Author Topic: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs  (Read 19694 times)

0 Members and 1 Guest are viewing this topic.

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 21606
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #50 on: November 15, 2021, 01:01:02 am »
It's certainly among the JFET jellybeans.  LF353, TL072 and a few others share a similar status.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #51 on: November 25, 2021, 07:52:25 pm »
I just made a connection with an obscure piece of jellybean operational amplifier history - George Erdi.

I knew that George Erdi had designed the OP-07 while at PMI, and was the uncredited designer (1) of the Fairchild 725 which was the first precision integrated operational amplifier:

http://analogfootsteps.blogspot.com/2014/02/guru-3-george-erdi.html

But he also designed the LT1006(single)/LT1013(dual)/LT1014(quad) while at Linear Technology which are an improved 358(dual)/324(quad) with precision and a class-AB output stage.  Where have I run across that before?

Fairchild also had a 358(dual)/324(quad) design improved with a class-AB output stage in the form of the 799(single)/798(dual)/3403(quad), (2) so I wonder if George Erdi was involved with those.

Does anybody know why the 324/358 design has a class-b output stage instead of class-ab?

(1) I do not know if anybody else has noticed, but Fairchild, National, and Texas Instruments made it policy not to credit engineers with their publications, while the companies started by dissident engineers from these companies like Analog Devices and Linear Technology did.  PMI did not either but George Erdi also moved from PMI to Linear Technology.

(2) The datasheets for these parts explicitly say the output operates down to the negative supply, but oddly enough the schematics (and specifications) show no such capability.  I wonder if that was left out of the schematics to foil competitors.  Maybe they never worked or were never finished.
« Last Edit: November 25, 2021, 08:17:56 pm by David Hess »
 

Offline Cerebus

  • Super Contributor
  • ***
  • Posts: 10576
  • Country: gb
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #52 on: November 25, 2021, 10:30:34 pm »
George Erdi was also a regular contributor to academic publications, so unlike his peers - who tended to publish mostly in the trade press or their employer's application notes - you tend to be able to find neutral information from him (i.e. designed to serve the purpose of communicating the technique rather than with the tacit understanding that the publication was to serve an employer's interests).

I've found papers of his in the IEEE Journal of Solid-state Circuits and in IEEE conference proceedings.
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #53 on: November 25, 2021, 11:04:52 pm »
George Erdi was also a regular contributor to academic publications, so unlike his peers - who tended to publish mostly in the trade press or their employer's application notes - you tend to be able to find neutral information from him (i.e. designed to serve the purpose of communicating the technique rather than with the tacit understanding that the publication was to serve an employer's interests).

I brought it up because it makes makes me wonder how much motivation that policy added to these critical employees leaving, and eventually forming their own company where their work would be more widely recognized.
 

Online mawyatt

  • Super Contributor
  • ***
  • Posts: 3192
  • Country: us
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #54 on: November 26, 2021, 02:14:45 am »
George Erdi's designs were extraordinary, in the same class as Widlar's IMO. Think George invented the cross-coupled quad diff amp, and his input bias current canceling scheme was one of the best analog designs I'd ever come across.

Agree, he deserved way more credit, same goes for Oliver Heaviside.

Best,
Curiosity killed the cat, also depleted my wallet!
~Wyatt Labs by Mike~
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #55 on: November 26, 2021, 02:12:33 pm »
Think George invented the cross-coupled quad diff amp, and his input bias current canceling scheme was one of the best analog designs I'd ever come across.

Are you referring to the thermally coupled cross-quad, like in the 725 and all later precision parts, or the linearized cross-quad?
 

Online mawyatt

  • Super Contributor
  • ***
  • Posts: 3192
  • Country: us
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #56 on: November 26, 2021, 02:22:42 pm »
Think George invented the cross-coupled quad diff amp, and his input bias current canceling scheme was one of the best analog designs I'd ever come across.

Are you referring to the thermally coupled cross-quad, like in the 725 and all later precision parts, or the linearized cross-quad?

Yes, the thermally cross coupled quad used in the 725.

Best,
« Last Edit: November 26, 2021, 02:25:12 pm by mawyatt »
Curiosity killed the cat, also depleted my wallet!
~Wyatt Labs by Mike~
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #57 on: November 26, 2021, 03:03:36 pm »
Think George invented the cross-coupled quad diff amp, and his input bias current canceling scheme was one of the best analog designs I'd ever come across.

Are you referring to the thermally coupled cross-quad, like in the 725 and all later precision parts, or the linearized cross-quad?

Yes, the thermally cross coupled quad used in the 725.

I asked because the first time I ran across the linearized cross-quad was in a PMI application note 105, and George Erdi worked there so might have invented it also.
 

Offline Rich S

  • Newbie
  • Posts: 9
  • Country: us
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #58 on: November 26, 2021, 11:51:15 pm »
Picking the top 5 jellybean op-amps is like picking the top 5 cars. The product has diversified into so many categories, most are specialized in some ways. 

Horowitz & Hill did a lot of work: their Art Of Electronics has many tables of their chosen favorites. As mentioned the LF411 is their preferred "teaching tool" op-amp. AoE 3rd ed., table 4.2a, is the starting place for choosing op-amps -- over 30 parts there alone. (Note: their focus is on extant parts, & not historically what has been 'popular'. So they omit the LM741. And only some of these are 'jellybeans').  Top of list, no surprise, is the LM358/LM324 - it just delivers a high bang-for-buck.

While mentioning the designers behind the op-amp winners, Like George Erdi, let's mention Tom Fredericksen. He designed the quad amps LM3900, LM324, and LM339. I don't know if he designed these quads first, then the dual/singles; or, someone else did those & he did the quads.

« Last Edit: November 26, 2021, 11:55:38 pm by Rich S »
 

Offline Cerebus

  • Super Contributor
  • ***
  • Posts: 10576
  • Country: gb
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #59 on: November 27, 2021, 12:49:16 am »
Picking the top 5 jellybean op-amps is like picking the top 5 cars. The product has diversified into so many categories, most are specialized in some ways. 

No it's not, it's like picking the top five economy cars. Jellybean remember, as in "inexpensive enough that you don't bother about the price of a single one, just the price of bowlfuls of them". So specialised doesn't come into it, jellybean op amps are, by their very nature, generalists. You can do the weekly shop in a Ferrari, just as you can use an OPA627 to build a low bandwidth, gain of 1, buffer - but in practice neither are in the running for everyday humdrum applications.
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #60 on: November 27, 2021, 05:51:50 am »
Horowitz & Hill did a lot of work: their Art Of Electronics has many tables of their chosen favorites. As mentioned the LF411 is their preferred "teaching tool" op-amp. AoE 3rd ed., table 4.2a, is the starting place for choosing op-amps -- over 30 parts there alone. (Note: their focus is on extant parts, & not historically what has been 'popular'. So they omit the LM741. And only some of these are 'jellybeans').  Top of list, no surprise, is the LM358/LM324 - it just delivers a high bang-for-buck.

National had a whole series of LF numbered JFET parts, and I prefer them to the Texas Instruments TL series, but they are less available now so more difficult to recommend.  They are or were second sourced however, and Linear Technology made improved replacements for many of them.
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14072
  • Country: de
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #61 on: November 27, 2021, 08:59:32 am »
The difference between LF411, LF351 and TL071 is rather small and for many uses it does not matter. Today the TL parts have the best availability. It is rare that a circuit defigned with the LF411 in mind would not work with one of the other variants.

With many of the yellybean parts there are also similar parts that could also get a comparable price point - the exact part name usually does not matter. The yellybean parts also come sometimes with slightly different names like CA324, KA324, BA324 and so on.
The yellybean parts often stand for a class of chips and there are sometimes similar parts that are not all the way identical. This is especially the case with the NE5532 and LMV321 with similar (but not all the way identical) alternatives (e.g. LM833, MC33078, MCP600x).
 

Online mawyatt

  • Super Contributor
  • ***
  • Posts: 3192
  • Country: us
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #62 on: November 27, 2021, 02:07:52 pm »
Horowitz & Hill did a lot of work: their Art Of Electronics has many tables of their chosen favorites. As mentioned the LF411 is their preferred "teaching tool" op-amp. AoE 3rd ed., table 4.2a, is the starting place for choosing op-amps -- over 30 parts there alone. (Note: their focus is on extant parts, & not historically what has been 'popular'. So they omit the LM741. And only some of these are 'jellybeans').  Top of list, no surprise, is the LM358/LM324 - it just delivers a high bang-for-buck.

National had a whole series of LF numbered JFET parts, and I prefer them to the Texas Instruments TL series, but they are less available now so more difficult to recommend.  They are or were second sourced however, and Linear Technology made improved replacements for many of them.

Recall some of those early National JFET LF type Op-Amps had a nasty feature where the output would invert when the input was significantly overdriven, not a good choice for feedback systems!

Best,
Curiosity killed the cat, also depleted my wallet!
~Wyatt Labs by Mike~
 

Offline Rich S

  • Newbie
  • Posts: 9
  • Country: us
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #63 on: November 27, 2021, 09:02:18 pm »
Re "Jellybean remember, as in "inexpensive enough that you don't bother about the price of a single one, just the price of bowlfuls of them""
Oh, I'm not sure the meaning is purely that.
But OK, if low price is the main criteria, then, the new contenders are...
(I did a quick look at Digi-key & Mouser; price in USD, thru-hole packages only)...
the '358 is at #4. still competitive.

#VendorMfr Part NumberPricingNumber of Channelsprice per chan.
1MouserNJM2059D$0.384$0.09
2DigikeyMCP6004-I/P$0.574$0.14
3MouserNJM13404D $0.342 $0.17
4MouserAS358P-E1 $0.362 $0.18
5DigikeyNJM2747D $0.764 $0.19
6DigikeyAD8534AN $0.784 $0.20
7DigikeyMCP6002-I/P $0.422 $0.21
8DigikeyTLV2765IN $0.844 $0.21
9DigikeyTLV2634IN$0.874 $0.22
10DigikeyTLV2635IN$0.974 $0.24
11MouserNJM12904D $0.492 $0.25
« Last Edit: November 27, 2021, 10:11:41 pm by Rich S »
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #64 on: November 28, 2021, 07:49:09 pm »
National had a whole series of LF numbered JFET parts, and I prefer them to the Texas Instruments TL series, but they are less available now so more difficult to recommend.  They are or were second sourced however, and Linear Technology made improved replacements for many of them.

Recall some of those early National JFET LF type Op-Amps had a nasty feature where the output would invert when the input was significantly overdriven, not a good choice for feedback systems!

That is called phase reversal and it is a common flaw with BiFET operational amplifiers, including the TL series form Texas Instruments, (1) and some bipolar ones like the 324/358 and related devices.  Analog Devices published details on exactly what is going on.

Modern improved parts (with different part numbers) may be designed to prevent or be resistant to phase reversal, like the Linear Technology LT1013/LT1014 and various LT BiFET parts.

(1) From Texas Instrument TL071 datasheet - Avoid input voltage values below 1 V to prevent phase reversal where output goes high.
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 6733
  • Country: pl
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #65 on: December 06, 2021, 01:32:08 pm »
This paper about phase reversal is highly unhelpful as it deals with only one peculiar input stage configuration and recommends a fix (series resistance) which AFAIK does absolutely nothing to solve phase reversal in TL072 and similar chips.

The problem with TL072 is simple: low common mode input voltage slides the input pair's operating point so low that it no longer is able to drive the (inverting) second stage so the output is pulled up by the second stage active load. And there are other chip with the same problem:
- RC4558/4560/etc - identical topology with PNP instead of PJFET inputs, unlike µA741 which has no phase reversal
- LM4562 ::) - a modern audio opamp, possibly similar to the above
- LM324/358 - uses PNP emitter followers which permit input down to ground but not much below
- TLC272/TS272 - the output doesn't really go up but its ability to sink current is compromised

A useful feature of TL072 and TI's CMOS TL072H is input range which includes the positive rail. One could say that they are "negative" single supply opamps.

Somebody asked about internal compensation of NE5534. Every datasheet, even TI, says this in the second paragraph:
Quote
These operational amplifiers are internally compensated for a gain equal to or greater than three.
As an aside, I'm yet to find a specimen which is unstable at unity gain driving 10kΩ resistive load. I tried Philips, Signetics, JRC, TI.
« Last Edit: December 06, 2021, 01:38:57 pm by magic »
 
The following users thanked this post: edavid

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #66 on: December 06, 2021, 09:17:27 pm »
A useful feature of TL072 and TI's CMOS TL072H is input range which includes the positive rail. One could say that they are "negative" single supply opamps.

The ancient LM301A also has a common mode input range which includes the positive supply.

I have often wished that there was positive ground version of the 324/358 with NPN inputs instead of PNP inputs.
 

Offline jwet

  • Frequent Contributor
  • **
  • Posts: 447
  • Country: us
Re: EEVBlog 1436 - The TOP 5 Jellybean OPAMPs
« Reply #67 on: March 28, 2022, 03:42:44 pm »
I've enjoyed your series on Jelly Bean parts.  I worked at Maxim as a new product definer and eventually worked my way up to managing all product definition for the standard products side of the company- this included amps, refs, switches, interface, non portable power, supervisors, comparators.  I later worked as an exec on business development for new markets- automotive and industrial automation.

Ironically in Maxim's story, these standard products were the real money making parts for the company- the margins were very good and the need was consistent.   The black eye that Maxim got in the early 90's for poor delivery performance was caused mostly by new very high volume parts that only a few customer bought.  At any given time, we had some surprise upside on some part.  A lot of this was specialized power solutions for Notebook PC's and high efficiency high current bucks with VID control for the then new Pentium. The market has a long memory for this, I still see people bashing Maxim over a delivery issue that happened 30 years ago.  Maxim ran all their own somewhat boutique fabs and trying to pump out a million parts a week of some exotic part that hadn't reach good yields can make your manufacturing pretty difficult.  I have to respect them for never backing down on quality and we eventually caught up with demand- in fact Maxim was the first analog company to fab 300 mm wafers- that's the big fix.

Back to Jelly beans.  One of the strategies (among many others) were to look at good jelly bean parts and expand out from them in one of more vectors without affecting price too much.  LTC did similar things.  Maxim's specialty was CMOS analog though all of our processes could be called BiCMOS with really good CMOS analog, fancy precision resistors and good caps.  LTC in contrast worked on mostly bipolar processes with some CMOS and included all the other goodies like SiChrome R's and poly caps.

We developed a lot of neat strategies for new products.  Most were market driven by customers telling us what they wanted and us extrapolating a bit on what they were really saying.  We also did continuous studies of markets and product segments to see where there were wholes.  Maxim couldn't (or shouldn't) compete with National making LM324's- we couldn't do anything with a part like that except make it cost more and it wouldn't use our strength.  One of the strategies that was ongoing in the product oriented organization and business management side was to look at comps often jelly beans and see what could be improved to capture some of that market using our strengths.

In Op-Amps specifically, this was pretty easy because the pinout of quads and duals were well established.  If you take an LM358 dual as the center of gravity- you could push bandwidth, precision, low power, common mode range and output drive.  If you look at Maxim's portfolio thought the 90's, you can see this taking place.  We pushed power and voltage so low that no one cared anymore- we had several parts that drew fractions of a microamp! and people bought them.  We made parts that could drive 80 mA of output, customers started using them as headset amps and little motor drivers this lead to a new product areas.  We made a standard dual that could do 28 Mhz on 1mA of Icc- LM324 territory- there was a lot of pent up demand and led us in that direction to a line of high speed low power parts.  Our rail to rail input parts led to high side current sensing apps which became an important product segment as everything was "portabilized".   Most of these parts sold (not all) but they gave us a lot of intelligence of what the market really wanted.  There was some misunderstanding in the market and inside our own managment that we were just carpet bombing the world- the truth was more strategic.

I'll give you one more example war story as this is getting long.  Analog switches existed mainly as the DG series from Siiiconix and AD.  Originally JFET and later CMOS versions- the only customers were precision instrument guys, ATE and military.  National created several jelly bean switches and muxes in the 4000 series- 4016, 4066, 405x series.  (The 4016 was terrible and was soon replaced by the 4066.)  Making useable analog switches on commodity CMOS process was genius and really opened things up.  Maxim was making DG parts and improved second sources in those pinouts but we really hit something big when we started making good switches in the 4000 series pinouts.  The vectors that you can push on with switches is on resistance, off leakage, capacitance, charge injection and HF isolation and cross talk.  One of the more subtle parameters of switches is Ron flatness- this is the variation in resistance over the voltage range.  Flatness turns directly into distortion- if you lower Ron and improve flatness, you can even drive low impedance loads with low distortion.  We pushed on resistance down really low to fractions of an ohm- customers started using them to switch speaker and headphone circuits rather than using a second amplifier.  This created a huge opportunity in cell phone as customers wanted to do things like use the main speaker as the ringer source but allow you to use headphones too- they couldn't use a simple swtiched source headphone jack.  This jelly bean stuff cuts in a lot of directions and its really fascinating.

I think every engineer should know these jelly bean parts and they should be the first thing they go to in a design.  Only when they prove to themselves that a jelly bean won't do the job should they use something "better".  They and their company will ultimately be more successful and have a lot less headaches.

Once again a thought provoking and well presented topic.
 
The following users thanked this post: thm_w, Fungus, mycroft, 1Ghz, cortex_m0


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf