EEVblog > EEVblog Specific

EEVBlog 1436 - The TOP 5 Jellybean OPAMPs

(1/14) > >>

Dave looks at his TOP 5 (plus change) Jellybean OPAMP's, and explains why you need to know them.

00:00 - Jellybean OPAMP's
01:47 - LM358
07:51 - FET Input TL071/72/74
11:28 - CMOS LMV358
15:17 - LM324
17:23 - The LM321 is NOT a thing
19:09 - Oh, all right, the LM741
19:41 - RC4558
21:11 - The Audiophiles go WILD! The NE5532
22:06 - OP07 Precision OPAMP

TLV2372 didn't make the list smh ;)


The Ti datasheet for the TL07x is confusing. It includes the TL07x H series and these are totally different (CMOS) parts. It is still a good OP but a bloody cunfusing part numbering  :rant:.  Looks like the want to carry over the old part number and sell cheaper to make CMOS chips. They are good in some repects, but also worse than the old in other aspects.  There are a few threads about this, e.g.:

The NE5534 is the single version, not the quad.  Again a bit confusing numbering (like the LM2904 LM2902 for the dual and quad). The note with the need for external compensation is good however.

The RC4558 is sometimes called equavalent to 741, but it is actually a quite differente OP: higher BW, lower noise,... more like half the way to the NE5532. Closer to an dual 741 is the MC1458 /1558. The RC4558 is still a good yellybeen part - wish there would be a single version (because of the lower power).

The LM321 is a tricky one:
 often they are LM358 dies with just 1 half used. The current per OP is twice. AFAIK a few are actually single dies. SO unleess you really need a small case one could as well use the 358.
There is another odd point: there was a seprate chip with LM321 part number and this was not a single OP.

AFAIK there is no direct dual or even quad OP07. With the relatively high power consumption this would also no be practical. There are duals and quads from later improved versions - e.g. as my new favorite replacement OPA202 / OPA2202 / OPA4202. In the old days this was more like AD704,705/706.

How did these end up being made by multiple manufacturers under the same model name? Does the mask mark expire and somehow another company either gets ahold of it or reverse engineers it from samples?

Also how does it make economic sense to use masks from the 70s?? For maximum cheap, you should balance cheapest process technology with maximum parts per wafer and minimum out-of-tolerance yield loss. But each of these parts has a separate lineage of improved versions from each manufacturer. Those derivatives are unsurprisingly rarely made by more than one manufacturer. Surprisingly though, such parts rarely compete with the ancient originals on cost and somehow never manage to obsolete them. Was copying one another's model names just a trendy thing that went away by the 2000s?

I certainly don't have a complete history of multiple sourcing, but it's my understanding that, a lot of that was driven by direct military requirement.  Keep in mind, silicon valley has always been closely integrated with the military-industrial complex; they are what enabled ever more advanced and compact missile guidance, RADAR, countermeasures, cryptography, etc. -- everything from bombs to bugs.  So, it pays to have one spec (e.g. JEDEC 2Nxxxx series) and everyone makes parts that fit them.  I don't know what the extent of systems like JEDEC really was/is, why they didn't put amplifiers in there for example (but some quite complex optos are, take 6N137 for example), but in any case, op-amps and comparators and everything else were certainly too important to pass up, plus everyone needed them for commercial purposes anyway, so why not make a copy and get your toe in the door?

As for origin, I don't know if that was by licensing, copying (reverse engineering (RE)), or parallel development, and how much of each.  Certainly if nothing else, some work was required to get one chip design working on another fab line, fabs were especially finicky back then (and, I would suppose, are only so much worse today -- they're just also better at abstracting it out via simulation and synthesis tools).

One famous case I do know was by licensing: the only reason AMD has x86 to this day (well, to this day, who knows what else might've happened through history, right, but this is how it started anyway), is because the military insisted upon second sourcing the CPUs and so made Intel and AMD do a cross-licensing deal.  That, or to address anticompetitive action?  Now I can't remember... anyway, it's a well recorded part of computer history.

(In contrast, Cyrix, or a couple of others I think, got the 486 by RE.  So, more than just the two giants in that era of x86 history.)



[0] Message Index

[#] Next page

There was an error while thanking
Go to full version