The 7805 Is Dead! Long Live The 7805!

The 78XX series of regulators are very handy to use. If you need, say, a 5V regulator, you grab a 7805, add a capacitor for stability, and send in enough voltage for the regulator to work with. Cheap and easy. However, the part is not without its faults.

A stock 7805 can’t convert 5.1V to 5V. You need to have a good bit more voltage coming in. But the more voltage you put in, the more the part is going to dump out as heat. So running from 9V is going to be cooler than running from 24V. All that heat isn’t very energy efficient on batteries, either. [Stefan] wanted to do better, so he made a drop-in replacement for these venerable regulators some time ago. But he’s recently made the board layouts available so you can build your own replacement, too.

The device accepts 4.5 to 16V, and you can select the output voltage using two resistors. You can draw up to 2A out of the regulator, which is more than you can say for a stock 7805.

The heart of the tiny board is an SD8942 buck converter. There are two voltage-setting resistors, one more resistor, a handful of capacitors, and — of course — an inductor. The schematic notes you can also replace the IC with an MT2492. With a right-angled header, you can swap the board in for a standard 78XX. Note the values on the schematic are set for 5V, but you can change them pretty easily.

This is a great example of how surface mount has changed so many things. We remember old 7805 “drop in” switching supplies that were much fatter and they weren’t inexpensive. Boards are cheap these days, and if you want the boards assembled, it looks like they’d cost $20 or $25 each in small quantities. But most of that is in setup and loading, so in larger quantities, the unit price would drop quickly. Or just solder them yourself.

If you want to see how the real McCoy works, we’ve watched [Ken] look inside the 7805 before. If you want to dive into the mechanics of buck converters, we did two simulations about that, too.

70 thoughts on “The 7805 Is Dead! Long Live The 7805!

    1. I think the same.
      Though I would prefer a real 7805 any time, if possible.
      The 78xx series is so nice because of its simplicity, clean output (no RF noise or ripple; remember the caps and protective diode though) and robustness, not its efficiency.

      With a generously sized heatsink, it will run for decades without failure.
      Sure, the higher the input voltage, the higher the higher dissipation. But that’s how it works, it’s a linear regulator (in Germany, we also call it “Stabi”, from Stabilisator = stabilizer). It must compensate somehow, which causes thermal energy.

      But on the other hand, the 78xx will keep the output voltage stable no matter what.
      Remember, it can handle voltages up to 25-35v (7805, 7812 resp). So you technically could power you device with two car batteries at 13,8v each.

      Btw, the recommended input voltage should be at least ~2,5v above the rated voltage.
      So an 7812 should be operated with 14,5v.
      Unfortunately, that’s not the case in practice.
      Many developers traditionally used 13,8v (stabilized) as the power source for their products running with an 7812 inside, a voltage used by car batteries and lab power supplies and CB radio power supplies supplies.
      That’s violating the datasheets, strictly speaking.
      Because it’s merely 1,8v above, not 2,5v.

      Anyway, I’m merely mentioning this because this article is about an 78xx replacement. Knowing that the reality didn’t stick to datasheets and specs might be helpful in troubleshooting.

      Best regards, vy73,

      1. exactly
        buck regulators and the likes generate tons of noise and cause every sort of problems that are difficult or impossible to trace.
        i have an engine monitor in my airplane that i designed myself. iintially i used a 7805 but it was running a bit hot and not much space for a heat sink so i went for a buck regulator.
        noise in the intercom, funny quirks on the analog inputs of the MC, and the autopilot (completely separate unit, factory built, very well shielded) occasionally would do really funny stuff.

          1. I had a hf transceiver and two 12v psu, one a linear regulated type, the other a switchmode. The switchmode gave birdies every 20khz on certain bands. it was so annoying that i delegated the switchmode to power the power amplifier, and the linear for the transceiver.

      2. You could make a 1.5A combined switching+linear regulator with 2V of headroom, easy. Getting it to fit in that volume would be tough, although there was an integrated one in a small package I used previously (which is sadly no longer manufactured).

          1. I used to work for a company that designed and built switchers, our designs used 78xx and 79xx all over the place as post regulators with the switch -mode convertors. Almost all of our supplies had multiple outputs, v1 was the regulated switcher, all the rest had the three terminal post regulators. These did a pretty good job, as most of these went into medical or telecom equipment.
            I used to do the design and verification testing….so I know how good those numbers were.

          2. Why in the heck would you use a 78xx as a cleanup regulator?? I meant a switcher with wide input range plus an LDO afterwards with high ripple rejection in the switcher’s frequency range.

          3. Given that the frequency of the switcher is known you only need to improve the performance of the output filter to hit whatever noise limits you need. Of course if you didn’t know what you needed before you started you are always going to be in trouble 😜

      3. There are a lot more tradeoffs to consider when dealing with mass produced electronics and not hobby stuff or classroom theory.

        Traditionally, a linear supply had a bad efficiency not because of the regulator, but due to everything upstream. You had to size the AC transformer (which is lossy btw, even when doing nothing) so the circuit would get appropriate voltage considering variations in your mains AC (usually +-5% depending on the country), then consider the transformers drop under load, plus the constant drop at the rectifier bridge, and at last, the ripple after the filter capacitors had to have its minimum voltage under load higher than the ~2V the linear regulator needed. You ended feeding the poor 7805 something like 10V or more most of the time.

        Then you have LDOs, which don’t need such a huge input, but if you choose to feed a 5v with a 0.1v minimum input, with 5.5v, that leaves you with a tiny margin of error before undervolting. You need a previous supply to stabilize that (which can be a SMPS). And that adds cost and complexity.

        Then you say, well, better go straight to a SMPS and do it right. And once again comes the tradeoffs, better filtering means more cost, and if it doesn’t work, you need to up the switching frequency, which means more losses and a LDO with a cheaper SMPS comes to mind again. Or you can use a more advanced high frequency IC, but cost gets in and EMI becomes something to care about.

        Obviously, one thing is how you feed X volts to a device within a specified noise/ripple level, another is retrofitting 1970’s hardware designed with a linear regulator in mind. And I haven’t even talked about stuff like transient response, voltage sag and others.

        I just will say modern computers run fully on switching supplies, with the larger kW-sized ATX ones putting amps in the 12v rail that could beat a welder, and that is fed to the motherboard’s (or GPU) voltage regulator (switching too), which take the voltage down to the 1s (1.3v is common now a days), feeding several hundred amps of current to the silicon, and constantly changing both current and voltage every millisecond with the dynamic clocks of today.

      4. I’m looking at a 1989 National Semiconductor General Purposes Linear Devices Databook right now. For the LM7800 series, no maximum dropout voltage is specified, typical is 2.0 volts at 1 amp, 25 degrees Tj. One graph shows that at 1.75 V at 500 mA, 1.5 V at 20 mA. A different graph on the same datasheet shows 1.75 V at 1 amp, 1.2 V at 100 mA.

        As you imply, running outside guaranteed or even typical specs is not good practice. However, if the designer needs much less than 1 amp from an LM78XX, relying on a lower dropout voltage usually works.

        Sometimes, a regulator is just used to protect load devices from excessive voltage: you don’t care if the regulator’s output drops to 10 V, as long as it never exceeds 12.6 V. An LM7812 might be useful then.

  1. I wonder why the mechanical layout was kept (that hole…) despite being not compatible at the mechanical level… (that heatsink…).
    he could have saved a few millimeters of height.
    definitely a nice project

    1. Mechanical stability? Even if no heatsink is required, a mounting hole isn’t unnecessary per se. It’s more a question of being professional. Just because consumer tech is poorly built with lose parts doesn’t mean we must follow this approach.

      1. This. I’ve seen plenty of circuits with TO220 and similar packages where the tab is used to ground the package to the board (sometimes fastened with a screw, others soldered down). In some cases, this is mainly for thermal dissipation on the board ground plane, but I’ve seen it even in very low current applications where thermal dissipation is a non-issue.

        In certain applications (for example, I’ve done some analog audio work), maximizing conductivity in power busses, especially ground, is critical to maintaining signal integrity. If your power capacitors and signal lines are grounded extremely well, THD from line impedance can be minimized. I’ve never had to do this with 78xx packages myself, but my applications were battery powered, and I designed my circuits around the raw battery voltage. I had to learn a ton about signal integrity though, and good grounding is critical in hi-fi audio applications. (Analog audio was actually my first foray into electronics. I now have an electric guitar I don’t know how to play, because I needed a reasonably stable signal source for testing, and I didn’t have a signal generator…)

    2. The hole can be used to connect a heatsink to ground. While the 7805 replacement might not need it, it might be shared heatsink, with other ICs or transistors bolted (or glued) to it. A TO220 case will have form a ~15pF capacitance with the heatsink, assuming a standard mica insulator (YMMV). It might be that the circuit relies on that extra 15pF to ground, or that the heatsink, if left ungrounded, might radiate EMI due to the capacitive coupling to some device. Okay, sinking EMI through the ground connection of your regulator sounds like a bad idea, but people did a lot of crazy things back then.

      Above explanation is just an assumption, a plausible way to explain the hole. I guess you’d need a silicone insulation (softer and thicker than mica) between the 7805 replacement to prevent the components from getting crushed and the pins from getting shorted. Or maybe even coat the while thing in epoxy. However, the 4.7 ohm resistor appears the get into the way of a potential screw head…and I don’t trust these clip-on fasteners.

      Note that the 7805 replacement’s hole is smaller; this is actually plausible, as it wouldn’t make sense to add an isolator to the hole if it’s there just for the ground connection. But…after all, I guess it’s just there for cosmetic reasons.

  2. 7805 is linear, this is an SMPS. That instantly raises the question: what about the added switching transients?

    It would be beneficial to see the input and output power spectrum at 10%, 50% and 90% loads.

        1. We manufacture stir plates and home brewery controllers and we build our own PWM controllers in-house. I’ve gone the route of active components direct from China and was bitten hard by counterfeit parts and parts that just didn’t perform to spec. Because of my history with Chinese suppliers the last thing I’d trust direct from China is a voltage regulator. So these days everything comes from DigiKey. Sure, most still come from China but at least if the part is in their catalog it’s vetted by an importer or manufacturer and others are buying the exact same parts.

          In one month I lost about $7,000 USD because of substandard SSRs. I even wrote an article about that one!
          We still buy SSRs direct from China but only from one manufacturer I have history with.

          And we still buy two different 12 VDC direct from China but these also come from one manufacturer I have history with. But any plug-in 12V power supply has poor regulation and I understand this. We also sell one of those 12V 1A power supplies on Amazon because they work well.

      1. So.. An SMPS is fine, *if* the manufacturer acts responsible and spends an reasonable amount of external parts for filtering?

        If so, do you guys/gals see the problem here? ;)

        (PS: To the newbies: Filtering is the most cost intensive part of a circuit design. And it should always be implemented, even if a calm 7805 is used. Filtering is also the number uno that’s omitted, because the circuit does not instantly fail if it gets removed. Chokes, DC blocking caps, coils, AF transformers, opto-coupler etc. They’re all helpful and would keep most headaches away from the user. Alas, they’re more expensive than 5 cent, so they have a high ranking on the cost-optimizing list.)

  3. Every “switcher” I have tried made freqs. below ~7Mhz unusable.
    I have tried 2 brand-name/well made supply’s and sold both within
    a week. Went back to Astron(RS-70 and RS-50), no noise!
    If you don’t use HF below ~7Mhz you may be OK??
    No code pencil neck geeks love the switchers.

    1. > No code pencil neck geeks

      Sir, this is HaD, toxic ham comments are not appropriate here. We cater to a broader spectrum than your particular niche, for which such projects may be entirely appropriate. I suggest instead complaining about how it should have been done with a 555.

    2. Yes, Asrton linear power supplies (mostly intended for amateur radio use) are much RF quieter on the HF bands if you can’t get far away from your antenna or your receiver doesn’t have good common noise filtering on its DC input. The Aston design is 40 year old technology based on an LM723 and pass transistors, but it works well.

      I have two complaints about Astron. The newer power trsnsformers have much higher magnetizing current than the old units. I have a 35 Amp from 1983 which idles under 10 Watts at no load. The same supply about five years old idles at 30 Watts and gets noticeably warm from the transformer and there is a little mechanical hum.

      The old units used stud rectifiers for the high current DC. The newer design is using those square brick bridge rectifiers as center tapped full waverectifiers with two sets of internal diodes in parallel. That is a bad design because they aren’t using ballast resistors to balance the current in the paralleled diodes. Because those brick rectifiers are meant to be operated as bridges instead of full wave, there is no particular matching of the internal diodes because the diodes normally operate in series, not parallel. You can’t consider two paralleled 25 Amp diodes as 50 Amp without resistors to balance the current
      One diode with the lower junction voltage will “win” and take all the current, burn out, leaving the other diode which will then get all the current and also burn out.

  4. What makes me wonder, how is latency?
    A plain 7805 can be used to down convert a 12v signal of the RS232 to TTL friendly +5v signal.
    Does this “hack” still work with such replacements, too? How much baud do they reach?

    1. It would not work. I don´t think they like much being driven by a square wave or similar.
      But look, since this “hack” we got tons of cheap, low Rds and very fast MOSFETs, so why would you use an SMPS to achieve what you want ?

    1. The Recom parts work well.

      I’ve had quite a few designs over the past few years where I used an R-78E5 or R-78E3, including some updates to legacy products that originally had a 78-series linear, and they were great, at least for powering logic.

      In fairness, I probably wouldn’t use any switcher to power a delicate linear circuit, but, you know, right tool for the job and all that.

    2. Indeed, I use RECOM “LDO Killers”, such as R-785.0-0.5 for far long time.

      The only caveat is related to EMI radiosity and susceptance. I found in some very dense applications (componentes too close to each other) with nearby RF antenas or other switched circuits, to affect the indutor core, which does not have much RF protection. Besides, ripple and noise must be your nemesis.
      But for most noise sensitive applications, LDO’s are incomparable.

  5. A lot of mentions above about how the “extra input voltage” is wasted as heat, but none of them addresses the current; the amount of heat you get is proportional to the product of the voltage drop by the current.

    So yeah, if you’re trying to pull a full 1A out of your 7805 and you’re feeding it 24V, that’s (24-5)*1 = 19W of heat. Otoh if all you need is 50mA that’s (24-5)*0.05 = 0.95W, which can be dissipated with a reasonable size heat sink.

    In other words, it’s a trade-off with these linear regulators – higher voltage drop across the unit vs higher current delivered from it. Fwiw, I generally only use them for low current situations.

    1. You’re right, my bad. I was expecting the worst case scenario under full load.. There are also low-power versions of the 7805 that look like a little transistor I vaguely remember (form factor of a BC548 or 2N2222A)..

  6. A replacement for the 309K TO-3 package would be a better idea as original stock is either fake or so rare its $75 each, but the demand is too low for the “big guys” to ship a commercial competitor.

    Just adjust the pinout to TO-3 pattern and make sure it runs up to 3+ amps, as “several” antique video game consoles rely on full limits (or more LOL)

  7. I really like the availability of small switching regulators. I have some concerns that they might fail sooner than a linear reg, and like others the noise from switching regulators often gives me problems. Linear regulators are quiet enough to use on a noise sensitive board; replacing one with a switching reg of the same footprint would cause problems. I usually put switching regulators into their own enclosure and try to quiet any switching noise there. Sometimes, the only way to quiet down a switching regulator is to follow it with a linear reg. ;-)

  8. As many have noted, these have been commercially available for like a decade now and are practically free, bonus points is that almost every design vomit RF like a frat boy mid party and anything that outputs video is basically destroyed … so why would I want to make one myself based on a hobby level design?

    1. “so why would I want to make one myself based on a hobby level design?”

      Hi! 🙂 Generally speaking? I don’t know. 🤷‍♂️

      However, some hobbyist works are better than commercial products.
      Same goes for prototypes developed in a lab or university.
      Simply because they’re designed without tradeoffs – commercial aspects don’t matter.

      That’s similar to old military products. They last. Because, they weren’t the victims of cost efficient planning. They had to meet the requirements above everything else.

      Especially in the amateur radio scene, hobbyists used to develop things that outperformed commercial products by several magnitudes.

      They did put all their heart into it, not caring about a bit of extra costs involved to invent a proper design. None of them would take something online/on aur that would possibly cause harm to others.

      (That’s in stark contrast to commercial manufacturers who build cheap plastic stuff without considering the customer’s safety. Every device and PSU should contain a fuse, for example, no matter the price tag.)

      – I’m speaking of real hams, here. Not those few “traitors” (imho!) that turned the hobby/service into a shady business idea and did start selling dubious stuff to their fellow hams. Or, included sabotaged schematics and sanded the chip markings off to prevent duplication of their products. Real hams are friendly and helpful, even allow repair or clone/improve their creations. They have values.

  9. Well, it was the typical component to always have in our drawers! A battery, an Arduino and this and it was ready to go on embedded systems. Of course, LDOs, hem DOs in the case of the 78xx are way better EMC-wise! For thermal issues, if we have a 12V battery, we’d better use a 7809 and a 7805 afterwards. The heat dissipation is shared.
    BTW, there is another emblematic circuit which is the so-called LM317. Has anyone used it? Apparently, it does a great job but I never dug into its meanders…

  10. Years ago, when I was into audio stuff, the audio maniacs preferred the Linear Technology 3 terminal regulators over the 78/79 series parts. I think it had something to do with better transient response, or maybe they were preferred because they cost more- there was a lot of that, too.

    I’m sure those folks would cringe at the thought of using a switching regulator in any audio circuit.

    1. Sure they will, but they also pay hundreds of $ for a single capacitor, and believe everything made of gold is better (even if copper is a better conductor of electricity).

      In fact, using a SMPS for audio with good results is pretty easy, as humans only hear up to ~20khz. CD audio is sampled at 44khz, so having a switching frequency waaaay above that (1mhz and up) is not that hard and thus, filtering is pretty easy. You just need to be careful with isolation so you don’t get any EMI from the supply into the amplifiers, either via the ground or by electromagnetic waves. Actually, that is how most consumer audio works nowadays. Top consumer brands (Bose, JBL, Bang & Olufsen, etc) aren’t placing linear supplies inside a stereo, even less inside portable speakers which are battery powered.

  11. i saw this in the projects page a few days ago. it’s interesting to me because probably about half the times i use a 7805 or LM317, i would prefer to have greater efficiency and i wouldn’t mind a bit of noise (i.e., a quick and dirty NiMH charger).

    what i really love about linear regulators is that i have a drawer with a dozen TO-220 7805s, a dozen TO-92 7805s, a dozen LM317s, and a few LM2930s or some other low-dropout regulator. i’m pretty sure there’s a couple 7905s. so i’ve got my bases covered pretty thoroughly, and at an absurdly low cost. and whenever i need to restock, i know i can always get another bundle of 7805s.

    the thing is, this doesn’t actually solve any of my problems. i have looked into these little switching ICs before, and they are kind of a pain to use compared to the 7805. you need more passives and you need to pay greater attention when picking them out. and at the end of the day, you put all the components together, it is actually pretty expensive for something to just have on hand without a specific project already in mind.

    if someone actually productized this, to where i could buy a heap of them and become familiar with their limitations (the tradeoffs forced by the passives that were selected, etc), then i probably wouldn’t mind the cost (within limits). so that’s my complaint, it is exciting enough to get my attention but there’s no ‘buy here’ link :)

  12. “It depends on what’s on the PCB” is a standard issue for all device specifications. Standard solution is to state the measurement conditions, then measure and report results.

    EMC info is necessary.

  13. It isn’t like asking the “how long is a piece of string”. Define conditions, define measurements, state results. That’s how the industry works, and it will work in this case too.

    Overall, if you want silly analogies, it is more like “how long is your piece of string?” And comparing that with my piece of string.

  14. 2020+
    Raspberry wants 5.3 V open circuit voltage input, Raspberry reports.

    Canakit power supply does this. We used Drok variable voltage supply because it has usb A output.

    5.3 w output from HF 12V AGM 35Ah worked great. Until poster connected battery to full sine wave inverter connected to Canakit power supply and killed [murdered] HF battery.

    What do want to do? Connect 7805 with 3 V diode drop to usb type A connector? Or buy a Drok?

  15. Off the shelf drop-in 3-lead TO-220 linear voltage regulator switching converter replacements have been around for years. The examples below for 3.3V [1], 5V [2], and 12V [3], are from Texas Instruments and according to [4] are based on the TPS54302 DC/DC converter IC.[5] The TPS54302 DC/DC converter IC page says the part can be simulated in TI’s free PSpice for TI design and simulation tool. Other switching replacements physically attach the exposed metal pad on the bottom of the switching IC to the TO-220’s metal tab to remove heat thereby allowing significantly more output current. Some versions allow for one or both of the feedback resistors to be used externally thereby making the output voltage adjustable like the LM317 adjustable linear regulator. Most drop-in replacement switchers are buck (step-down) converters. But I have seen boost (step-up) versions as well as buck-boost or SEPIC adjustable versions capable of stepping both up and down. The disadvantages with using these drop-in switching replacements are higher cost, higher output noise, and (perhaps) lower reliability. Usually these TI replacement switchers are rather popular and normally well stocked. But as you can see below for the 5V and 12V versions, that was before the forever chip-drought we’re in these days:

    * References:

    1. TPSM84203EAB, 3-lead Linear Regulator Replacement DC DC Converter 1 Output 3.3V, 1.5A 4.5V – 28V Input, Qty.-1 $8.05 each, 1,135 in-stock.

    2. TPSM84205EAB, 3-lead Linear Regulator Replacement DC DC Converter 1 Output 5V, 1.5A, 7V – 28V Input, Qty.-1 $7.39 each, 0 in-stock.

    3. TPSM84212EAB, 3-lead Linear Regulator Replacement DC DC Converter 1 Output 12V, 1.5A 14.5V – 28V Input, Qty.-1 $7.46 each, 0 in-stock.

    4. Upgrade your TO-220 linear regulator with a pin-compatible buck power module, Anthony Lang, Sep 22, 2017

    5. TPS54302 4.5-V to 28-V Input, 3-A Output, EMI Friendly Synchronous Step-Down Converter

  16. For several years, I have used the TI TPS62160 on a self-designed 78xx style PCB (tho no mounting hole – here, a mica insulator might work, though it depends on the pinout having the inductor on the “back” side or not). 17Vin max, 2.4MHz switching, Vin down to Vout (i.e. 100% duty if necessary), 1A. The key resistor for the voltage divider feedback is situated where it can be easily installed after primary assembly – so one can assemble several to 90% to bin, and when you need a particular output voltage, drop that resistor in. Granted 3V3 and 5V0 are what I generally use. One liability of that particular switcher is that the output voltage range doesn’t go very high – it’s not suitable for a 12V regulator, but I don’t need that. The filter caps are, just as in the project on this page, part of the PCB, itself, though are much larger values. The inductor though is much smaller (a TDK VLCF4018T-2R2N1R4-2)

    Admittedly, I don’t use the TO-220 style PCB package a lot (it is easier to integrate the switcher design into the other PCBs), but the drop-in switcher is darned handy for experimentation, or rework.

    For those who have applications that are so sensitive they can’t cope with even the possibility of noise (whether or not that is really coming from the switcher), stick with linear regs and their heat – nobody is telling you that you have to use switchers. There remains a lot of projects that don’t need the coddling, and the boost in efficiency is beneficial.

  17. If you don’t need a lot of current, Texas Instruments has several linear LDO (low drop out) surface mount regulators. I’d have to look up the part number. I’m using one to regulate the battery voltage on a 1930s era Bulle electromechanical clock. The LDO regulator output is adjustable and quiescent current is in the 10s of nanoamp range.

  18. I would agree that this might cause noise related problems in many circuits. But keeping efficiency and counterfeit parts in mind (Recently I got many fake ones with bad tolerances) I would say that is a nice solution.

  19. It’s not like asking “how much does a piece of string cost”. Define events, interpret measurements, and interpret results. This is how business works and it will work in this area as well. It’s like “how long is your leash?” if you want a bad all round comparison. And compare it to my channel.

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.