Are your jellybeans getting stale? [lcamtuf] thinks so, and his guide to choosing op-amps makes a good case for rethinking what parts you should keep in stock.
For readers of a certain vintage, the term “operational amplifier” is almost synonymous with the LM741 or LM324, and with good reason. This is despite the limitations these chips have, including the need for bipolar power supplies at relatively high voltages and the need to limit the input voltage range lest clipping and distortion occur. These chips have appeared in countless designs over the nearly 60 years that they’ve been available, and the Internet is littered with examples of circuits using them.
For [lcamtuf], the abundance of designs for these dated chips is exactly the problem, as it leads to a “copy-paste” design culture despite the far more capable and modern op-amps that are readily available. His list of preferred jellybeans includes the OPA2323, favored thanks to its lower single-supply voltage range, rail-to-rail input and output, and decent output current. The article also discussed the pros and cons of FET input, frequency response and slew rate, and the relative unimportance of internal noise, pointing out that most modern op-amps will probably be the least thermally noisy part in your circuit.
None of this is to take away from how important the 741 and other early op-amps were, of course. They are venerable chips that still have their place, and we expect they’ll be showing up in designs for many decades to come. This is just food for thought, and [lcamtuf] makes a good case for rethinking your analog designs while cluing us in on what really matters when choosing an op-amp.
Ahh the LM324, my first op amp.
At JLCPCB a version of LM324 is available as a basic part (no assembly fee). In applications where it is good enough why pay more?
You have to ask, why is it free? Are they throwing in some random clones or whatever back-alley crap they can get for a penny?
it’s not free
it is just used so commonly that reel with these is mounted on PCBA machine all the time anyway, so theres no extra fee on top of parts price itself
Their basic part is an ATM LM374DT quad in soic14. Its datasheet looks near on identical to TI’s LM324-MIL. I thoroughly doubt JLCPCB would swap it out for no-name clones
STM*, silly autocorrect.
The old school “better than LM324” choice would have been TL071 family or CA3140 … TL071 family has a nasty caveat though: It will latch up with inputs driven even momentarily too close or beyond the rail.
That TL07x is a nice op amp, but very power-hungry compared to the LM324 or LM358.
Fairly sure people are still using 2n2222’s and 2n3904’s because they were in all the textbooks 20+ years ago and no-one can be bothered to change.
20 years ago? 40 years ago, and further back. But the issue isn’t “can’t be bothered to change”, but well know characteristics, suitable for the job, and cheap.
These (and their pnp siblings) are still pretty darn good general purpose BJT’s, and in a lot of applications, a BJT is the most appropriate choice.
Also there are complementary pairs. 2N3904/2N3906 etc.
If it ain’t broke, don’t fix it. If you’re not driving it beyond its limits, it ain’t broke.
… and when working with ancient designs, it’s nice that someone else already marched over all of the potential land mines for you.
When it is ACTUALLY broke, of course you should fix it. But modernizing to gain unused capability isn’t usually a win, when it comes with newer and less-well-understood land mines. Also, hindsight is better than foresight. Certain legacy parts have been in production “forever”, and are widely enough depended upon that that’s unlikely to change any time soon, while many newer parts aren’t profitable enough to affect decisions about their fab nodes.
A modern replacement may be wonderful, but it’s much less wonderful if it depends on more complex, more niche fabrication processes that more easily lose profitability. Many of the modern parts depend on advanced fabrication processes that have high cost and complexity. These parts are not profitable enough (by themselves) to justify continuing operating a foundry on that process node just for them.
When the profit-driver products shift off that process node, many of the cool parts don’t justify a re-design for the new fab node, so they go EOL. In comparison, the classical nodes are much more limited but can turn contaminated spit and semiconductor trash into commercially profitable parts, and all of the costs are long since sunk (and recovered), so they’ll keep churning out the parts as long as there’s demand, or until something really expensive to fix breaks at the factory.
Neither class of part is “wrong”: there are projects with needs that can only be met one way. But it’s a good idea to spend at least a little bit of time considering the bigger picture, at least if your project isn’t a one-off vanity project, or is somewhat mission-critical. Occasionally even HaD-type projects benefit from considering these factors, even though they’re usually peripheral to the needs of a quick hack or an art piece.
(I recently got contacted by an old employer because a quick hack I threw together to tide us over until a vendor got a module back in stock had finally failed. It was originally scheduled to absolutely be replaced within six months, but it failed after 27 years… The (already paid-for) custom vendor module got delayed several times, and when it was finally delivered, I was no longer involved with that department, so they apparently just warehoused the part and left the temporary fix in place. Sometimes nothing’s as permanent as a temporary fix…)
Here in Europe, everybody started with BC547 and BC557 as “small signal transistors”, and I used to buy them in bags of 100 for cheap. These have been common from the ’80-ies. I still use them when TO92 is convenient for breadboarding or prototypes on matrix board. There is nothing “wrong” or “obsolete” about these.
If I want to test / verify SOT-23 resistors, I start by soldering two pins to a 0.1″ single row header. The third pin is connected to the header with a wire. This gives quick and easy access both for measurement equipment as for use on breadboards.
“Everybody”? Try starting with OC71! If you scraped the paint off it, it would also work just fine as a phototransistor.
Do any of you remember the red spot and green and yellow spot Germanium transistors.
There isn’t anything particularly wrong with 2N2222. The only real sin with discrete transistors is using long-obsolete parts. Doubly so if the circuit depends on some exotic property of the obsolete component (I’m looking at you, BF862).
if you push the limits or if you are using them particularly often then of course you want something better. the specs that you can get if you’re willing to pay more than a penny per transistor are really something these days!
but i use on average probably like 1-3 transistors per year. i love that i’m familiar with the transistors i have…there’s no surprises in their datasheet, at least not in the range i use them. i love that i have a bunch sitting around, so if i get an idea i don’t have to wait for shipping. i can’t really imagine stocking a wide variety of more specialized transistors (though i have a tiny assortment of ‘power supply’ transistors and a few FETs too).
and that’s how i feel about the 555 and the 7805 and LM317 as well. though i do have a handful of low-dropout regulators, LM2940 or something like that. i also have a few op amps but i don’t use them often. but there’s nothing like having parts you’re familiar with. for a real pro that’s going to be a wide variety of parts but for me it’s mostly the beginner parts i learned in the 90s.
It is driving me crazy that the picture of the µa741 is upside down.
And it is so tough to correct when my tablet has auto-rotate engaged!
The reason why I often resort to ua741’s is that I have such a large amount of them that they have their own drawer in my parts drawers. I have many others opamps, including CA3140, LM324, TL071, NE5534, and what have you. But less of each, and they all fit into one drawer. Hence, it’s easier to open the ua741 drawer and know what I have when I take one out, than to open the other drawer and rummage for a type that is most suitable. Basically the 741 can almost always be made suitable for anything that I need.
It’s called ‘laziness’, I know. ;) But I’m a hobbyist, not a professional. And my design have only one requirement: they have to do their thing more or less fine.
As long as you don’t design circuits that use > about 5 Volts power and you don’t mind paying 4 to 10 times as much per part, I guess those are good substitutes for hobbyists.
However if you are designing for a living, care about BOM cost and an LM324 will work fine, you’d be foolish to use those suggested parts.
Oops, didn’t notice the replacements are all duals at first, so the price increase vs a 324 is 8X to 20X on a per OpAmp basis
I’m the original author. The point is that the vast majority of modern electronics are nominally single-supply 3.3 V or 5 V. Trying to use obsolete op-amps in this setting increases complexity and cost.
The rest of the article is a discussion of how to select op-amps if your requirements diverge from that. There’s plenty of modern chips that work in the rare case you need higher voltages, etc.
Of course, if it works fine, go with the cheapest part. But stuff like LM324 usually doesn’t “work fine” without a lot of caveats that are absent in more modern designs. Crossover distortion, etc.
Pro audio equipment isn’t using 3.3 V or 5 V single supplies. And input noise matters, a lot.
I seem to remember some pro audio stuff I worked on years ago, like doing repairs on a Neve studio console, used ±15V. People who only think of battery-powered consumer items get tunnel-visioned into thinking everything has to be 5V (or lower).
MZ, the 324’s crossover distortion you mention can be remedied by biasing the output. Yes, the output. For good audio, I have found that for example if I have a 30K load on the output, putting a 10K from output to ground will adequately get rid of the crossover distortion. There are other tricks like this that took me years of designing products with them to figure out, things that often aren’t even in the data sheets, which is why I do not take lightly the matter of switching to another part, since there may be another cycle of learning the nuances with the new part (although if cost were no object, I might be putting LT1125’s in everything).
It says above,
“For readers of a certain vintage, the term ‘operational amplifier’ is almost synonymous with the LM741 or LM324, and with good reason. This is despite the limitations these chips have, including the need for bipolar power supplies at relatively high voltages and the need to limit the input voltage range lest clipping and distortion occur.”
I’ve designed a lot of stuff for private-aircraft communications where we’re bound to old standards of things like voltage swing, meaning we cannot use something that only swings 5V. Further, we always have at least a 12V supply available, and in many cases it’s 28V. (I regulate the latter down to 12V, and have to tolerate voltage spikes of up to about 50V, without things shutting down.) The 324 does not need bipolar power supplies though, and its inputs operate just fine all the way to the negative rail, and output can too if the load is super light. The inputs and outputs are generally kept a couple of volts away from the positive rail. I usually bias it to make the center of voltage swing to be (Vcc-2V)/2, or maybe just a tad higher.
I wrote above, “…putting a 10K from output to ground…” I mean to the negative rail; but I was thinking of a single power supply when I wrote it. If there were ± power supplies, you’d run that resistor to the negative supply, not ground.
LM324 also works with a single supply rail, and with input and output voltages pretty close to the negative real. All “rail to rail” opamps have some limits in this regard, although some can have input voltages that are outside of the power supply rails.
I’d say, LM324 and LM358 are still good opamps to use when starting in electronics. They’re cheap and available everywhere, and they are still “good enough” for many circuits. Also, when you bump into the limits of the capabilities of these opamps, It is a good moment to start learning of what these limits exactly are, and either how to work with or around them, or when to start looking for other opamps.
LM308 quickly replaced the 741 in any app that can use lower power and much better performance. Among old parts I would go with LM308 and LM324. LM308 can make a basic integrator with a time constant of seconds, versus a fraction of a second with 741.
In a design at work 30 years ago, I did a delay that was several minutes, using an LM324.
This is a great write up. And a nice list of single supply op amps. Although I generally only work with dual supply systems in my hobbies. Most projects are mixed signal, so the single supply requirement is very appropriate in those use cases. Still, I ordered a tube of MCP6022 recently, which made the list.
That said, a quad op-amp like TL074 is very economical for projects that have a lot of analog stages that aren’t particularly sensitive to distortion. (like an envelope generator for a modular synthesizer)