Help with Discrete Op-Amps for each frequency range.

[KSTR]

Edit: this not an opinion, but a fact proven by measurements. It is impossible to build a discrete opamp which is better than an integrated, because the devices are not matched due to not being on the same die.

This is not correct. One can certainly build a discrete opamp that has better specs than any integrated one suitable for audio. The most critical part, the input transitors or FETs, can be had in dual matched versions, but even with non-matched parts you can get along with manual trimming.

 

[LTig]

You said digital audio was solved 39 years ago, I said what do you mean!

A typo, I meant 30 years.

30 years ago any decent implemented DAC was transparent regarding domestic use, which means they just deliver the analog signal such that nobody is able to hear distortion when playing music, or noise when silent, all this with a perfectly flat frequency response. That's all one needs, and a better DAC with respect to distortion or noise won't sound better.

 

[LTig]

This is not correct. One can certainly build a discrete opamp that has better specs than any integrated one suitable for audio. The most critical part, the input transitors or FETs, can be had in dual matched versions, but even with non-matched parts you can get along with manual trimming.

What about DC offset, if you don't want to use coupling caps in audio applications? And manual trimming runs away with temperature and/or time.

 

[DSJR]

So, but some audio enthusiasts and others say they sound better to the ears, but as you said discrete op-amps can be designed for higher output voltage and/or current, I don't know if that's all, but I'm taking that reason into account and not ruling out integrated op-amps.

Please take in mind the suggestion above regarding 'is better' or 'sounds better.' Many of us old hands here have been duped too many times as to 'sounds better' and had our initial opinions reversed quite often on a different day. Some 'golden ears' have heard a clear A-B difference when absolutely NOTHING was changed. Like 'reformed (now ex) smokers,' we tend to get a bit preachy to those who haven't yet experienced this and insist on audible differences that almost certainly don't exist..

Subjectivist audiophiles don't usually seem to realise that so many senses, moods, health, weather/humidity play a part in our perceptions of 'differences' and our hearing is arguably one of the least important taken as a whole - honestly guv'nor, I ain't making it up

You know, I'd start with a bog-standard and very old-fashioned 'audio' chip (oh I dunno, OPA134 or 5534, the latter all over the place in audio for decades now). Not too wideband (into many MHz) to go unstable as I gather some hugely wideband chips apparently can if inappropriately used and utterly transparent to the audio signals fed through them. Get your circuit stable and working right with these (they're cheap, even 'proper official ones' from the likes of Mouser) and then, if you really want to, you could try replacing them with discrete ones - maybe the extra noise from the latter may appeal, I don't know...

My old Crown IC-150 preamp had initially OPA134's to replace the originals (I forget the originals now) and, after reading Ken Rockwell's review of this fifty year old model, I got some 'pukka' 5534s to replace them and 'swore' I could tell them apart. Like hell I could, and I got myself royally confused as to which pair of chips were fitted (I have two just-about working samples of this preamp and one had one pair of chips and the other the second. I think they now have 5534s in them and I can't be a**ed any more to mess around They easily reproduce recording and production differences in the music I play and the phono stage isn't bad either...

One or two of the electronics engineers here may tell more and with much better first hand technical authority. Just please go into this kind of thing with eyes and mind wide open

 

[KSTR]

What about DC offset, if you don't want to use coupling caps in audio applications?

Parts with low differential offset are available, and the rest can be trimmed.
EDIT: MAT12 for example, 1nV/rtHz voltage noise, and typical offset of 10uV

 

[DVDdoug]

So, but some audio enthusiasts and others say they sound better to the ears

And they (usually) aren't doing proper controlled blind listening tests and they often use meaningless words to describe what they think they hear. And often they can't measure what they think they hear.

There is a LOT of nonsense in the "audiophile" community. This is one of the few rational-scientific audio resources.

MOST electronics is better than human hearing unless there is audible noise (hum, hiss or whine in the background) or unless an amplifier is over-driven into clipping (distortion).

I want something that sounds good and warm and natural,

Again, "warm" is not defined. I used to think it was mid-bass boost but to some people it means slight "pleasing" distortion. So I now try to avoid that kind of terminology unless I explain what I'm talking about. Unless you are adjusting EQ/tone controls, or unless you have a tube amplifier with distortion you won't get "warm" sound. (Not all tube amps have audible distortion and if they do, they aren't all the same.)

And if you're talking about op-amps, the whole idea is high open-loop gain with negative feedback ("corrective feedback") which magically makes everything better! Some "audiophiles" think negative feedback is a bad thing but it's actually a very good thing. It's easy to make a good amplifier with op-amps, especially if you don't need a lot of gain. If you are making a microphone preamp you need lot's of gain so it's more important to choose a low-noise op-amp.

I don't want the usual boring stuff

"High fidelity" means faithful to the original. Yes, that can be boring if the original recording/production is boring, and it can be boring to talk about.

and I also don't want that added colorful digital sound.

Assuming nothing is wrong, digital (at 16-bits, 44.1kHz or higher) is better than anything analog and better than human hearing. ("Better" meaning low, distortion, low nose, and flat frequency response.)

See Audiophoolery.

 

[Robson Santos]

A typo, I meant 30 years.

30 years ago any decent implemented DAC was transparent regarding domestic use, which means they just deliver the analog signal such that nobody is able to hear distortion when playing music, or noise when silent, all this with a perfectly flat frequency response. That's all one needs, and a better DAC with respect to distortion or noise won't sound better.

Yes, I understand, but I see that I will have to look at one like Hypex, Icepower, but they are very expensive compared to Brazil, anyway, but I am researching other cheaper ones that are good, this forum is helping me a lot.

 

[Robson Santos]

Please take in mind the suggestion above regarding 'is better' or 'sounds better.' Many of us old hands here have been duped too many times as to 'sounds better' and had our initial opinions reversed quite often on a different day. Some 'golden ears' have heard a clear A-B difference when absolutely NOTHING was changed. Like 'reformed (now ex) smokers,' we tend to get a bit preachy to those who haven't yet experienced this and insist on audible differences that almost certainly don't exist..

Subjectivist audiophiles don't usually seem to realise that so many senses, moods, health, weather/humidity play a part in our perceptions of 'differences' and our hearing is arguably one of the least important taken as a whole - honestly guv'nor, I ain't making it up

You know, I'd start with a bog-standard and very old-fashioned 'audio' chip (oh I dunno, OPA134 or 5534, the latter all over the place in audio for decades now). Not too wideband (into many MHz) to go unstable as I gather some hugely wideband chips apparently can if inappropriately used and utterly transparent to the audio signals fed through them. Get your circuit stable and working right with these (they're cheap, even 'proper official ones' from the likes of Mouser) and then, if you really want to, you could try replacing them with discrete ones - maybe the extra noise from the latter may appeal, I don't know...

My old Crown IC-150 preamp had initially OPA134's to replace the originals (I forget the originals now) and, after reading Ken Rockwell's review of this fifty year old model, I got some 'pukka' 5534s to replace them and 'swore' I could tell them apart. Like hell I could, and I got myself royally confused as to which pair of chips were fitted (I have two just-about working samples of this preamp and one had one pair of chips and the other the second. I think they now have 5534s in them and I can't be a**ed any more to mess around They easily reproduce recording and production differences in the music I play and the phono stage isn't bad either...

One or two of the electronics engineers here may tell more and with much better first hand technical authority. Just please go into this kind of thing with eyes and mind wide open

Yes, I have an open mind, and I am considering everything that is being said. In short, I agree with what you said. I am taking note of the suggestions. Thank you. It is always good to hear from someone who has experience.

 

[Speedskater]

The only useful place for discrete parts in an op-amp circuit is the output stage to:
a] in effect turn the op-amp into a small power amplifier to drive low impedance loads.
b] with a second power supply, to swing high voltage signals.

 

[Bruce Morgen]

This is not correct. One can certainly build a discrete opamp that has better specs than any integrated one suitable for audio. The most critical part, the input transistors or FETs, can be had in dual matched versions, but even with non-matched parts you can get along with manual trimming.

I wouldn't bet on the folks in today's discrete audio op-amp business doing any of that. Moreover, one would have to hand select every passive component in order to get anywhere the precise values that are routinely achieved by decent silicon foundries.

 

[KSTR]

I wouldn't bet on the folks in today's discrete audio op-amp business doing any of that. Moreover, one would have to hand select every passive component in order to get anywhere the precise values that are routinely achieved by decent silicon foundries.

Well, that might well be true because IHMO those designers of "audio" discrete opamps have much different goals than creating an amplifier with stellar specs on the edge what's physically possible (not that this would be needed for the usual audio tasks, of course). And quite a few simply lack the expertise to do so. But that does not defeat the concept.

Discrete precision resistors and capacitors are not that much of a problem. Also note that absolute value precision of on-chip caps and resistors is usually quite bad, easily 10% or worse. Only matching between same values can be made extremely good. Parasitics of on-chip caps and resistors (voltage and temp coefficients, etc) are also not quite as good as for the best discrete components.

It's only because of the ingeniuity of the designers of monolithic chips that they are able to reach excellent and consistent performance. And just until recently many precision opamp required costly on-chip laser trimming and/or selection and binning of the finished products.

You can rest assured when some of the industry's chip designer giants like Scott Wurcer (rip) were given the chance to do a no-holds-barred discrete opamp the results would dwarf the best monolithic designs he did, like the legendary AD797. Discrete design has far fewer constraints than monolithic, that's were the real advantage is.

What counts as "discrete" is another question not so easy to answer. Anything what is not comprised of just one single monolithic silicon chip? Do composite designs, like using two monolithic chips in a master-slave design count as discrete? Or do we restrict ourselves to only low/zero-integration basic level parts and semiconductors?

 

[LTig]

What counts as "discrete" is another question not so easy to answer. Anything what is not comprised of just one single monolithic silicon chip? Do composite designs, like using two monolithic chips in a master-slave design count as discrete? Or do we restrict ourselves to only low/zero-integration basic level parts and semiconductors?

I vote for the latter.

 

[AnalogSteph]

Discrete opamps may have an advantage if you need any of the following:
1. Operating voltage exceeding +/-15-18 V
2. Major output driving capability
3. Ultralow voltage noise
Though the latter two are also achievable by augmenting IC parts with additional surrounding circuitry like buffer stages or discrete input stages.

In the context of an active speaker design, I don't see why anything overly exotic would be required. (I can pretty much guarantee you that among the best-measuring DACs on Amir's graph, almost all are going to be using off-the-shelf IC opamps. Often OPA1612s.) You probably won't find anything fancier than NJM2068s (NJM8068s) in the vast majority of off-the-shelf examples (maybe the odd OPA1678 or something these days), and the balanced input may still be handled by a TL071. It's all shockingly utilitarian, generally. Integration, reliability, power consumption and BOM cost tend to be important considerations aside from circuit performance. The latter only needs to be good enough for the task at hand.

 

[KSTR]

As for "suitable" standard opamps besides the NE5532 industry standard, OPA1602 is one my favorites (very good, and affordable), and it does pretty well in most crossover/filter type of circuits. For really high circuit node impedances OPA1642 is better suited, though.
And for quite some time now we have the OPA1656 (and its twin-brother, the OPA2156) which are also pretty much excellent for almost everything.

 

[GXAlan]

So, in your opinion, integrated op-amp (IC) is more superior than discrete op-amp?

Think about discrete circuits for computers. Stuff like

EDVAC - Wikipedia

en.m.wikipedia.org

Versus integrated circuits.

There is no doubt that ICs can be superior because it takes $1B to build the factories that make ICs.

Discrete amps can be tuned for sound, which may not be transparent but somehow occasionally different.

If you want to tune the sound, discrete op amps may introduce non linearities that are difficult to predict. Personally, you are likely better off going for something like the TA-ZH1ES as are processing box or a Marantz SA-10 to impart a subtle sound to your audio. Not everyone will agree that the SA-10 has a “sound” but I have tried showing some PKMetric differences.

In any case, as you recognize, speakers and room make the biggest difference in sound.

 

[Cbdb2]

The most critical part, the input transitors or FETs, can be had in dual matched versions, but even with non-matched parts you can get along with manual trimming.

You cant seperate the LTP from the rest of the input circuit. All the current mirrors and current sources consistancy depends on matching junction temperatures as well as parameters. The IC is obviously better at this. You say laser triming is expensive, its still orders of magnitude cheaper than hand trimming each OA.
The cost of R&D for ICs must also be orders of magnitude higher, and the designers are the cream of the crop, TI, ADI, etc. have thousands of EEs to choose from and the teams (not one or two guys) they put together have experts in ALL of the different specialties needed.