This is the crux of it.
Op amps are designed to have many uses and most conform to similar design standards (30V split supply, A/B output stage, etc). When used within their parameters they work very well, but they are limited.
Do you need an opamp that will source/sink 100 mA at 20 volts p-p? Build a discrete opamp. How about an opamp that can deliver 80 volts p-p? Build a discrete opamp. Do you want a class A output? You get my drift.
Is it really justified to build a discrete op amp for a unity gain tone control circuit or a low current buffer? If you think so, please state why.
Virtually everything built after 1990 that I have repaired or stripped for parts has at least one 5532 (cheapies have TL072 instead).
While I agree this might be a factor in some cases, two of the opamp-based preamp designs I'm talking about are in the public domain and I'm quite sure are competently designed and implemented: the D. Self Precision Preamp '96 (EW&WW 1996, built on the beautiful Signal Transfer Co PCB), and the Silicon Chip Studio Series Preamp (Silicon Chip magazine, October 2005). I do not have access to AP gear at the moment, but when I built the SC preamp I measured the THD+N curve using a Neutrik A1 and it was below the noise floor (0.001%) at 1V in and out from 20Hz - 20kHz so I see no reason to think it wasn't working properly. I completed the D. Self preamp recently and no longer have access to any audio-specific test gear, but there are no signs of oscillation or visual waveform distortion using a 200MHz scope.
I did "single blind" testing of both opamp preamps against a discrete design by getting my wife to connect them up through another (RC) preamp's input and tape output section so I didn't know which was which. I knew straight away which was which - it was obvious on most tracks I listened to, and I confirmed I was right by checking the connections afterwards. To make it a true double blind test would involve some seriously tedious listening sessions and I really don't see why the above test isn't adequate to prove there is an audible difference (even though not everyone might hear it).
The point is, this thread was started to find out which are the better sounding discrete opamp designs. If you happen to BELIEVE that opamps are always better because your measurements say so and that listening is a waste of time because our hearing is too unreliable, good luck to you. Carry on designing equipment that is technically competant or whatever it is you do that makes you so arrogant. But please don't waste the time of those that are here to build equipment that improves our enjoyment of music. This pursuit should not be mutually exclusive with the quest for technical perfection, but by insisting on this argument you are tending to make it so, rather than trying to find out the reasons why the usual measurements don't relate well to what some people hear.
I did "single blind" testing of both opamp preamps against a discrete design by getting my wife to connect them up through another (RC) preamp's input and tape output section so I didn't know which was which. I knew straight away which was which - it was obvious on most tracks I listened to, and I confirmed I was right by checking the connections afterwards. To make it a true double blind test would involve some seriously tedious listening sessions and I really don't see why the above test isn't adequate to prove there is an audible difference (even though not everyone might hear it).
The point is, this thread was started to find out which are the better sounding discrete opamp designs. If you happen to BELIEVE that opamps are always better because your measurements say so and that listening is a waste of time because our hearing is too unreliable, good luck to you. Carry on designing equipment that is technically competant or whatever it is you do that makes you so arrogant. But please don't waste the time of those that are here to build equipment that improves our enjoyment of music. This pursuit should not be mutually exclusive with the quest for technical perfection, but by insisting on this argument you are tending to make it so, rather than trying to find out the reasons why the usual measurements don't relate well to what some people hear.
You believe that it somehow makes a profound and positive difference the scale something is built at? (IC opamps arent really very different, many designs will be bread-boarded with discrete parts first, before committing to manufacture).
the problem is, that in most cases the likely cause (apart from psychological factors...) for people preferring discrete, is the higher distortion, or different distortion profile, yet the anti IC brigade very rarely accept this may be a factor. where the friction arises is the much less likely theory that it must be some mysterious never before measured by anyone in history, technical advantage. the idea that some special/exclusive hi-end audio transparency exists, that somehow dodges analysis designed for more demanding, critical wider bandwidth signal applications. basically its an I want my cake and eat it too thing
also choosing to believe that essentially perfectly measuring devices (not just THD+N) can somehow produce abhorrent non-linearities that are obviously audible.
these beliefs are contrary to any design technique, discrete or IC based. building things discretely doesnt mean you shouldnt have to look at a meter as well as listening, to make sure it matches your design goals
its funny that you used THD+N alone as an example above to prove that it was behaving correctly....
the problem is, that in most cases the likely cause (apart from psychological factors...) for people preferring discrete, is the higher distortion, or different distortion profile, yet the anti IC brigade very rarely accept this may be a factor. where the friction arises is the much less likely theory that it must be some mysterious never before measured by anyone in history, technical advantage. the idea that some special/exclusive hi-end audio transparency exists, that somehow dodges analysis designed for more demanding, critical wider bandwidth signal applications. basically its an I want my cake and eat it too thing
also choosing to believe that essentially perfectly measuring devices (not just THD+N) can somehow produce abhorrent non-linearities that are obviously audible.
these beliefs are contrary to any design technique, discrete or IC based. building things discretely doesnt mean you shouldnt have to look at a meter as well as listening, to make sure it matches your design goals
its funny that you used THD+N alone as an example above to prove that it was behaving correctly....
Last edited:
Nope, I don't believe that at all - hope you don't either. Scale probably has very little if anything to do with it. I seriously doubt that semiconductor companies breadboard their latest opamp designs with discrete devices before committing to fab - but I don't know and unlike some am not about to pretend I'm an expert on things I know very little about. Besides, the topic under discussion (or supposed to be) is how they sound, and I'm just stating what I think - makes no difference what arguments are presented as to why ICs are or aren't better to my ears.
I utterly reject the suggestion and generalisation that discrete designs only sound better because of higher distortion levels - I have measured the distortion and examined the residual of one of the discrete designs I referred to (similar to D.Self's published discrete opamp) several years ago on some borrowed gear and at 10kHz the only harmonics visible above the noise floor were 2nd and 3rd at 0.0005% and 0.0003% respectively - hardly less linear than any opamp is likely to be in a series feedback line preamp circuit. Besides, even if discrete is preferred by some because of higher or different distortion profiles, so what? Why do you feel the need to criticise this?
Re your comment on my use of THD+N to verify that the preamps were working properly - are you just missing the point and being a smart a** or are you suggesting that a badly designed circuit that was oscillating could give <0.001% THD+N across the board and I should have measured something else? If so what? The designers of those preamps seemed to have relied on THD+N pretty heavily... please tell us your "secret".
You're making plenty of assumptions and speaking as though you are an expert in these areas. Even if you are it wouldn't hurt to try a little more humility and civility in future.
I utterly reject the suggestion and generalisation that discrete designs only sound better because of higher distortion levels - I have measured the distortion and examined the residual of one of the discrete designs I referred to (similar to D.Self's published discrete opamp) several years ago on some borrowed gear and at 10kHz the only harmonics visible above the noise floor were 2nd and 3rd at 0.0005% and 0.0003% respectively - hardly less linear than any opamp is likely to be in a series feedback line preamp circuit. Besides, even if discrete is preferred by some because of higher or different distortion profiles, so what? Why do you feel the need to criticise this?
Re your comment on my use of THD+N to verify that the preamps were working properly - are you just missing the point and being a smart a** or are you suggesting that a badly designed circuit that was oscillating could give <0.001% THD+N across the board and I should have measured something else? If so what? The designers of those preamps seemed to have relied on THD+N pretty heavily... please tell us your "secret".
You're making plenty of assumptions and speaking as though you are an expert in these areas. Even if you are it wouldn't hurt to try a little more humility and civility in future.
pot kettle?
nope i'm not an expert, not by the measure of this forum
nope, I applaud this if its admitted, you really oughta read more carefully before reacting, starting to type in anger (I can just see the keys being struck with resolve) without digesting whats written fully.
low order distortion products can be quite pleasing and there is nothing wrong with that if thats what you prefer, I like it to generally come in the music signal rather than the reproduction chain, but thats just my preference. people like what they like, theres no accounting for taste and there is no right or wrong taste. thats called being a realist who knows what they like and that is perfectly acceptable, even admirable.
the problem is it almost never is admitted, people usually view this as a negative, like theyve lost somehow; even when at the same time spinning the party line that THD and measurements dont really matter that much to them. I notice you keep on pulling out THD+N numbers and I havent mentioned it once.
usually, like yourself people prefer to think there is some sort of mysterious technical advantage or mechanism that means they have been very smart in choosing and building discrete/elite and possess items that spew forth unheard of musical wonders that white bread opamp lovers can never dream of.
I have no preference one way or the other, it depends on the application and my mood. discrete can be wonderful and is fun to build, somewhat easier to tweak for specific purposes etc, but there is absolutely no advantage to being discrete vs being an IC except maybe power handling; depends, even that could be matched.
if the same circuit is adapted to being made lithographically it will have equal or better performance due to tighter device coupling, shorter signal paths and better thermal matching/coupling
You are the latter type I mentioned by the sounds of it, you seem to believe there is some magical unknown factor that nobody knows about and nobody has measured, a special thing that relates to listening to audio and resists all attempts to uncover it...
well guess what, often they do indeed breadboard designs for the simpler things like opamps. Scott Wurcer, Jim Williams (RIP) to name a couple, have indicated such. Scott is a member here, so I dont need to be an expert, he has indicated he does such things in his thread about his discrete opamp design, development. Scott designed the legendary AD797 among others and he has been documenting the design process for a discrete opamp design here on the forum. I have been following it with interest, I like discrete opamps too.
ICs are simply collections of discrete components etched into thin film and then fabricated. making a semiconductor master die is not cheap and simulation, though very good now, may not tell the whole story, so yes sometimes they will build a prototype with discrete parts to test some variations and see if they are on target. Simulation and the device models were not always so good as it is now, of course they built them up discretely...
what do you think ICs are made of if not collections of very small transistors, fets, resistors etc? have you looked at the simplified schematics in datasheets?
the civility starts to wear out after a while of reading the same anecdotes and assumptions and by extension insults that anyone happy with a well designed opamp based circuit is listening to trash and simply doesnt know it, so obvious are the clearly audible issues
Last edited:
I think this is entirely possible, but would generally be the result of environmental or implementation issues.
For example a chip that measures near perfect under ideal lab conditions may react very badly to the slightest whiff of RFI at it's input. Too bad for you if you live in a neighborhood frequented by radio taxis or truckers with CB radios. The sound quality of the preamp (or whatever you used the chip for) would be genuinely degraded at your house and the problem may be easily measured at your house. Send it to someone else for testing, and they won't find a problem.
Another problem can arise with chips that have very low output stage quiescent current. The chip may offer 0.0000x distortion under ideal conditions, but it's injecting half-wave rectified signal current into the supply rails. Depending on layout, grounding, supply decoupling etc, some of that may find it's way to ground or the signal path, causing audible distortion.
That would be easy to measure but the point is that the distortion of a preamp may be orders of magnitude greater than what's claimed in the datasheets of the devices it uses.
Then there's stability. Again, a chip that measures near perfect under ideal conditions won't sound good (or measure good) when it's oscillating. Fair enough to say that avoiding oscillation just requires competent design and good layout, but sometimes the challenges aren't trivial. for example the TPA6120A2 is absolutely fiendish in this regard, IMO. Then there's Mooly, demonstrating in his thread on the subject that sometimes problems can arise even with a circuit that should theoretically be stable.
Bottom line for me is it's no surprise that some people can get good results with one chip but not another. Not everybody has advanced test equipment and a PHD in layout, grounding etc.
hey in that line i'm not just talking about the devices in isolation, sorry I can see how my wording might lead you to that misunderstanding and I completely agree with you regarding implementation or environmental effects being unpredictable (but completely measurable).
What I mean is essentially perfect (for our purposes) measuring finished designs using IC opamps that are described/maligned as being obviously flawed, despite measuring to their near perfect standards.
often the word 'fatiguing', or 'hyped' is used, perhaps because its sufficiently vague and has a somewhat emotional link... (yes i'm pretty jaded, cynical about this by now)
one area of course that can create will be external loading via the speaker and its interaction with the electrical damping, but again this is user error, nothing to do with the opamp being an IC vs discretely made and easily measured.
What I mean is essentially perfect (for our purposes) measuring finished designs using IC opamps that are described/maligned as being obviously flawed, despite measuring to their near perfect standards.
often the word 'fatiguing', or 'hyped' is used, perhaps because its sufficiently vague and has a somewhat emotional link... (yes i'm pretty jaded, cynical about this by now)
one area of course that can create will be external loading via the speaker and its interaction with the electrical damping, but again this is user error, nothing to do with the opamp being an IC vs discretely made and easily measured.
Godfrey, these are all good points, but I wonder, should not the same arguments and caveats be made for discrete circuitry?
Discrete circuitry generally is larger and has larger loop area in the circuit thus more susceptability to things like RFI.
They also generally have less thermal/gain/etc stability so may become unstable at shifting conditions that would not occur in an IC.
I guess the bottom line is that both IC and discrete designs will only be successful if done competently.
jan
Discrete circuitry generally is larger and has larger loop area in the circuit thus more susceptability to things like RFI.
They also generally have less thermal/gain/etc stability so may become unstable at shifting conditions that would not occur in an IC.
I guess the bottom line is that both IC and discrete designs will only be successful if done competently.
jan
You took the words right out of my mouth. Any circuit that picks up RFI and "rectifies the signal current"
is poorly designed and implemented, discrete or not.I have a hardwired EQ circuit wired into the tape monitor of the reciever in my automotive/metal shop. All kinds of power tools are used every time I work in there. Sometimes they're plugged into the same circuit as the audio system. Although the circuit is a simple active filter with a couple of 5532s
in a small plastic box, and although it has 47K input impedance, the operation of this homemade circuit is completely unobtrusive. You can't hear hash from the bench grinder right next to it, and you can't hear pops or RFI coming in through the power line. You would never know it's there.OK, we're on the same page then - no argument.
Maybe so but now they're advertising the dang thing as a headphone amp (see link in previous post).
I first heard about it in a thread on another forum:
step 1: Noob wants to build a simple headphone amp.
step 2: Someone suggests TPA6120.
step 3: Noob is delighted, thinks TPA6120 looks ideal.
Rest of the thread was like watching a train wreck in slow motion.
Nope - I don't, nor have I noticed a single post here from someone in favour of the SQ of discrete making those sort of remarks. Seems to be you that's thinking that way.
Even Douglas Self would disagree with that statement according to his section on discrete design in Small Signal Audio Design. Sounds as though you think you know more about it though.
I don't need to know much about IC design to know that this is a ridiculous generalisation - it would obviously depend on the circuit design and implementation.
A repetition of your previous nonsense with no basis. If anyone is being closed-minded about it....
Well I don't see either of those members here condescendingly correcting our delusions and dismissing our OPINIONS on what sounds good. Interested to hear this if it's true - I doubt it is always the case though and probably not in the direct way you were saying. I suppose Intel breadboard their latest CPU designs using billions of discrete fets too?
Really? I would never have imagined it. I thought they were full of smoke, as it seems to escape when you short the pins...
What was that about pot-kettle? It would appear that all the assumptions here are yours. If you're so sure discrete can never be better than IC and like to generalise everything, good for you - I envy you, life must be much simpler. But why do you feel the need to spoil what could have been an interesting discussion with all this twaddle? Do you really think anyone who prefers the sound of the discrete designs they've tried to the opamp designs they've tried is going to be the least bit influenced by your rude manner, sloppy reasoning and ridiculous assumptions?
yeah I know, unfortunately, set up well, things designed to drive cables, usually make pretty good headphone amps too, but necessitate wide bandwidth, so care needs to be taken.OK, we're on the same page then - no argument.
Maybe so but now they're advertising the dang thing as a headphone amp (see link in previous post).
I first heard about it in a thread on another forum:
step 1: Noob wants to build a simple headphone amp.
step 2: Someone suggests TPA6120.
step 3: Noob is delighted, thinks TPA6120 looks ideal.
Rest of the thread was like watching a train wreck in slow motion.
owdeo, qusp
It pains me to see such two good friends of mine hacking at each other in such a fashion.
I can think of 2 advantages discrete components have offhand. In the current marketplace. Voltage and low noise. You can argue about the low-noise designs given that there is some room for paralleling opamps, but most people, even D.Self, will turn to discretes when designing an MC pre. And you won't be building any Stax headphone drivers using ICs, unless you're thinking of driving an OPT backwards.
owdeo, that is a low motivation to ascribe to an engineer. And this 'totally kills my enjoyment of music played through it' is pretty strong, too strong for most of us. Perhaps you could moderate it to... 'I can always identify it'? I mean, given that most of us once enjoyed AM radio and some of us go back as far as clockwork phonographs. Which I can tell you were prized posessions in some cases.
I'm puzzled as to exactly what's going on here. I used to think that you were an unashamed self-publicist and celebrity-worrier. But since you insist that you have taken 'nearly-blind' tests, I'm forced to wonder if maybe you aren't detecting the preamp reliably. Obviously though, given the seriousness of the issue, and the time given over to it's discussion, some more rigorous testing is becoming a necessity in order to move the conversation forward. Given the undercurrent (hardly an undercurrent) of hostility you have displayed toward designers, at least some of them that is, I'm not entirely sure that your dislike of the Self preamp has more to do with the origins of the design than the sound of the amplifier, so I'd suggest that merely identifying it would be a good first step.
Given, however, that you are convinced that you can detect the amplifier reliably, I would like some more substantial evidence that you can, in fact, do so. And if you can, I would like to have it identified how you can tell the difference, whether it be due so some malfunction either of the amplifier or of those to which you are comparing it, or some other factor. Is there any possibility that you could arrange for some more substantial evidence that you can, at least in the first instance, reliably distinguish between this amplifier and another amplifier of your choice? Obviously you would need to choose a second amplifier which informed opinion would reasonably expect on the basis of measurement to be indistinguishable from the amplifier in question.
Is there any possibility of your undertaking such an exercise?
It pains me to see such two good friends of mine hacking at each other in such a fashion.
I can think of 2 advantages discrete components have offhand. In the current marketplace. Voltage and low noise. You can argue about the low-noise designs given that there is some room for paralleling opamps, but most people, even D.Self, will turn to discretes when designing an MC pre. And you won't be building any Stax headphone drivers using ICs, unless you're thinking of driving an OPT backwards.
owdeo, that is a low motivation to ascribe to an engineer. And this 'totally kills my enjoyment of music played through it' is pretty strong, too strong for most of us. Perhaps you could moderate it to... 'I can always identify it'? I mean, given that most of us once enjoyed AM radio and some of us go back as far as clockwork phonographs. Which I can tell you were prized posessions in some cases.
I'm puzzled as to exactly what's going on here. I used to think that you were an unashamed self-publicist and celebrity-worrier. But since you insist that you have taken 'nearly-blind' tests, I'm forced to wonder if maybe you aren't detecting the preamp reliably. Obviously though, given the seriousness of the issue, and the time given over to it's discussion, some more rigorous testing is becoming a necessity in order to move the conversation forward. Given the undercurrent (hardly an undercurrent) of hostility you have displayed toward designers, at least some of them that is, I'm not entirely sure that your dislike of the Self preamp has more to do with the origins of the design than the sound of the amplifier, so I'd suggest that merely identifying it would be a good first step.
Given, however, that you are convinced that you can detect the amplifier reliably, I would like some more substantial evidence that you can, in fact, do so. And if you can, I would like to have it identified how you can tell the difference, whether it be due so some malfunction either of the amplifier or of those to which you are comparing it, or some other factor. Is there any possibility that you could arrange for some more substantial evidence that you can, at least in the first instance, reliably distinguish between this amplifier and another amplifier of your choice? Obviously you would need to choose a second amplifier which informed opinion would reasonably expect on the basis of measurement to be indistinguishable from the amplifier in question.
Is there any possibility of your undertaking such an exercise?
Last edited:
the subtlety that was missed by both if you, was that I was not speaking of only the products in the marketplace, but the technology itself. if, someone with sufficient budget actually set out to make an IC to do either of those things, I feel very confident it could be made to happen. noise is often higher than a single pair of jfets for a phono pre of course, there is more series devices in opamps and they have supporting circuitry inside for other purposes than keeping noise the absolute lowest, they maintain their low noise over wider bandwidth and with greater stability vs environmental conditions too.
also if you read an earlier post, (one that was also objected to), I admitted that voltage and power were current limitations of available devices.
the actual dies in discrete devices are made using much the same technologies these days and are tiny, what gives the power handling and adds the bulk is the packaging to get rid of the heat and stabilize tempco somewhat. you just need to look at the latest run of GaN SMD transistors (fets, discrete fets) for power conversion and class d output, they are tiny little passivated die packages 0.9 x 1.7mm that will handle 200V (with 600V devices coming online shortly) that will provide 6A with short duty cycle, but pretty limited by the package for linear.
Last edited:
The problem with monolithic opamps isn't the active devices, but the passives. There's no way to fabricate a resistor on the die that can match the linearity of the cheapest 0.125W axial metal-film resistor that you can lay your hands on. There are many reasons for this, most of them unchanged since the '60s, but basically it's hard to make a controlled thickness metal-film of sufficiently small size that can be thermally isolated from adjacent circuitry on a monolithic die. This has been trivial for decades on a discrete PCB. It's also been hard to make a purely ohmic (non-rectifying) contact between metal and poly (the bulk resistance material) on die. You're also limited in the range of resistance values available to a few kohms, but it's possible to design around that, as Widlar demonstrated in the '60s.
Then we come to capacitance - limited to a few 10s of pF with oxide dielectric, nowhere near as linear or thermally stable as the cheapest Chinese rare-earth C0G/NP0 discrete ceramic capacitor. Don't even think about high-quality plastic films like polypropylene, polystyrene or Teflon - impossible on die.
Finally, inductances are limited to a few nH on die and will mutually couple with everything else nearby, so they're rarely used. That's not that much of a problem, since inductances are rarely used in discretes also. But when they are used, like in the JE990 input stage, they're indispensable.
There are ways to make better resistors on die, but they're too expensive to be used in commodity monolithic opamps. You may find them on MMIC ICs and other high-performance niche devices that cost upwards of $100 per part.
On the subjective front, anybody who has heard even a simple discrete with say a dozen actives, isn't likely to prefer a jelly-bean monolithic over it. The situation is slightly different with a well-designed circuit using a premium opamp like an OPA627, but that's about as hard as designing a fully-discrete circuit.
Then we come to capacitance - limited to a few 10s of pF with oxide dielectric, nowhere near as linear or thermally stable as the cheapest Chinese rare-earth C0G/NP0 discrete ceramic capacitor. Don't even think about high-quality plastic films like polypropylene, polystyrene or Teflon - impossible on die.
Finally, inductances are limited to a few nH on die and will mutually couple with everything else nearby, so they're rarely used. That's not that much of a problem, since inductances are rarely used in discretes also. But when they are used, like in the JE990 input stage, they're indispensable.
There are ways to make better resistors on die, but they're too expensive to be used in commodity monolithic opamps. You may find them on MMIC ICs and other high-performance niche devices that cost upwards of $100 per part.
On the subjective front, anybody who has heard even a simple discrete with say a dozen actives, isn't likely to prefer a jelly-bean monolithic over it. The situation is slightly different with a well-designed circuit using a premium opamp like an OPA627, but that's about as hard as designing a fully-discrete circuit.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.