Discrete Components Vs. Analog Integrated Circuits

[living sounds]

Do a search, this question has been covered before.

[pointsource]

Originally Posted by living sounds

Do a search, this question has been covered before.

Sorry, I used the serach engine, always do that. Maybe I'm doing it wrong? If this questions has been answered you could point me to the thread please.

[living sounds]

Here's one:

op amps - IC's or Discrete or "real" whats better?

[tINY]

There are more ways to screw up discrete designs, but people get lazy when using integrated solutions.

Each design is really it's own beast.

If you are building something where you want distortion or "warmth" (or "sparkle" "focus" etc), there are big advantages to discrete topologies. So, in things like guitar amps and mic pre-amps tend to use more discrete designs.

When it comes to power amps, discrete designs can be made more reliable.



-tINY

[mjrippe]

From your "oscillators or filters" I am going to venture a guess that you are talking about analog synths. If this is the case, many different circuits have been used and all have their ups and downs. A fully discrete oscillator *may* have more drift and inaccuracy (or warmth and organic sound, depending on your ear). For a filter, the analog fiends may tell you that a discrete one has more balls or better resonance. In truth, it is what you are looking for that matters. Don't buy the hype, buy what sounds good to you.

[pointsource]

alright guys, thx fot all inputs

[Frank_Case]

Originally Posted by pointsource

I was wondering the difference between using discrete components and IC's (appart from size and lower cost).

Here's some reasons people still build with discrete components...

(1) Discrete components can run in class A Single Ended (or class A PP for that matter) more easily, especially in the final audio output stage where a SS device can get quite hot. Class A sounds different, always runs hotter, and some subjectively say sounds better. On the other hand, opamps tend to run in class B or class AB just because they don't have the ability to dissipate heat as well and have to be designed to run cool due to their small size.

(2) Discrete can be designed to operate off higher voltages. Anybody here wanna try 100v+ PS rails? Then go discrete.

(3) Discrete can be made to deliver higher output current depending on the device, which often means lower output impedance under some circumstances. It's a power dissipation issue again.

(4) Opamps often sound different than discrete for the following reasons: opamps use (a) extensive internal feedback, (b) internal current mirrors for biasing, (c) internal differential amplification, (d) internal SS current source devices as substitutes for internal resistors in the chip, (e) internal direct coupling between stages to eliminate coupling capacitors within the chip, among other things. Because of this, some opamps can sometimes sound a bit clinical and boring. When you design with discrete, you tend to use more resistors and capacitors in the circuit which add their own sound.

(5) Specialty discrete bipolar transistors can be made to have lower actual base resistance, referred to as R(bb), not to be confused with intrinsic emitter resistance R(e). This means that discrete devices can be made to have less Johnson noise and therefore can be quieter if chosen correctly.

(6) Opamps are designed to objectively meet spec in terms of low THD and noise and are not necessarily designed to subjectively deliver the best sound. Some opamps sound great, others sound sterile. Some opamps sound bright, others sound dark. With discrete, designers can tweak the subjective nature of the sound to a finer degree since they have more control over all parameters.

Of course there are exceptions to each of these points. Some opamps sound quite good. Listen to a Neve 1073 preamp which runs in class A discrete, and then listen to an opamp based pre, and ask yourself which you like better. This is not a trick question because there is no real answer. You will like what you like. But regardless, some of the difference in tone will be because of the discrete Class A circuitry in the 1073.

[Matt Syson]

Hi
ICs have limitations of power dissipation and 'by design' voltage ratings. They COULD be designed with higher supply rails but there is insufficient market for them so you don't see them.
As with all things you can design a 'good' sounding piece of gear using either 'technology' and equally well, a bad sounding piece of gear. Even using the SAME components but a different physical layout will affect things so without considerable experience it is impossible to tell just by looking.
Matt S

[JohnRoberts]

Originally Posted by pointsource

Hi,

being a complete newb to electronics, I was wondering the difference between using discrete components and IC's (appart from size and lower cost).

For example, building an oscillator or filter, there will be any difference on the sound quality? Cause I see some advertise like "100% discrete components".

Thx

This question is confused by a broader debate or argument over which is better. Your question A) is there any difference? and B) (unasked) why are they advertising use of discrete?

A) In general there doesn't have to be a difference for most applications, but since ICs can usually squeeze the equivalent of hundred or thousands of discrete devices in a single chip. IC designs will generally be more linear or accurate.

B) Advertising discrete components is implying, or perhaps overtly claiming superior performance. This is not supported by the science, so caveat emptor.

Ironically the older low cost products were based on discrete components because they were cheaper, nowadays ICs are cheaper because the labor to insert the parts can be more than the cost of the parts, so one pop is cheaper than 10x or 20x pops for same functionality.

JR

[Matt Syson]

Hi
If you are constructing something like an oscillator or filter which has various constraints on it (defined overall gain (typically about 3) and frequency response) it is likely that swapping circuitry will not have anywhere near as much influence on it as a 'flat, wideband' amplifier stage.
Comparing a 'class A' 1073 and an op amp based circuit is too far fetched as they are different animals and not because one is 'class A' and the other AB or any other permutation. There is the amount of feedback applied, and WHERE it is applied and the internal gain structures, all have vast implications on the 'sound', likewise layout and the presence of a transformer or two make it a comparison between 2 dissimilar things.
Most people 'like' audio (and other sensations) that 'gradually tail off at the extremes' rather than sawing your head off at 21KHz as it is a natural phenomenon. Continued 'squalking' is just tedious.
As John said, advertising on the basis that so many don't actually understand the real processes. Digital broadcasting is supposed to be 'fabulous' according to some (with vested interests) but so far it is 'poo' and not up to the perceived quality of decent FM radio.
Matt S

[JohnRoberts]

Originally Posted by Jim Williams

A discrete transistor is also a chip. It's made the same way, a disc of silicon. The only real difference is if you have one or more transistors on that die.

?? not exactly. While many modern small signal transistors use similar technology for mass production, there is a distinction for "integrated" circuits where resistors, multiple semiconductor devices arranged to perform specialized functions, and wiring interconnects are all "integrated' into a printing press like master design that can be stepped and repeated and mass produced.

You can by IC's with two transistors matched on one die or thousands of transistors (millions in a modern CPU chip). Both are considered integrated circuits.

Indeed, but the highly parallel IC as transistor (like old LM394), or simpler 4x or more transistor array, is in fact a special class of "integrated" circuit.

You can also buy high voltage IC opamps from companies like Apex, they have opamps running at + - 150 volts available. Some have amps of output current too, a power device like the old LM12.

Some of the old high power opamps used "hybrid" technology where multiple IC and/or discrete device dies were bonded to one common substrate and wired together using bonding wires, usually used to attach die pads to package leads. Kind of like a tiny PCB inside a package.

The major IC manufacturers also developed high voltage IC process technologies where isolated regions of the chips could tolerate much higher voltage than the rest of the die (like old power amp front end ICs). These could handle the voltage but not the current and power of full discrete designs.

You can have a chip with one or two transistors, or thousands. You can have discrete or integrated opamps running up to hundreds of volts.

Yes, those old lines are blurred these days.

The lines are still there, while all transistors made for the last several decades use similar technology to ICs. Very very old transistors (contact junctions) were quite different, but that is TMI for this discussion.

JR

[zvukofor]

Originally Posted by Jim Williams

I've never heard of any euphoria from analog chip designers. Mostly, they set a design perameter and get handshakes when they achieve it. Maybe the team will go out and get a beer.

They don't amuse themselves thinking like an audio guy. They are not interested in audio fashion or style. They are EE's out to manufacture a product that meets the design criteria.

They leave the audio permutations of errors to the users.

I might say this is the best answer to this question ever. May i quote it sometimes when asked about discrete vs Ics?

[gman76]

Originally Posted by JohnRoberts

?? not exactly. While many modern small signal transistors use similar technology for mass production, there is a distinction for "integrated" circuits where resistors, multiple semiconductor devices arranged to perform specialized functions, and wiring interconnects are all "integrated' into a printing press like master design that can be stepped and repeated and mass produced.

Indeed, but the highly parallel IC as transistor (like old LM394), or simpler 4x or more transistor array, is in fact a special class of "integrated" circuit.

Some of the old high power opamps used "hybrid" technology where multiple IC and/or discrete device dies were bonded to one common substrate and wired together using bonding wires, usually used to attach die pads to package leads. Kind of like a tiny PCB inside a package.

The major IC manufacturers also developed high voltage IC process technologies where isolated regions of the chips could tolerate much higher voltage than the rest of the die (like old power amp front end ICs). These could handle the voltage but not the current and power of full discrete designs.


The lines are still there, while all transistors made for the last several decades use similar technology to ICs. Very very old transistors (contact junctions) were quite different, but that is TMI for this discussion.

JR

The LM12 could not run at rails greater than +/-35V, as JR implied. Its power was primarily limited by the package, TO3 case. I did apps engineering on this device many yrs ago. Blew one the leads off after trying to drive a nasty inductive load. That's the thing about linear power circuits - the output transistors need to satisfy the case where your current/voltage is 90deg (or more) out of phase, which means the top or bottom must stand off >60V while driving amps of current. Widlar did quite a novel thing at the time, sense thermal runaway right in the output stage and shut down the base drive immediately (well 10us or so). The claim was that you could short the output to ground or one of the power rails, and the chip would survive. Quite impressive at the time. I remember taking a screwdriver to the test board and looking away - didnt want to get shrapnel in the face/eyes. And yes, it was monolithic, not hybrid.

Discrete power amps have the luxury of a couple things - spreading out the heat over many power devices (lower operating junction temp) and having fewer interactions between transistors (very low parasitics). To stand off high voltages, you need deep junctions (special process).

[Jim Williams]

Apex also manufactures a large line of power opamps. Some run on + - 150 volts to drive servo systems. I wouldn't consider any of them a good choice for audio.

[gman76]

Originally Posted by Jim Williams

Apex also manufactures a large line of power opamps. Some run on + - 150 volts to drive servo systems. I wouldn't consider any of them a good choice for audio.

Do u mean their class-D parts? Dont think I'd use classD for full range apps. Apex does build some power amps for the industrial market that could be used for audio, but the price of say a PA05 on digikey is over $400. Yikes!

[Jim Williams]

Those are high voltage bipolar opamps. They have made them for years. Not class D.