Who is the main manufacturer of jelly bean parts?

Question

Why do some parts like opamps and mosfets of different companies have same part number but they just change the initial letters. Does it mean that these are actually manufactured by just one company and different companies buy it and white label it.

Examle-

mosfet 2n7002 : make available are nexperia,dioides inc,onsemi,cdil,vishay,micrchip,toshiba,etc.

opamp lm4558 : make available are ti , htc korea , adi , etc.

If there is one main manufacturerthat sells without branding and if anyone knows their name can you please share. Looking for mosfets,opamps,transistors manufacturer.

Answer 1

There are some industry standards, and a lot of de-facto standards, and more ad-hoc imitation.

Back in the day, JEDEC registered parts could be produced by anyone, and sold as that part number so long as they met the spec sheet. (And many other provisions, I'm sure; I've certainly not read any official standards, or agreements, that might've been signed between such parties.) The thing you're buying is actually the spec, not, what's inside, how the part is made; this flexibility allowed multiple sources for standard components, which benefited the major customers at the time: growth of the semiconductor industry is closely tied with the military-industrial complex. (Which is more-or-less to say: they're one of the tentacles of exactly that complex.)

This is the origin of 2Nxxxx series parts, for example; the "2N" is actually JEDEC nomenclature, "2" being one less than the number of pins or electrodes (so, a tri-ode), "N" being a semiconductor device (hence 1N for diodes, 3N for certain MOSFETs, some optocouplers with 4N and 6N, and so on). This dates all the way back to vacuum tube days in fact (for example, I have a 2C53, a high-voltage triode; "C" must be hard vacuum type... although, 2E21 is a pentode, surely it should be 4-something? maybe there was a change of convention in there, dunno), so the early days of semiconductors were already accommodated by existing schema.

There are other basic-device nomenclatures in use around the world, too; or have been. 2SAxxxx (A, B, C, D, J, K) in Japan, and BCxxx (and other letters) in Britain/Europe (another schema dating back to vacuum tube days: EC91 is a 6V-heater (E) triode (C), for example), for two prominent cases.

As time went on, more and more advanced and proprietary devices were developed, and second sources were still stipulated; this forced negotiation of licensing terms between competitors, which explains, for example, how AMD obtained Intel's 8086-family designs.

JEDEC is -- I don't want to say less important, let alone unimportant, these days, but, outside of the industry at least, you don't hear much about them in regards to specific parts, and that's probably more to the effect that, the circumstances present in those earlier days, simply aren't around anymore, and standard parts have drifted away from industry-standard families towards proprietary ones.

Thus, we get de facto standards, like TIP31s made by everyone (literally, TI (Texas Instruments), 'P'lastic transistor, type 31); D44H11 is old GE nomenclature (later Harris, or maybe the discretes went separate ways, not sure offhand); MJE350 is old Motorola (I think, 'M'otorola, 'J'unction transistor, 'E' series plastic package, type 350); IRF540 (International RectiFier); etc. etc.

As for IC naming conventions, generally it's a few-letter prefix; µA for Fairchild, MC for Motorola (probably "C" for integrated Circuit), TL for TI Linear (I think??; also SN for digital, for some reason), LM and DM for National (linear/digital), etc. Many manufacturers second-sourced similar numbers with new prefixes, like how we have µA741, LM741; CD4013, MC14013; etc.

Probably, these days, there are few restrictions or limitations on who can make a part, or sell something as a given part number. You always have to be careful choosing substitutes: 3Peak makes a "LM2903A" for example, which looks to be a largely compatible imitation, but is most definitely a CMOS part, entirely unrelated to the original bipolar type. It doesn't even have to be a second source; the first-party can pull such tricks on you too. I recall a particular TI part offering a new CMOS version in the main-line datasheet, which could make for some awkward realizations when one is depending on typical bipolar behaviors.

Even just "ordinary" updates, revisions, die stepping, can cause issues; several STM32s, I think, have gained notoriety for having done this.

In summary: while there are some common parts, sometimes it's entirely coincidence (there's "2N3055" and "MTP3055", although maybe with ballpark-similar voltage and current ratings, but one is an NPN BJT and the other an N-ch MOSFET), sometimes it's imitation (2N2222A (TO-18 metal can) is a boutique part these days, but PN2222A (plastic TO-92) and MMBT2222 (SOT-23; Motorola lettering convention I think) are widely available), and you always need to check the datasheet to be sure.

Another example: µA/LM/etc. 723 regulator/controller ICs, a staple way back in the day; 2nd sources quickly dropped the original's buried zener voltage reference for a bandgap reference of comparable performance -- but noticeably poorer noise level.

Answer 2

I won't repeat what Spehro Pefhany has said, but will make an additional point. Sometimes when different manufacturers use the same part numbers, they also use an essentially identical datasheet as well. But one needs to check! Parts with the same part number but different manufacturer are, in general, pin compatible. However, it is often the case that there are subtle differences between the parts manufactured by different suppliers. These differences might include ranges for suggested operating conditions. Never rely on company X's datasheet for company Y's part, even though they have identical part numbers.

Answer 3

There are many manufacturers of those 'jellybean' parts. That's part of the attraction- multiple sourcing. They may be very similar to each other or there may be differences, subtle or not. In some cases there may even be multiple sources from a single manufacturer because they've acquired another maker and continue to supply both products.

They keep similar part numbers so that they can sell into the same market. A supplier who called their dual op-amp a NXY340012 rather than an LM358NXY will likely get fewer sales for the same effort. Sometimes they are even tricky and give completely different parts a similar part number such as LMV358. The designer knows what to expect in general- an economical dual op-amp with modest performance (but obviously there will be other differences - V suggests voltage is one of them).

2N7000 (and VN2222) are examples of trying to sell MOSFETs in place of BJTs by giving them more familiar-looking part numbers.

Answer 4

TL;DR Once patents expire, many different manufacturers

There are 3 issues that come to mind:

• Trademarks - It is possible, though not always so easy, to trademark a part number. A famous example of this is where Intel produced 8086, 80186, 80286, 80386, 80486 and then Pentium. Why the change? Because you can easily trademark a largely made-up name such as Pentium. It is much harder to trademark (if at all possible) a number. If a manufacturer chooses to trademark a particular part number then, provided they enforce the trademark, only companies that license the design (or at least the name) can use the same part number.
• Patents - Patents are for a limited time - 20 years is the usual US time. While a patent is active, it may be problematic for an alternate, unlicensed, manufacturer to market a similar product with the same part number as the patented product. Either it works identically and infringes the patent or it works "similar" and confuses people and causes problems. But after all relevant patents have expired, it makes sense for an alternate manufacturer to use the original manufacturer's part number if they are producing a compatible product.
• Copyright - This gets complicated. A small IC has little to no copyright issues as long as the new manufacturer doesn't literally copy an image mask. But more complex circuits, particularly modern microprocessors, have microcode in them which can be copyrighted. Producing unlicensed versions of those chips legally typically requires clean-room reverse engineering - i.e., figure out the functionality and then write microcode from scratch to do the same thing. Unlike patents, copyrights are, at least in the semiconductor era, essentially forever.

This is similar to "generic" products in many other industries, with the notable exception of pharmaceuticals (due to significant safety-related regulations).

For example, Oreo is a trademark and true Oreo cookies are made only by the trademark holder and official licensees. But anyone can make a chocolate cookie with a creamy filling and sell it in a supermarket right next to actual Oreo cookies. For many uses there will be no functional difference, maybe not even in a blind taste test. But for some uses - getting a cranky child exactly the cookie they have seen advertised or eaten at their friends' homes, etc. - nothing else will do.