gazillion of them in parallel.
    And then build an amplifier by parralleling twice as many NE5534’s.
    And hook it up to an octo-core XU4.

    1. I use lots of 8 bit uCs and shift registers, the problem these people seem to be having is that they’re struggling. My advice, don’t struggle.

      What amuses me is these people who use giant chips with large pin counts and then they have a bunch of pins whose entire purpose is just to drive a single leg of a 7.1 segment display. And they are struggling because they’re never going to have enough pins. If they used shift registers, they wouldn’t even need to redesign for a new uC when they want to add another display segment. :)

  18. … push-pull … ? I think not – try “complementary-symmetry”. Otherwise, excellent!
    re:741: same sort of circuit (long-tail pair) as the 386’s input, but none of the problems with output-stage distortion, power consumption instability etc. I wouldn’t even attempt 386s for the stereo 10-band graphic equaliser I cobbled together ‘rat’s-nest as an excercise years ago using scrap 741s and slider pots (cost: nil, nada, zip!) that’s still functional.

    1. A lot of the chips suggested here low voltage and single rail supply. From memory the 741 was split rail with a min of about 9v. So yes the 741 is cleaner but it wont work in some of the situations the others will.

  19. The LM386 is still superior to Class D chips designed to run off of LiPO batteries. I compared a cheap PAM8302 amplifier powered by a USB powerbank against a Ruby with the bass boost mod into the same 12″ guitar speaker. At a rated 2.5W output, you’d expect the Class D chip to get twice as loud before distorting horribly. In practice, it got really hashy at about half the volume as the Ruby, and while the Ruby is still usable for grunge when it distorts, the Class D amp is useless when it starts clipping.

  20. Looks like JRC is end-of-lifing the 386 (and the 5534!), at least in DIP-8 package; Digikey just sent me the notice a couple of days ago. Their last-time buy is listed as end of 2019. Gotta stock up!

    1. Hey, this would be a good story. Can you send a link or the notice itself?

      I’m looking at NJRC’s website, and it still reads active, and they say they’ll give 2 years’ warning.

      If they really do stop, I can send you a couple. The photos in this article are real. :)

      1. I’ll send the e-mail they sent me in a few minutes. If you go to either’s site or, they both list the LM386 as “End-of-Life.” says in particular: “Part Status: End of Life; Last Time Buy Date: 12-20-2019. Minimums may apply.”

        And thanks for the kind offer, but I’ll be buying a tube or two of those babies before they’re gone forever. As well as the 5534s. According to Doug Self, “The 5532/5534 is made by several companies, but they are not all created equal. Those by Fairchild, JRC and ON-Semi have significantly lower THD at 20 kHz and above, and we’re talking about a factor of two or three here.” (Douglas Self, Small Signal Audio Design [2nd edition], p. 145.)

  21. Not sure if anyone has ever used the LM3900 series of CURRENT Amplifiers. They are interesting parts and are useful for optical purposes. Very high resistances and/or impedances are used. Takes a while to wrap your head around them as they are not your typical Op Amp. and are called Norton Amplifiers. They have been around forever, but are rarely used. – – – As a side note ; what is the next best replacement for an LM386 that is still Class AB ?

    1. I’m not sure there *is* a replacement for the LM386 / NJM386 that’s Class AB. Searches on Digikey and Mouser (as well as Googling various sites for insight) seem to suggest that more modern “all-on-the-chip” audio amps tend to use far less external components and don’t have the same pinouts, even if they’re available in a DIP-* package (which is becoming more and more uncommon). Just as an example, TI’s OPA134/2134 series are available in through-hole designs and have vanishingly small distortion numbers, but even if they were the same pinout as the 386, they wouldn’t use all of the external components. They’re also between 5 to 7 times more expensive than the 386 in single quantities, slight variations between different vendors noted. Last but not least, TI doesn’t even state the class of the amplifiers in this series. Some online designs using them call their resulting circuits “Class A,” but I remain uncertain as to why the designers believe this. Mebbe I’m just a thickie, but scanning through the datasheets and reference designs hasn’t yielded any insight, at least so far. The LM/NJM386 appears to be ‘sui generis.’

  22. Nothing wrong with the LM386. Some might say 250 mW is not a lot of output but it was more than adequate when I operated radio gear outside this year and powered the 386 with a PP3. It worked fine in that application. I’ve never tried using the inverting input.

  23. One of my first LM386 projects worked fine. Well, it worked. There was some distortion. I soon found that it was oscillating, at 14MHz! If I had wanted to make a 20m transmitter out of an LM386 I don’t think I could ever have gotten it to work, yet here I had one without even trying (lending credence to the addage: “To get an oscillator, make an amplifier…to get an amplifier, make an oscillator…”). Since then, I’ve found that the snubber is important, as are good layout techniques. Pay attention to those and the LM386 works just FINE!

  24. I’m currently building an analogue synthesizer on breadboards. I’ve got an LM386N-4 on there to drive half a pair of headphones (27 ohms) but it does some strange things. When I was running it off 5V, there was a LOT of 9kHz noise on the output (though the headphones didn’t seem to care). I’ve just switched to running off the unregulated NiMH PP3 and the noise is now at 386kHz! (Curiously, there’s less of that noise the bigger the input to the 386.) I tried a 4 ohm speaker but then the 386 oscillates very loudly (no matter the supply voltage).

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.