Audiophile quality for transistors...

[EBK]

Ok, here's a question that has been bugging me.  Why can't these fancy discrete op amps be simply shrunk down and turned into IC op amps instead? Is there something legitimate to their mojo that requires larger parts or wider part spacing?  Is it just for showing off?

Here is something I've found, but it leaves me less impressed than I should be, perhaps:
"Class A biasing and high output current capability coupled with a proprietary compensation scheme requiring multiple NPO dielectric capacitors make these discrete op amps impossible to fabricate as a monolithic IC."

[reddesert]

I'm waiting for someone to build a circuit that approximates a transistor's current amplification, using an arrangement of IC op amps.

IC op-amps are very sophisticated - look at the equivalent schematic of one. IC op amps are/were designed by electrical engineers and semiconductor process experts who are/were bigger wizards than any of us will ever be (Bob Widlar, but also many others), who exploited various features and limitations of semiconductor manufacturing to make differential amplifiers that are linear, stable, resistant to variations in operating temperature and other conditions, resistant to process variation, and so on. They used features such as having the transistors on the same substrate to control variations, that you can't do with discrete circuitry.

Discrete op amps are fun to talk about, like the kit that allows you to build a giant 741 out of individual transistors, but claims for their superiority are made by people who are not fit to carry the slide rules of IC op-amp designers.

[noisette]

Discrete op amps are fun to talk about, like the kit that allows you to build a giant 741 out of individual transistors, but claims for their superiority are made by people who are not fit to carry the slide rules of IC op-amp designers.

I´d say that´s wrong. There is no and will probably never be an off-the-shelf part that can do, what the Jensen 990 can do.
The complexity in integrated opamp design often solves problems, that audio processing doesn´t have in the first place, and is
therefore in that case overcomplexity.  And then there is the "sound" factor which is totally subjective, API2520?). 
IC´s are mostly classAB, discrete can be classA, which will sound...different.       
Of course reddesert has a point there with a lot of BS and badmouthing of ICs going on. Everybody wants a special fetish,
that works better and is more special than everybody else´s.

There´s no such thing as a good or bad sounding transistor in itself, I have seen R.G. again and again explaining why, check the search.

[noisette]

Discrete op amps are fun to talk about, like the kit that allows you to build a giant 741 out of individual transistors, but claims for their superiority are made by people who are not fit to carry the slide rules of IC op-amp designers.

I´d say that´s wrong. There is no and will probably never be an off-the-shelf part that can do, what the Jensen 990 can do.
The complexity in integrated opamp design often solves problems, that audio processing doesn´t have in the first place, and is
therefore in that case overcomplexity.  And then there is the "sound" factor which is totally subjective, API2520?). 
IC´s are mostly classAB, discrete can be classA, which will sound...different.       
Of course reddesert has a point there with a lot of BS and badmouthing of ICs going on. Everybody wants a special fetish,
that works better and is more special than everybody else´s.

There´s no such thing as a good or bad sounding transistor in itself, I have seen R.G. again and again explaining why, check the search.

www.johnhardyco.com/pdf/990.pdf

[R.G.]

Ok, here's a question that has been bugging me.  Why can't these fancy discrete op amps be simply shrunk down and turned into IC op amps instead? Is there something legitimate to their mojo that requires larger parts or wider part spacing?  Is it just for showing off?

Here is something I've found, but it leaves me less impressed than I should be, perhaps:
"Class A biasing and high output current capability coupled with a proprietary compensation scheme requiring multiple NPO dielectric capacitors make these discrete op amps impossible to fabricate as a monolithic IC."

Discrete opamps have some advantages and some disadvantages compared to integrated circuit opamps. In fact, almost all modern solid state audio power amplifiers are in fact discrete opamps. Discretes can have a much higher power supply voltage, and have vastly larger current and power outputs. However, they do not have the temperature stability, low drift, DC accuracy, or other special features that can be implemented with identical transistors closely coupled on a single substrate. There are ways to get around the issues with discretes, but they all involve much more complicated circuitry and much higher costs. It may or may not be possible to equal the noise, dsitortion and/or some other characteristics of IC opamps in a discrete design, but it will be a LOT of work, and may sacrifice other characteristics of the opamp's performance. DC performance and offsets are one common area where discretes must be hand tweaked on every unit, and may not even then be able to compete with tweaking.

Hifi tweakos like discrete opamps and other such things because it adds a layer of complicating complexity to tweak.

[Tony Forestiere]

Hifi tweakos like discrete opamps and other such things because it adds a layer of complicating complexity to tweak.

Hifi tweakos like discrete opamps and other such things because it adds a coating of complicating complexity to contemplate.

Sorry. Was feeling in an alliterative mood and heard some music there. "Complicating complexity" was my Muse. 
Apologies again. Carry on.

[ElectricDruid]

Hifi tweakos carefully choose components to compound the coatings of complicating complexity to contemplate?

Or feasibly philistines forget fine formulas for fitting fidelity?

[antiuser]

Hifi tweakos carefully choose components to compound the coatings of complicating complexity to contemplate?

Or feasibly philistines forget fine formulas for fitting fidelity?

Alliterations are awesomely adaptable and appropriate argument artifices. Approved!

[Tony Forestiere]

"Hifi tweakos carefully choose components to compound the coatings of complicating complexity to contemplate?"

Dang! I bow to ED. 
Well played.

[thermionix]

Alliterations are awesomely adaptable and appropriate argument artifices. Approved!

adds antiuser.

[antiuser]

Alliterations are awesomely adaptable and appropriate argument artifices. Approved!

adds antiuser.

Allegedly acting as an amicable acquaintance, although always ascribing asinine assumptions at audiophiles. Apt at avoiding angry altercations and advancing along an accordingly advantageous ambition.

Ok now this is getting as silly as audiophile snake oil...

[PRR]

> Why can't these fancy discrete op amps be simply shrunk down and turned into IC op amps instead? Is there something legitimate to their mojo that requires larger parts or wider part spacing?

* If you can buy it for 19 cents, it can't be AudioPhile.

* There's a mind-set in IC shops. The high cost of the first wafer means it "MUST" do many jobs with the LEAST amount of Silicon. This leads to devices too small for some audio tasks. The '5532 is an anomaly in that the input devices are "huge" by IC standards (it comes from a troubled history and unexpected uses). Even then it is not hiss-optimum below ~~1K.

However you can buy "large" devices individually. Just a matter of paying for area. Now the bean-counters switch from functions/wafer to selling wafers by the square inch.

* IC devices can be *strange*. If you know devices, you should read Hans' book.
http://www.designinganalogchips.com/
True, ICs have changed, maybe improved, since he worked and wrote. But the improvements have costs and the MBAs have hammers to keep costs down. You can break the "all NPN" rule, but good isolation eats-up die area, reduces chips per wafer.

[wavley]

> Why can't these fancy discrete op amps be simply shrunk down and turned into IC op amps instead? Is there something legitimate to their mojo that requires larger parts or wider part spacing?

* If you can buy it for 19 cents, it can't be AudioPhile.

* There's a mind-set in IC shops. The high cost of the first wafer means it "MUST" do many jobs with the LEAST amount of Silicon. This leads to devices too small for some audio tasks. The '5532 is an anomaly in that the input devices are "huge" by IC standards (it comes from a troubled history and unexpected uses). Even then it is not hiss-optimum below ~~1K.

However you can buy "large" devices individually. Just a matter of paying for area. Now the bean-counters switch from functions/wafer to selling wafers by the square inch.

* IC devices can be *strange*. If you know devices, you should read Hans' book.
http://www.designinganalogchips.com/
True, ICs have changed, maybe improved, since he worked and wrote. But the improvements have costs and the MBAs have hammers to keep costs down. You can break the "all NPN" rule, but good isolation eats-up die area, reduces chips per wafer.

I will point out one upside to the selling of wafers by square inch, those of us doing R&D don't have to run a whole wafer to test a design, sometimes we can get as little as four MMIC chips from a run, several times a year folks do development runs of GaAs, InP, SiGe...  I actually wrote a paper about the SiGe amp I designed for the Phased Array Feed on the Green Bank Telescope, I benefited from a development run of SiGe transistors Caltech/JPL had done. http://iopscience.iop.org/article/10.1088/1538-3873/aa7115  Even so, there is very little being done to develop discrete devices anymore, bare die transistors require a level of assembly skill that isn't really that common of cheap when it comes to GHz and above technology so the industry is moving more and more to MMIC because the technology is now really competitive and sometimes better than discrete chip and wire units and MMIC technology allows us to build multi chip modules that combine things like amps, mixers, AD conversion, fiber transmission, and LO distribution into one module saving on space and power.  Basically, MMIC is doing the same thing to our field that the packaged op amp did 40 years ago, but it wasn't until just recently the noise was good enough to compete with discrete transistors.

[marcelomd]

I posted it somewhere else, but hey...

Nice analysis of a TL084 die: http://www.righto.com/2018/06/silicon-die-analysis-op-amp-with.html

Can you spot the FET inputs?

[stringsthings]

I posted it somewhere else, but hey...

Nice analysis of a TL084 die: http://www.righto.com/2018/06/silicon-die-analysis-op-amp-with.html

Can you spot the FET inputs?

Why would I want to do that?  Is there a prize involved? 

[amptramp]

I used to work in a Remote Sensing department of a company that made rotating infrared scanners for shipboard use to detect incoming aircraft.  One of the better detector formulations for a photoresistive detector for normal ambient temperatures was HgCdTe, mercury - cadmium - telluride which was sensitive to the 8 to 14 micrometre range of normal ambient temperaturres.  The sensors were about 50 ohms output impedance and we found the best transistor for use at this impedance was the lowly 2N4405.  The way you select a transistor is to use noise maps that are specified for some transistors in terms of current and voltage with noise in db but where these are not available, use a transistor where the current noise matches the voltage noise for the circuit impedance you are using.

Another detector formulation was designed for the 3 to 5 micrometre wavelength range which could detect the heated air from a jet engine even when the aircraft was at nose aspect.  This was a photovoltaic detector meaning is was like a solar cell optimized for that waveband.  Since it was photovoltaic, we used op amps configured as transresistance amplifiers where the detector was connected directly to the inverting input of the op amp and the feedback resistor was large, usually 25 megohm low-noise resistors.  These were JFET op amps optimized for minimum current noise since the input impedance was so high.  Having the detector forced to virtual ground gave the best linearity so although the detector was photovoltaic, forcing the voltage to remain at zero gave the most linear video.

[PRR]

> clearly above average silicon transistor sounds.