There are so many op amps to choose from…. which one should I use?
The synth project is full of op amps. Along with the occasional OTA or comparator, operational amplifiers are the workhorse of the system. Most of the time, I use a basic TL072H (or TL074H; the 2- or 4-element ICs have the same characteristics) and just keep moving. Every once in a while, though, you need something more specialized and it’s worth paying up for better performance.
That said, “performance” isn’t a single axis. Op amps with the most stable tempco often have higher input bias current than the leading CMOS- or JFET-input amplifiers. Or low-offset op amps might have lower bandwidth. And of course, the more parameters are improved, the more it will cost.
The synth runs on ±12V DC power, so I’m mostly using “wide” op amps that have a total DC voltage range of 36-40V. The lower-power op amps that are limited to 5V range aren’t especially useful. I recently went through TI’s catalog of general-purpose op amps, filtered to 36V+ range, and compiled a bunch of the parameters I care about most into a single spreadsheet that covers over a dozen options.
I highlighted any particular strengths (i.e., the “winners” in a given column) in bold. Prices listed for each op amp were at MOQ 1 on DigiKey for SOIC on cut tape at the time I put the spreadsheet together (August–September, 2024).
The spreadsheet also contains tabs that have some logic in “flowchart” or “rule” formats to walk through the decision-making of which op amp to use when. I found myself too often spending way too much time reworking the same criteria into essentially the same decisions each time I found myself needing something more specialized than TL072H.
TL072H? Where’s TL072?
TL072H is the next-generation version of TL072. You shouldn’t use the old TL072. The “H” variant has better Vos, better tempco, an impressively low input bias current, faster slew rate, improved common mode input range (alas, still not rail-to-rail), and a common mode output range that’s almost rail-to-rail, requiring just 210mV of headroom.
It also costs the same or less than the original; at $0.45, it’s about the same price as a postage stamp.
There’s just no good reason to use the older generation chip in a PCBA any more. The only “advantage” of TL072 over TL072H is that it’s still offered in PDIP. That does make breadboarding mildly easier, but it’s also not too much work to solder an SOIC onto a breakout / adapter board.
What happened to LM324?
Honestly, this op amp just isn’t worth it imo. With the “B” upgrade and “A” binned ICs (i.e., LM324BA) it almost gets to be something impressive enough worth considering, but there’s very little it can do that TL072H can’t.
Typical V_os is actually better in LM324BA. dV/dT temperature stability is much worse. Noise is comparable but THD is 10x worse. Input bias current is 10,000x worse! CMRR is significantly worse. It’s a very slow opamp with a GBW of 1.2 MHz and a slew rate of just 0.5 V/μs. That’s acceptable in a chopper-stabilized zero-drift amplifier, but in something general-purpose? Tech has moved on from the 70’s. Oh, and for some absurd reason it’s a few pennies more expensive to boot.
Pretty much the only situation where I’ve found myself consciously selecting LM324BA over TL072H is that whereas the TL072H input range includes the V+ rail, it needs about 1.5V of headroom above the V- rail; on the other hand LM324BA needs about 2V of input headroom away from V+ but the input and output range both include the V- rail. Which means you can hold one input at true GND and the other at virtual ground in an op amp with a single-sided power supply by using the LM324BA, whereas TL072H would require you to bias the inputs at something like V+ / 2.
If you just want the “what do I use, when” decisions, here’s a copy of the “flowchart” from the spreadsheet. You start in the upper-left corner, presuming you’ll use TL072H. Then as you want to optimize one parameter or another, you follow the arrows to where that leads you. In most cases, further progression deeper into the chart leads you to a more expensive option, so you can also stop when it hits your budget. Sometimes that’s not the case though, so it’s worth it to look up pricing before making a decision. (One or two edges are also marked “budget” which leads to a nearly-as-good but explicitly cheaper option as well.)
Lately I’ve used several OPA2991 in a project where I wanted lower input offset voltage but more importantly, lower temperature dependency in Vos. Price-wise, it’s only the first rung of the upgrade ladder after the starting point budget-friendly TL072H (at a little under two bucks per ‘2991), but it’s already a significant improvement and I’m really impressed at what’s possible as soon as you get out of bargain-town.
When I’ve really needed to dial up the capabilities, OPA2145 is an incredibly impressive device (although almost twice the cost, at $3.65 each). One thing to note is that the OPA19x, OPA99x and OPA156 have rail-to-rail inputs, but the very low-dV/dT options like OPA145 and OPA18x have a more restricted common-mode input range. Rather than follow the chart blindly, check where it leads you on the spreadsheet, as most op amps are better in some dimension but sacrifice something else (often headroom). Unless you wind up at OPA192. That thing is just fantastic in every way. Of course, you’ll have to pay up for that.